The Philosophy of Physics

Why has philosophy evolved in the way that it has? How have its
subdisciplines developed, and what impact has this development exerted
on the way the subject is now practiced? Each volume of "The Evolution
of Modern Philosophy" will focus on a particular subdiscipline of
philosophy and examine how it has evolved into the subject as we now
understand it. The volumes will be written from the perspective of a
current practitioner in contemporary philosophy, whose point of departure
will be the question: How did we get from there to here? Cumulatively,
the series will constitute a library of modern conceptions of philosophy
and will reveal how philosophy does not in fact comprise a set of timeless
questions but has rather been shaped by broader intellectual and scientific
developments to produce particular fields of inquiry addressing particular
issues.
Roberto Torretti has written a magisterial study of the philosophy of
physics that both introduces the subject to the nonspecialist and contains
many original and important contributions for professionals in the area.
Unlike other fields of endeavor such as art, religion, or politics, all of which
preceded philosophical reflection and may well outlive it, modern physics
was born as a part of philosophy and has retained to this day a properly
philosophical concern for the clarity and coherence of ideas. Any
introduction to the philosophy of physics must therefore focus on the
conceptual development of physics itself. This book pursues that
development from Galileo and Newton through Maxwell and Boltzmann
to Einstein and the founders of quantum mechanics. There is also
discussion of important philosophers of physics in the eighteenth and
nineteenth centuries and of twentieth-century debates. In the interest of
appealing to the broadest possible readership the author avoids
technicalities and explains both the physics and the philosophical terms.

Roberto Torretti is Professor of Philosophy at the University of Chile.

" . . . a rich work full of fascinating material on the history of the

interaction between physics and philosophy."
Lawrence Sklar, author of Physics and Chance
THE EVOLUTION OF MODERN PHILOSOPHY

General Editors:
Paul Guyer and Gary Hatfield (University of Pennsylvania)

Roberto Torretti: The Philosophy of Physics

Forthcoming:
Paul Guyer: Aesthetics
Gary Hatfield: The Philosophy of Psychology
Stephen Darwall: Ethics
T. R. Harrison: Political Philosophy
William Ewald & Michael J. Hallett: The Philosophy of
Mathematics
Michael Losonsky: The Philosophy of Language
David Depew & Marjorie Grene: The Philosophy of Biology
Charles Taliaferro: The Philosophy of Religion
The Philosophy of Physics

ROBERTO TORRETTI
University of Chile

CAMBRIDGE
UNIVERSITY PRESS
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge, United Kingdom

CAMBRIDGE UNIVERSITY PRESS

The Edinburgh Building, Cambridge CB2 2RU, UK www.cup.cam.ac.uk
40 West 20th Street, New York, NY 10011-4211, USA www.cup.org
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
Ruiz de Alarcon 13, 28014, Madrid, Spain

© Roberto Torretti 1999

This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 1999

Typeface Sabon 10.25/13 pt. System QuarkXPress [BTS]

A catalog record for this book is available from the British Library.

Library of Congress Cataloging-in-Publication Data is available.

0 521 56259 7 hardback

0 521 56571 5 paperback

Transferred to digital printing 2004

For Carla
 Contents

Preface xiii

The Transformation of Natural Philosophy in the

Seventeenth Century 1
1.1 Mathematics and Experiment 2
1.2 Aristotelian Principles 8
1.3 Modern Matter 13
1.4 Galileo on Motion 20
1.5 Modeling and Measuring 30
1.5.1 Huygens and the Laws of Collision 30
1.5.2 Leibniz and the Conservation of "Force" 33
1.5.3 Romer and the Speed of Light 36
Newton 41
2.1 Mass and Force 42
2.2 Space and Time 50
2.3 Universal Gravitation 57
2.4 Rules of Philosophy 69
2.5 Newtonian Science 75
2.5.1 The Cause of Gravity 75
2.5.2 Central Forces 80
2.5.3 Analytical Mechanics 84
Kant 97
3.1 Leibniz and Berkeley on the Scope of
Mathematical Physics 98
3.1.1 The Identity of Indiscernibles 98
3.1.2 Mentalism and Positivism 101
3.2 Kant's Road to Critical Philosophy 104

IX
 Contents

3.3 Kant on Geometry, Space, and Quantity 113

3.4 The Web of Nature 120
3.4.1 Necessary Connections 120
3.4.2 Conservation of Matter 122
3.4.3 Causality 128
3.4.4 Interaction 134
3.5 The Ideas of Reason and the Advancement
of Science 138
The Rich Nineteenth Century 147
4.1 Geometries 147
4.1.1 Euclid's Fifth Postulate and Lobachevskian
Geometry 147
4.1.2 The Proliferation of Geometries and Klein's
Erlangen Program 152
4.1.3 Riemann on the Foundations of Geometry 157
4.2 Fields 168
4.3 Heat and Chance 180
4.3.1 Heat as Motion 181
4.3.2 The Concept of Entropy 187
4.3.3 Molecular Chances 195
4.3.4 Time-Reversible Laws for Time-Directed
Phenomena? 205
4.4 Philosophers 215
4.4.1 William Whewell (1794-1866) 216
4.4.2 Charles Sanders Peirce (1839-1914) 222
4.4.3 Ernst Mach (1838-1916) 234
4.4.4 Pierre Duhem (1861-1916) 242
Relativity 249
5.1 Einstein's Physics of Principles 250
5.2 Minkowski's Spacetime 260
5.3 Philosophical Problems of Special Relativity 271
5.3.1 The Length of a Moving Rod 271
5.3.2 Simultaneity in a Single Frame 273
5.3.3 Twins Who Differ in Age 277
5.3.4 Kinematical Determinism 280
5.3.5 The Quantities We Call 'Mass' 283
5.4 Gravitation as Geometry 289
5.5 Relativistic Cosmology 299
 Contents xi

Quantum Mechanics 307

6.1 Background 309
6.1.1 The Old Quantum Theory 309
6.1.2 Einstein on the Absorption and Emission
of Radiation 313
6.1.3 Virtual Oscillators 316
6.1.4 On Spin, Statistics, and the Exclusion
Principle 318
6.2 The Constitution of Quantum Mechanics 321
6.2.1 Matrix Mechanics 321
6.2.2 Wave Mechanics 325
6.2.3 The Equivalence of Matrix and Wave
Mechanics 329
6.2.4 Interpretation 331
6.2.5 Quantum Mechanics in Hilbert Space 336
6.2.6 Heisenberg's Indeterminacy Relations 348
6.3 Philosophical Problems 349
6.3.1 The EPR Problem 349
6.3.2 The Measurement Problem 355
6.4 Meta-Physical Ventures 367
6.4.1 Complementarity 368
6.4.2 Hidden Variables 373
6.4.3 Quantum Logic 378
6.4.4 "Many Worlds" 387
6.5 A Note on Relativistic Quantum Theories 393
Perspectives and Reflections 398
7.1 Physics and Common Sense 398
7.2 Laws and Patterns 405
7.3 Rupture and Continuity 420
7.4 Grasping the Facts 431
Supplements 443
I On Vectors 443
II On Lattices 453
III Terms from Topology 455

References 458

Index 493
 Preface

Like other volumes in "The Evolution of Modern Philosophy" series,

this book is meant to introduce the reader to a field of contemporary
philosophy - in this case, the philosophy of physics - by exploring its
sources from the seventeenth century onward. However, while the
modern philosophies of art, language, politics, religion, and so on seek
to elucidate manifestations of human life that are much older and prob-
ably will last much longer than the philosophical will for lucidity, the
modern philosophy of physics has to do with modern physics, an intel-
lectual enterprise that began in the seventeenth century as a central
piece of philosophy itself. The theory and practice of physics is firmly
rooted in that origin, despite substantial changes in its informational
contents, conceptual framework, and explicit aims. A vein of philos-
ophical thinking about the phenomena of nature runs through the four-
century-old tradition of physics and holds it together. This philosophy
in physics carries more weight in the book than the reflections about
physics conducted by philosophers. Our study of the evolution of the
modern philosophy of physics will therefore pay much attention to the
conceptual development of physics itself.
The book is divided into seven chapters. The purport and motiva-
tion of the first six are summarily described in the short introductions
that precede them. The seventh and last chapter - "Perspectives and
Reflections" - does not have an introduction, so I shall say something
about it here. I had planned to close the book with a survey of current
debate on the philosophy of physics in general (beyond the special
philosophical problems of relativity and quantum mechanics studied in
Chapters Five and Six). But the series editors asked me to give instead
my own vision of the subject. Now, my imagination is too weak to
encompass a vision of anything so vast, so I sketch instead what I

Xlll
xiv Preface

regard as a coherent way of tackling the main issues. I believe that to

do this ought to be more fruitful and will agree better with the con-
temporary spirit of philosophy than to erect some new idol of the
forum for others to practice their markmanship on.
It is a welcome feature of contemporary societies that educated
people have very different educational backgrounds. However, it makes
it difficult to find a common denominator of prerequisites for poten-
tial readers of a book like this one. I assume that:

(a) The readers will know the names of great philosophers, such as
Descartes, Spinoza, and Kant, and will be vaguely acquainted with
some philosophical ideas, such as mind-body dualism, but, for the
most part, they will have no professional training in philosophy. I
have therefore avoided philosophical jargon and explained all
essential philosophical notions.
(b) They are interested in physics and have a good recollection of high-
school physics. Some college physics will make many things easier
to understand, but it is not indispensable. Previous acquaintance
with popular and semipopular books on twentieth-century physics
can also be useful.
(c) They enjoyed their high-school mathematics and remember the gist
of it; or they have later developed a taste for it and studied it again.
I take this to include elementary Euclidean geometry, high-school
algebra, and the rudiments of calculus. Mathematics beyond this
level is needed only in §4.1.3 on Riemannian geometry; §§5.4 and
5.5 on general relativity and relativistic cosmology; and §§6.2, 6.3,
and 6.4 on quantum mechanics. This is supplied in the Supplements
at the end of the book and in some footnotes to §4.1.3. They are
written in the standard prose style of mathematical textbooks and
probably will be inaccessible to someone wholly unacquainted with
this form of English. Like all idiolects, this one can only be acquired
by practice, for example, by taking a good undergraduate course
in modern algebra. Readers who find that they cannot understand
the Supplements should just omit the sections listed above; they can
also omit §2.5.3, "Analytical Mechanics", which mainly serves as
an antecedent to §6.2.
(d) Except in §6.4.3, on quantum logic, readers are not required
to know any formal logic. However, philosophy students who
have taken a couple of courses in this area will - I expect - be
enabled thereby to read and understand the mathematical
supplements.
 Preface xv

References are usually given by author name and publication year.

Multiple works published by the same author in the same year are dis-
tinguished by lowercase letters. The choice of letters is arbitrary, except
in the case of Einstein, in which, for papers published before 1920, I
follow the lettering of the Collected Papers. In a few cases - usually
"collected writings" - in which the publication year would be unin-
formative, I use acronyms (mostly standard ones). All coded references
are decoded in the Reference list at the end of the book.
Translations from other languages are mine unless otherwise noted.
The English translations that I have consulted (e.g., of Kant) are men-
tioned in the Reference list. In translations from continental languages,
I treat Nature and Reason as feminine, when this use of gender con-
tributes to dispel ambiguities.

I warmly thank the friends and colleagues who have assisted me

with advice, comments, preprints, offprints, and photocopies while I
wrestled with the book: Juan Arana, Harvey Brown, Jeremy Butter-
field, Werner Diedrich, John Earman, Bruno Escoubes, Miguel
Espinoza, Alfonso Gomez-Lobo, Gary Hatfield, Christian Hermansen,
Bernulf Kanitscheider, David Malament, Deborah Mayo, Jesus
Mosterin, Ulises Moulines, Michel Paty, Massimo Pauri, Michael
Redhead, and Dudley Shapere. I acknowledge with special gratitude
the contribution of Francisco Claro, who detected a serious error in
the first version of §6.3.2, on the problem of measurement in quantum
mechanics. This led me to reformulate that section in a way that cor-
rected the error and placed the whole matter in what I think is a better
light. Of course, if new errors turn up in the new version, Dr. Claro
bears no responsibility for them.
I also owe thanks to the University of Puerto Rico, where I taught
until 1995. Although the book was written after my retirement, it is
based on reading done while I was there using its library facilities and
the free time that the University generously granted to me for research.
A shorter Spanish version of §3.4 was published in 1996 as "Las
analogias de la experiencia de Kant y la filosofia de la fisica" in Anales
de la Universidad de Chile. Parts of Chapter Seven were included, in
Spanish translation, in my paper "Ruptura y continuidad en la histo-
ria de la fisica", which appeared in 1997 in Revista de Filosofia (Uni-
versidad de Chile). I thank the editors of these journals for permission
to use those materials here.
xvi Preface

I am very grateful to Alexis Ruda, Gwen Seznek, and Rebecca

Obstler, of Cambridge University Press, for promptly replying to all my
queries and requests while the book was been written and produced,
and to Elise Oranges for her careful and accurate editing.
As in everything I have done during my adult life, my greatest debt
is to Carla Cordua. It is with great joy that I dedicate to her this, my
latest book, just as I did the first.

Santiago de Chile, 25 December 1998

 CHAPTER ONE

The Transformation of Natural Philosophy in

the Seventeenth Century

Physics and philosophy are still known by the Greek names of the
Greek intellectual pursuits from which they stem. However, in the
seventeenth century they went through deep changes that have condi-
tioned their further development and interaction right to the present
day. In this chapter I shall sketch a few of the ideas and methods that
were introduced at that time by Galileo, Descartes, and some of their
followers, emphasizing those aspects that I believe are most significant
for current discussions in the philosophy of physics.
Three reminders are in order before taking up this task.
First, in the Greek tradition, physics was counted as a part of phi-
losophy (together with logic and ethics, in one familiar division of it)
or even as the whole of philosophy (in the actual practice of "the first
to philosophize" in Western Asia Minor and Sicily). Philosophy was
the grand Greek quest for understanding everything, while physics or
"the understanding of nature (physis)" was, as Aristotle put it, "about
bodies and magnitudes and their affections and changes, and also
about the sources of such entities" (De Caelo, 268 a l-4). For all their
boasts of novelty, the seventeenth-century founders of modern physics
did not dream of breaking this connection. While firmly believing that
nature, in the stated sense, is not all that there is, their interest in it
was motivated, just like Aristotle's, by the philosophical desire to
understand. And so Descartes compared philosophy with a tree whose
trunk is physics; Galileo requested that the title of Philosopher be
added to that of Mathematician in his appointment to the Medici court;
and Newton's masterpiece was entitled Mathematical Principles of
Natural Philosophy. The subsequent divorce of physics and philoso-
phy, with a distinct cognitive role for each, although arguably a direct
consequence of the transformation they went through together in the

1
2 Natural Philosophy in the Seventeenth Century

seventeenth century, was not consummated until later, achieving its

classical formulation and justification in the work of Kant.
Second, some of the new ideas of modern physics are best explained
by taking Aristotelian physics as a foil. This does not imply that the
Aristotelian system of the world was generally accepted by European
philosophers when Galileo and Descartes entered the lists. Far from it.
The Aristotelian style of reasoning was often ridiculed as sheer ver-
biage. And the flourishing movement of Italian natural philosophy was
decidedly un-Aristotelian. But the physics and metaphysics of Aristo-
tle, which had been the dernier cri in the Latin Quarter of Paris c. 1260,
although soon eclipsed by the natively Christian philosophies of Scotus
and Ockam, achieved in the sixteenth century a surprising comeback.
Dominant in European universities from Wittenberg to Salamanca, it
was ominously wedded to Roman Catholic theology in the Council of
Trent, and it was taught to Galileo at the university in Pisa and to
Descartes at the Jesuit college in La Fleche; so it was very much in their
minds when they thought out the elements of the new physics.
Finally, much has been written about the medieval background of
Galileo and Descartes, either to prove that the novelty of their ideas
has been grossly exaggerated - by themselves, among others - or to
reassert their originality with regard to several critical issues, on which
the medieval views are invariably found wanting. The latter line of
inquiry is especially interesting, insofar as it throws light on what
was really decisive for the transformation of physics and philosophy
(which, after all, was not carried through in the Middle Ages). But here
I must refrain from following it.1

1.1 Mathematics and Experiment

The most distinctive feature of modern physics is its use of mathematics
and experiment, indeed its joint use of them.
A physical experiment artificially produces a natural process under
carefully controlled conditions and displays it so that its development

1
The medieval antecedents of Galileo fall into three groups: (i) the statics of Jordanus
Nemorarius (thirteenth century); (ii) the theory of uniformly accelerated motion devel-
oped at Merton College, Oxford (fourteenth century); and (iii) the impetus theory of
projectiles and free fall. All three are admirably explained and documented in Claggett
(1959a). Descartes's medieval background is the subject of two famous monographs
by Koyre (1923) and Gilson (1930).
 1.1 Mathematics and Experiment 3

can be monitored and its outcome recorded. Typically, the experiment

can be repeated under essentially the same conditions, or these can be
deliberately and selectively modified, to ascertain regularities and cor-
relations. Experimentation naturally comes up in some rough and ready
way in every practical art, be it cooking, gardening, or metallurgy, none
of which could have developed without it. We also have some evidence
of Greek experimentation with purely cognitive aims. However, one of
our earliest testimonies, which refers to experiments in acoustics, con-
tains a jibe at those who "torture" things to extract information from
them.2 And the very idea of artificially contriving a natural process is a
contradiction in terms for an Aristotelian. This may help to explain why
Aristotle's emphasis on experience as the sole source of knowledge did
not lead to aflourishingof experiment, although some systematic exper-
imentation was undertaken every now and then in late Antiquity and
the Middle Ages (though usually not in Aristotelian circles).
Galileo, on the other hand, repeatedly proposes in his polemical
writings experiments that, he claims, will decide some point under dis-
cussion. Some of them he merely imagined, for if he had performed
them, he would have withdrawn his predictions; but there is evidence
that he did actually carry out a few very interesting ones, while there
are others so obvious that the matter in question gets settled by merely
describing them. Here is an experiment that Galileo says he made. Aris-
totelians maintained that a ship will float better in the deep, open sea
than inside a shallow harbor, the much larger amount of water beneath
the ship at sea contributing to buoy it up. Galileo, who spurned the
Aristotelian concept of lightness as a positive quality, opposed to heav-
iness, rejected this claim, but he saw that it was not easy to refute it
by direct observation, due to the variable, often agitated condition of
the high seas. So he proposed the following: Place a floating vessel in
a shallow water tank and load it with so many lead pellets that it will
sink if one more pellet is added. Then transfer the loaded vessel to
another tank, "a hundred times bigger", and check how many more
pellets must be added for the vessel to sink.3 If, as one readily guesses,
the difference is 0, the Aristotelians are refuted on this point.

2
Plato (Republic, 537d). The verb potaocvi^eiv used by Plato means 'to test, put to the
question', but was normally used of judicial questioning under torture. The acoustic
experiments that Plato had in mind consisted in tweaking strings subjected to varying
tensions like a prisoner on a rack.
3
Benedetto Castelli (Risposta alle opposizioni, in Galileo, EN IV, 756).
4 Natural Philosophy in the Seventeenth Century

Turning now to mathematics, I must emphasize that both its scope

and our understanding of its nature have changed enormously since
Galileo's time. The medieval quadrivium grouped together arithmetic,
geometry, astronomy, and music, but medieval philosophers defined
mathematics as the science of quantity, discrete (arithmetic) and con-
tinuous (geometry), presumably because they regarded astronomy and
music as mere applications. Even so, the definition was too narrow,
for some of the most basic truths of geometry - for example, that a
plane that cuts one side of a triangle and contains none of its vertices
inevitably cuts one and only one of the other two sides - have precious
little to do with quantity. In the centuries since Galileo mathematics has
grown broader and deeper, and today no informed person can accept
the medieval definition. Indeed, the wealth and variety of mathematical
studies have reached a point in which it is not easy to say in what sense
they are one. However, for the sake of understanding the use of mathe-
matics in modern physics, it would seem that we need only pay atten-
tion to two general traits. (1) Mathematical studies proceed from
precisely defined assumptions and figure out their implications, reach-
ing conclusions applicable to whatever happens to meet the assump-
tions. The business of mathematics has thus to do with the construction
and subsequent analysis of concepts, not with the search for real
instances of those concepts. (2) A mathematical theory constructs and
analyzes a concept that is applicable to any collection of objects, no
matter what their intrinsic nature, which are related among themselves
in ways that, suitably described, agree with the assumptions of the
theory. Mathematical studies do not pay attention to the objects them-
selves but only to the system of relations embodied in them. In other
words, mathematics is about structure, and about types of structure.4
With hindsight we can trace the origin of structuralist mathematics
to Descartes's invention of analytic geometry. Descartes was able to
solve geometrical problems by translating them into algebraic equa-
tions because the system of relations of order, incidence, and congru-
ence between points, lines, and surfaces in space studied by classical
geometry can be seen to be embodied - under a suitable interpretation
- in the set of ordered triples of real numbers and some of its subsets.
The same structure - mathematicians say today - is instantiated by geo-

4
For two recent, mildly different, philosophical elaborations of this idea see Shapiro
(1997) and Resnik (1997).
 1.1 Mathematics and Experiment 5

metrical points and by real number triples. The points can be put - in
many ways - into one-to-one correspondence with the number triples.
Such a correspondence is known as a coordinate system, the three
numbers assigned to a given point being its coordinates within the
system. For example, we set up a Cartesian coordinate system by arbi-
trarily choosing three mutually perpendicular planes K, L, M; a given
point O is assigned the coordinates (a,b,c) if the distances from O to
K, L, and M are, respectively, \a\9 \b\, and |c|, the choice of positive or
negative a (respectively, b, c) being determined conventionally by the
side of K (respectively, L, M) on which O lies. The origin of the coor-
dinate system is the intersection of K, L, and M, that is, the point with
coordinates (0,0,0). The intersection of L and M is known as the x-
axis, because only the first coordinate - usually designated by x - varies
along it, while the other two are identically 0 (likewise, the y-axis is
the intersection of K and M, and the z-axis is the intersection of K and
L). The sphere with center at O and radius r is represented by the set
of triples (x,y,z) such that (x - a)2 + (y - b)1 + (z - c)2 = r2; thus, this
equation adequately expresses the condition that an otherwise arbi-
trary point - denoted by (x,y,z) - lies on the sphere (O,r).
By paying attention to structural patterns rather than to particular-
ities of contents, mathematical physics has been able to find affinities
and even identities where common sense could only see disparity, the
most remarkable instance of this being perhaps Maxwell's discovery
that light is a purely electromagnetic phenomenon (§4.2). A humbler
but more pervasive and no less important expression of structuralist
thinking is provided by the time charts that nowadays turn up every-
where, in political speeches and business presentations, in scientific
books and the daily press. In them some quantity of interest is plotted,
say, vertically, while the horizontal axis of the chart is taken to repre-
sent a period of time. This representation assumes that time is, at least
in some ways, structurally similar to a straight line: The instants of
time are made to correspond to the points of the line so that the rela-
tions of betweenness and succession among the former are reflected by
the relations of betweenness and being-to-the-right-of among the latter,
and so that the length of time intervals is measured in some conven-
tional way by the length of line segments.
Such a correspondence between time and a line in space is most nat-
urally set up in the very act of moving steadily along that line, each
point of the latter corresponding uniquely to the instant in which the
6 Natural Philosophy in the Seventeenth Century

mobile reaches it. This idea is present already in Aristotle's rebuttal of

Zeno's "Dichotomy" argument against motion. Zeno of Elea claimed
that an athlete could not run across a given distance, because before
traversing any part of it, no matter how small, he would have to tra-
verse one half of that part. Aristotle's reply was - roughly paraphrased
- that if one has the time t to go through the full distance d one also
has the time to go first through 1/2 d, namely, the first half of t (Phys.
233a21 ss.). In fact, Zeno himself had implicitly mapped time into space
- that is, he had assigned a unique point of the latter to each instant
of the former - in the "Arrow", in which he argues that a flying arrow
never moves, for at each instant it lies at a definite place. Zeno's
mapping is repeated every minute, hour, and half-day on the dials of
our watches by the motion of the hands, and it is so deeply ingrained
in our ordinary idea of time that we tend to forget that time, as we
actually live it, displays at least one structural feature that is not
reflected in the spatial representation, namely, the division between past
and future. (Indeed, some philosophers have brazenly proclaimed that
this division is "subjective" - by which they mean illusory - so one
would do well to forget i t . . . if one can.)
There is likewise a structural affinity between all the diverse kinds
of continuous quantities that we plot on paper. Descartes was well
aware of it. He wrote that "nothing is said of magnitudes in general
which cannot also be referred specifically to any one of them," so that
there "will be no little profit in transferring that which we understand
to hold of magnitudes in general to the species of magnitude which is
depicted most easily and distinctly in our imagination, namely, the real
extension of body, abstracted from everything else except its shape"
(AT X, 441). Once all sorts of quantities are represented in space, it is
only natural to combine them in algebraic operations such as those that
Descartes defined for line segments.5 Mathematical physics has been
doing it for almost four centuries, but it is important to realize that at
one time the idea was revolutionary. The Greeks had a well-developed
calculus of proportions, but they would not countenance ratios
between heterogeneous quantities, say, between distance and time, or
between mass and volume. And yet a universal calculus of ratios would
seem to be a fairly easy matter when ratios between homogeneous

5
Everyone knows how to add two segments a and b to form a third segment a + b.
Descartes showed how to find a segment ax b that is the product of a and b: ax b must
be a segment that stands in the same proportion to a as b stands to the unit segment.
 1.1 Mathematics and Experiment 7

quantities have been formed. For after all, even if you only feel free to
compare quantities of the same kind, the ratios established by such
comparisons can be ordered by size, added and multiplied, and com-
pared with one another as constituting a new species of quantity on
their own. Thus, if length b is twice length a and weight w is twice
weight f, then the ratio bla is identical with the ratio w/v and twice
the ratio w/(v + v). Euclid explicitly equated, for example, the ratio of
two areas to a ratio of volumes and also to a ratio of lengths (Bk. XI,
Props. 32, 34), and Archimedes equated a ratio of lengths with a ratio
of times {On Spirals, Prop. I). Galileo extended this treatment to speeds
and accelerations. In the Discorsi of 1638 he characterizes uniform
motion by means of four "axioms". Let the index i range over {1,2}.
We denote by s, the space traversed by a moving body in time U and
by Vi the speed with which the body traverses space st in a fixed time.
The body moves with uniform motion if and only if (i) Si > s2 if
t\ > t2, (ii) tx > t2 if Si > s2, (iii) Si > s2 if V\ > v2, and (iv) vx > v2 if Si >
s2. From these axioms Galileo derives with utmost care a series of rela-
tions between spaces, times, and speeds, culminating in the statement
that "if two moving bodies are carried in uniform motion, the ratio of
their speed will be the product of the ratio of the spaces run through
and the inverse ratio of the times", which, if we designate the quanti-
ties concerning each body respectively by primed and unprimed letters,
we would express as follows:

-=f-Y- (1.1)
v' Wit
By taking the reciprocal value of the ratio of times, the ratio of speeds
can also be expressed as a ratio of ratios:

If we now assume that the body to which the primed quantities refer
moves with unit speed, traversing unit distance in unit time, eqn. (1.1.)
can be rewritten as:

r(i)Ai)
which, except for the pedantry of writing down the l's, agrees with the
familiar schoolbook definition of constant or average speed.
8 Natural Philosophy in the Seventeenth Century

1.2 Aristotelian Principles

The most striking difference between the modern view of nature and
Aristotle's lies in the separation he established between the heavens and
the region beneath the moon. While everything in the latter ultimately
consists of four "simple bodies" - fire, air, water, earth - that change
into one another and into the wonderful variety of continually chang-
ing organisms, the heavens consist entirely of aether, a simple body that
is very different from the other four, which is capable of only one sort
of change, viz., circular motion at constant speed around the center of
the world. This mode of change is, of course, minimal, but it is inces-
sant. The circular motion of the heavens acts decisively on the sublu-
nar region through the succession of night and day, the monthly lunar
cycle, and the seasons, but the aether remains immune to reactions
from below, for no body can act on it.
This partition of nature, which was cheerfully embraced by medieval
intellectuals like Aquinas and Dante, ran against the grain of Greek
natural philosophy. The idea of nature as a unitary realm of becom-
ing, in which everything acted and reacted on everything else under
universal constraints and regularities, arose in the sixth century B.C.
among the earliest Greek philosophers. Their tradition was continued
still in the Roman empire by Stoics and Epicureans. Measured against
it, Aristotle's system of the world appears reactionary, a sop to popular
piety, which was deadly opposed to viewing, say, the sun as a fiery rock.
But Aristotle's two-tiered universe was nevertheless unified by deep
principles, which were cleverer and more stimulating than anything put
forward by his rivals (as far as we can judge by the surviving texts),
and they surely deserve no less credit than the affinity between Greek
and Christian folk religion for Aristotle's success in Christendom.
Galileo, Descartes, and other founding fathers of modern physics were
schooled in the Aristotelian principles, but they rejected them with sur-
prising unanimity. It will be useful to cursorily review those principles
to better grasp what replaced them.
Aristotle observes repeatedly that the verb 'to be' has several mean-
ings ("being is said in various ways" - Metaph. 1003b5, 1028a10). The
ambiguity is manifold. We have, first of all, the distinction "according
to the figures of predication" between being a substance - a tree, a
horse, a person - and being an attribute - a quality, quantity, relation,
posture, disposition, location, time, action, or passion - of substances.6
6
'Substance' translates oixrioc, a noun formed directly from the participle of the verb
eivoci, 'to be'. So a more accurate translation would be 'being, properly so called'.
 1.2 Aristotelian Principles 9

Aristotle mentions three other such spectra of meaning, but we need

only consider one of them, viz., the distinction between being actually
and being potentially. This is the key to Aristotle's understanding of
organic development, which is his paradigm of change (just as organ-
isms are his paradigm of substance - Metaph. 1032a19). Take a corn
seed. Actually it is only a small hard yellow grain. Potentially, however,
it is a corn plant. While it lies in storage the potentiality is dormant; yet
its presence can be judged from the fact that it can be destroyed, for
example, if the seed rots, or is cooked, or if an insect gnaws at it. The
potentiality is activated when the seed is sown and germinates. From
then on the seed is taken over by a process in the course of which the
food it contains, plus water and nutrients sucked up from the environ-
ment, are organized as the leaves, flowers, and ears of a corn plant. The
process is guided by a goal, which our seed inherited from its parents.
This is none other than the morphe ('form') or eidos ('species') of which
this plant is an individual realization. The form is that by virtue of which
this is a corn plant and that a crocodile. If a substance is fully and invari-
ably what it is, its form is all that there is to it. Such are the gods. But a
substance that is capable of changing in any respect is a compound of
form and matter (hyle, literally 'wood'), a term under which Aristotle
gathers everything that is actively or dormantly potential in a substance.
Only such substances can be said to have a 'nature' (physis) according
to Aristotle's definition of this term, that is, an inherent principle of
movement and rest.
Although the development of organisms obviously inspired Aristo-
tle's overall conception of change, it is not acknowledged as a distinct
type in his classification of changes. This is tailored to his figures of
predication. He distinguishes (a) the generation and destruction of sub-
stances, and (b) three types of change in the attributes of a given sub-
stance, which he groups under the name kinesis - literally, 'movement'
-, viz., (bi) alteration or change in quality, (b2) growth and wane or
change in quantity, and (b3) motion proper - phora in Greek - or
change of location. We need consider only (a) and (b3), the former
because it was believed to involve a sort of matter - in the Aristotelian
sense - that eventually came to be conceived as matter in the un-
Aristotelian modern sense and the latter because change of location
was the only kind of change that this new-fangled matter could really
undergo.

But the modern thinkers we are presently interested in were taught to say 'substance'
(Lat. substantia) for Aristotle's ovcia, so we better put up with it.
10 Natural Philosophy in the Seventeenth Century

Motion (phora) was viewed by Aristotle as one of several kinds of

movement (kinesis). Organic growth was another, somehow more
revealing, kind. He wrote that kinesis is "the actuality of potential
being as such" (Phys. 201 a ll), a definition that Descartes dismissed as
balderdash (AT X, 426; XI, 39) but that surely makes sense with regard
to a corn seed that grows into a plant when its inborn potentialities
are actualized. Movement is thus conceived by Aristotle as a way of
being. Zeno's arrow surely is at one place at any one time, but it is
moving, not resting there, for it is presently exercising its natural poten-
tiality for resting elsewhere, namely, at the center of the universe,
where, according to Aristotle, it would naturally come to stand if
allowed to fall without impediment.
I mentioned previously Aristotle's doctrine of the four simple bodies
from which everything under the moon is compounded. They are char-
acterized by their simple qualities, one from each pair of opposites,
hot/cold and wet/dry, and their simple motions, which motivate their
classification as light or heavy. Thus fire is hot, dry, and also light, in
that it moves naturally in a straight line away from the center of the
universe until it comes to rest at the boundary of the nethermost celes-
tial sphere; earth is dry, cold, and also heavy, that is, it moves natu-
rally in a straight line toward the center of the universe until it comes
to rest at it; water is wet, cold, and heavy (though less so than earth);
and air is hot, wet, and light (though less so than fire). Aristotle's notion
of heaviness and lightness can grossly account for the familiar experi-
ence of rising smoke, falling stones, and floating porous timber.7
But what about the full variety of actual motions? To cope with it,
Aristotle employs some additional notions. Although the natural
upward or downward straight-line motion of the simple elements is
inherent in their compounds, the heaviness of plants and animals -
which presumably consist of all four elements, but mostly of earth and
water - can be overcome by their supervenient forms. Thus ivy climbs

7
A reflective mind will find fault with them even at the elementary level. Imagine that
a straight tunnel has been dug across the earth from here to the antipodes. Aristotelian
physics requires that a stone dropped down this tunnel should stop dead when it
reaches the center of the universe (i.e., of the earth), even though at that moment it
would be moving faster than ever before. Albert of Saxony, who discussed this thought
experiment c. 1350, judged the Aristotelian conclusion rather improbable. He
expected the stone to go on moving toward the antipodes until it was stopped by the
downward pull toward the center of the earth (which, after the stone has passed
through it, is exerted, of course, in the opposite direction).
 1.2 Aristotelian Principles 11

walls and goats climb rocks. The simple bodies and their compounds
are also liable to forced motion (or rest) against their natures, through
being pushed/pulled (or stopped/held) by other bodies that move (or
rest) naturally. Thus a heavy wagon is forced to move forward by a
pair of oxen and a heavy ceiling is stopped from falling by a row of
standing pillars. But Aristotelian physics has a hard time with the
motion of missiles. This must be forced, for missiles are heavy objects
that usually go higher in the first stage of their motion. Yet they are
separated from the mover that originally forces them to move against
their nature. Aristotle (Phys. 266b27-267a20) contemplates two ways
of dealing with this difficulty. The first way is known as antiperistasis:
The thrown missile displaces the air in front of it, and the nimble air
promptly moves behind the missile and propels it forward and upward;
this process is repeated continuously for some time after the missile has
separated from the thrower. This harebrained idea is mentioned
approvingly in Plato's Timaeus (80al), but Aristotle wisely keeps his
distance from it. Nor does he show much enthusiasm for the second
solution, which indeed is not substantially better. It assumes that the
thrower confers a forward and upward thrust to the neighboring air,
which the latter, being naturally light, retains and communicates to
further portions of air. This air pushes the missile on and on after it is
hurled by the thrower.8
Despite its obvious shortcomings, Aristotle's theory of the natural
motions of light and heavy bodies is the source of his sole argument
for radically separating sublunar from celestial physics. It runs as
follows. Simple motions are the natural motions of simple bodies.
There are two kinds of simple motions, viz., straight and circular. But
all the simple bodies that we know from the sublunar region move nat-
urally in a straight line. Therefore, there must be a simple body whose
natural motion is circular. Moreover, just as the four familiar simple
bodies move in straight lines to and from the center of the universe,
the fifth simple body must move in circles around that center. The
nightly spectacle of the rotating firmament lends color to this surpris-

John Philoponus, commenting on Aristotle's Physics in the sixth century A.D.,

remarked that if this theory of missile motion were right, one should be able to throw
stones most efficiently by setting a large quantity of air in motion behind them. This
is exactly what Renaissance Europe achieved with gunpowder. However, the unex-
pectedly rich and precise experience with missiles provided by modern gunnery has
not vindicated Aristotle.
12 Natural Philosophy in the Seventeenth Century

ing argument. Its conclusion agrees well, of course, with fourth-century

Greek mathematical astronomy, which analyzed the wanderings of the
sun, moon, and planets as resulting from the motion of many nested
spheres linked to one another and rotating about the center of the uni-
verse with different (constant) angular velocities.9
Changes of quality, size, or location, grouped by Aristotle under
the name of kinesis, suppose a permanent substance with varying
attributes. But Aristotelian substances also change into one another in
a process by which one substance is generated as another is being
destroyed. The generation of plants and animals can be understood as
the incorporation of a new form in a suitable combination of simple
bodies behaving pliably as matter. But the transmutation of one simple
body into another - which, according to Aristotle, occurs incessantly
in the sublunar realm - cannot be thus understood. However, the tra-
ditional reading of Aristotle assumes that in such cases the change of
form is borne by formless matter, an utterly indeterminate being that
potentially is anything and yet, despite its complete lack of definition,
ensures the numerical identity of what was there and is destroyed with
what thereupon comes into being. Recent scholars have questioned this
interpretation and the usual understanding of the Aristotelian expres-
sion 'prime matter' (prote hyle) as referring to the alleged ultimate sub-
stratum of radical transformations.10 Their view makes Aristotle into
a better philosopher than he would otherwise be, but this is quite irrel-
evant to our present study, for the founders of modern science read
Aristotle in the traditional way. Indeed, what they call 'matter' appears
to have evolved from the 'prime matter' of Aristotelian tradition in the
course of late medieval discussions. Ockam, for example, held that
prime matter, if it is at all real - as he thought it must be to account
for the facts of generation and corruption -, must in some way be
actual: "I say that matter is a certain kind of act, for matter exists in
the realm of nature, and in this sense, it is not potentially every act for

9
It is important to realize that the celestial physics of Aristotle was deeply at variance
with the more accurate system of astronomy that was later developed by Hipparchus
and Apollonius and which medieval and Renaissance Europe received through
Ptolemy. Each Ptolemaic planet (including the Sun and the Moon) moves in a circle
- the epicycle - whose center moves in another circle - the deferent - whose center
is at rest. But not even the deferent's center coincides with the Aristotelian "center of
the universe", that is, the point to or from which heavy and light bodies move nat-
urally in straight lines.
10
King (1956), Charlton (1970, appendix). For a defense of the traditional reading see
Solmsen (1958), Robinson (1974), and C. J. F. Williams (1982, appendix).
 1.3 Modern Matter 13

it is not potentially itself."11 Such matter "is the same in kind in all
things which can be generated and corrupted".12 Moreover, although
the heavenly bodies are incorruptible, Ockam was convinced that they
too were formed from that same kind of matter:
It seems to me then that the matter of the heavens is the same in kind as
that of things here below, because as has been frequently said: one must
never assume more than is necessary. Now there is no reason in this case
that warrants the postulation of a different kind of matter here and there,
because every thing explained by assuming different matters can be
equally accounted for, or better explained by postulating a single kind.13

1.3 Modern Matter

'Matter' is just the anglicized form of materia^ a Latin word meaning
'timber' that Cicero deftly chose for translating Aristotle's hyle.
'Matter' may thus in all fairness be seen as a contribution of philoso-
phy to ordinary English. Yet its everyday meaning is a far cry from
Aristotle's. In fact, the term could hardly keep its Aristotelian sense in
a Christian setting. Christian theologians cherished Plato's myth of the
divine artisan who molds matter14 as potter's clay, but their God did
not encounter matter as a coeval 'wet-nurse of becoming' (Timaeus
52d) but created it out of nothing. As an actual creature of God's will,
Christian matter cannot be purely potential and indeterminate, but
comes with all the properties required for God's purpose. Indeed, some
seventeenth-century authors thought it most fitting that the world
created by an all-knowing, all-powerful God should consist of a single
universal stuff that develops automatically into its present splendor
from a wisely chosen initial configuration, with no further intervention
on His part. Be that as it may, surely the Deity of Christian philoso-
phy knew exactly what He wanted when He created the world and
could bring forth a material thoroughly suited to His ends.
Both Plato and Aristotle held that an exact science of nature was

11
Ockam (Summulae in libros Physicorum, Pars I, cap. 16, fol. 6ra) quoted in Wolter
(1963, p. 134).
12
Ockam (Expositio super octo libros Physicorum, Lib. I, com. 1) quoted in Wolter
(1963, p. 130).
13
Ockam (Reportatio in Sentent. II, q. 22, D) quoted in Wolter (1963, p. 146) (my
italics).
14
Plato's word is %copa, which in ordinary Greek meant 'extension, room, place', and
often 'land, country'.
14 Natural Philosophy in the Seventeenth Century

precluded by matter's inherent potentiality for being otherwise. Just as

a geometer admires an excellent drawing but does not expect to estab-
lish true geometrical relations by studying it, so a "real astronomer"
will judge "that the sky and everything in it have been put together by
their maker in the most beautiful way in which such works can be put
together, but will - don't you think? - hold it absurd to believe that
the metrical relation (symmetrian) of night to day, of these to month,
and month to year, and of the other stars to these and to each other
are ever the same and do not deviate at all anywhere, although they
are corporeal and visible" (Plato, Rep. 530a-b). The predictive success
of Eudoxus's planetary models caused Plato to recant, and his
spokesman in Laws (821b) asserts that "practically all Greeks now
slander those great gods, Sun and Moon", for "we say that they and
some other stars besides them never go along the same path, and we
dub them roamers (planeta)." For Aristotle, heavenly motions are exact
because they are steered directly by gods, but even gods could not
achieve this if the heavens did not consist of aether, which admits no
change except rotation on the spot. All other matter is incapable of
such unbending regularity, and therefore sublunar events are not liable
to mathematical treatment. Therefore, according to Aristotle, physics
should not rely on geometrical notions such as 'concave', but rather
use concepts like 'snub', which is confined to noses and involves a
reference to facial flesh (Phys. 194a13; cf. De an. 431b13, Metaph.
1025b31, 1064a24, 1030b29).
The idea of created matter does away with all such limitations.
Indeed, the conception of universal matter professed with compara-
tively minor variations by Galileo, Descartes, and Newton seems
expressly designed for mathematical treatment, or, more precisely, for
treatment with the resources of seventeenth-century mathematics. The
gist of it is concisely stated by Robert Boyle: "I agree with the gener-
ality of philosophers, so far as to allow that there is one catholic or
universal matter common to all bodies, by which I mean a substance
extended, divisible, and impenetrable" (1666, in SPP, p. 18). Matter
being one, something else is required to account for the diversity we
see in bodies. However, this additional principle need not consist of
immaterial Aristotelian forms, but simply of the diverse motions that
different parts of matter have with respect to each other. As Boyle puts
it: "To discriminate the catholic matter into variety of natural bodies,
it must have motion in some or all its designable parts; and that motion
must have various tendencies, that which is in this part of the matter
 1.3 Modern Matter 15

tending one way, and that which is in that part tending another" (Ibid.).
Indeed, the actual division of matter into parts of different sizes and
shapes is "the genuine effect of variously determined motion"; and
"since experience shows us (especially that which is afforded us by
chemical operations, in many of which matter is divided into parts too
small to be singly sensible) that this division of matter is frequently
made into insensible corpuscles or particles, we may conclude that
the minutest fragments, as well as the biggest masses, of the universal
matter are likewise endowed each with its peculiar bulk and shape"
(SPP, p. 19). "And the indefinite divisibility of matter, the wonderful
efficacy of motion, and the almost infinite variety of coalitions and
structures that may be made of minute and insensible corpuscles, being
duly weighed, I see not why a philosopher should think it impossible
to make out, by their help, the mechanical possibility of any corporeal
agent, how subtle or diffused or active soever it be, that can be solidly
proved to be really existent in nature, by what name soever it be called
or disguised" (Boyle 1674, in SPP, p. 145).
This view justifies Galileo's assertion that the universe is like a book
open in front of our eyes in which anyone can read, provided that she
or he understands the "mathematical language" in which it is written
- for "its characters are triangles, circles, and other geometric figures
without which it is humanly impossible to understand a single word
of it" (1623, §6; EN VI, 232). It also implies the notorious distinction
between the inherent "primary" qualities of bodies, viz., number,
shape, motion, and their mind-dependent "secondary" qualities, viz.,
all the more salient features they display to our senses.15 As Galileo
explains further on in the same book:

15
The distinction can be traced back to Democritus's dictum "By custom, sweet; by
custom, bitter. By custom, hot; by custom, cold. By custom, color. In truth: atoms
and void" (DK 68.B.9). But there are deep conceptual differences between ancient
atoms and modern matter. Greek atomism is a clever and imaginative reply to Eleatic
ontology: Being cannot change, but if allowance is made for Non-Being in the guise
of the void, there is room for plurality and motion, and this is enough to account for
the variety of appearances. Each atom is indivisible (a-tomos) precisely because it is
a specimen of Parmenides's changeless Being. Modern matter is not subject to such
ontological constraints. Descartes explicitly rejects atoms and denies the possibility
of a void. Even Boyle, who invested much ingenuity and effort in pumping air out
of bottles, was not committed to the existence of a true vacuum absolutely devoid of
matter of any sort. And, although Boyle believed that matter is stably divided into
very little bodies, he did not think that these corpuscles were indivisible in principle.
16 Natural Philosophy in the Seventeenth Century

As soon as I conceive a matter or corporeal substance I feel compelled

to think as well that it is bounded and shaped with this or that figure,
that it is big or small in relation to others, that it is in this or that place,
in this or that time, that it moves or rests, that it touches or does not
touch another body, that it is one, or few, or many; and I cannot sepa-
rate it from these conditions by any stretch of the imagination. But that
it must be white or red, bitter or sweet, sonorous or silent, of pleasant
or unpleasant smell, I do not feel my mind constrained to grasp it as nec-
essarily attended by such conditions. Indeed, discourse and sheer imag-
ination would perhaps never light on them, if not guided by the senses.
Which is why I think that tastes, odors, colors, etc. are nothing but
names with regard to the object in which they seem to reside, but have
their sole residence in the sensitive body, so that if the animal is removed
all such qualities are taken away and annihilated. But since we have
bestowed on them special names, different from those of the other
primary and real attributes, we wish to believe that they are also truly
and really different.
(Galileo 1623, §48; EN VI, 347-48)

Less notorious but no less remarkable is the similarity between

Galileian matter and the Aristotelian aether: Both are imperishable and
unalterable, and capable only of change by motion. (Indeed, Galileo
held at times that the natural motion of all matter is circular, though not
indeed about the center of the universe.) Looking at a museum exhibit
of lunar rocks we admire Galileo's hunch that they would turn out to
be just sublunar stuff. But the very familiarity of that sight may cause
us to overlook the main drift of his work. His aim was not to conquer
heaven for terrestrial physics (a dismal prospect, given the state of the
latter c. 1600), but rather to apply right here on earth exact mathemat-
ical concepts and methods such as those employed successfully in
astronomy. The modern concept of matter made this viable.
The modern concept of matter also conferred legitimacy on experi-
mental inquiry in the manner I shall now explain. If natural change
involves the supervenience and operation of "forms" that scientists do
not know, let alone control, and there is moreover an essential dis-
tinction between natural and forced changes, it is highly questionable
that one can learn anything about natural processes by experiment.
But if all bodies consist of a single uniform and changeless stuff,
and all variety and variation results exclusively from the motion and
reconfiguration of its parts and particles, the distinction between
natural and forced changes cannot amount to much. If all that ever
happens in the physical world is that this or that piece of matter
 1.3 Modern Matter 17

changes its position and velocity, a scientist's intervention can only

produce divisions and displacements such as might also occur without
him. Experiments can be extremely helpful for studying - and master-
ing - the ways of nature because they are just a means of achieving
faster, more often, and under human control, changes of the only kind
that matter allows, viz., by "local motion, which, by variously divid-
ing, sequestering, transposing, and so connecting, the parts of matter,
produces in them those accidents and qualities upon whose account
the portion of matter they diversify comes to belong to this or that
determinate species of natural bodies" (Boyle 1666, in SPP, p. 69).
The most resolute and forceful spokesman for the modern idea of
matter was Descartes (1641, 1644). He asked himself what constitutes
the reality of any given body, for example, a piece of wax fresh from
the honeycomb - not its color, nor its smell, nor its hardness, nor even
its shape, for all these are soon gone if the piece of wax is heated, and
yet the piece remains. But, says Descartes, when everything that does
not belong to it is removed and we see what is left, we find "nothing
but something extended, flexible, mutable" (AT VII, 31). Indeed,
"extension in length, breadth and depth", its division into parts, and
the number, sizes, figures, positions, and motions of these parts (AT
VII, 50) are all that we can clearly and distinctly conceive in bodies
and therefore provide the entire conceptual stock of physics. Obviously,
motion is the sole idea that Cartesian physics adds to Cartesian geom-
etry. Moreover, it is defined by Descartes in geometric terms:

Motion as ordinarily understood is nothing but the action by which a

body goes from one place to another. [. . .] But if we consider what must
be understood by motion in the light not of ordinary usage, but of the
truth of the matter, we can say that it is the transport of one part of
matter or of one body, from the vicinity of those bodies which are imme-
diately contiguous to it, and are regarded as being at rest, to the neigh-
borhood of others. By one body or one part of matter I understand all
that is transported together, even if this, in turn, consists perhaps of many
parts which have other motions. And I say that motion is the transport,
not the force or action which transports, to indicate that motion is
always in the mobile, not in the mover [. . .]; and that it is a property
(modus) of it, and not a thing that subsists by itself; just as shape is a
property of the thing shaped and rest of the thing at rest.
(Descartes 1644, II, arts. 24, 25; AT VIII, 53-54)

Matter as extension being naturally inert, the property of motion is

bestowed on several parts of it by God; indeed the actual division of
18 Natural Philosophy in the Seventeenth Century

matter into distinct bodies is a consequence of the diverse motions

of its different parts. While Descartes is emphatic that motion is just
change of relative position, he was well aware that in collisions it
behaves like an acting force. To account for this, Descartes developed
the concept of a quantity of motion, which resides in the moving body
and is transferred from it to the bodies it collides with according to
fixed rules. According to Descartes, the immutability of God requires
that the quantity of motion He conferred on material things at creation
should remain the same forever. Descartes computes the quantity of
motion of a given body by multiplying its speed by its quantity of
matter. This notion led to the classical mechanical concept of momen-
tum {mass x velocity), so I shall call it Cartesian momentum. Two
important differences must be emphasized: (i) If extension is the sole
attribute of matter, the quantity of matter can only be measured by its
volume, so there is no room in Cartesian physics for a separate concept
of mass, (ii) Cartesian momentum is the product of the quantity of
matter by (undirected) speed, not by (directed) velocity, so, in contrast
with classical momentum, it is a scalar, not a vector. This raises ques-
tions to which I now turn.
(i) A ball of solid gold can cause much more damage on impact that
a ball of cork of the same size moving with the same speed. How does
Cartesian physics cope with this fact? If matter coincides with exten-
sion, there are no empty interstices in the cork. However, the quantity
of motion borne by either ball depends on their respective quantity of
matter, that is, the volume of all the matter that moves together, and
within the outer limits of each ball there are interstices filled with
matter that does not move with the rest. This Cartesian solution is
quaint, for one normally expects a moving sponge to drag the air in
its pores, but it is not altogether absurd.
(ii) The only principle of Cartesian physics that still survives is the
principle of inertia: A moving body, if not impeded, will go on moving
with the same speed in the same direction. Here we have a universal
tendency that - one would think - underlies the conservation of motion
in a system of two or more bodies that impede each other by collision.
Now, if direction is one of the main determinants of the persistent
motion contributed by each colliding body, why did Descartes exclude
it from his definition of the quantity of motion conserved in the system?
This is not an easy question, as we shall now see.
The principle of inertia is embodied in two "natural laws" that
Descartes derives "from God's immutability":
 1.3 Modern Matter 19

The first law is: Each thing, in so far as it is simple and undivided,
remains by itself always in the same state, and never changes except
through external causes. Thus if a piece of matter is square, we shall
easily persuade ourselves that it will remain square for ever, unless some-
thing comes along from elsewhere which changes its shape. If it is at rest,
we do not believe it will ever begin to move, unless impelled by some
cause. Nor is there any more reason to think that if a body is moving it
will ever interrupt its motion out of its own initiative and when nothing
else impedes it.

The second natural law is: each part of matter considered by itself does
not tend to proceed moving along slanted lines, but only in straight lines.
[. . .] The reason for this rule, like that for the preceding one, is the
immutability and simplicity of the operation by which God conserves
motion in matter. For He conserves it precisely as it is at the moment
when He conserves it, without regard to what it was a little earlier.
Although no motion can take place in an instant, it is nevertheless
evident that every thing that moves, at every instant which can be indi-
cated while it moves, is determined to continue its motion in a definite
direction, following a straight line, not any curved line.
(Descartes 1644, II, arts. 37, 39; AT VIII, 63-64)
With hindsight we scoff at Descartes for overlooking that an instant
tendency to move in a particular direction may well be coupled with
an instant tendency to change that direction in a given direction and
still with a third tendency to change the direction of change, and so
on, so that any spatial trajectory could result from a suitable combi-
nation of such directed quantities.16 But this does not detract from the
novelty and significance of his insight: Although motion cannot be
carried out in an instant, it can exist at an instant, not as "the actual-
ity of potential being" (whatever this might mean), but as a fully real
directed quantity. Still, how could this insight be entirely forgotten in
the definition of Cartesian momentum? It is true that vector algebra
and analysis are creatures of the nineteenth century. But the addition

16
For example, a particle moving at instant t with unit velocity v(t) in a particular direc-
tion can also be endowed at that instant with, say, unit acceleration v'(£). If v'(£) is
perpendicular to \(t) and its rate of change v"(£) = 0, the particle moves with uniform
speed on a circle of unit radius. In Newtonian dynamics, acceleration is always due
to external forces and the Principle of Inertia is preserved, but this is not the outcome
of a logical necessity, let alone a theological one, as Descartes claims (see §2.1).
20 Natural Philosophy in the Seventeenth Century

of directed quantities by the parallellogram rule dates at least from the

sixteenth century, and Descartes used it in his Dioptrique and subse-
quently discussed it at length with Fermat in 1637 in correspondence
mediated by Mersenne (AT I, 357-59, 451-52, 464-74).17
Why then did he not resort to it for adding the motions of collid-
ing bodies? In §1.5 we shall consider some devastating criticism of
Cartesian physics by Leibniz and Huygens, which ultimately results
from this omission. Some scholars think that Descartes could not
combine motions by the parallellogram rule because he shared the
Aristotelian belief that "each individual body has only one motion
which is peculiar to it". 18 1 cannot go further into this matter here, but
there is one interesting consequence of the definition of Cartesian
momentum as a scalar that I must mention. According to Descartes the
human mind is able to modify - he does not say how - the direction
of motion of small particles in the pineal gland although it cannot alter
their quantity of motion. This ensures that a person's behavior can
depend on her free will. This escape provision for human freedom is
not available if the unalterable quantity of motion is a vector instead
of a scalar.

1.4 Galileo on Motion

Galileo was 32 years older than Descartes and was already philo-
sophizing about motion when the latter was born in 1596. Galileo's
early writings criticize some Aristotelian tenets and show the influence
of the impetus theory. According to this view, which was fathered by
John Philoponus in late Antiquity and revived and further elaborated
in the fourteenth century, the violent motion of missiles continues after
they separate from the mover because the initial thrust impresses

17
By this rule, if v and w are two directed quantities represented by arrows with a
common origin p, their sum v + w is a directed quantity represented by an arrow from
p to the opposite vertex of the parallellogram formed by the arrows representing v and
w. Compare Newton's rule for the composition of forces, illustrated in Fig. 7 (§2.2).
18
Descartes (1644, II art. 31), as cited in Damerow et al. (1992, p. 105). Taken in
context, the passage does not, in my view, seem to support their opinion. Descartes
wrote: "Etsi autem unumquodque corpus habeat tantum unum motum sibi proprium,
quoniam ab unis tantum corporibus sibi contiguis et quiescentibus recedere intellig-
itur, participare tamen etiam potest ex aliis innumeris, si nempe sit pars aliorum cor-
porum alios motus habentium" (AT VII, 57; I have italicized the sentence quoted by
Damerow et al.).
 1.4 Galileo on Motion 21

a force or "impetus" on the missile that keeps it going until it gradu-

ally wears out. Although this conception is superficially similar to
Descartes's idea of momentum transfer, it is not linked to a conserva-
tion principle and therefore is of little use for the quantitative study of
motion. Galileo's fame as Newton's forerunner in the foundation of
modern dynamics does not rest however on his early writings, but on
the Latin treatise "On local motion" inserted in the Third Day of the
Discorsi (1638). It begins with the analysis of uniform motion to which
I referred in §1.1. It then sets up a mathematical model of free-fall, and
finally tackles the motion of missiles, viewed as a combination of
uniform motion and free-fall. Compared with the modern treatment
of these matters, Galileo's suffers from some obvious limitations, for
(i) in the proof of some key theorems he must make do without
the infinitesimal calculus invented thirty years later by Newton and
Leibniz, and (ii) he did not postulate, like Descartes and Newton, that
unimpeded motion continues indefinitely at the same speed on a
straight line in every case, but he argued for and used a "Law of
Inertia" restricted to horizontal motion, that is, to motion near the
surface of the earth that neither falls nor rises and therefore remains
perpendicular to the local radius of the earth. But despite these short-
comings, "On local motion" provided a paradigm of mathematical
physics that inspired the next generations and in a general way is still
alive today.
Galileo defines uniform motion as one "in which the parts run
through by the mobile in any equal times whatever are equal to one
another" (EN VIII, 191). Since the impact of a freely falling body
increases with the height from which it falls, it is clear that free-fall is
not uniform motion. Galileo guesses that it is the simplest conceivable
sort of accelerated motion (for free-fall is natural, and nature "habit-
ually employs the first, simplest and easiest means" - EN VIII, 197),
so he offers a precise definition of a type of motion meeting this require-
ment, produces geometrical proofs of several properties that this
type of motion must have according to its definition, and leaves it to
experiment to show that free-fall exhibits these properties. The
definition is this: "We shall call that motion equably or uniformly
accelerated which, departing from rest, adds on to itself equal momenta
of swiftness {momenta celeritatis) in equal times" (EN VIII, 205). The
expression 'momenta of swiftness' is later used interchangeably with
'degrees of speed', so one may assume that 'momentum' here simply
means 'amount' or 'increment'; however, its Latin meaning, 'impulse,
22 Natural Philosophy in the Seventeenth Century

c B

Figure 1

push', conveys the idea of a physical quantity that translates into

impact on collision and whose actual presence in the mobile is what
carries it forward at any given time.
This idea lurks in the proof of the fundamental Theorem I: "The
time in which a space is traversed by a mobile in motion uniformly
accelerated from rest is equal to the time in which the same space would
be traversed by the mobile carried in uniform motion with a degree of
speed (velocitatis gradus) equal to one-half the final and greatest degree
of speed of the said uniformly accelerated motion". To prove it, we
draw the lines AB, representing the duration of the uniformly acceler-
ated motion, and BE, perpendicular to AB, representing the final speed
v attained in that motion (Fig. 1).
Let F be the midpoint of BE, and draw the rectangle ABFG. Each
parallel to GA drawn from a point on AB and contained in this rec-
tangle represents the speed of the mobile moving uniformly with speed
ji/, at the time represented by that point. On the other hand, the speed
attained at that time by the uniformly accelerated mobile is represented
by the perpendicular to AB drawn from the said point to the line AE.
Let I be the midpoint of FG and C the point where the perpendicular
from I meets AB. The speed of the uniformly accelerated mobile equals
jv only at the instant represented by C. For each instant t before C in
which the speed of the uniformly accelerated mobile falls short of \v
by a certain amount there is exactly one instant f(t) after C in which
the speed of the uniformly accelerated mobile exceeds \v by precisely
the same amount. Thus, "the deficit of the momenta in the first half of
the accelerated motion - represented by the parallels in triangle AGI -
is made up by the momenta represented by the parallels in triangle IEF"
 1.4 Galileo on Motion 23

(EN VIII, 209). Hence, the two mobiles will run through equal spaces
in the time represented by AB.
It is worth noting that this argument does not rest only on the stated
one-to-one correspondence between speed deficits at instants before C
and speed excesses at instants after C. It holds good because, under the
circumstances, the following metric condition is satisfied: given any
time interval T before C, if T denotes the set of speeds less than \v
that the uniformly accelerated mobile successively sports during T, the
matching set f(Y) of speeds greater than \v is sported by the mobile
after C during an interval f(T) of the same length as T. Evidently, if
f(T) ^ T, the contribution of f(V) to the mobile's displacement cannot
balance the deficit in the contribution of T. The said contributions
depend not only on the "momenta of swiftness" contained in T and
/"(°IO, but also on the length of time during which their push is at work.
A dolt would probably need the calculus to grasp this, but Galileo
obviously did not.
By establishing a precise relation between uniformly accelerated
motion and uniform motion, Theorem I enables us to use the mathe-
matically manageable properties of the latter to calculate quantitative
features of the former. Equation (1.2) entails that if mobiles M and M'
run, respectively, through spaces s and s' in times t and f with con-
stant speeds v and z/, sis' - vt/v't'. From this elementary relation and
Theorem I, Galileo infers that, "if a mobile descends from rest in uni-
formly accelerated motion, the spaces run through in any times what-
ever are to each other . . . as the squares of those times" (Theorem II;
EN VIII, 209). Let v{ denote the speed attained and s,- the space run
through by the uniformly accelerated mobile in time U. According to
Theorem I, s{ would also be run through in time ti if the mobile were
carried in uniform motion with speed \v{. The concept of uniformly
accelerated motion entails that vx has to v2 the same ratio that t\ has
to t2. (This can be read from Fig. 2, where the speeds attained in times
AB and AC are represented, respectively, by the perpendiculars BE and
CI, and BE:CI::AB:AC.) So

v2t2j t\
As a corollary we have that the spaces traversed by a uniformly accel-
erated mobile in the first, second, third, fourth,. . . , nth units of time
are to each other as the successive odd numbers 1:3:5:7: . . . : In - 1.
(See Fig. 2, or apply the formula n2- (n - I) 2 -In - 1.)
24 Natural Philosophy in the Seventeenth Century

s= 1 » ( / = l )

= 16 --(t =

Figure 2

Galileo devised a clever way of testing this corollary.19 He prepared

a tilted plane with a groove down which a polished metal or marble
ball could roll freely. He tied moveable gut frets around the plane as
frets are tied on the neck of a lute. The rolling ball would make a sound
as it passed each fret. The positions of the frets were adjusted until the
sounds were on beat - this was something that Galileo's trained musical
ear could ascertain to within 1/64 of a second. He could then verify
that the distances between the frets - the spaces run through by the
ball in equal times - satisfied the corollary. Galileo's plane had approx-
imately a 3% tilt and was about 2 meters long. On a steeper plane the
fall would be too fast for this kind of chronometry. However, Galileo's
results are extended to any, even vertical, inclination by the following
postulate introduced right after the definition of uniformly accelerated
motion:

19
See Drake (1978, pp. 86 ff.). Barbour (1989, p. 371) claims that in fact Galileo first
discovered the corollary through the experiment that will be described in the main
text, and then developed the theory from which he subsequently derived it. Although
he does not cite any hard evidence for this claim, he may well be right. We ought
however to bear in mind that, as Barbour himself emphasizes, "meaningful mea-
surement is hardly possible without an underlying theoretical conception of what it
is significant to measure" (1989, p. 376).
 1.4 Galileo on Motion 25

B X

Figure 3

/ assume that the degrees of speed acquired by a given mobile over dif-
ferently inclined planes are equal whenever the heights of those planes
are equal.
(Galileo EN VIII, 205)20
This postulate does not add any new qualifications to the concept of
uniformly accelerated motion put forth by Galileo, but it does place a
constraint on the physical realizations of this mathematical concept.
Galileo claims that this concept is applicable to all cases of unimpeded
free-fall, by which he means fall on inclined planes near the surface of
the earth with all impediments removed.21 The postulate prescribes a
definite quantitative relation that must hold between such applications.
From Theorems I and II and the said postulate, Galileo derives
numerous, in part testable, propositions concerning motion on
diversely inclined planes. The postulate plays an essential role in
Galileo's argument for his "Law of Inertia". Let us take a look at the
curious reasoning leading to it. Consider a mobile freely falling from
rest at A on an inclined plane AB of height h (Fig. 3). If it attains speed
v in time t, the distance s it traverses in this time equals \vt. If the
mobile then continues to move with constant speed v, it will traverse
in time t the distance vt = 2s. Suppose however that during this second
interval t the mobile is made to go up a plane BC with the same incli-

20
The second, posthumous edition of the Discorsi (1655) contains a "proof" of this
postulate that Galileo dictated before his death in 1642. It requires additional assump-
tions - which Galileo smuggled in from statics - for, as a moment's reflection shows,
the behavior of the mobile on differently inclined planes cannot be extracted from
the definition of uniformly accelerated motion.
21
Note the difference between the classical notion of free fall and Galileo's. Classical
physics treats the inclined plane as a partial impediment to fall.
26 Natural Philosophy in the Seventeenth Century

nation as AB. "As soon as the ascent begins there naturally supervenes
that which happens to [the mobile] from A on the plane AB, namely,
a certain descent from rest according to those same degrees of accel-
eration by force of which it descends the same amount in the same time
on [plane BC] that it descended on AB" (EN VIII, 244).
Thus, in time £, the mobile will move up plane BC a distance 2s
thanks to its initial speed v, and down the same plane BC a distance s
due to the plane's inclination. Therefore, it will run upward through a
net distance s, thus reaching the same height h from which it fell. Had
the mobile fallen from height h on a differently inclined plane DB, it
would climb to height h on plane BE with the same tilt as DB. By the
postulate the mobile attains at B the same speed v whether it falls on
plane AB or DB. And, as we have just seen, once it has reached this speed
it will climb back to height h, on plane BC or on plane BE, regardless
of their particular inclinations. So, Galileo concludes, a mobile moving
up a tilted plane with initial speed v will move forward, if all impedi-
ments are removed, until it reaches the same height h by falling from
which it would have attained that speed. The smaller the tilt, the greater
the space run through by the mobile and the longer the time it will
employ to decelerate until it stops. In the limiting case in which there is
no tilt at all, the mobile will move forever with the same speed.
This is the law of rectilinear uniform motion - the "Law of Inertia"
- that Galileo will use in his theory of missiles. It is clear, from its
derivation, that it is only meant to hold within the narrow range in
which this theory is applicable, namely, within a few miles from a
cannon's muzzle, under circumstances in which gravity is the only
accelerating or retarding factor that counts. When the Aristotelian Sim-
plicio points out that a surface which neither rises nor falls cannot be
a plane, for its points must be equidistant from the center of the earth,
Galileo's spokesman Salviati promptly concedes that he is right but also
cites the example of Archimedes, who took it "as a true principle that
the arm of a balance or steelyard lies in a straight line equidistant at
all points from the common center of heavy things, and that the cords
to which weights are attached hang parallel to one another" (EN VIII,
274). Such blatantly false assumptions lead only to negligible errors
because "in actual practice our instruments and the distances we
employ are so small in comparison with the great distance to the center
of our terrestrial globe that we may well regard one minute of a degree
of the meridian as if it were a straight line, and two verticals hanging
from its extremities as if they were parallel" (Ibid., 275-76). The great-
est distances occur in artillery shots, the longest of which "do not
 1.4 Galileo on Motion 27

exceed four miles", ca. 1/1000 of the distance to the center. Indeed, the
Galilean gunner is likely to err much more by considering air resistance
to be negligible than by assuming the horizontal to be straight.
The keystone of Galileo's theory of missiles is his proof that a mobile
endowed with uniform horizontal and uniformly accelerated vertical
motions describes a semiparabola. The parabola had been studied by
Apollonius in his treatise on conic sections (third century B.C.), but
it had no place in a seventeenth-century gentleman's education. By
appealing to such curves, the new physics wedded itself from the outset
to what was then perceived as higher mathematics. For the benefit
of Sagredo and Simplicio, his dialogue partners, Salviati defines a
parabola as the curve at which the surface of a right cone is cut by a
plane parallel to one of the cone's sides and orthogonal to the plane
through that side and the cone's axis of symmetry. Such a curve has its
own axis of symmetry, which it meets at the apex. The apex divides
the parabola into two semiparabolas. Let P be a point on one of them.
P's amplitude x(P) is the length of the perpendicular from P to the axis.
P's altitude y{V) is the distance from the apex to the said perpendicu-
lar.22 Galileo utilizes only two properties of the parabola, viz., (Ill) if
Q is another point on the semiparabola, y(P)/y(Q) = x2(P)/x2(Q) - the
altitudes are to one another like the amplitudes squared -, and (n2)
if Q is a point on the axis, on the convex side of a parabola, at a dis-
tance from the apex equal to the altitude of P, the straight QP is tangent
to the parabola.23
The proof in question follows at once from property (FIi). Let a
mobile move from a point O with constant horizontal velocity v, while
falling vertically from rest with uniformly accelerated motion. Let ad-
measure its horizontal advance and y{ its vertical descent after time tt.
Since Xi/x2 = txlt2 and yjy2 = t\lt\ for any pair of times t\ and t2, the
successive descents of the mobile are to one other like its advances

22
Galileo speaks of the amplitude and altitude of a given semiparabola, by which he
means, in my jargon, the amplitude and altitude of the point on which it meets the
base of the cone. However, any parabola T forming the intersection of a finite cone
K with a plane <E> is part of a longer parabola V that forms the intersection of C> with
a cone K' that includes K as a proper part. Since the height of the cone employed for
determining a given semiparabola is utterly irrelevant to Galileo's dynamics, it seems
to me that my terminology, with mild anachronism, conveys Galileo's thinking more
clearly than his.
23
Salviati proves both. The modern reader will recognize in property (111) the definition
of a parabola in analytic geometry by the equation y = ax1 (in terms of the coordi-
nates of an arbitrary point in a Cartesian system originating at the apex).
28 Natural Philosophy in the Seventeenth Century

squared. Descents and advances thus stand to one another, respectively,

as the altitudes and amplitudes of the points on a semiparabola with
its apex at O and a vertical axis. Therefore, the successive positions of
the mobile must lie on such a semiparabola.
This beautiful piece of reasoning rests on the assumption - adopted
by Galileo as a matter of course - that the mobile's constant forward
thrust and its uniformly increasing downward thrust coexist and do
their respective work without interfering.24 In the above formulation
it only applies to horizontal ("point-blank") shots, but it can be readily
extended to missiles shot at any angle above the horizontal by analyz-
ing the muzzle velocity into a horizontal and a vertical component
and subtracting the constant magnitude of the latter from the growing
downward velocity of fall (see Polya 1977, pp. 102-5). Galileo,
however, does not handle such cases in this way but invokes instead
considerations of symmetry. He tackles the gunner's problem of reach-
ing a desired range by adjusting the gun's angle of elevation. After
proving a few theorems he produces tables that display, for angles
from 1° to 89°, the ranges and the maximum heights attainable by mis-
siles shot with the same initial speed, and the different initial speeds
required for attaining a given range. Such tables provide a large supply
of testable predictions. The arguments leading to these results rest, as
heretofore, on Galileo's ability to translate physical relations into ratios
between geometrical magnitudes. To this effect he introduces a geo-
metrical measure of a missile's initial speed, which he calls 'sublimity'.
The sublimity a(v) is the height through which a mobile must fall from
rest to attain speed v. Let x and y denote the amplitude and the alti-
tude of an arbitrary point in the parabolic trajectory of a missile
thrown horizontally with initial speed v. Galileo proves that \x is the

24
When Sagredo and Simplicio say that they are confused by the composition of
motions, Salviati takes the said assumption for granted and merely discusses its
algebra. He reaches the following precise statement of the rule for adding orthogo-
nal vectors (cf. note 17): "When one must indicate the quantity of impetus resulting
from two given impetuses, one horizontal and the other vertical, both being equable,
one must take the squares of both, add them together, and extract the square root of
their sum; this will give us the quantity of the impetus compounded from both."
However, "when a motion enters the mixture, which starts from the greatest slow-
ness and increases its speed as time goes by, it is necessary that the quantity of time
shall manifest to us the quantity of the degree of speed at the given point. As for the
rest, the impetus compounded from these two is (as in uniform motions) equal in the
square to both components." (EN VIII, 289).
 1.4 Galileo on Motion 29

Figure 4

mean proportional between y and o(v). This enables him to calculate

amplitudes from altitudes, for any given initial horizontal speed.
Proposition VII and its remarkable "Corollary" pave the way for infer-
ences concerning nonhorizontal shots.
PROPOSITION VII. In missiles which describe semiparabolas of the same
amplitude, less impetus is required by the one which describes that whose
amplitude is twice its altitude, than by any other.
(Galileo, EN VIII, 295)
Say that B and D are, respectively, the apex and an arbitrary point of
a parabola, and let C denote the foot of the perpendicular from the
axis to D (Fig. 4). Then, by Proposition VII, the requisite horizontal
thrust for a missile to go from B to D is minimal if CD = 2BC. But
2BC is the distance from the horizontal through C (and D) to the point
A where the axis meets the tangent to the parabola at D (by property
n 2 of parabolas). Thus the right triangle ACD is isosceles, and the
tangent at D makes an angle of 45° with the horizontal. Coupled with
a tacit but extremely bold appeal to symmetry, this simple reasoning
yields the following
COROLLARY. From this it is clear that, conversely, the missile thrown
from point D requires less impetus to describe the semiparabola DB,
which has the tangent AD making one-half a right angle with the hori-
zontal, than any other semiparabola having greater or smaller elevation
than semiparabola DB. Hence it follows that if missiles are thrown from
30 Natural Philosophy in the Seventeenth Century

point D with the same impetus but with different elevations, the
maximum throw, or amplitude of semiparabola .. . will be achieved with
the elevation of half a right angle. The other throws, made at larger or
smaller angles, will be shorter.
(Galileo, EN VIII, 296)
The title 'Corollary' bestowed on this key proposition is apparently
meant to silence questioners. For such a symmetric correspondence
between the semiparabola described by a missile thrown horizontally
and the trajectory of a missile thrown upward along a tangent to the
former's path can hold only if unimpeded rectilinear motion persists in
every direction - and not just on the horizontal - in accordance with
Descartes's principle of inertia, which Galileo never quite managed to
embrace.25

1.5 Modeling and Measuring

In this section I shall propose three examples of seventeenth-century
physics in the style of Galileo. In each one of them his method of math-
ematical representation of essentials while neglecting the negligible is
brought to bear on a well-defined physical question under assumptions
that were standard after Galileo and Descartes. In the first two exam-
ples, important theses of Cartesian physics are demolished and
replaced. In the third one, the power of mathematical modeling yields
the first estimate - to the right order of magnitude - of a speed so fast
that its direct measurement by some commonsense method was utterly
impracticable.

1.5.1 Huygens and the Laws of Collision

According to Descartes's definition of motion (cited in §1.3), a body B
moves with speed v if (i) its place relative to its immediate surround-
ings S changes at that speed and (ii) S is regarded as being at rest. Let
T denote the immediate outer surroundings of S. Suppose that the place
of S relative to T is also changing with speed v, but in the opposite
direction. One can then obviously regard S as moving and its total

25
He comes close but stops short of it, for example, in the following statement: "What-
ever degree of speed is found in the mobile, this is by its nature indelibly impressed
on it when external causes of acceleration or retardation are removed, which happens
only on the horizontal plane" (EN VIII, 243; my emphasis - Galileo's text is all in
italics).
 1.5 Modeling and Measuring 31

Figure 5

immediate surroundings T u B a s being at rest. Descartes's concept of

motion is therefore completely relativistic (to a degree that Einstein's,
despite his verbal allegations, is not). Indeed, his referral of motion
to immediate surroundings is philosophically idle and only serves his
desire of avoiding ecclesiastical odium (Descartes's earth is at rest in
the atmosphere, although it circles the sun). Christiaan Huygens took
this relativism seriously and used it for deriving true laws of collision
to replace Descartes's rules.
Huygens assumes the validity of Descartes's principle of inertia. He
postulates that two equal bodies colliding head-on at the same speed
rebound after collision with the same speed. This is Descartes's first
rule of collision (1644, II, art. 46) and should be evident on grounds
of symmetry. By virtue of this postulate, Huygens's inquiry is restricted
to perfectly elastic collisions. Descartes stated six more rules for colli-
sions between bodies of different sizes or moving against each other
with different speeds. By paying due attention to Descartes's definition
of motion, Huygens shows that at least two of these rules must be false
if the first one is true. He invites us to reflect on some imaginary exper-
iments performed on a boat gliding along a quay - a daily occurrence
in Huygens's Holland. The experiments are analyzed from the point of
view of two persons who hold hands while standing, respectively, on
the boat and on the quay, and assign different speeds to the colliding
objects (Fig. 5). For the sake of brevity, I shall ignore this romantic
setting.
32 Natural Philosophy in the Seventeenth Century

Two equal balls, E and F, collide head on while moving in opposite

directions with speed v relatively to boat B. Let v stand for the
(directed) velocity of E, so that of F is -v. According to Descartes's first
rule, when they collide the balls in effect exchange velocities: E
rebounds with velocity -v and F with velocity v. Suppose that at the
instant of collision B moves along Q with velocity v. What are the
velocities of E and F relative to Q before and after colliding? Huygens's
commonsense answer is that E - which moves on B with the same
velocity v that B has relative to Q - moves relatively to Q with veloc-
ity 2v, while F is at rest. Upon collision, the balls exchange velocities,
as their motion relative to Q is computed by subtracting their new
velocities on B from the latter's velocity v along Q (so E is at rest and
F moves with velocity 2v). This is Huygens's first theorem on collisions:
If a body collides with an equal body at rest, after contact it will be at
rest, while the latter will have acquired the velocity of the body that
collided with it. It openly contradicts Descartes's sixth rule.26
Suppose now that E and F collide while moving on Q with veloci-
ties v and &v, respectively (where k is any positive or negative real
number). Let B move along Q with velocity v' =y(v + kv) =j(l + k)y.
Then, right before collision, the velocity of E relative to B is v - v' =
v - j ( l + k)y = j ( l - &)v, while the velocity of F relative to B is
k\ - v' = k\ - 1 ( 1 + k)\ = -j (1 - k)\. Thus, relative to B, balls E and
F collide head-on with equal speeds. Descartes's first rule applies: E and
F exchange velocities. So after collision, E moves relative to Q with
velocity - j ( l - k)y + v' = -\ (1 - k)y + j(l + k)y - &v, and F moves
relative to Q with velocity j ( l - k)y + v' = j ( l - k)y + j ( l + k)y = v.
We have thus proved Huygens's second theorem on collisions: If two
equal bodies collide head-on with different velocities they shall move
after collision with their velocities exchanged. It contradicts Descartes's
third rule. 27

26
Descartes (1644, II, art. 51). Here is a paraphrase: Rule 6: If a body collides while
moving with speed v with an equal body at rest, it will rebound with speed \v while
the second body moves with speed \v in the direction the first one had initially.
27
Descartes (1644, II, art. 48). Here is a paraphrase. Rule 3: If two equal bodies collide
head on while moving with speeds v and v' such that v > v\ after colliding they both
move together with speed v - v' in the original direction of the faster body. Huygens's
second theorem also conflicts with Descartes's Rule 7 (Ibid., art. 53), insofar as it
applies to equal bodies. This is a very complicated and seemingly arbitrary rule applic-
able to bodies moving in the same direction with different speeds that collide when
one catches up with the other.
 1.5 Modeling and Measuring 33

Huygens's second theorem entails both his first theorem and

Descartes's first rule as special cases. By his masterful appeal to the rel-
ativity of motion Huygens succeeded in deriving the general rule from
Descartes's first rule. One could also derive the latter and Huygens's
second theorem from the first theorem, by using as a term of compar-
ison a frame in which one of the balls is momentarily at rest at the
time of collision. All three laws require the idealizing assumption of
perfect elasticity, but will certainly be verified, within a plausible
margin of error, on a billiard table. On the other hand, Descartes's rules
2-7 probably made him the laughing stock of billiard players who
became acquainted with them. Note, by the way, that Huygens does
not in any way assume the constancy of the velocities under discus-
sion: The instantaneous velocity of each body right before and right
after collision, with regard to each of two frames, is all that matters.
It would seem however that the frames in question - a boat and a quay
- do indeed move uniformly past each other. Not much later, Newton
stated fundamental Laws of Motion under which this condition turns
out to be indispensable.

1.5.2 Leibniz and the Conservation of "Force"

The German lawyer, philosopher, and mathematician, G. W. Leibniz,
raised numerous objections against the physics and metaphysics of
Descartes. In particular, he repeatedly argued that the 'quantity of
motion' defined by Descartes as mass x speed cannot be conserved
in nature. Strictly speaking, this quantity does not even exist, "for a
whole never exists unless its parts coexist" (Leibniz, GM VI, 235), and
motion, "being successive, perishes continually" (Leibniz, GP VII,
402). "Thus there is nothing real in motion save that which must
consist of a force striving towards change", and "whatever there is in
corporeal nature besides the object of geometry, or extension, must be
reduced to this force" (GM VI, 235). "Force", in this sense, is the
causal agent in motion. Its quantity must remain unchanged, for if it
increased, "there would be an effect more powerful than its cause, that
is, mechanical perpetual motion, capable of reproducing its cause plus
something in addition, which is absurd"; and if it could decrease, "it
would in the end perish utterly, because not being able to grow and
yet being liable to diminish, it would always be set on a path of decay,
which is no doubt contrary to the order of things" (GM VI, 220). The
latter explanation reminds one of the bland optimism popularly asso-
34 Natural Philosophy in the Seventeenth Century

ciated with Leibniz since Voltaire's Candide. On the other hand, by his
unhesitant dismissal of mechanical perpetual motion - the engineering
equivalent of the economist's proverbial "free lunch" - Leibniz stands
out clearly as the first one to adumbrate the modern concept of energy:
"No machine, and thus not even the whole world can stretch its force
without a new impulse from outside" (GM VI, 117).28
To prove his point Leibniz examines a particular example, but, he
says, "one will find the same in any other example one might choose"
(GM VI, 222). He assumes that the same "force" is required to lift one
pound four feet as to lift four pounds one foot. (This is obvious, insofar
as each operation can be analyzed into four operations of lifting one
pound one foot.) Then, in the style of Galileo, he demands that all
"external impediments shall be excluded or neglected, as if there
weren't any".29 He explains that, since we are here pursuing "les
raisons des choses" and not any practical business, we may conceive
of motion as taking place in a void, so there is no resistance of the
medium; we may imagine that planes and spheres are perfectly smooth,
so there is no friction, and so on. All this "in order to examine each
thing separately with a view to combining them in practice". In virtue
of this idealizing requirement Leibniz can sensibly assume that the
entire force of a body can be transferred to another body without loss.
The same, of course, must hold for Descartes's quantity of motion
if this quantity is conserved. Finally, Leibniz accepts as "proven by
others" that if any two bodies attain speeds vx and v2 by falling freely
from rest through heights bx and h^ h1/h2 = v\lv\ (this is a straighfor-
ward consequence of Galileo's Theorem I on free-fall). And, of course,
a pair of bodies moving with initial speeds vx and v2 and meeting no

28
Leibniz distinguishes between physical and mechanical perpetual motion. The former
would be performed by a perfectly free pendulum "but this pendulum will never
surpass its initial height, nor indeed will even return to it if on the way it operates
or produces the slightest effect or overcomes the slightest obstacle; otherwise this
would be a case of mechanical perpetual motion". The latter occurs when a system
of bodies in motion "returns after some time to a state that is not only as violent as
it was initially but even more so, because it is required that the device, besides restor-
ing the first state, produces some mechanical effect or utility, without any external
force contributing to it" (Leibniz in Costabel 1960, p. 98). Leibniz's use of 'mechan-
ical' agrees with the original sense of the Greek verb |i£%avdo|jm, 'to contrive,
procure for oneself.
29
This quotation and the next are from Leibniz's "Essay de dynamique" of 1692, pub-
lished in Costabel (1960, p. 99).
 1.5 Modeling and Measuring 35

air resistance will climb along perfectly smooth inclined planes to

heights hi and h2 such that h1/h2 = v\lv\.
From these premises we can reason as follows. Choose the unit of
time so that a body falling from rest at a height of one foot acquires
a speed of one foot per time unit. Suppose that we have the "force" to
lift one pound four feet. If Descartes's quantity of motion is conserved,
we can "stretch" this "force" to lift the same pound 16 feet! For we
can certainly employ it to lift four pounds one foot. As they fall back
to the starting point, our four pounds acquire a speed of one foot per
time unit and thus a quantity of motion of 4 x 1 = 4. Under our
idealizing assumptions, Descartes's conservation principle implies that
this quantity of motion can be integrally transferred to a body of one
pound. This body will then acquire a speed of four feet per time unit
and will climb along a smooth inclined plane to a height of 42 = 16
feet. Hence, Descartes's principle cannot hold. The quantity conserved
in collisions and other such processes in which some bodies gain
"force" at the expense of others is not proportional to mass x speed
but rather to mass x speed squared. This quantity was usually referred
to in eighteenth-century literature as "live force" (vis viva). If the factor
of proportionality is set equal to j , the live force equals our kinetic
energy, T = jmv2.30
However, if the "quantity of motion" is equated, not with "the
motion taken absolutely (without regard to its direction)", but with
"the advance in a certain direction", then
the total advance . . . will be the sum of the particular quantities of
motion, when the two bodies go in the same direction. But when they
go in opposite directions, it will be the difference of their particular quan-
tities of motion. And one shall find that the same quantity of advance is

30
In 1807 Thomas Young introduced 'energy' as a term for "the product of the mass
or weight of a body, into the square of the number expressing its velocity". However,
it did not gain currency in physics until the 1850s, when it was adopted by Rankine
and others as a general term for a body's capacity to do mechanical work. See the
Oxford English Dictionary (OED), s.v. 'Energy', 6. Rankine distinguished between
'actual energy' or the power possessed by a body of doing work by virtue of its
motion, and 'potential energy' or the power of doing work a body possesses by virtue
of its position relative to other bodies. The latter name has stuck, but the former gave
way to 'kinetic energy' (i.e., 'energy of motion'). Still, Einstein used 'lebendiger
Kraft", - i.e., 'live force' - for 'kinetic energy' as late as 1902 (p. 433, line 7). As to
the factor y, Kuhn (1977, p. 87, n. 47) says that it was introduced by Coriolis in
1829 to make the vis viva numerically equal to the work that it can produce.
36 Natural Philosophy in the Seventeenth Century

conserved. But this should not be confused with the quantity of motion
in the ordinary [i.e., Cartesian - R. T.] sense. The reason for this rule
about advance can be shown in some way and it is reasonable that if
nothing supervenes from outside the whole (composed of bodies in
motion) will not impede its advancing as much as it did earlier.
(Leibniz in Costabel 1960, p. 105)
Thus, Leibniz recognized the conservation of momentum or mass x
velocity, besides the conservation of the quantity he called "force",
which he deemed proportional to mass x speed squared, and also to
mass x height of fall.

1.5.3 Romer and the Speed of Light

To the untutored eye light instantaneously fills any space exposed to
it. However, if light is a corporeal substance flowing from its source,
it is reasonable to believe, with Empedocles, that it takes time to spread
(DK 31.A.57). So Aristotle, ever anxious to reconcile natural philoso-
phy with common sense, denied that light is an effluence of any sort
and conceived it as "the actuality of the transparent qua transparent"
(418b10). The latter is a certain nature (physis) common to air, water,
and other things that possess it at least potentially even in darkness. It
becomes actual at a stroke due to the presence of fire and the like.
Therefore, Empedocles
was wrong in saying that light travels and arrives at some time between
the earth and that which surrounds it, without our noticing it. For this
is contrary to the clear evidence of reason and also to the apparent facts
(ta phainomena); for it might escape our notice over a short distance,
but that it does so over the distance from east to west is too big an
assumption.
(Aristotle, De anima II.7 418b20-25, Hamlyn's transl.)
The instantaneous propagation of light remained the prevailing
opinion for two millenia, until almost the end of the seventeenth
century. Descartes explained it on the analogy of an impulse instanta-
neously transmitted from one end to another of a rigid body (AT VI,
83-85).31 Galileo too expressed approval for it in 17 Saggiatore (EN VI,
352); but in the Discorsi, Salviati questions the Aristotelian belief

31
Surprisingly, this view did not prevent Descartes from trying to explain the disper-
sion of light on the analogy of a rain of pellets, whose angular velocities change in
different ways on penetrating a new medium.
 1.5 Modeling and Measuring 37

voiced by Simplicio (EN VIII, 87). He proposes a simple experiment

for measuring the speed of light:
I would have two men each take one light, inside a lantern or other
vessel, which each could conceal and reveal by interposing his hand, in
sight of his companion. Facing each other at a distance of a few fathoms,
they could practice revealing and concealing the light from each other's
view, so that when either man saw a light from the other, he would at
once uncover his own. After some mutual exchanges, this signaling
would become so adjusted that without any sensible variation, either
would immediately reply to the other's signal, so that when one man
uncovered his light he would at the same time see the other man's light.
This practice having been perfected at a short distance, the same two
companions could place themselves with similar lights at a distance of
two or three miles and resume the experiment at night, observing care-
fully whether the replies to their showings and hidings followed in the
same manner as near at hand. If so, they could surely conclude that the
expansion of light is instantaneous, for if light required any time at a
distance of three miles, which amounts to six miles for the going of one
light and the coming of the other, the delay ought to be quite noticeable.
And if it were desired to make such observations at yet greater distances,
of eight or ten miles, we could make use of the telescope.
(Galileo, EN VIII, 88)
The idea behind this experiment underlies the very accurate methods
for measuring the speed of light employed in the twentieth century
(until it was fixed at 299,792,458 meters per second [m/s] by interna-
tional convention in 1983), as well as Einstein's famous rule for the
synchronization of distant clocks (§5.1). But Galileo's proposal for its
actual performance grossly underestimates the speed of light compared
with that of nervous signals from the experimenters' eyes to their
hands. His spokesman acknowledges that no delay was observed when
the experiment was tried over a distance of less than one mile.
In 1676, Ole Romer reported to the French Academy a phenome-
non which showed, in his analysis, that the speed of light is finite and
provided a way of measuring it. At the time there was a desperate need
for a reliable method of ascertaining longitude on a ship at sea.
Someone figured out that Jupiter's satellite, Io, periodically hiding
behind the planet once in every 40 hours or so, could be used for that
purpose.32 So Romer, a junior astronomer in the Paris observatory,

32
If I read in the nautical almanac that Io will emerge from hiding at midnight GMT
and see it appear three hours before midnight, I can calculate that I am placed at Ion-
38 Natural Philosophy in the Seventeenth Century

Figure 6

began painstakingly timing Io's eclipses. Prompted by observed irreg-

ularities or by his understanding of kinematics (or, more probably, by
a combination of both), he lighted on the very bright idea that I shall
now explain with the aid of Fig. 6. Let the circle through EFGHLK
represent the orbit of the earth about the sun at A, while the circle
through CD represents the orbit of Io about Jupiter at B. At C, Io enters
Jupiter's shadow, from which it emerges at D.
Romer estimated Io's period to be 42.5 hours. He realized, however,
at that if Io is seen to emerge when the earth is at L and its next emer-
gence occurs when the earth is at K, the interval between these two
events must be greater than 42.5 hours for one must add the time that
the light from Io takes in traversing the chord LK. Likewise, if two suc-
cessive emergences are observed from F and G, the interval will be less
than 42.5 hours. Romer reckoned that if light took one second to tra-
verse a distance equal to the diameter of the earth, it would need 3.5
minutes to go from L to K or from G to F, so the satellite's observed
period would be seven minutes shorter when the earth approaches
Jupiter than when it recedes from that planet. Since no such difference
was recorded he concluded that light takes less than one second to
travel through one terrestrial diameter (so its speed is greater than
12,000km/s).33 This does not mean, however, that it takes no time at

gitude 45°W. Indeed in the seventeenth century clocks were not accurate enough for
such methods to work properly.
33
This is the only numerical estimate of the speed of light ever published by Romer.
But our teaching profession has rushed in to fill the gap. Andrzej Wroblewski (1985)
 1.5 Modeling and Measuring 39

all.34 Although Romer could not detect a time difference when com-
paring two single satellite periods, he found that 40 consecutive periods
observed from side F are sensibly shorter than 40 consecutive periods
observed from side L - no matter what the position of Jupiter - and
that the difference amounts to 22 minutes for the earth's orbital diam-
eter HE.
On the strength of his idea and of eight years of observations, Romer
announced in September 1676 that in November of that year Io would
be emerging 10 minutes later than one would have expected from
observations performed in August. When Io's emergence was observed
in Paris on 9 November 1676 at 17h 34m 45s, confirming his pre-
diction, he submitted to the French Academy a paper containing the
above analysis of the delay. He stressed that the inequality could not
be explained by Io's excentricity or any of the other causes usually
adduced to account for the irregularities of the moon and the planets.
For although he had observed that Io is excentric and that its periods
are shorter or longer depending on whether Jupiter approaches the sun
or recedes from it, the inequalities arising from these causes were not
sufficient to filter out the inequality due to the finite speed of light.
Romer's idea can be concisely expressed as follows: A signal issued
at fixed intervals and propagating with the same finite speed will be

has compiled an amusing table of values of the speed of light, as diverse as 193,120
km/s and 351,000 km/s, for which Romer is held responsible in standard physics text-
books. The dates given for Romer's discovery in these textbooks range from 1656 -
when he was a mere child - to 1876 - when, if still alive, he would have been long
past the age of discovery - but they tend to cluster in the interval 1673-78.
34
Robert Hooke - a distinguished English experimentalist remembered mainly for
Hooke's law of elastic force, for his admirable drawings of tiny creatures seen under
the microscope, and for his priority dispute with Newton over the discovery of uni-
versal gravitation - thought otherwise. Commenting on Romer's work he wrote in
1680: "So far he thinks indubitable that [light] moves a Space equal to the Diame-
ter of the Earth, or near 8000 Miles, in less than one single Second of the time, which
is as short time as one can well pronounce 1, 2, 3, 4: And if so, why it may not be
as well instantaneous I know no reason" ("Lectures on Light" in Hooke, PW, p. 78;
cited in Wroblewski 1985, p. 625). Other contemporary scientists, including Romer's
boss, Cassini, rejected his discovery. But both Huygens and Newton embraced it
wholeheartedly. The former reproduced Romer's reasoning in his Traite de la lumiere
(submitted to the French Academy in 1678 and published in 1690); based on it, he
estimated that light traveled more than 600,000 times faster than sound. Using for
the speed of sound Huygens's value of 180 toises/s ~ 330 m/s, this gives a speed of
light > 198,000km/s.
40 Natural Philosophy in the Seventeenth Century

perceived at shorter intervals by an observer moving toward its source

and at longer intervals by one moving away from it. This is known as
the Doppler effect, for the German physicist who rediscovered it in the
nineteenth century. It can be heard as a sudden tone change when a
police siren speeds by you. The variation depends on the velocities of
the signal and the observer (relative to the source), and if one of them
is known, we can calculate the other. Let v denote the fixed frequency
with which the signal is issued and Av the difference, positive or neg-
ative, between the observed frequency and v; let u and c be, respec-
tively, the observer's speed and the signal's speed relative to the latter's
source. If we straightforwardly add velocities in the manner of Huygens
(§1.5.1) and classical kinematics, Av/v = ±(u/c), with the plus sign
applying if the observer approaches the source and the minus sign if it
recedes from it. If any three of the four quantities in this equation are
given, it is an easy matter to compute the fourth one.
 CHAPTER TWO

Newton

"Natural Philosophy consists in discovering the frame and operations

of Nature, and reducing them, as far as may be, to general Rules or
Laws, - establishing these rules by observations and experiments, and
thence deducing the causes and effects of things . . ."-1 Sir Isaac Newton
wrote this in the program he proposed to the Royal Society after he
became its president in 1703. It is likely that by "the frame of nature"
he and his readers meant its ultimate ingredients, the original compo-
nents of bodies.2 One is tempted, however, to take the phrase as refer-
ring to the conceptual frame required for the mathematical description
and explanation of natural phenomena. Newton himself put forward
one such frame in the introductory sections of his masterpiece, the Prin-
cipia of 1687 (Newton 1726, pp. 1-27). Without it one cannot make
sense of the single, simple mathematical law by which he accounts
in one breath for heavenly motions and free-fall. Newton's resolve
to make explicit the structure underlying his physics sets him apart
from the other founding fathers of modern physics; we must go back
to Aristotle to find something of comparable breadth and depth. But
Newton's conceptual frame, in stark contrast with Aristotle's, involves
quantifiable, measurable attributes of things and was designed to fit

1
"Scheme for establishing the Royal Society" (University Library Cambridge, Add. MS
4005.2); quoted in Westfall (1980, p. 632).
2
That is apparently the meaning of the phrase in the following passage: "Perhaps the
whole frame of Nature may be nothing but various Contextures of some certaine
aethereall Spirits or vapours condens'd as it were by praecipitation, much after the
manner that vapours are condensed into water or exhalations into grosser Substances,
though not so easily condensible; and after condensation wrought into various formes,
at first by the immediate hand of the Creator, and ever since by the power of Nature"
(Newton, Correspondence, I, 364; quoted by Westfall 1971, p. 364).

41
42 Newton

the needs of mathematics and experiment. It remained the universal

frame of physical inquiry until the advent of Einstein's Relativity, and,
as we shall see in Chapter Three, it supplied a substantial part of the
subject matter of Kant's critique of reason. It is still a paradigm of what
such frames must be, which all who seek to outdo Newton bear in
mind. It is a moot question whether a "frame of nature" in this sense
is not always an invented frame for nature, but Newton and his imme-
diate successors tended to think that he had discovered it.3
The elements of Newton's conceptual frame are mass, force, time,
and space. Let us take them up by pairs.

2.1 Mass and Force

'Mass' is short for the quantity of matter, which, according to
Definition I in the Principia, "arises jointly from its density and its mag-
nitude".4 The mass of a body is proportional - at a given location - to
its weight. If m is the mass of a body moving with velocity v, its quan-
tity of motion is measured by the product ms (Def. II). This definition
recalls Descartes's yet differs from it in two essential respects: the
body's mass is not equated with its bulk, and velocity is consistently
treated as a directed quantity.
The word 'force' (Lat. 'vis', It. 'forza', Fr. 'force') was loosely used
by seventeenth-century natural philosophers to mean an effective
source of activity. Thus we hear about the force of oars and the force
of a magnet, the force of a spring and the force of a blow, "the force
of the wheels pulled by the weight" of a pendulum clock and "the law
. . . that bodies conserve the force which causes their center of gravity
to rise to the height from which it descended". Descartes protested that
by "force" he did not mean "the power called the force of a man when

3
In his reply to Cotes's letter of 18 February 1713, Newton wrote that "the first Prin-
ciples or Axioms which I call the laws of motion . . . are deduced from Phaenomena
and made general by Induction" (quoted in Koyre 1965, p. 275). But, of course, phe-
nomena must be described and therefore conceived if they are to be used as premises
in a deduction; and, as we shall see in §§2.1 and 2.2, Newton's description of the phe-
nomena of motion essentially depends on the analysis expressed in the said "Princi-
ples or Axioms".
4
Some have objected that this definition is circular, for 'density', they say, must be
understood in terms of 'mass'. However, I do not see why Newton, who invented the
calculus, could not have conceived density as a scalar field (a real-valued function of
location in space) and mass as the integral of density over volume. I dare say that
Definition I furnishes eloquent evidence that he did so, even if there is no other.
 2.1 Mass and Force 43

we say that such a man has more force than another" (AT II, 432); but
quite obviously he often meant just that.5 We saw in §1.5.1 that Leibniz
sought to fix the meaning of the word in a physically significant
way and designated by 'force' what we now call 'kinetic energy' (or
perhaps 'energy' in general). Newton too struggled to attach a precisely
quantifiable meaning to 'force' or 'vis'. He succeeded so well that the
concept of force that is taught to fledgling students of physics even to
this day can be traced directly to his.6
In an unfinished manuscript, "On the gravity and equilibrium of
fluids" (c. 1668), Newton proposes the following definition:
Def. 5. Force is the causal principle of motion and rest. And it is either
an external principle that generates or destroys or in some way changes
the motion impressed in a body; or an internal principle by which the
motion or rest vested (inditam) on a body is conserved, and by which
any being endeavors to persevere in its state and fights back if hindered
[et impeditum reluctatur).
(Hall and Hall 1962, p. 114)
This is echoed in Principia, Definitions III and IV, which character-
ize the vis insita {inherent force) of matter as that "by which every
body, insofar as it is up to it, perseveres in its state of rest or of uniform
straightline motion", and the vis impressa (impressed force) as "an
action, exercised on a body, to change its state of rest or of uniform
straightline motion" (1726, p. 2). Newton's remarks on these two kinds
of force deserve close attention. The vis insita in a body is always pro-
portional to its quantity of matter or mass (which Newton, in opposi-
tion to Descartes, distinguishes from its bulk). It differs from the inertia
of mass only in the way in which it is conceived. "Due to the inertia
of matter it is difficult to perturb the state of rest or motion of
any body." Hence, the inherent force may also be called vis inertiae
('force of sluggishness'). A body exercises this force only when its state

5
All the examples are culled from the appendices on the use of the term 'force' by
Galileo, Descartes, and Huygens in Westfall (1971). The first two quotations are from
Huygens (OC, XVIII 95) and (IX 456), respectively. Westfall says that "'force' in
Borelli's usage was more an intuitive term, which referred to any apparent source of
activity, than a precise technical one" and that it is often impossible to distinguish
'force' from 'strength' (1971, pp. 542-43); I should say that this is also true of the
other authors mentioned. Of course, the core meaning of Latin vis and its equivalents
in romance languages comprises both 'force' and 'strength'.
6
For a concise and most illuminating study of seventeenth-century concepts of force,
see de Gandt (1995, Ch. 2).
44 Newton

is changed by another force impressed on it. The exercise of vis iner-

tiae is, under different respects, both resistance and impulse: "resis-
tance, insofar as the body, to preserve its state, fights (reluctatur) the
impressed force; impulse, insofar as the same body, yielding with
difficulty to the force of a resisting obstacle, tends to change the state
of this obstacle." As to the vis impressa, it "consists in the action only,
and does not persist in the body after the action." Hence, the new state
induced by the impressed force is maintained in the body solely by the
force of inertia. Vis impressa has diverse origins, such as blows, pres-
sure, or "centripetal force", that is, that force "by which bodies are
drawn or impelled or in any way tend from any place towards a certain
point, as to a center" (1726, p. 3).
Newton's remarks may not seem altogether consistent, for first he
says that a body's vis inertiae is exercised only when it resists other
things and reacts on them, and later he adds that the vis inertiae is
what continually keeps the body in its current state of rest or motion.
But perhaps such uneventful state preservation is not what Newton
meant by 'exercising a force'. If, for a moment, one ignores the refer-
ence to centripetal force, one may feel inclined to understand Newton's
basic scheme as follows: Each body is endowed with an inherent force
that is a principle of motion and rest that keeps it moving in a fixed
direction with constant speed > 0; when two bodies collide, the inher-
ent force of each translates into an impulse that is impressed on the
other. This is reminiscent of the Aristotelian world in which some
bodies are compelled to move in a forced way as a consequence of the
inherent motion of others; although, of course, for Newton as for
Descartes, any inherent motion is uniform and rectilinear. But if
Newton ever toyed with such an idea, he must soon have dropped it,
for it is beset with difficulties. The resistance that a Newtonian body
opposes any attempt to change its state of motion or rest depends on
the body's mass, not on its velocity. On the other hand, the effect that
such a body has on another body that stands in its way depends both
on its mass and on the velocity with which it moves relatively to the
other - indeed, as Leibniz showed, some effects are proportional to the
squared velocity. In the light of these facts it is very difficult to view
even the impulsive forces that show up in collisions as an expression
of the vis inertiae of the colliding bodies. Moreover, the definition of
force in "On the gravity . . . " suggests that already at its fairly early
date Newton thought of the external principle of motion as primary,
while the internal principle, the vis inertiae, merely preserves what a
 2.1 Mass and Force 45

body has received from the former. This is clearly stated in one of the
queries added by Newton to the 1717 edition of his Opticks:

The Vis inertiae is a passive Principle by which Bodies persist in their

Motion or Rest, receive Motion in proportion to the Force impressing
it, and resist as much as they are resisted. By this Principle alone there
never could have been any Motion in the World. Some other Principle
was necessary for putting Bodies into Motion.
(Newton, Opticks, p. 397)

Indeed, only by regarding impressed forces as primary - and not nec-

essarily rooted in some inherent property of matter - was Newton free
to conceive his revolutionary idea of a centripetal force that pulls a
distant body to a mere point in space.
If inertia is "a passive principle", it is no wonder that physicists
stopped referring to it as a force. 'Force' - in current parlance, 'New-
tonian force' - was promptly equated with what he called 'impressed
force'. And even this was not understood by his followers exactly as
in his text. To see the difference we must turn to the "Axioms, or Laws
of Motion" in which Newton specifies the relation between force and
motion. In Principia these Axioms are very sensibly placed after the
Scholium in which he explains the notions of time and space, for
without these notions the Axioms are unintelligible. However, after
200 years of relentless public education, Newtonian space and time
have become part and parcel of common sense, at least in the coun-
tries where the present book will circulate, so I may well wind up my
discussion of force before tackling the Scholium. Indeed, as I hope
to show, the Scholium itself only makes the right sense in the light of
the Axioms, so there is some advantage in stating them first. Here
is Newton's text, in Clifford Truesdell's translation:

LAW I. Every body continues in its state of rest, or of uniform motion

straight ahead, unless it be compelled to change that state by forces
impressed upon it.
LAW II. The change of motion is proportional to the motive force
impressed; and it takes place along the right line in which that force
is impressed.
LAW III. To an action there is always a contrary and equal reaction; or,
the mutual actions of two bodies upon each other are always equal
and directed to contrary parts.
(Truesdell 1968, pp. 88-89; cf. Newton 1726, pp. 13-14)
46 Newton

Except for the "unless" clause, Law I is none other than the princi-
ple of inertia that we met in §1.3. The "unless" clause is a timely
reminder of a remarkable fact: In a Newtonian world no body can ever
continue, even for a very short while, in a state of rest or of uniform
motion in a straight line,7 unless indeed it is infinitely distant from all
other bodies. For every body is attracted toward every other body by
a force inversely proportional to their distance squared. Law I does not
therefore reflect observations of the uniform rectilinear motion of
bodies that are not acted on by forces - according to Newton, there
are no such bodies to be seen - but rather a decision to analyze every
actual motion into two contributing factors: the present velocity along
the tangent to the observed trajectory, and the change that it is under-
going. The former is conceived as a state that the mobile possesses and
will naturally keep unless disturbed, while the latter is to be accounted
for by an external force. But why should the present velocity be recti-
linear? Descartes argued that at any given instant a body can be deter-
mined to move in a particular direction but not along a particular curve
(1644, II §39; AT VIII, 64; quoted in §1.3). Commenting on Descartes's
text, I noted that a body endowed with a suitable set of instantaneous
tendencies to move in a given direction, to change that tendency at
a certain rate, to change that rate of change at a certain rate, and so
on, can thereby be determined to move along a fixed trajectory of any
conceivable shape. Newton's incisive analysis disposes of this objec-
tion: All those tendencies may indeed be alive at a given instant, but
only the first is to count as the body's actual state of motion; the second
displays the present action of external forces, while the third, fourth,
and so on, must be accounted for by the changing configuration of such
forces. The decision to see things in this way committed physics to
find those forces or, at the very least, the equations linking their pres-
ence and evolution to observable physical quantities. To endorse
Newton's analysis was therefore tantamount to signing an enormous
promissory note. But Newton's first installment payment was so

7
Descartes asserted that the same was true in his world; but the reason he gave does
not hold water in his physics. He noted that, since there is no void, a body B1 cannot
move unless another body B2 yields its place to it. B2 in turn will take the place yielded
by £3, etc. Such replacements ultimately require that a body Bn (for some finite index
n) takes up the place that Bl is vacating. Therefore, concludes Descartes, in the real
world all motion is by circulation. But this conclusion is quite unnecessary if matter,
as postulated by Descartes, extends indefinitely in every direction.
 2.1 Mass and Force 47

magnificent and subsequent contributions by his followers so success-

ful that his scheme of analysis was generally accepted as the naked
truth of the matter.
Except in history books, Newton's Law II is never stated today in
his own terms. Consider a body of mass m that is now moving with
velocity v. Let p stand for the (quantity of) motion mv. According to
Newton's text, if the motion presently changes to p + Ap, the change
is proportional to and collinear with the force impressed on the body:
Ap oc F. This approach to motion change is well suited to the case
of collisions in which a body's motion is altered by the seemingly
instantaneous action of another body. However, it works rather poorly
if the change is brought about by a centripetal force, acting continu-
ally. Evidently, the magnitude of the change in motion effected by
a force of this kind depends on the length of time At during which
it acts. It is more reasonable, therefore, to put F = {At and to con-
ceive the force as the (directed) quantity represented by the vector f.
In these terms, f °c Ap/At. Indeed, since the centripetal force on a body
B changes its direction from instant to instant as B moves, the stated
relation is ambiguous or perhaps meaningless unless one puts on the
right-hand side, not the quotient Ap/At, but the limit p to which that
quotient converges as At decreases indefinitely: thus alone will the left-
hand side stand for the actual force acting on B at any given moment
(cf. Newton 1726, p. 38; pis written dp/dt in the now current nota-
tion invented by Leibniz). If, by a good choice of units, we make the
constant of proportionality equal to 1, we obtain the now current for-
mulation of Law II:
f = p = mv (2.1)
The last term displays matter's resistance to change in motion: The
acceleration v due to a given force f is inversely proportional to the
mass m. Equation (2.1) is appropriate also in the case of collisions, as
a moment's reflection will show. Since Newton and his contemporaries
did not assume that a colliding body's motion p can suffer a finite vari-
ation in no time at all, Ap must be regarded as the cumulative effect
of a force acting during a finite interval At, be it ever so short. So one
should write Ap = f At or, making allowance for a time-dependent force,
Ap = Jo'f(£)d£. The latter quantity - which I previously designated by F
- is commonly known as impulse. But there is no question that Newton
himself called it force (vis). Take, for instance, this passage from the
Scholium to the Laws of Motion:
48 Newton

When a body is falling, its uniform gravity (gravitas uniformis), acting

equally in equal time intervals, impresses equal forces (vires) on that
body, and generates equal velocities; and in the whole time impresses
a total force (vim totam) and generates a total velocity proportional to
the time.
(Newton 1726, p. 21)8
Newton's choice of words suggests that the force generating the body's
velocity is gradually impressed on the body and stored inside it. But
this suggestion is clearly incompatible with Newton's remark that the
impressed force "consists in the action only, and does not persist in the
body after the action" (1726, p. 2; quoted above). This remark suits
well the quantity I called f, made manifest by the instantaneous rate of
change of the motion as in eqn. (2.1). There are passages in Principia
where 'force' (vis) can only mean f. Take the first sentence of the proof
of Book II, Prop. XXIV: "For the velocity which a given force (vis) can
generate in a given matter in a given time is directly proportional to
the force and the time and inversely proportional to the matter" (1726,
p. 294). This entails that the force is directly proportional to the matter
and the velocity generated and inversely proportional to the time.
Obviously here 'force' denotes f, not F. With the symbols we used
above, the last statement can be written as f <* mAv/At, which yields
eqn. (2.1) at the limit At -» 0.9
The argument leading to eqn. (2.1) also motivates the standard
understanding of Law III as a statement about force (f) and counter-
force (-f). Note that Law III speaks of actions exerted on bodies by
bodies.10 Thus it is devoid of application if, as some learned commen-

In Motte's translation, 'gravitas uniformis' is rendered as 'the uniform force of its

gravity'. Since 'vires' and 'vim totam' are rendered by 'forces' and 'whole force',
respectively, Newton is made to appear using the one word 'force' to name the two
quantities f and F in a single sentence.
Here is another example. In the corollaries to Lemma X of Book I, Sect. I (1726, p.
34) we read that the distances traversed by bodies urged by different forces are among
themselves, "at the very beginning of the motion, as the products of the forces and
the squares of the times" (Corol. 3); so "the forces (vires) are directly as the spaces
described at the very beginning of the motion and inversely as the squares of the
times" (Corol. 4). Proportionality to s/t2 holds of course for |f| (the magnitude of f)
but not for |F|.
Cf. the following draft of Law III, first published by Herivel (1965, p. 307): "Corpus
omne tantum pati reactione quantum agit in alterum" ("As much as a body acts on
another so much does it experience reaction").
 2.1 Mass and Force 49

tators argue on the basis of other texts, Newtonian matter is incapable

of acting. Anyway, gravity, the one force of nature that Newton was
able to conceive at work in the phenomena of motion, shows up
according to him as an action of each piece of matter on all other pieces
of matter, which in turn react on the former in accordance with Law
III.11 If, as Newton's text openly suggests, Laws II and III hold in the
same way for all forces, every observed acceleration a must be matched
by an observable acceleration b collinear with a but in the opposite
direction. However, b might not be easy to see if a is due to the action
of many scattered bodies, so that b is in turn the acceleration of the
system composed of the latter, resulting from the reactions suffered by
each of them. Let mk be the mass of a body labeled k and a^ the accel-
eration experienced by body k due to the action of body h. Then, by
Laws I and II, in the case of two mutually interacting bodies labeled 1
and 2, we have that

mia21 = -ra2a12 (2.2)

In particular, if a2i = -a 12 , evidently m\ = m2. Thus, in Newton's scheme
of things, two bodies are said to have the same mass if and only if,
when they interact, they impart on each another quantitatively equal
accelerations. If a given body b is agreed to have unit mass, another
body B has mass m if and only if, when interacting with fe, it imparts
on it an acceleration that is equal to the acceleration imparted by b on
B multiplied by the scalar factor -m. Ernst Mach (1868) regarded
this as the only viable definition of the Newtonian concept of mass
(see §4.4.3).
After the Laws of Motion, Newton states six "Corollaries". The
name suggests that they can be easily inferred from the Laws, but
this is not wholly obvious. Corollary I says that "a body, acted on by
two forces simultaneously, will describe the diagonal of a parallelo-
gram in the same time as it would describe the sides by those forces
separately" (1726, p. 14). Of course this can hold only if the lines along
which the forces act lie on the same plane, but Newton does not
mention this condition, perhaps because he tacitly assumed that the
forces are both applied at the same point, as indeed they must be if
they act on a dimensionless body (a "particle" in the sense defined in

I shall therefore occasionally say that a gravitational force is exerted on a body by

another body. Readers who feel uncomfortable with this manner of speech should
substitute 'from' for 'by'.
50 Newton

Figure 7

§2.2).12 It follows that two such forces, represented by vectors f and g,

have the same effect on the body as the force represented by their vector
sum f + g, constructed as in Fig. 7.
Corollaries V and VI are quoted and discussed in §§2.2 and 2.3,
respectively.

2.2 Space and Time

In a much quoted passage of the Preface to Principia (1687), Newton
says that the whole task of philosophy consists in this: "From the phe-
nomena of motion to investigate the forces of nature, and then from
these forces to demonstrate the other phenomena". The Laws of
Motion set up the links required for performing the first part of this
task: Any body that deviates from rest or uniform motion in a straight
line bears witness to the action of external forces, as well as to their
magnitude and direction. However, without further specifications, it is
impossible to ascertain, let alone to measure, such deviations.
For the sake of clarity and conciseness I shall introduce a few terms
that became familiar in subsequent discussions of Newtonian mechan-
ics. By a particle I mean a body of negligible length, width, and depth.
By a rigid body I mean a collection of at least four particles, not all on
the same plane, whose mutual distances do not vary. Newton's Laws
of Motion must be understood as referring primarily to particles;
indeed, they cannot be applied without more ado to nonrigid bodies.
A particle is said to be free when it is not subject to impressed forces.

12
Lagrange (1788, p. 6) states the principle of composition of forces without saying
that for it to make sense the forces must be coplanar. But then on p. 50 he explains
that "up to now we have considered bodies as points".
 2.2 Space and Time 51

So Law I says in effect that a free particle which is not at rest moves
uniformly in a straight line.
Talk of uniform motion makes no sense without a standard of time
by which successive intervals can be pronounced equal. The Greeks
found such a standard in the rotation of the firmament about the
poles;13 but for Newton and his contemporaries that motion merely
reflected the actual rotation of the earth, which - if governed by New-
tonian mechanics - achieves very nearly but not perfectly uniform
speed.14 So Newton made it clear that time, as understood in his Laws
of Motion, is "absolute, true, and mathematical time," which "in itself,
and from its own nature, flows equably without relation to anything
external," and should be carefully distinguished from "any sensible and
external (whether accurate or inequitable) measure of duration by
means of motion" (Newton 1726, p. 6).15 Still, "true time" cannot play
a role in our physics unless it is displayed in a definite way by the phe-
nomena of motion. Now, Law I provides just that: A free particle in
motion traverses equal distances in equal times. This does not mean,
however, that the uniform motion clause of Law I is merely a conven-
tion bestowing physical meaning on Newton's mathematical time. The
Law is supposed to hold for all free particles. Hence, if one free parti-
cle is conventionally chosen to display real time, the other free parti-
cles will bear witness to the Law's validity. This view of Law I was put
forward by Carl Neumann (1870). It raises an important question,
which was noted by James Thomson. The arrival of a free particle
at successive equidistant points on its path marks the completion of
successive equal time intervals. This information, however, should be
available everywhere. How can we collect it, say, at the origin of the

13
Greek scientists were so sure about it that when they realized that the circumpolar
motions of the firmament and of the Sun do not keep a constant but a periodically
fluctuating ratio, they unhesitatingly concluded that the Sun's apparent motion was
not uniform.
14
The angular momentum of the earth is being very slowly but steadily eroded as a
consequence of tidal friction. In the short run, of course, the earth's angular momen-
tum is practically constant, so its angular velocity must fluctuate if there are small
changes in the distribution of the earth's mass about its axis; this happens all the time,
as a significant part of the earth's water gets reapportioned among the polar caskets,
the oceans, and the atmosphere.
15
Note, however, that Newton's absolute time, although free from any link to particu-
lar motions, is nevertheless structurally indistinguishable from the time of Greek
mathematical astronomy and so, from the Euclidian straight line, the covering space
of "the circle above" (Sophocles's term for the firmament at Philoctetes 815).
52 Newton

Figure 8

motion? In other words, how can one determine which events at the
origin are respectively simultaneous with the particle's arrival at the
said points? Evidently, this can only be done by signals transmitted
between distant points. Thomson comments: "The time required in the
transmission of the signal involves an imperfection in human powers
of ascertaining simultaneity of occurrence in distant places. It seems,
however, probably not to involve any difficulty of idealising or imag-
ining the existence of simultaneity" (1884, p. 569). Newton, at any
rate, took it for granted. "Every moment of time - he wrote - is dif-
fused indivisibly throughout all spaces" (Hall and Hall 1962, p. 104).16
Talk of straight line motion must also be referred to some standard
of rest or spatial frame of reference, apart from which it is meaningless.
Consider a rigid body Bu consisting of three mutually perpendicular
straight edges, xu yu Zu that meet at one point. Let B2 be a copy of Bu
with edges x2, Ji, z2, such that Zi is always aligned with Z\* Suppose that,
when viewed from Bu B2 appears to rotate about the Zi-£2-axis (Fig. 8).
Obviously, if Law I holds when referred to one of our bodies it does not
hold if referred to the other. For if a free particle moves with constant
speed, say, along the straight line containing xu it cannot describe a
straight line at rest relative to B2. Newton resolved such uncertainties
by assuming that his Laws of Motion describe displacements in
"absolute space", which, "without relation to anything external,
remains always similar and immovable". Absolute space must not be
confused with any relative space "defined by our senses by its position
with respect to bodies" (1726, p. 6). Newton claims that, when thus
understood, his Laws provide of themselves the means for distinguish-
ing between "true" motions in absolute space and "relative" motions in

16
Curiously enough, when Thomson raised the question of distant simultaneity, Michel-
son had just invented a laboratory device that, by increasing beyond all contempo-
rary expectation the precision of optical measurements, would wreak havoc with
Thomson's facile solution. See §5.1.
 2.2 Space and Time 53

relative spaces. Specifically, absolute and relative motion are distin-

guished from one another "by the forces of receding from the axis of
circular motion", which do not exist if the circular motion is merely rel-
ative, but are proportional to the quantity of the motion if it is true and
absolute (1726, p. 10; cf. the quotation in §4.4.3). Thus, if our body Bx
is at rest in absolute space, B2 will suffer stress along x2 and y2 while
X\ and ji remain unstressed, although they are rotating about the
Z\-z2-axis in the relative space defined by B2.
Newton's clever appeal to his own Laws tofixthe space for which they
are meant to hold does not work quite as he says. Suppose that B2 rotates
about the Zi-z2-axis as above while this axis is parallelly translated with
constant speed in absolute space. Then, according to Newton's Laws, x2
and y2 will suffer exactly the same stresses as before, and xx and yx will
be unstressed even though they are moving together with zx. This follows
at once from Corollary V to Newton's Laws of Motion:
The motions of bodies included in a given space are the same among
themselves, whether that space is at rest, or moves uniformly forwards
in a straight line without any circular motion.
(Newton 1726, p. 20)
Corollary V embodies Newton's principle of relativity,17 by virtue of
which the Laws of Motion hold equally well when referred to any
member of an infinite family of frames - known as 'inertial frames' -
moving uniformly in straight lines past one another. Ludwig Lange
(1885; see also J. Thomson 1884), defined them on the analogy of
Neumann's time standard: Three free particles, A, B and C, traveling
from a point in three noncollinear directions, determine an inertial
frame 8F on which one may define polar coordinates with origin at A,
axis through A and B, and meridian cp = 0 on the plane ABC.18 The First
Law holds then by definition for particles A, B, and C but as a matter
of testable fact for every other free particle. Moreover, it can be proved

17
Better known as the Principle of Galileian Relativity, presumably because Galileo elo-
quently argued that by observing the phenomena of motion in a closed room one
cannot tell whether one is standing on land or on a ship smoothly sailing on a tranquil
sea (EN VII, 212-14). However, an essential feature of Galileo's comparison - as he
saw it - is that the ship moves uniformly in a constant direction on the spherical surface
of the earth. Cf. the various references to Galileo's "Law of Inertia" in §1.4.
18
Lange does not specify that the trajectories of the three particles must not be collinear,
but his scheme does not work without this requirement (made explicit by Robertson
and Noonan 1969, p. 13). On the other hand, the condition that the three trajecto-
ries be noncoplanar - stipulated by von Laue (1955, p. 3) - is needlessly strong.
54 Newton

by sheer mathematical argument that any frame moving uniformly and

rectilinearly with respect to 2F is also inertial by this definition.
It appears therefore that Newton's notion of absolute space, which
has raised so much dust in philosophical debates, is actually otiose.
However, in Newton's time mathematical discourse was not nimble
enough to articulate the idea of inertial frames, that is, of relative spaces
definable by their positions with respect to rigid bodies but extending
to infinity and moving uniformly in one another with every conceiv-
able velocity. On the way to this idea, the quaint conception of absolute
space, which is traceable to late medieval thinkers19 and was well
entrenched in Cambridge c. 1660, served in effect as an excellent
crutch. Indeed, what Newton himself had to say about the nature of
space is, as we shall now see, very helpful for understanding frames
(and much else that is essential to mathematical physics).
In the medieval ontology taught in seventeenth-century schools,
everything was either a substance or a so-called 'accident', that is, an
attribute or relation of substances. A substance in the world could be
created or annihilated by God without regard to the existence of other
worldly substances. On the other hand, attributes rested on the sub-
stances to which they belonged. Since a real interdependence between
created substances would restrict God's power to act on one without
touching the others, there was a strong tendency to think that relations
among such substances existed only in the eye of the beholder. Con-
trary to all expectations, Newton maintains that space "has its own
manner of existence, which fits neither substances nor accidents" (Hall
and Hall 1962, p. 99). For space is incapable of acting like a substance
- that is, of thinking like a mind or moving like a body - and yet it is
not an attribute of a particular substance but rather something that is
shared by all, "a disposition of being qua being"; for "no being exists
or can exist which is not related to space in some way: God is every-
where, created minds are somewhere, and body is in the space that it

19
Bradwardine (fl349), Oresme (fl382), Crescas (cA340~cA4U). R M. Cornford
(1936) equated modern absolute space with the Greek atomists' void, which, as he
noted, made its appearance simultaneously with the science of geometry. However,
Greek geometry is not the science of infinite space, but offinitefigures.And there are
some deep differences between the atomists' void, existing from eternity outside
bodies, and the space of Newton (and medieval theologians) in which God placed all
bodies at creation. Anyway, it is likely that cinquecento natural philosophers like
Bruno - who wholeheartedly embraced absolute space - drew inspiration from the
atomist Lucretius, who referred to the void as 'spatium inane'.
 2.2 Space and Time 55

fills" (p. 103). A deeper intimation of the ontological peculiarity of

space is contained in Newton's proof of its immobility:

Just as the parts of duration are individuated by their order, so that (for
example) if yesterday could change places with today and become the later
of the two, it would lose its individuality and would no longer be yester-
day, but today; so the parts of space are individuated by their positions, so
that if any two could exchange their positions, they would also exchange
their identities, and would be converted into each other qua individuals.
It is only through their reciprocal order and positions (propter solum
ordinem et positiones inter se) that the parts of duration and space are
understood to be the very ones that they truly are; and they do not have
any other principle of individuation besides this order and position.
(Hall and Hall 1962, p. 103)

Newton's conception of a multitude of entities that are precisely the

individual entities they are only by virtue of their mutual relations flies
in the face of medieval ontology.20 On the other hand, it is essential to
modern mathematical physics, for any realization of a mathematical
structure will in a way satisfy this conception. Now, a structure's rela-
tional system does indeed individuate its elements, but only up to iso-
morphism. This provides all of the individuation that we are likely
to need if there are no internal isomorphisms besides the identity.21

20
Of mundane things, that is, for the Christian Trinity, if at all conceivable, surely must
come under some such conception.
21
Let S and T be two realizations of the same type of structure. An isomorphism 0:
S -> T is a structure-preserving one-to-one mapping of S onto T, that is, roughly
speaking, an assignment of an element §(x) of T to every element x of S such that (i)
each element of T corresponds to one and only one element of S, and (ii) if x, y, z,
. . . e S have a certain structural property or hold a certain structural relation to one
another, 0(x), 0(y), 0(z),. . . e T have the homologous property or hold the homolo-
gous relation. The isomorphism is internal if T = S. The standard term for internal
isomorphism is 'automorphism'. Obviously, the identity mapping x \-^ x, which
assigns each x e S to itself, is an automorphism of S. As a simple example of a struc-
ture that admits no automorphisms besides the identity consider the ring of integers
modulo 3, i.e., the set {0, 1, 2}, with addition defined by0 + a = a + 0 = a, 1 + 1 =
2, 1 + 2 = 2 + 1 = 0, 2 + 2 = 1, and multiplication defined byOx<z = <2xO = O,
1 x 1 = 1, 1 x 2 = 2 x 1 = 2 , 2 x 2 = 0. To see this, note that if 0 is an isomorphism
of this ring onto itself, <|>(0) = 0 (if <|>(0) = 1, <|>(1) * 1, and <|>(0) x <|>(1) = 0(1) * §(0),
so 0 is not an isomorphism; and the same would hold, mutatis mutandis, if 0(0)= 2).
Thus, unless 0 is the identity, 0(1) = 2 and 0(2) = 1, in which case 0(2) x 0(2) = 1
* 0(1), so 0 is not an isomorphism. On the other hand, the group of integers modulo
3, i.e., the set {0, 1, 2}, with addition defined as above but without multiplication,
admits the automorphism 0(0) = 0, 0(1) = 2, 0(2) = 1, as the reader can easily verify.
56 Newton

However, Newtonian - that is, Euclidian - space admits an infinity of

distinct internal isomorphisms, namely, by translation (of each point
by the same distance along parallel directions), rotation (of each point
by the same angle about the same center), reflection (mirror-imaging
respect to a given plane), or any combination of mappings of one or
more of these three kinds. Therefore, a single realization of Newton-
ian space contains infinitely many copies of itself and its points are
individuated by their mutual relations only in the context of one or the
other of those copies. In particular, if we designate one of these copies
by % and we represent by the vector v a translation of each point of %
in the direction of v by a distance equal to v's length, then, if the para-
meter t ranges over the real numbers, the translations ts yield the suc-
cessive positions of a frame %y moving through % with constant velocity
v. If % is inertial in Lange's sense, so is %y. Since % and v are arbitrary,
in this - admittedly anachronistic - approach all inertial frames are
equivalent.22
Newton's remarks on the ontology of space are contained in the
manuscript "On the gravity . . .", first published in 1962. That is why
his understanding of space as a self-contained relational system has
been traditionally ascribed to Leibniz.23 Newton did however commu-
nicate to the public his thoughts on the link of space to God. In the
General Scholium appended to the second edition of Principia he says
that God "endures forever, and is everywhere present; and, by existing
always and everywhere, he constitutes time and space" (1726, p. 528).
In Queries 28 and 31, at the end of the Opticks, he is more specific.
He resorts to the medieval notion of the sensorium of animals, which
he describes as "that place to which the sensitive Substance [the per-
ceiving soul - R. T ] is present, and into which the sensible Species
[aspects - R. T ] of Things are carried through the Nerves and Brain,

22
We reach a more natural understanding of inertial frames and their equivalence in
Newtonian physics if, with even bolder anachronism, we start not from Newtonian
space and time, but from neo-Newtonian spacetime. See, for example, Ellis and
Williams (1988, pp. 6-13), Friedman (1983, pp. 71-86), or Torretti (1983, pp.
20-31).
23
But Leibniz would not dare to think that a relational system has "its own manner of
existence which fits neither substances nor accidents", and maintained that space "is
a mere ideal thing", which men conceive by forgetting bodies and paying attention
only to their abstract order of coexistence. See Mr. Leibnitz's Fifth Paper, §47, in The
Leibniz-Clarke Correspondence (Alexander 1956, pp. 69-72).
 2.3 Universal Gravitation 57

that there they may be perceived by their immediate presence to that

Substance"; and he goes on to speak of "a Being incorporeal, living,
intelligent, omnipresent, who in infinite Space, as it were in his Sensory,
sees things themselves intimately, and throughly (sic) perceives them,
and comprehends them wholly by their immediate presence to himself"
(Opticks, p. 370). This "powerful, ever-living Agent [.. .] being in all
Places, is more able by his Will to move the Bodies within his bound-
less uniform Sensorium, and thereby to form and reform the Parts of
the Universe, than we are by our Will to move the Parts of our own
Bodies" (Opticks, p. 403). I cannot say that I understand these theo-
logical pronouncements. I mention them only because we shall find
them echoed, with an anthropic twist, in the philosophy of Kant.

2.3 Universal Gravitation

Newton's most celebrated achievement is his discovery of universal
gravitation, or, in plain English, the heaviness of everything. But
Newton does not understand heaviness (Lat. gravitas) in the Aris-
totelian sense, as the natural tendency of a body to move with accel-
erated motion toward the center of the universe. Newtonian gravity is
doubly universal in that, by virtue of it, every body is accelerated
toward every other body at once. Thus bluntly expressed the idea
sounds incredibly confusing, but placed in the Newtonian frame
described in §§2.1 and 2.2 it is not so. By Law II, the acceleration a2i
of a given particle pi toward another particle p2 discloses the action of
a force f2i impressed on px and pulling it toward p2, in accordance with
eqn. (2.1):
f2i = w i a 2 i (2.1*)
Newton unhesitatingly treats f2i as a force issuing from p2 that is there-
fore matched, according to Law III, by a counterforce f12, issuing from
pi and impressed on p 2 , such that
fi2 = m2a12 = -m1a21 = -f21 (2.2*)
Still, this (anachronistic) display of vector algebra would be a false pre-
tence had Newton not produced a "law of force" stating precisely how
fi2 depends on the positions and the masses of pi and p2. Equation (2.3),
commonly employed for expressing this law, is not found in Newton's
writings in any even remotely similar form, but there is no doubt that
58 Newton

it faithfully conveys the sense of his words, as illustrated by his prac-

tice, if Laws II and III are correctly represented by eqns. (2.1) and (2.2).
Let m{ denote the mass of particle pi and r{(t) its position at time t in
the chosen inertial frame of reference (/ = 1,2). For simplicity's sake I
formulate eqn. (2.3) for a time t at which p2 is momentarily located at
the origin of our frame, so that t2{t) = 0. ti(t) is therefore a vector point-
ing from p2 to pi and its length \ti(t)\ equals the distance between the
two particles at t. Then, the gravitational force exerted by p2 on px at
time t is given by:

in(t) = -Gm1m2^J (2.3)24

kiwi
G is a constant of proportionality that can be made equal to 1 by a
clever choice of units (but, of course, if natural phenomena are subject
to several distinct "laws of force", one cannot expect all such constants
to be simultaneously eliminable in this way). Substituting the right-
hand side of eqn. (2.3) for the left-hand side of (2.2*) we see at once
that mi occurs on both sides and may be canceled. In this way,
Newton's law of gravity does justice both to the ordinary feeling that
heaviness goes hand in hand with massiveness and to the experimen-
tal fact that all bodies fall with the same acceleration at any given place
on earth.
Using Corollary I to the Laws of Motion (§2.1 ad finem), we can
infer from eqns. (2.3) and (2.2) the acceleration ak(t) experienced by
each member pk of a system of n particles due to the gravitational forces
exerted on it at time t by the other n - 1 particles:

fs

rt(t)\
If the only forces impressed on each particle are the gravitational forces
issuing from the others, ak(t) is the total acceleration of pk, and is there-

24
We write Ti in the numerator to convey the direction of the force f2i; but by so
doing we make this force proportional to the distance Ir^ between the particles.
So, to express that i2i is inversely proportional to |rj|2 we must write |ri|3 in the
denominator.
25
The Greek letter X (sigma) indicates summation; the summands are obtained by
replacing the index / in the term on the right of 2 by each number in the range indi-
cated under X, which in this case is every number from 1 to n except k (so there are
n — 1 summands).
 2.3 Universal Gravitation 59

fore equal to vk(t), the second time derivative of rk(t) (that is, the instan-
taneous rate of change of the instantaneous rate of change of p&'s posi-
tion at t). Equation (2.4) holds, of course, for any time t in which the
said conditions hold. Letting k range over {1,. . ., n}, we obtain a
system of n second order ordinary differential equations in n unknowns
rl9 . . . , rn:

(2.5) 26

It is a property of such systems that they have at most one solution for
each set of initial conditions. In other words, if solutions exist on a
time interval I for certain given values of the masses m, and of the posi-
tions and velocities rt{t) and £,•(£) at a particular time t e I, eqns. (2.5)
uniquely determine the values of the r/s and the r,'s at all times in that
interval. Of course, such determinations can be translated into exact
predictions and retrodictions only if the initial data are exact and
eqns. (2.5) are solved exactly.
For all their power, eqns. (2.5) would be of little use - and it would
have been virtually impossible for anyone to light on them - were it
not for the following features of our natural environment as it is
captured in the Newtonian frame:
1. The nongravitational forces acknowledged by Newton and his suc-
cessors (to this day) do not make a contribution to the observable
accelerations of the major bodies - planets, planetoids, satellites,
comets - that circulate around the sun.
2. The action of nongravitational forces on bodies falling near the
earth can be attributed to air resistance or else brought under the
concept of constraints (e.g., to roll along an inclined plane or to
swing at the end of a string; more is said on constraints in §2.5.3);
Galileo taught how to filter out these two kinds of action in the
study of terrestrial gravitation.
3. In a remarkable theorem (Principia Bk. I, Prop. LXXVI), Newton

26
Roughly speaking, a second-order ordinary differential equation in one unknown is
an equation relating known quantities to the second derivative d2cp/d£2 of an unknown
function (p of a single independent variable t and possibly also to its first derivative
dq>/dt and to q> itself. The equation is supposed to hold for the entire domain of the
function, that is, for every value of t for which (p is defined. In the notation that
Newton began to use in his manuscripts c. 1690, if cp is a function of time, the first
and second derivatives of q> are denoted by cp and (p, respectively.
60 Newton

established that, under his Law of Gravitation, the total attractive

force between any two disjoint, spherically symmetric masses Mi
and M2 is equal to the force with which the masses Mi and M2
would attract each other if each was concentrated at the center of
the respective sphere. Thus, if the density of a spherical body
depends only on the distance from the center and not on the direc-
tion, one can study its gravitational action on an external particle
near its surface without paying any attention to the fact that some
parts of the body are closer to the particle than others. Fortunately,
the earth, the sun, and other major bodies of interest for the study
of gravitation can, to a good approximation, be treated as spheri-
cally symmetric in this sense.
4. Although there are much more than two bodies in the world, the
overwhelming preponderance of the earth over the bodies that fall
on it, and of the sun over the bodies that move around it makes the
study of gravitation in two-body systems a realistic option. Thus,
the acceleration of gravity at a particular latitude can be studied on
a pendulum while ignoring the gravitational attraction of small
bodies nearby. Thus, also, Kepler's results on planetary motion can
be accounted for by Newton's Law of Gravity if - and only if - each
planet is viewed as forming with the sun an isolated two-body
system (see notes 28 and 29). This was essential for the discovery
and initial confirmation of the Law, for even today eqns. (2.5) can
only be solved exactly if n = 2.
5. Corollary VI to the Laws of Motion entitles one to deal with a two-
body system as if it were isolated even when it is subjected to the
strong gravitational action of other bodies, so long as this action is
constant in intensity and direction over the region occupied by the
two bodies. For Corollary VI states that
If bodies, moved in any manner among themselves, are urged in the
direction of parallel lines by equal accelerative forces, they will all
continue to move among themselves, after the same manner as if they
had not been urged by those forces.
(Newton 1726, p. 21)

Thanks to Corollary VI, the simple Keplerian laws derivable for iso-
lated two-body systems from the Law of Gravity apply to the system
formed by the earth and the moon, even though they are jointly
accelerated toward the sun, and to the system formed by the sun
and any single planet or comet, even though they jointly suffer the
gravitational pull of distant stars.
 2.3 Universal Gravitation 61

Newton claimed that the Law of Gravity was found by "deduction

from phenomena" and generalized by "induction" (1726, p. 530), in
compliance with the Rules of Philosophy set forth in Principia at the
beginning of Book III. Before discussing these Rules (in §2.4; see the
text quoted there), it will be good to watch him follow them. In
the first place we must note that the six "phenomena" listed right after
the Rules and on which Newton's induction rests are not such as one
may observe by opening one's eyes and paying attention to what one
sees.27 They describe the behavior of diverse components of the Solar
System as reconstructed by Kepler and his successors by laborious
analysis - supplemented with geometrical interpolation - of numerous
astronomical observations performed by Tycho Brahe with the naked
eye and by seventeenth-century astronomers with the telescope. Kepler
elicited from Tycho's data his so-called laws:28
I Every planet travels on an ellipse with the sun at one focus (law of
elliptical orbits).29

27
The title "Phenomena" was first used in the second edition (1713), which also intro-
duced Phenomenon II. In the first edition, the other five phenomena were listed under
the label "Hypotheses", together with the first two Rules of Philosophy and two other
statements that later were either moved or removed. 'Hypothesis' is used here, as in
Euclid, for the assumptions invoked as premises in the proof of the subsequent the-
orems. This usage differs from that of the General Scholium of the third edition,
where - after proudly declaring: "I do not contrive hypotheses" - Newton defines
the term as follows: "Whatever is not deduced from phenomena is to be called an
hypothesis" (1726, p. 530; quoted in context in §2.5.1).
28
Although Kepler's laws merely describe planetary motion, they are a far cry from
sense appearances, as can be seen by the fact that astronomers from Eudoxus to Tycho
did not even dream of them. Indeed, in Newton's own time not all astronomers
acknowledged them. Huygens only accepted Law III and Newton himself was prob-
ably unaware of Law II - or rejected it - as late as 1679 (de Gandt 1995, pp. 84,
283). "Hence it was an unusual and very daring step to erect an astronomical system
encompassing Kepler's three laws, as Newton did"; and they "gained a real status in
exact science" only through him (Cohen 1980, p. 229). For an illuminating account
of Kepler's achievement, see Barbour (1989, pp. 264-351).
29
Newton was well aware that, because the Solar System comprises many interacting
bodies, the law of elliptical orbits cannot hold exactly. In the tract De Motu, con-
taining draft versions of definitions and propositions of Principia, he notes that, even
if one treats the Solar System as isolated, so that its common center of gravity is either
at rest or moves inertially in real space, this center does not coincide with the Sun,
although it lies within it or is always very close to it. "Due to this deviation of the
Sun from the center of gravity the centripetal force does not always tend to that immo-
bile center, and hence the planets neither move exactly in ellipse nor revolve twice in
the same orbit. There are as many orbits to a planet as it has revolutions, as in the
62 Newton

II The straight line joining the planet to the sun sweeps out equal areas
in equal times (area law).
Ill For each planet, the square of the time of revolution is proportional
to the cube of its mean distance to the sun (harmonic law).30
Newton's Phenomena I-IV are reformulations of the area law and
the harmonic law as applying (a) to Jupiter's moons in relation to
Jupiter (I); (b) to Saturn's moons in relation to Saturn (II); and (c) to
Mercury, Venus, Mars, Jupiter, and Saturn in relation to the sun (III
and IV). Phenomenon VI is that the area law is fullfilled by the moon
with respect to the earth, and Phenomenon V is that neither the area
law nor the harmonic law are obeyed by any of the named planets with
respect to the earth. Newton does not mention the law of elliptic orbits
among the phenomena, but it was evidently in Newton's mind as he
developed the mathematical theory of motion under the action of a
centripetal force in Principia, Book I.
Modern textbooks often show how to derive Kepler's three Laws
from eqn. (2.3).31 Newton proceeded in the opposite direction. We
cannot go into details here, but I shall sketch the general drift of his
argument.32
Consider a body B moving around a fixed point O, to which it is

motion of the Moon, and each orbit depends on the combined motion of all the
planets, not to mention the action of all on one other. But to consider all these causes
of motion at once and to define these motions by exact laws allowing of convenient
calculation exceeds - if I am not mistaken - the force of every human intellect. Ignor-
ing those minutiae, the simple orbit which is the mean among all errors will be the
ellipse previously discussed." (Herivel 1965, p. 297; transl. on p. 301).
Kepler's harmonic law fails even for two-body systems, but it too can be recovered
by ignoring minutiae. The situation is as follows: If two bodies ?! and P2 move under
Newton's Law of Gravity around a third body 5, then, if we neglect the interaction
between the first two bodies and denote by jXi and \i2 the ratio of their respective
masses to the mass of S, their periods of revolution 7\ and T2 stand in this relation
to the average distances rx and r2 between each body and S:
372
^ (r2/T2)3/2

The harmonic law holds exactly only if fii = |i 2 .

31
See, for example, Courant (1936, pp. 422 ff.) or Pollard (1976, Ch. 1).
32
Chandrasekhar (1995) provides a detailed commentary of Newton's argument for
universal gravitation, designed for today's reader. Unfortunately, this splendid book
was not available when I planned and wrote the present chapter, so I have used it
only to improve my exposition here and there.
 2.3 Universal Gravitation 63

drawn by a force. Newton proves that the areas swept by the moving
radius OB all lie in the same plane and are proportional to the times
it takes to sweep them; in other words, the body B obeys Kepler's law
of areas (Bk. I, Prop. I; see the appendix at the end of this section). He
proves next that if O is fixed or moves inertially and B moves around
it in accordance with Kepler's law of areas, B is urged by a centripetal
force directed to O (Prop. II). Moreover, if O is the center of another
body C, moving in any way, and B moves around O according to the
law of areas, B is urged by a force compounded of a centripetal force
directed to O plus the sum of all the accelerative forces acting on C
(Prop. Ill; a consequence of Cor. VI). If B moves with constant angular
velocity in a circle, the centripetal force is directed to the center of the
circle; and if several bodies move in this way and their periods of rev-
olution are as the fth power of the radii of the respective circles (cf.
Kepler's harmonic law), the centripetal forces will be inversely as the
squares of the radii, and conversely (Prop. IV and its Cor. 6). Suppose
now that B moves in an ellipse, urged by a centripetal force directed
to a point O inside it. Can we say something about the magnitude
of that force if O is (i) the center of the ellipse or (ii) one of its foci?
Newton shows that in case (i) the force is proportional to the distance
OB (Prop. X), and in case (ii) the force is inversely proportional to OB
squared (Prop. XI). These results are extended to parabolic and hyper-
bolic trajectories. In particular, if the center of the ellipse of case (i) is
removed to an infinite distance, "the ellipse degenerates into a
parabola, the body will move in this parabola, and the force, now
tending to a center infinitely remote, will become constant: this is
Galileo's theorem," that is, his law of free-fall (Newton 1726, p. 54).
Take several bodies moving in ellipses around a point O and urged by
centripetal forces directed to O and inversely proportional to the
square of their respective distances from O; these bodies obey Kepler's
harmonic law, or, as Newton puts it, "the periodic times in the ellipses
are as the fth power of their greater axes" (Prop. XV). "Therefore, the
periodic times in ellipses are the same as in circles whose diameters are
equal to the greater axes of the ellipses" (1726, p. 61; cf. Prop. IV and
its Cor. 6).
In the middle sections of Book I, Newton proves a series of propo-
sitions on the motion of bodies attracted toward a motionless or iner-
tially moving center of force. Many of them tackle from diverse angles
the important astronomical problem of finding a trajectory from a few
given points. Others draw conclusions from hypothetical assumptions
64 Newton

to eventually show contrapositively that such assumptions do not hold

in nature (e.g., Prop. XLV Cor. 1, on the motion of the apsides of the
elliptical trajectory of a body urged by a centripetal force varying
inversely as the rth power of its distance to the center, where r ^ 2).
Then, at the beginning of Section XI, he notes that "very probably there
is no such thing in nature" as a motionless center of attraction, "for
attractions are made towards bodies, and the actions of the bodies
attracted and attracting are always reciprocal and equal, by Law III;
so that if there are two bodies, neither [one] is truly at rest, but both,
being as it were mutually attracted, revolve about a common center of
gravity;" and if there are several bodies, "which either are attracted by
one body, which is attracted by them again, or which all attract each
other mutually, these bodies will be so moved among themselves, that
their common center of gravity will either be at rest, or move uniformly
forwards in a straight line" (1726, p. 160). However, as Newton now
goes on to prove, the abstract theory of centripetally urged motion is
applicable in more realistic settings. Among other things, he proves the
following: (a) If two bodies move in any way under mutual attraction,
"their motions will be the same as if they did not at all attract each
other, but were both attracted with the same forces by a third body
placed in their common center of gravity; and the law of the attract-
ing forces will be the same in respect of the distance of the bodies from
the common center, as in respect of the distance between the two
bodies" (Prop. LXI). (b) Bodies urged by centripetal forces that are
inversely proportional to the square of their mutual distances "can
move among themselves in ellipses, and by radii drawn to the foci
describe areas very nearly proportional to the times" (Prop. LXV). (c)
If three bodies attract each other with forces that are inversely pro-
portional to the square of their mutual distances and the two lesser
ones revolve about the largest one, the lesser body closest to the largest
one describes areas around it that are more proportional to the times
and a figure more approaching to that of an ellipse having one focus
at the largest body if the mutual attractions between any two bodies
are equal and opposite, in accordance with Law III, than if the largest
body is not attracted at all by the lesser ones, or is attracted very much
more or very much less than it attracts them (Prop. LXVI). (d) If any
two bodies A and B attract each other and a system of other bodies C,
D, and so on, with an acceleration that is inversely proportional to the
square of the distance from the attracted to the attracting body, the
 2.3 Universal Gravitation 65

forces issuing from A and B are proportional to their respective masses

(Prop. LXIX).
Section XII proves several theorems on the attraction of spherical
bodies, including these: If to the several points of a given sphere there
tend equal centripetal forces decreasing as the square of the distances
from the points, a particle placed within the sphere is attracted to
the center of the latter by a force that is proportional to its distance
from it (Prop. LXXIII), and a particle situated outside the sphere is
attracted to the center of the latter by a force that is inversely propor-
tional to the square of its distance from it (Prop. LXXIV). On the other
hand, if a particle placed outside a homogeneous sphere is attracted to
the center of the latter by a force that is inversely proportional to the
square of the distance, and the sphere consists of attractive particles,
the force of each particle will vary inversely as the square of the
distance from it (Cor. 3 to Prop. LXXIV). I referred previously to
Prop. LXXVI.
The astronomical phenomena listed at the beginning of Book III can
be readily viewed as instances of the mathematical theory of Book I.33
The phenomena are described and the theory itself was developed with
a view to just this application. As noted earlier, the phenomena concern
what we may call the Jovial system (Jupiter and its moons), the Satur-
nal system, the SP system (the sun and the five major planets known
from Antiquity), and the EM system (earth and moon). By applying
Props. II or III and Cor. 6 of Prop. IV (in Bk. I) to Phenomena I, II and
V, Newton concludes that the Jovial, Saturnal, and SP systems are held
together by centripetal forces tending, respectively, to Jupiter, Saturn,
and the sun, which are inversely proportional to the squared distances
from the circumambulating bodies to the respective centers. In the case
of the SP system, the proportion between the centripetal forces and the
squared distances can also be inferred from the stability of the orbital
apsides, for "the slightest deviation" from the said proportion "should
produce - by Cor. 1 of Prop. XLV in Bk. I - a motion of the apsides
detectable in a single revolution, enormous in many" (1726, p. 395).
From Phen. VI and Props. II or III in Bk. I it follows that the force by

33
Book II of Principia develops a mathematical theory of the motion of bodies in resist-
ing mediums, partly to show that Descartes's vortex theory of gravity is physically
impossible (at any rate within the Newtonian frame). It marks the beginning of fluid
mechanics.
66 Newton

which the moon is retained in its orbit tends to the earth. That this
force is inversely as the square of the distance from the position of the
moon to the earth's center is inferred through a detailed calculation
from the very slow motion of the moon's apogee (i.e., the orbital apside
furthest from the earth). Newton takes next the bold step that joined
heaven and earth: he claims that the force that continually draws the
moon off from a rectilinear motion and retains it in its orbit is none
other than gravity, the force that causes ordinary heavy bodies to fall.
The claim rests on a calculation: Given the moon's distance from the
center of the earth R and its centripetal acceleration G, the earth's
radius r, and the gravitational acceleration of a heavy body on the
earth's surface g, Newton shows that, to a satisfactory approximation,
gIG ~ R2/r2. To better appreciate the import of this calculation, Newton
bids us imagine that the earth has several moons, and that the lowest
of them is very small and so near the earth as almost to touch the tops
of the highest mountains. The calculation implies that the centripetal
force that retains this little moon in its orbit would be very nearly equal
to the weight of any terrestrial bodies found on those mountain tops.
"Therefore if the little moon should be deprived of the motion by
which it advances in its orbit, then, lacking the centrifugal force by
which it persists therein, it would descend to the earth; and that with
the same velocity, with which heavy bodies actually fall on the tops of
those mountains, due to the equality of the forces that make them
descend. [. . .] Therefore since the force on heavy bodies and the one
on the moons are directed to the center of the earth, and are similar
and equal between themselves, they both have (by Rules I and II [see
§2.4]) the same cause" (1726, p. 398).
From this point on, Newton's reasoning flies unhampered. Since the
revolutions of Jupiter's and Saturn's moons about Jupiter and Saturn,
and of the planets about the sun, "are phenomena of the same kind as
the revolution of the moon about the earth", and it has been shown
that the forces that keep the said bodies in their orbits tend, respec-
tively, to the centers of Jupiter, of Saturn, and of the sun, and decrease
with distance in the same way as the terrestrial gravity decreases in
receding from the earth, he concludes that there is a force of gravity
tending to all the planets, that all planets gravitate toward one another,
and that every body gravitates toward each planet (1726, pp. 399,
400). By filling wooden boxes with equal weights of gold, silver, lead,
glass, sand, common salt, wood, water, and wheat, and letting them
oscillate as pendulums from equal threads, 11 feet long, Newton had
 2.3 Universal Gravitation 67

satisfied himself that terrestrial gravity acts equally on all materials.34

Having established that it is no different from the gravity that tends to
the other heavenly bodies, he took these experimental results as
sufficient evidence that "the weights of bodies towards any one planet,
at equal distances from the center of the planet, are proportional to
the quantities of matter which they severally contain" (Bk. Ill, Prop.
VI). The next proposition proclaims universal gravitation:
PROP. VII. Gravity is exerted towards all bodies; it is proportional to
the quantity of matter in each. [. . .] Cor. 2. Gravitation towards the
several equal parts of a body is inversely proportional to the square of
the distance from the particles.
For this twofold conclusion Newton argues as follows: He has proved
that all planets gravitate toward each other and that the gravitational
attraction toward each, considered separately, is inversely proportional
to the square of the distance from the center of the planet to the place
where the pull is exerted. Hence, by Bk. I, Prop. LXIX, the gravity
tending toward the planets is proportional to their respective masses.
Now, all the parts of any planet A gravitate toward any other planet
B, and the gravity of each part is to the gravity of the whole as the
mass of the part is to the mass of the whole. Since every action is
matched by an equal reaction (by Law III), planet B will gravitate
toward all the parts of planet A, and its gravity toward each part will
be to its gravity toward the whole as the mass of the part is to
the mass of the whole. Corollary 2 flows immediately from Bk. I,
Prop. LXXIV, Cor. 3 (quoted on p. 65).

Appendix
To give a foretaste of Newton's style I shall paraphrase here his proof
of Proposition I of Principia, Book I.35 It says that the areas which a
revolving body describes by radii drawn to an immovable center of

34
According to Newton, his results could be wrong at most by 1 part in 1,000.
Braginsky and Panov (1972) confirmed them for solar gravity to 1 part in
1,000,000,000,000. For a recent survey of this matter, see Ciufolini and Wheeler
(1995, pp. 91-97).
35
I wish I could also give the proof of Prop. XI but I lack the room. I heartily encour-
age all readers to try to work their way through it. This has now been made easy
by de Gandt's lucid explanation (1995, pp. 38-41; Prop. XI of Principia Book
I = Problem 3 of Newton's De Motu).
68 Newton

Figure 9

force lie in the same immovable plane and are proportional to the times
in which they are described. To prove it, Newton bids us suppose the
time to be divided into equal parts, in the first of which the body by
its innate force describes the straight segment AB (Fig. 9). In the second
part of the time it would, if unhindered, proceed directly to c along the
segment Be = AB. Choose an arbitrary point S, outside the straight line
through A, B, and c. AABS = ABcS (they have equal bases and equal
heights). Thus, if the body moves inertially, the radii joining it to S
sweep equal areas in equal times. Assume now that when the body
reaches B a centripetal force pulls it suddenly toward S, so that the
body instantly deviates from its original trajectory and continues to
move along the straight line toward C. Draw cC parallel to BS. Then,
by Cor. I to the Laws of Motion, at the end of the second part of the
time the body will be at C, in the same plane with AABS. ASBC = ASBc
= AABS, so the radii joining the body to S sweep equal areas in equal
times. By a similar argument, if the centripetal force drawing the body
toward S acts suddenly when it reaches C, D, E, and so on, causing it
to describe the straight segments CD, DE, EF, and so on, its trajectory
will lie on the same plane and ASEF = ASDE = ASCD = ASBC. There-
fore, in equal times, equal areas are described in one immovable plane;
and, by composition, any sums SABCDS, SABCDEFS, of those areas
are to each other as the times in which they are described. These con-
clusions do not depend on the length of the equal parts into which we
 2.4 Rules of Philosophy 69

divided time in the preceding argument and therefore hold good if we

make such parts arbitrarily short. In the limit, the trajectory of the body
becomes a curved line and the centripetal force by which it is drawn
back from the tangent to this curve acts continually. Still, the areas
described by the radii joining the body to the center of force S lie on
one plane and are proportional to the times of description.

2.4 Rules of Philosophy

Newton could only jump to such stupendous conclusions by following
inference rules of his own devising. He called them 'Rules of Philoso-
phy'.36 Only the first two appeared in the first edition of Principia; Rule
III was added in the second and Rule IV in the third. In Newton's par-
lance, 'philosophy' designates what we would now call 'physics'; on
the other hand, to excogitate and to phrase such rules as his would be
described, in current usage, as a philosophical activity. In all likelihood
Newton found them not by musing on truth and reason in the abstract
but by reflecting on his own intellectual practice. So the statement of
Newton's Rules of Philosophy, which was in a way the founding act
of the modern philosophy of physics, followed on the foundation of
modern physics itself. The Latin text of the Rules can be rendered into
English as follows:

RULE I. One shall allow no more causes of natural things than are both
true and sufficient to explain their phenomena.
RULE II. Therefore, to natural effects of the same kind (ejusdem generis)
one shall - as far as possible - assign the same causes.
RULE III. The qualities of bodies which cannot intensify or weaken and
belong to every body on which it is possible to perform experiments,
shall be held to be qualities of all bodies (corporum universorum).
RULE IV. In experimental philosophy, any propositions gathered by
induction from phenomena shall be held to be true - either accurately
or to the best available approximation (quamproxime) - notwith-

36
In Latin, Regulae philosophandi, that is, properly, 'rules for doing philosophy'. This
title was introduced in the second edition; in the first, as I pointed out in note 27,
Rules I and II are labeled 'hypotheses' and lumped together with five statements later
labeled 'phenomena' and two more statements, one of which was subsequently
deleted, while the other - "That the center of the system of the world is immovable"
- was renumbered as 'Hypothesis V and placed after Prop. X of Bk. III.
70 Newton

standing any contrary hypotheses, until other phenomena occur by

which they may either be made more accurate or liable to exceptions.
(Newton 1726, p. 387)37
These Rules are admirably tailored to vindicate the boldest steps in
Newton's argument. After proving from his mathematical theory and
detailed computation from astronomical data that the moon M is accel-
erated continually by a centripetal force directed toward the center of
the earth E and proportional to (rEM)~2 (where rEM stands for the dis-
tance from E to M), Newton calculated that if rEM were the earth's
radius, the moon's acceleration would be equal to that of heavy bodies
close to the earth's surface (remember his thought experiment with
"little moons" - which is now performed every time an orbiting space-
ship is braked in preparation for landing). So the moon's acceleration
is "of the same kind" as the acceleration of heavy bodies and must
therefore, by Rules I and II, have the same cause, that is, gravity (heav-
iness). But Newton had also shown - again by combining astronomi-
cal data with his mathematical theory of motion under centripetal
forces - that the Jovial, Saturnal, and SP systems are held together by
forces "of the same kind" as the one acting on the moon, that is, by
gravity. By ably wielding the equality of action and reaction (Law III),
he concluded that the force of gravity must issue from every part of
these bodies, since every part experiences it. Since this is shared by all
the bodies on which we can perform experiments,38 we must say, by
Rule III, "that all bodies whatsoever gravitate towards each other"
(1726, p. 388). And, by Rule IV, this result of induction must be held
true in the face of any speculative hypotheses contrived to account dif-
ferently for the same phenomena, until it is improved or overturned by
further inductions.
Newton's Rules of Philosophy usually make a poor impression
on philosophically trained readers. Yet modern physics could only

A draft for "Rule V" has been found in Newton's manuscripts. It is reproduced in
Koyre (1965, p. 272). Newton probably desisted from publishing it because it is not
really a rule of inference, but a characterization of "hypotheses" such as - by Rule
IV - one ought not to allow to prevail over propositions gathered from phenomena.
An equivalent although much shorter characterization was printed in the General
Scholium (see note 27).
In Book III, Newton studies two additional grounds for his induction, viz., ocean
tides and the behavior of comets (but his gravitational account of tides is not alto-
gether satisfactory).
 2.4 Rules of Philosophy 71

get going on some such principles. Philosophers' misgivings about

Newton's Rules arise, I dare say, from mistaking their purpose. If, fol-
lowing the lead of the Councils of the Christian Church, we consider
truth to be a matter of dogma, that is, of neat definitive pronounce-
ments not liable to correction or qualification, we cannot sensibly
expect that Newton's Rules will direct us to find the truth about nature.
But, although Newton very probably shared this dogmatic conception
of truth with almost everybody else in his century, the text of the Rules,
especially of the fourth one, clearly indicates that they are not meant
to yield statements which in that sense are true, but statements that
one should for the time being hold to be true. Of course Descartes,
under the spell of the said dogmatic conception, had decided not to
admit anything as true for which the evidence was not as overwhelm-
ing as the evidence he had of his own existence. But Descartes himself
was not completely faithful to this stringent standard, under which
physics would have been stillborn. So Newton, inspired, as I have
already suggested, by his own practice, developed rules for holding true
- and therefore for building on - statements of whose truth "in God's
eye" he could not be sure.
Before looking more closely into this, I must try to clarify some key
ideas involved in the Rules. As we have seen, the term 'phenomena' is
not used here by Newton in the sense made familiar by "phenomenal-
ist" philosophers. A Newtonian phenomenon is not an eddy in
someone's stream of thought, let alone a congeries of so-called sensa-
tions, but rather a class of processes or states of affairs that, through
the intelligent reading of many observations, we have come to recog-
nize as a real feature of our environment. The six phenomena proposed
by Newton himself under this designation were discovered after long,
painstaking, intellectual work. But some conceptual ordering of per-
ceptions and some judgments are also required for establishing more
obvious phenomena, for example, that the point in the horizon where
the sun rises every morning at a given place moves year after year back
and forth between two fixed limits. Clearly, if the constitution of phe-
nomena proceeds according to some general principles - concerning,
for example, the structure of space and time - these principles are not
reached by inference from phenomena pursuant to Newton's Rules.
Rules I and II concern the assignment of causes to natural effects.
Here again we have a term that students of philosophy, standing in the
shadow of David Hume, normally take in a sense that is very different
from Newton's. In Hume's sense, an object c is the cause of an object e
72 Newton

only if objects of the same kind as e are always contiguous with and
immediately preceded by objects of the same kind as c. But the univer-
sal force of gravity - the one cause that emerges in Principia from the
application of Rules I and II - neither precedes the acceleration of falling
bodies nor can be properly described as contiguous to it. Naturally,
Newton would not use the term 'cause' in 1687 in a sense developed
fifty years later by Hume to fit his own narrow outlook. Newton took
it either (i) in the common anthropomorphic sense in which the cause
of e is the agent that is to blame for e's existence, or (ii) in the traditional
Aristotelian sense in which a cause of e is anything that contributes to
explain e. (Aristotle counted among the causes of a statue not just the
sculptor who made it, but also the purpose for which, the material from
which, and the design after which it was made.) Both (i) and (ii) fit
Newton's force of gravity in its relation to the phenomena of motion,
but the unqualified reference to explanation in Rule I suggests that
Newton understood 'cause' in sense (ii) (but not that he expected every
explanation to fit into one of the four pigeonholes - agent, goal, matter,
form - of Aristotle's scheme).
Rule I states a ban: No more causes should be allowed than are true
and sufficient for explaining the phenomena. Now, it is clear that if a
cause is known to be untrue, that is, not really operative in the case
one desires to explain, it cannot be allowed. On the other hand, if a
cause is known to be true, one should certainly allow it, even if a
sufficient set of causes is already available without it. But the rules are
supposed to guide us in ascertaining the true causes, or at any rate
those that may be held to be true, so the inclusion of truth among the
criteria for this rule's application seems somehow incongruent. What
we have here is really an injunction to stop our inquiry when the expla-
nation achieved is sufficient, coupled with a warning that it must restart
if the explanation turns out to be false. To justify this rule, Newton
invokes the old saying that Nature does nothing in vain, which he takes
to mean that she will not do with more what can be done with less.
One wonders why the omnipotent God whom Newton believed to be
the author of Nature should submit to this principle of economy.
Indeed, even if He chooses to follow it - say, because overexpenditure
is so vulgar - we have no inkling of what is actually less for Him; it
might well be that He finds it costlier to restrain His exuberance than
to go on effortlessly multiplying things beyond necessity. Although
useless for determining which are the true causes in God's eye, Rule I
is perfectly apt as a guide for physical research: If we have found an
 2.4 Rules of Philosophy 73

explanation that, as far as we can see, leaves nothing to be desired, it

would be wrong and confusing for us to seek any further.
Rule II is presented as an obvious consequence of Rule I, but it
brings an additional idea into play: if we are not to accept any more
causes than are sufficient to account for a given effect, we must ascribe
a single (kind of) cause to all effects of the same kind. To use this rule
one must certainly be able to classify effects into kinds. Now this
appears to be purely a question of tact, or - if you wish - of genius,
for there are plainly no clear, firm rules by which to do it, nor can we
simply abide by the classifications inherited from the Stone Age. At any
rate Newton does not abide by them, but analyzes in a novel way a
body's state of motion into inertial velocity and acceleration, and pro-
ceeds to argue, in a fantastic feat of scientific imagination, that the
actual acceleration of any moon or planet does not differ in quantity
or direction from the acceleration that would be experienced by any
heavy body placed in its stead. From this he concludes that the accel-
erations of planets, moons, and falling bodies are effects of the same
kind, so that, by Rule II, all should be assigned the same cause. Rule
II is flawless from a human standpoint, but it cannot lead to incorri-
gible truth if the classification of phenomena is open to revision.39
Rule III is the mainstay of physical induction: Certain properties
must be attributed to all bodies if they are found in every body within
our reach. The properties in question must not be liable to intensify or
weaken. This condition is not so restrictive as it seems. Of course, most
observable physical properties can be graded or quantified. But any
such property P displayed by a body B may be conceived as the par-
ticular value taken by a function f in the circumstances of B, and the
property of being a seat - or, as mathematicians say, an argument - of
fis not liable to increase or decrease. Thus, while the property of being
attracted with a given force to a given body at a given distance cannot
be generalized to all bodies by Rule III for it admits variation, the prop-
erty of being susceptible to attractive forces governed by a given func-
tional relation - for example, eqn. (2.3) - is not a matter of degree and
therefore should be held to be a universal physical property if it is seen
to "belong to every body on which it is possible to perform experi-

39
As we shall see in §5.4, when Einstein saw that it was impossible to accommodate
Newton's Theory of Gravity within the new kinematics he had developed to cope
with electromagnetic phenomena, he reclassified free-fall as a form of inertial motion,
imaginatively anticipating the now familiar spectacle of weightless astronauts.
74 Newton

ments". This last condition confirms, by the way, that Rule III cannot
be relied on in a quest for certainty. The bodies on which we experi-
ment and the properties our experiments disclose in them are not a
fixed, or a large, or even a random sample of the whole, and no induc-
tion can be secure which rests on such a ground. Rule III, like the
preceding two, yields, not glimpses of God's worldview, but stepping
stones for human inquiry.
If the reader still has any doubt about this, the text of Rule IV should
dispel it. By virtue of it, the conclusions of our inductive inferences
must be held to be true aut accurate aut quamproxime ("either exactly
or as close to it as possible") until newly discovered phenomena compel
us to revise them. Thus, Newton himself makes it clear that his rules
are not meant to lead us to timeless truth, but to a provisional fixation
of belief, such as we need to keep research going. Central to Rule IV
is the adverb quamproxime - literally 'as near as possible' - which is
also found in other passages of Principia. Evidently, the generalizations
obtained by induction from experiment can only be as good as our
measurements, which are accurate only within a margin of error. Let
me illustrate with a few examples some of the implications of this fact.
Take eqns. (2.5). You will not clash with any of the observations that
support Newton's Law of Gravity if in the denominator of the right-
hand side you substitute 3 + 8 for the exponent 3, provided that you
take the arbitrary real number 8 sufficiently close to 0. You can also
replace the constant G with a function of time G(£), provided that the
derivative dG/dt is close to 0, so that G(t) practically behaves like a
constant. Moreover, you may add to the left-hand side a polynomial
in the time derivatives of rk up to the rcth, for some arbitrarily large
integer n, provided that you pick sufficiently small coefficients to
multiply each term. That no such thing is ever done can of course be
explained by the physicists' faith in the so-called simplicity of nature
(or by their belief that, as Einstein put it, "the Lord is sophisticated,
but not nasty"). But a more likely explanation is that the actual prac-
tice of physics will gain nothing by such changes if their overall result
still agrees within the admissible margin of error with the simpler orig-
inal formula.40

40
On the other hand, changes in the Law of Gravity like the first two described above
have at times been proposed to deal with otherwise unexplained phenomena. A slight
change in the exponent of the denominator in the right-hand side of eqn. (2.5) was
tried as one way of coping with Mercury's anomalous perihelion advance. A varying
 2.5 Newtonian Science 75

2.5 Newtonian Science

2.5.1 The Cause of Gravity

From the phenomena of planetary motion and free-fall, set in the New-
tonian frame, it is possible to infer by the Rules of Philosophy that
every bit of matter attracts and is attracted by every other bit of matter
with a force obeying Newton's Law of Gravity. It would seem, there-
fore, that the property of exerting such a force and of responding to
it is a universal property of matter. Such, at any rate, was the reading
of Newton's contemporaries, both friend and foe. It was reasonable
to expect that hitherto unsuspected properties should accrue to the
concept of matter with the progress of inquiry, if matter was designed
and created by God and can therefore comprise everything that God
judged useful for the fulfillment of His plans (cf. §1.3). Thus John
Locke insinuated in his Essay that matter could even have the prop-
erty of thinking, which Cartesians so studiously ruled out ([Locke]
1690, IV.iii.6). And yet in the same book, published three years after
Principia, Locke asserted that it is "impossible to conceive that body
should operate on what it does not touch (which is all one as to imagine
it can operate where it is not), or when it does touch, operate any other
way than by motion" (II.viii.ll). But later he deleted this passage
because, as he explained to Stillingfleet,
I have been convinced by the judicious Mr. Newton's incomparable book
that there is too much presumption in wishing to limit the power of God
by our limited conceptions. The gravitation of matter toward matter in
ways inconceivable to me is not only a demonstration that God, when

gravitational "constant" was suggested in the 1930s to bring the age of rocks calcu-
lated from the statistics of radioactive minerals into harmony with the time available
for the Earth to form in our expanding, formerly very hot universe. My last example
apparently has no counterpart in history, perhaps because raising the order of their
differential equations is the last thing that physicists would wish to do. When Ein-
stein was searching for a new law of gravity he explicitly decided not to look for
equations of a higher order than Newton's because "it would be premature to discuss
such possibilities in the present state of our knowledge of the physical properties
of the gravitational field" (Einstein and Grossmann 1913, p. 234). However, a term
multiplied by a devilishly small coefficient X was added by Einstein (1917b) to the
field equations of gravity (1915i) to secure a model of the universe that was both
finite and static. (When he became convinced that the universe was not static but
expanding he described this move as his greatest mistake ever.)
76 Newton

it seems to Him good, can put into bodies powers and modes of acting
which are beyond what can be derived from our idea of body or
explained by what we know of matter; but it is furthermore an incon-
testable instance that He has really done so.
(Locke 1699, p. 468)
This was generally the position in England. Indeed, Roger Cotes,
who prepared the second (1713) edition of Principia under Newton's
supervision, placed gravity among the "primary" properties of bodies,
on a par with extension, mobility, and impenetrability.41 But this was
precisely what the older generation of continental savants, educated
in Cartesian austerity, would not countenance. Thus Huygens, while
expressing admiration for Newton's achievement in brushing aside
difficulties previously associated with Kepler's laws and in destroying
Descartes's planetary vortices, said that he did not agree
with a Principle according to which all the small parts that we can
imagine in two or several different bodies mutually attract each other or
tend to approach each other. That is something I would not be able to
admit because I believe that I see clearly that the cause of such an attrac-
tion is not explainable by any of the principles of Mechanics, or of the
rules of motion.
(Huygens, Discours sur la cause de la pesanteur [1690];
OC, XXI 471)
And on 18 November 1690 he wrote to Leibniz that Newton's "Prin-
ciple of Attraction" seemed "absurd" to him (Huygens OC, XXI 538).
Leibniz's own opposition to Newtonian attraction was just as strong.
In his view, Newton had revived, with a vengeance, the discredited
occult qualities of medieval and renaissance science. For, as he wrote
to Hartsoeker on 6 February 1711,
the ancients and moderns, who admit that gravity is an occult quality,
are right, if they mean by it that there is a certain mechanism unknown
to them, whereby all bodies are pushed towards the center of the earth.
But if their opinion is that the thing is performed without any mecha-
nism, by a simple primitive quality, or by a law of God, who produces
that effect without employing any intelligible means, it is an unreason-
able occult quality, and so very occult, that it is impossible that it should

41
Preface to Newton 1713 (in Newton 1934, p. 27). In a draft Cotes had written that
gravity is an "essential property" of bodies, but he weakened the epithet in response
to criticism by Clarke (see Koyre 1965, p. 159).
 2.5 Newtonian Science 77

ever become clear, though an Angel - not to say God himself - should
wish to explain it.
(Leibniz GP, III, 519)

And again, in a letter to Conti (November or December 1715):

It is not sufficient to say: God has made such a law of Nature, therefore
the thing is natural. It is necessary that the law should be capable of
being fulfilled by the nature of created things. If, for example, God were
to give to a free body the law of revolving around a certain center, he
would have either to join to it other bodies which by their impulsion
would make it always stay in a circular orbit, or to put an Angel at its
heels; or else he would have to concur extraordinarily in its motion.
(Quoted in Koyre 1965, p. 144)

Newton was infuriated by the suggestion that he traded in occult

qualities. Among his several pronouncements on this point, the fol-
lowing (from Opticks, Query 31) is particularly clear. In his view, the
particles of matter do not just have "a vis inertiae, accompanied with
such passive Laws of Motion as naturally result from that Force, but
[are also] moved by certain active Principles, such as is that of Gravity,
and that which causes Fermentation, and the Cohesion of bodies".

These Principles I consider, not as occult Qualities, supposed to result

from the specifick Forms of Things, but as general Laws of Nature, by
which the Things themselves are form'd; their Truth appearing to us by
Phenomena, though their Causes be not yet discover'd. For these are
manifest Qualities, and their Causes only are occult. [. . .] To tell us that
every Species of Things is endow'd with an occult specifick Quality by
which it acts and produces manifest Effects, is to tell us nothing: But to
derive two or three general Principles of Motion from Phenomena, and
afterwards to tell us how the Properties and Actions of all corporeal
Things follow from those manifest Principles, would be a very great step
in Philosophy, though the Causes of those Principles were not yet dis-
cover'd: And therefore I scruple not to propose the Principles of Motion
above-mention'd, they being of very general Extent, and leave their
Causes to be found out.
(Newton Opticks, pp. 40If.; my italics)

But Newton emphatically denies that the power of attraction is an

"inherent" property of matter. On this he appears to be much closer
to Huygens than to Locke or Cotes. As early as 1693, he wrote to the
great Hellenist Richard Bentley:
78 Newton

It is inconceivable that inanimate brute matter should, without the medi-

ation of something else which is not material, operate upon and affect
other matter without mutual contact [...]. That gravity should be innate,
inherent, and essential to matter, so that one body may act upon another
at a distance through a vacuum, without the mediation of anything else,
by and through which their action and force may be conveyed from one
to another, is to me so great an absurdity that I believe no man who has
in philosophical matters a competent faculty of thinking can ever fall
into it. Gravity must be caused by an agent acting constantly according
to certain laws, but whether this agent be material or immaterial I have
left to the consideration of my readers.
(Newton 1974, p. 54)

This stance probably inspired Newton's dignified words in the General

Scholium he added to Principia in 1713:

Thus far I have explained the phenomena of the heavens and of our sea
by the force of gravity, but have not yet assigned the cause of gravity.
This force must arise in any case from some cause that penetrates to the
very centres of the sun and planets, without suffering the least diminu-
tion of its power; which acts not according to the area of the surfaces
of the particles acted upon (as is usual with mechanical causes), but
according to the quantity of solid matter; and whose action extends on
all sides to immense distances, decreasing always as the inverse square
of the distances. [. . .] But hitherto I have not been able to deduce from
phenomena the reason for these properties of gravity, and I do not con-
trive hypotheses. For whatever is not deduced from the phenomena is to
be called an hypothesis; and hypotheses, whether metaphysical, or phys-
ical, or of occult qualities, or mechanical, have no place in experimen-
tal philosophy. In this philosophy propositions are inferred from the
phenomena, and afterwards rendered general by induction. Thus it was
that the impenetrability, the mobility, and the impulsive force of bodies,
and the laws of motion and of gravitation, became known. And it is
enough that gravity really exists, and acts according to the laws which
we have explained, and sufficiently accounts for all the motions of the
celestial bodies and of our sea.
(Newton 1726, p. 530)

The cause of gravity that Newton has in mind is certainly not a type
of event that is contiguous with and immediately precedes every exer-
cise of gravitational attraction (and thus not something that a Humean
philosopher would recognize as a cause). As I noted in connection with
 2.5 Newtonian Science 79

Rule I (§2.4), on the meaning of 'cause' Newton oscillates between old

common sense and Aristotle: The cause of gravity is either the agent,
material or immaterial, that brings about the mutual attraction of all
bits of matter (thus at the end of the above quotation from the letter
to Bentley), or it is the reason, broadly conceived, why matter behaves
as it does, that is, in agreement with Newton's mathematical law of
attraction (thus in the famous abjuration of hypotheses, halfway
through the last quotation from Principia).
With his talk about the cause of gravity, did Newton mean to say
that something was wanting in his theory, which another more fortu-
nate scientist might find? I do not think so. His remark that the force
of gravity does not operate like the usual mechanical causes seems
designed to warn us that the phenomena effectively preclude the kind
of explanation that his adversaries foolishly demanded. And the curt
satis est ("it is enough") in the last sentence quoted does not encour-
age any further search for the missing cause. In this matter - as in his
endorsement of tenability quamproxime, in contrast with Descartes's
commitment to incontestable certainty - Newton quite resolutely sets
the path of future science while still paying lip service to the notions
of his time. A century later, Auguste Comte classified the search for
causes - in the sense here at stake - as typical of the prescientific "meta-
physical" age of intellectual history, as opposed to the "positive"
scientific age, busy with the search for laws.42 Another century had not
wholly passed when Bertrand Russell cheekily asserted that "the reason
why physics has ceased to look for causes is that, in fact, there are no
such things".43

42
Comte was strongly influenced by Joseph Fourier, whose masterpiece on heat begins
with the sentence: "The primary causes are unknown to us; but they are subject to
simple and constant laws that can be discovered by observation, the study of which
is the purpose of natural philosophy" (1822, p. i). A similar sentiment is found in
Ampere (1827, p. 177): "To establish the laws of [electrodynamic] phenomena I have
only consulted experience, from which I have deduced the formula that alone can
represent the forces to which they are due. I have not inquired into the cause one
might assign to these forces, being convinced that every inquiry of this sort should
be preceded by purely experimental knowledge of the laws, and by the determina-
tion, based solely on the laws, of the value of the elementary forces . ..".
43
Russell (1917, p. 180). The quoted sentence immediately precedes Russell's famous
bon mot: "The law of causality, I believe, like much that passes muster among
philosophers, is a relic of a bygone age, surviving, like the monarchy, only because
it is erroneously supposed to do no harm."
80 Newton

2.5.2 Central Forces

Newton's Theory of Gravity was criticized not only because it postu-
lated forces that acted at a distance without intermediaries, but also
because it vested those forces on centers of mass, that is, dimension-
less points, which could be located in a void. A Newtonian might
indeed reply that the gravitational pull toward the center of mass of a
body or of a system of bodies was always the resultant of forces issuing
from pieces of matter. But a critic could still counter that the real forces
that actually manifest themselves through measurable effects are
always the resultants, not the components into which our thinking ana-
lyzes them.
A stronger line of defense, and one with which the Newtonians
finally won the day, is that the little we know about the workings of
nature does not entitle us to dismiss a theory whose predictive accu-
racy and breadth of coverage so much exceed that of every earlier
product of natural philosophy. As a matter of fact, we do not under-
stand how bodies act on each other by contact any better than we
understand their interaction at a distance. We are not surprised when
a billiard ball communicates its motion to another because we have so
often seen it happen, but we do not have a clearer notion of impulsive
force and the mutual exclusion of bodies than of Newtonian attrac-
tion.44 Why, it could even be that the impenetrability of solid bodies
and the transmission of motion in collisions are themselves the mani-
festation of a repulsive force. Indeed, Boskovic (1758, 1763) argued
that it could not be otherwise, if, as everyone assumed, Nature never
jumps and all transitions are continuous. For let a body A moving with
speed u catch up and collide with a body B moving in the same direc-
tion as A with speed v < u. After colliding, both bodies continue to
move together with speed (u + v)/2. Immediately upon collision the
speed of A has decreased and the speed of B has increased by u - v.
This change can be gradual only if it begins to occur before the two
bodies come into contact, that is, if the acceleration of B and the decel-
eration of A are caused by repulsive forces acting at a distance.
Boskovic's matter consists of dimensionless particles that act on one
another with a force that, as the distance between the particles varies,
alternatively becomes repulsive and attractive.

44
This point was eloquently made by Maupertuis (1732); see Arana (1990, p. 141).
 2.5 Newtonian Science 81

A theory of matter along similar lines had been proposed by

Immanuel Kant in his Monadologia physica (1756) as an example of
the "joint use of metaphysics and geometry in natural philosophy". His
declared aim was to reconcile the philosophical conception of bodies
as ultimately composed of indivisible elements with the geometric truth
of the infinite divisibility of space. But there are indications that he was
also motivated by a desire to tackle the intractable problem of the rela-
tions between mind and body that modern philosophy had inherited
from Descartes. Kant describes the physical world as an aggregate of
simple substances - designated by the Leibnizian term monads -
located each at a point in space, some of which are human souls. Each
monad exerts on all the others a repulsive and an attractive force. Both
forces depend on the distance between the locations of the interacting
monads. Over short distances, the repulsive force prevails over the
attractive force, but it decreases with distance at a faster rate than the
latter. In Kant's view, the interplay of both kinds of forces ensures that
each monad takes up - or, as he says, "occupies" - a definite volume
in space, which cannot be penetrated by the volumes "occupied" by
other monads. However, Kant's claim is not backed by precise mathe-
matical arguments, nor does he articulate any known phenomenon into
a testable model of his theory. Kant lost all hope that his monads could
be used for solving the mind-body problem when he realized that, by
the said interplay of forces, they would be liable to be amassed into
balls; his philosophical good sense would not let him countenance a
"clod of souls" (Kant 1766, Ak. II 321). Nevertheless, the conception
of the human mind as open to modification by the direct physical
action of other created entities persisted as a subtext in his critical writ-
ings, although in these writings he forbade all inquiry into the ultimate
constitution of reality.
The speculations of Kant and Boskovic had little impact on the
development of physics. But the concept of central forces acting at
a distance found effective application outside the field of planetary
motion and free-fall. It was known since the thirteenth century that the
north pole of a magnetized body (i.e., the extreme of it that tends to
point to the north) repels the north pole and attracts the south pole of
any other such body. In 1733, Charles du Fay published his discovery
that electrified bodies - that is, bodies that, after being rubbed or being
brought into contact with a body already electrified, behave like briskly
rubbed electron (Greek for 'amber') - fall into two classes, so that those
belonging to each class repel one another and attract those of the other
82 Newton

class. In the 1780s Coulomb showed, by means of his torsion balance,

that these phenomena of electric and magnetic attraction and repulsion
can be accounted for - within the Newtonian frame - by forces directly
proportional to certain quantities characteristic of the interacting
electrified or magnetic bodies and inversely proportional to their dis-
tance squared.45
Volta's invention of the pile c. 1800 made it possible to experiment
with steady electric currents and led to Oersted's discovery, in 1820,
of the following action of electricity on magnets: If a straight wire is
suspended above a magnetic needle at rest and parallel to it, the flow
of electric current through the wire causes the needle to turn so that
the pole under that part of the wire which receives electricity most
immediately from the negative end of the pile declines toward the west,
the declination angle being smaller if the distance between the needle
and the wire is greater, and also if the battery is less efficient.46 The
phenomena suggest the presence of a magnetic force that is perpen-
dicular both to the direction of the wire and to the shortest line from
the wire to the magnet. This is hardly the orientation one would expect
a central force to have. Nevertheless, Andre-Marie Ampere, after an
extraordinary bout of experimentation and mathematical theorizing,
succeeded in explaining - or so it was thought - electromagnetic inter-
action by central forces acting at a distance. The explanation involves
Ampere's experimental discovery - communicated one week after he

45
The torsion balance measures the angle through which a fine metal wire is twisted,
and enables one to calculate very exactly the torque on the wire. One such device
had been built by John Michell c. 1750 and used by him to establish the inverse
square law of magnetostatic force before Coulomb. An improved version of Michell's
torsion balance was employed by Cavendish to measure the Newtonian attraction of
large leaden balls on small bodies (and thereby to calculate the values of the gravi-
tational constant and the mass of the Earth). Cavendish published this result in 1798.
He had also used the torsion balance to establish Coulomb's law of electrostatic force
before Coulomb, but this and many other important findings of Cavendish were not
published until the late nineteenth century. Coulomb's electrostatic law had also been
anticipated by Priestley, who inferred it from the fact that a hollow electrified sphere
exerts no action on bodies placed inside it; this shows that electric force satisfies the
proposition proved in Book I, Section xii of Newton's Principia (quoted in §2.3) and
must therefore obey an inverse square law.
46
Oersted published his findings in a pamphlet in Latin, privately distributed to scien-
tists and scientific societies on 21 July 1820. An English translation is reproduced in
Shamos (1959, pp. 123-27).
 2.5 Newtonian Science 83

learned of Oersted's - that two parallel wires attract each

other if they carry electric currents in the same direction and repel each
other if they carry currents in opposite directions. It also resorts to the
hypothesis that magnets consist of well-aligned, minuscule, current-
carrying circuits (such circuits are also postulated in nonmagnetic
bodies, where they neutralize each other because they are not aligned).
Ampere accounts for all electrodynamic and electromagnetic phenom-
ena known to him by a force acting at a distance between infinitesimal
elements of electric current, along the line joining them. According to
Ampere's Law, any two such elements interact with a force directed
along the line joining them and proportional to the product of their
current intensities multiplied by a function of their directions and
divided by their distance squared. Thus, Ampere believes, he has
satisfied in this field the Newtonian requirement that "all motions
in nature must be reducible by calculation to forces acting always
between two material particles along the straight line that joins them,
so that the action exercised by one upon the other is equal and oppo-
site to that which the latter exercises at the same time upon the former"
(Ampere 1827, pp. 175f.). The compliance must be taken with a pinch
of salt, however, for Ampere's force depends on the direction of the
currents and thus, in effect, on the relative motion of electrified parti-
cles. One naturally expects this of a force generated by such motion.
But real forces dependent on relative motion do not fit well in the New-
tonian frame, and their admission will eventually wreak havoc with it
(§5.1). Still, Ampere's "Mathematical theory of electrodynamic phe-
nomena deduced from experience alone" (1827) was perceived as a
major triumph of the Newtonian paradigm of central forces acting at
a distance. Acceptance of the paradigm peaked two decades later, in a
paper by Hermann von Helmholtz, who describes it as "the condition
of the complete intelligibility of nature" (1847, p. 6; this passage is
quoted in context in §4.3.1). That was 160 years after the publication
of Principia, and only 10 before the emergence of a completely differ-
ent style of physical explanation in Maxwell's paper "On Faraday's
lines of force" (1855/56).47

47
In the 1870s Helmholtz led the last big fight for an action-at-a-distance theory of
electrodynamics in the style of Ampere, against Maxwell's field theory. Ironically, it
was Helmholtz's assistant, Heinrich Hertz, who secured Maxwell's triumph by build-
ing the first rudimentary radio transmitter.
84 Newton

2.5.3 Analytical Mechanics

During the century following the publication of Principia several
remarkable mathematicians - most notably Leonhard Euler - con-
tributed to the progress of mathematical analysis based on the differ-
ential and integral calculus invented by Newton and Leibniz, and used
it for the formulation and solution of mechanical problems within
Newton's conceptual framework. Their work led to the analytical
mechanics of Joseph-Louis de Lagrange, which proposes general
methods for the solution of all mechanical problems. This is not
the place for a detailed, historically accurate, abstract of Lagrange's
Mecanique analitique (1788), but I shall try to explain, with mild
anachronism, the gist of his approach.48 This will enable me to intro-
duce some notions that we shall need later.
In a Newtonian setting one naturally conceives a material system as
a collection of finitely or infinitely many particles acted upon by forces
issuing from the remaining particles and perhaps also from sources
external to the system. Taking dynamis as Greek for 'force', a system
thus conceived is called dynamical. Suppose that we have n particles
labeled with the positive integers from 1 to n. As before, I denote by
r, the radius vector from the origin of an inertial frame of reference to
the location of the zth particle and by raz the niass of this particle. Let
iki be the force with which the &th particle acts on the z'th particle and
ff the resultant of all external forces acting on the z'th particle. The
workings of the system are then completely described by n second-
order ordinary differential equations, the system's equations of motion:

The conceptual simplicity of these equations is very impressive, but as

n grows they soon become unmanageable. Also, while the external
forces on a dynamical system under study are often well known, espe-
cially when they are introduced by the experimenter, one is usually
ignorant of the detailed interaction between the system's parts. More-
over, the subsequent evolution of physics has shown that the interac-

48
For a judicious sketch of the development of analytical mechanics in the eighteenth
and early nineteenth centuries see Dugas (1955, pp. 232-85, 323-408). Before
attempting my own presentation I consulted a few recent treatises. In the end, I think
I profited most from Pars (1965).
 2.5 Newtonian Science 85

tion between small parts of matter at close distances cannot even be

described properly within the Newtonian frame.
However, Lagrange and his predecessors realized that it was not
always necessary to know the internal forces f& to predict the evolution
of a dynamical system under the Laws of Motion. In most cases of inter-
est the system's freedom of motion is drastically curtailed by constraints,
which we believe are due to the interaction of the system's parts, but
which are described and taken into consideration in the study of the
system's motion without our having the slightest inkling of how that
interaction works. Consider, for instance, a billiard ball of unit radius
confined to the top of a table 30 units long and 20 units wide. Choosing
one of the corners of the table as the origin of our Cartesian coordinate
system (§1.1) and identifying the £-axis with the vertical through the
origin, the constraints are fully expressed by the inequalities 1 < xc < 29,
1 < yc < 19,1 < zc (where xC9 yc, zc denote the position coordinates of the
ball's center). If a slab that is parallel to the surface of the table prevents
the ball from jumping ever so little, the third inequality must be replaced
by the equation zc = 1. As a second example, consider a bead that is free
to slide along a thread hanging between two walls and swinging back-
ward and forward. In this case the exact description of the constraints is
more troublesome but still can be given in purely kinematic terms,
without making any assumptions about the forces that maintain them.
Among the forces acting on a dynamical system Lagrange distin-
guished the forces of constraint, which hold the constraints in place,
from all the rest, which we shall call the impressed forces. Then, by
deftly wielding an idea of d'Alembert's, he eliminated the forces of
constraint from the equations of motion. To explain how this is done,
it is convenient to drop the vector notation used in eqns. (2.6) and to
refer our n particles to a Cartesian coordinate system. To simplify nota-
tion, I write the coordinates of the rth particle as (x3r-2,x3r-.ux3r). Its
mass will, depending on context, be represented by one of the three
symbols m3r_2, tn3r.u or m3r (all of which stand of course for the same
quantity). Each force acting on a particle will be viewed as the resul-
tant of three component forces, parallel to the coordinate axes. If xk is
the ith coordinate of a given particle (i = 1, 2, 3), I denote, respectively,
by X* and Ek the impressed force and the force of constraint which act
on it in the direction of the ith axis.49 With this notation we may replace

49
For greater precision, let me note that, if [q] denotes the integral part of the positive
rational number q - i.e., if [q] stands for the greatest integer less than or equal to q
86 Newton

the n equations of motion (2.6) with the following system of N = 3n

equations:
mkxk=Xk+Ek (& = 1,...,N) (2.7)
Let x = {xu . . . , xN} be the coordinates of the system at a given moment
and consider the variations they may simultaneously experience given
the constraints. As we gathered from the previous examples, such vari-
ations are restricted in definite ways, no matter what the impressed
forces. If (as in our first example) the constraints do not depend on
time, any set of such variations Sx = {8xu . . ., 8xN} corresponds to a
possible displacement of the system. We call it a virtual displacement.50
However, if the constraints change with time (as in our second
example), we reserve this name for such displacements as would be
possible if the constraints were frozen at the moment in question. More
precisely: In a constrained system with n particles, the possible infinites-
imal displacements dx = {dxl9. . ., dxN} are subject to K (< N = 3n)
conditions of constraint:

£ ^ Ahk(x, t)dxk + A , ( x , t)dt = 0 (h = l,...,K) (2.8)

where the coefficients Ahk(x,t) and Ah(x,t) are real-valued differentiable

functions defined on a suitable region of Un+1. A definite value of (x,t)
specifies a particular system configuration x at a time t. Any solution
{8%!,. . ., SxN} of the equations

J^k=iAhk8xk=0 (h = l , . . . , K ) (2.9)

for the corresponding values of the A^'s is a virtual displacement avail-

able from that configuration at that time. If conditions (2.8) do not
involve time, they agree with (2.9) for each configuration x, and the
class of virtual displacements coincides with the class of possible
displacements.
Lagrange assumed that in virtual displacements the forces of con-

- , then, by our conventions, X^ acts on the [(k + 2)/3]-th particle in the direction of
the (k -3[(k- l)/3])-th axis.
The term 'virtual displacement' originated in statics. If a system is in equilibrium, it
experiences no real displacements. However, we may inquire what simultaneous dis-
placements will be permitted by the constraints on the system in case the equilibrium
is broken. Thus, in a balance with equal sides one scale will climb from height h to
h + dh i( and only if the other simultaneously descends from h to h - dh. This com-
bination of movements constitutes a virtual displacement.
 2.5 Newtonian Science 87

straint do no work; in other words, he assumed that for any virtual

displacement xk \-> xk + §xk (k = 1 , . . . , N),

X"=iH*5x*=0 (2.10)
If we now multiply by bxk both sides of eqns. (2.7), add up all the
resulting equations, substitute from (2.10), and reshuffle terms, we
obtain:
y N
£ kJmkxk-Xk)&xk=0 (2.11)
Equation (2.11) is often called, somewhat improperly, d'Alemberfs
Principle.51 It provides a general framework for the formulation and
solution of all mechanical problems concerning systems for which eqn.
(2.10) holds. It must be emphasized that not every conceivable system
that is subject to Newton's Laws of Motion falls under this category.52
Lagrange's theory is strictly stronger than Newton's, so its scope is
strictly narrower. It is a sign of Lagrange's genius that he was ready to
give up the universality of eqns. (2.6) or (2.7) in exchange for the real
problem-solving power of eqn. (2.II). 53 Indeed, the best known and
most impressive corollaries of Lagrange's theory are even more
restricted in scope, as we shall now see.
To derive them, we recall first that when the constraints do not
depend on time, every possible displacement of the system counts as a
virtual displacement, so its actual velocity is a virtual velocity. One may
therefore substitute xk for bxk in eqn. (2.11), and obtain

^k=1MkXkXk = ^k=1xk*k (2.12)

Now, the left-hand side of this equation is the time derivative dT/dt of
the kinetic energy T = jS^slmkXk (cf. §1.5.2), so

§ = K>*4 (2.13)

51
Compare d'Alembert (1758, pp. 72ff). Lindsay and Margenau (1957, pp. 103ff.)
discuss d'Alembert's original approach and explain its connection with Lagrangian
mechanics.
52
Of course, all exceptions vanish if we take eqn. (2.10) as the definition of 'forces of
constraint'. But then some of the unperspicuous internal forces that hold our system
together may still turn u p in eqn. (2.11).
53
This power is beautifully displayed in Pars (1965), and also in more elementary text-
books of analytical mechanics (e.g., Goldstein 1950).
88 Newton

We now return to eqn. (2.11) and focus our attention on cases in

which the force components Xi,. . ., XN depend only on the position
coordinates xu . . ., xN, not on their time derivatives xu . . . , xN or on
the time t (remember the quotation from Helmholtz at the end of
§2.5.2). In many situations of this type there exists a scalar function
V (i.e., a real-valued function depending on position alone), such that
y N
£ k=iXkdxk=-dV (2.14)

V is called the potential energy function or simply the potential. For

reasons that will now become apparent, such systems - and the im-
pressed forces acting on them - are said to be conservative. Equation
(2.14) implies that
dv N a v ^ = _ y N x^ {2A5)

If the constraints on a conservative system do not depend on time, eqn.

(2.13) holds. By adding eqns. (2.13) and (2.15) we obtain the energy
equation,

A ( T + V) = 0 (2.16)
dt
Thus, in a conservative system subject to time-independent constraints
the sum of the kinetic and the potential energy does not change: Total
energy is conserved.
In a system subject to constraints the position coordinates {xi,. . .,
xN] are linked to each other and cannot vary independently. In many
cases the K conditions of constraint (2.8) can be integrated to yield K
equations:
/i(*i,..., *N, 0 = 0 (fc=l,...,K) (2.17)
If the constraints meet this condition, the system is said to be holo-
nomic.54 Equations (2.17) can then be used to eliminate K coordinates
by expressing them in terms of the remaining N - K. More generally,
the N interlinked coordinates {xu ..., xN} can be expressed as differ-
entiable functions of N - K = m independent variables {qu . . . , qm}:

54
Heinrich Hertz, who apparently invented the term 'holonomic', explains it as follows:
"The name indicates that such a system obeys integral (oXoc) laws (vouxx;), while
material systems in general are subject only to differential laws" (1894, §123).
 2.5 Newtonian Science 89

xk=xk(ql9...,qm9t) (ft = 1,..., N) (2.18)

The g/s are called generalized coordinates; in contrast with the x^'s,
they do not usually measure physical distances, m is known as the
number of degrees of freedom of the system.
It follows that, for each k e { 1 , . . . , N},

_,, > + i) (2 19)

-

and

bxk = yy\m ^r^-Sqj (2.20)

/=1
dq.
The second term of eqn. (2.11) becomes

The quantities Qf = X£iX^(3X^/3^;) are called components of the

generalized force.55
Substituting eqn. (2.20) into the first term of (2.11), it is clear that:

Yak=lfnkxkbxk = Y^=l ^k=1mk*k y ± 8 < ? / [121]

Now
.. dxk =
~ L i b : m^^ T ~ " mkXk x ^~
In the light of eqn. (2.19) it is clear that

dt{dqj
Moreover,
dxk

inasmuch as Xk depends only on qu . . ., qm and t, and dqjdqk - 0 if

i * k. Substituting from eqns. (2.24) and (2.25) into (2.23) we obtain:

55
Note that the Q, need not have the dimensions of force. However, each product Q/5^
has the dimension of work, i.e., mass x (space/time)2.
90 Newton

= 2 4 * r X4 ad~ m ^^;J (2-26)

Evidently

(2 27)
W^rafei^HJ -
and

As we know, the polynomial l^^mkx\ that occurs in both eqns.

(2.27) and (2.28) represents the system's kinetic energy T. Substituting
from (2.27) and (2.28) into (2.26), and from (2.21) and (2.26) into
(2.11), we obtain "d'Alembert's Principle" for holonomic systems:

Since the 8g/s are arbitrary and not tied by constraints, eqn. (2.29)
can only hold if all the factors of the 5g;'s vanish, that is, if, for
/ = 1,. . ., m,

WMYZ]«"
If the system is conservative, there is a potential V such that, by
eqns. (2.21) and (2.14),

Hence,

and
| ^ =0 (2.32)
6q
Substituting from eqns. (2.32) in (2.30) and putting T- V= L we derive
the Lagrange equations of motion:56
56
As a brief calculation shows, eqns. (2.33) also hold for a holonomic system that is
conservative in the following extended sense: There is a function U, depending on
the q, and their time derivatives qn such that

o =^L+±(K
W/
dq, AtUq,
 2.5 Newtonian Science 91

= l,...,m) (2.33)

The function L is the Lagrangian of the system.

A few remarks are in order before closing this subject.

(A) A smooth one-one mapping of the domain of coordinates
q = {qu . . . , qm} onto a suitable subset of Um yields new generalized
coordinates q' = {q'u . . . , q'm}. If V stands for the Lagrangian
expressed as a function of the new coordinates, it is fairly easy to show
that eqns. (2.33) entail the following:

(/ = V - > » 0 (2330

This property is described by saying that the Lagrange equations are

preserved under arbitrary coordinate transformations, or that they are
generally covariant.
(B) The generalized coordinate set q(t) = {qi(t),. . . , qm(t)} and its
time derivative q(t) = {qi(t)9. . . , qm{t)} provide a full description of the
system's state of motion at time t. As t varies, say, from a to fo, q(t)
describes a path in Rm. The mapping t •-» q(t) is a curve in the configura-
tion space Rm, which provides an accurate representation of the evo-
lution of the system in the time interval (a,b). This method of
representing the successive states of a dynamical system with m degrees
of freedom (m > 3) as tracing out a trajectory in m-dimensional space
prepared the way for even bolder modes of representation, like the 2m-
dimensional phase space (see D) and the infinite-dimensional Hilbert
space of Quantum Mechanics (§6.2.4).
(C) Readers acquainted with the calculus of variations will recall
that eqns. (2.33) state necessary and sufficient conditions for the action
S = \ Ldt to be stationary. In other words, the Lagrange equations of
motion are logically equivalent to the variational principle:

0 (2.34)

What this means is that the curve y:[?o?^i] -» Rw? which represents the
evolution of our system in configuration space from time t0 to time tu

and L = T - U. In Maxwellian electrodynamics forces depend on positions and veloc-

ities but are derivable from such a generalized potential function U, so the Lagrange
equations - in the extended sense - can also be used in the context of this theory.
92 Newton

is a solution of eqns. (2.33) if and only if the integral JY Ldt along y is

either less than or greater than the integral J^ Ldt along each neigh-
boring curve r|: [£0,£i] -> Rm that coincides with y at the endpoints but
differs slightly from it in between. Equation (2.34) is Hamilton's Prin-
ciple, one among several variational principles of mechanics proposed
in the eighteenth and early nineteenth centuries. Like the others, it sug-
gests that mechanical processes are governed by final causes, inasmuch
as the value reached by the integrals J Ldt along different trajectories
at the end of a time interval determines right from the beginning the
choice of one of these trajectories. But the matter can, of course, also
be viewed in this way: The actual trajectory is determined gradually,
at each instant, by the differential equations (2.33), so that in the end
- by virtue of a mathematical theorem - the integral J Ldt takes an
extremal value along the trajectory thus generated.
(D) The Lagrangian L can of course be expressed as a function of
our original Cartesian coordinates xu . . ., xN. Clearly,

\ \~mkxl - 0 = mkxk (2.35)

dxk dxk dxk d 2 J
So, if xk is the ith coordinate of a given particle (/ = 1,2,3), dL/dxk is
that particle's momentum component along the ith Cartesian axis. On
this analogy, the quantity

ft" (2.36)
Mi
is called the generalized momentum canonically conjugate with the gen-
eralized coordinate g;. From eqn. (2.33) we see at once that

d dL dL
dt dqj dqf
Let (q, p) stand for the list of numbers (ql9. . ., qm9 pu . . . , pm). The
2m-dimensional (q, p)-space is the system's phase space to which I
alluded above. The m second-order Lagrange equations (2.33) trans-
late into an extremely elegant set of 2m first-order equations in the g's
and p's, which determine the system's trajectory in phase space. We
introduce first the Hamiltonian function H:
 2.5 Newtonian Science 93

For an arbitrary variation {8qu... ,5gm,8<ji,... ,84m}? or, equivalently,

{8<ji,... ,8g m ,8pi,.. . ,S/?m}, with dt = 0, the corresponding variation of
H is given by:

For each i e {1,. . ., ra}, the partial derivative of H with respect to q{

- or pi - is calculated by allowing only q{ - or, respectively, pi - to vary
while all the other (q,p)-coordinates remain fixed. Therefore, eqn.
(2.39) implies that

l?= * f^= -|^

dpi dqi dq{
(2 40)
*
So, by eqns. (2.37),

qt=^~ i>i=-^r- (/ = l,...,m) (2.41)

dpi dqj
Equations (2.41) are the Hamilton equations of motion, which were
introduced in 1835 by W. R. Hamilton. They can also be derived
directly from Hamilton's principle (2.34), which, by using eqn. (2.38),
can be rewritten as:

=0 (2.42)

Equation (2.42) also entails the Hamilton-]acobi equation, which I give

here for future reference:

— + H(q,p,0 = 0 (2.43)

Dynamical systems governed by the Hamilton equations (2.41) - or,

equivalently, by the Lagrange equations (2.33) - generally meet the
mathematical conditions under which these equations have unique
solutions. When this is so, each point in the system's phase space lies
on one and only one solution of eqns. (2.41). If the system's state at
any given moment t0 is represented, say, by point (qo,Po)5 the unique
solution through this point is a curve representing the complete history
94 Newton

of the system before and after t0. This is ground enough for Laplace's
renowned vision of universal determinism (if the entire universe is in
effect a conservative dynamical system and the equations that govern
it meet the conditions for the existence and uniqueness of solutions):
An intelligence that knew, for a given instant, all the forces acting in
nature, as well as the positions of all the things that constitute it, and
who was capable of subjecting these data to analysis, would embrace in
a single formula the motions of the largest bodies and those of the light-
est atom. For her nothing would be uncertain, and the future, like the
past, would be present to her eyes.
(Laplace 1795, in OC, vol. VIII, pp. vi-vii)

Appendix
The following simple example illustrates some of the foregoing ideas
and the use of vector notation to convey them. Consider a particle of
mass m moving in a potential V. We denote its position by the Carte-
sian coordinates {x\,X2,x$), or, in vector notation, by x. The Lagrangian
is

L(x,x,f) = -mx2 - V(x9t) (2.44)

so the momentum components are, by eqns. (2.36),

dL
px——— — mx\ p2=mx2 p3=mx3 (2.45)

or, in vector notation, p = mx

By eqn. (2.38), the Hamiltonian H = x-p - L, or, substituting from
eqns. (2.44) and (2.45),

H(x,p,f) = — + V(x9t) (2.46)

2m
Clearly H is the total energy of the particle.
The Lagrange equations are

d/aM aL=0 (|. = w ) (233*)

dt\oXjJ dXj

or, substituting from eqns. (2.44) and (2.45),

 2.5 Newtonian Science 95

dt 3*i dt 3*2 dt 3*3

In vector notation, and by using the "nabla" operator

) 3 3
*1 ' 3*2 ' 3*3 /

eqns. (2.47) can be compressed into one:

p = -VV (2.48)
which is, of course, none other than Newton's Second Law (2.1)
with the force expressed, on the right-hand side, as the gradient of the
potential.
Substituting from eqn. (2.46) into the Hamilton-Jacobi equation
(2.43) we have that

^ ^ 0 (2.49)
dt 2m
which - as I will show next - is tantamount to

f + ^ + V=0 ,2.50,
dt 2m
This is the Hamilton-Jacobi equation for a single particle to which I
refer in §6.4.2.
To prove that eqn. (2.49) is equivalent to (2.50) it is enough to show
that p = VS, that is, that pi = dS/dx; for / = 1, 2, and 3. To evaluate the
derivatives dS/dXj we compare the action S along neighboring trajecto-
ries that begin at the same point x0 at time t0 but pass through differ-
ent points at time t\. The change in S is given by

We note that 8x, = 5(dx/dt) = d(8x)/dt. Integrating

9L d(5x,-)
h dt
-dt

by parts, we obtain:
96 Newton

Since the paths of actual motion satisfy the Lagrange equations (2.33 *),
the integrand in the last term equals 0. In the other term, of course,
8xi(t0) = 0 for all indices / (the paths compared begin all at the same
point), so, writing simply $Xj for 8x/(£i), we have that

t!^ i (2.53)
Therefore, by the same reasoning that took us from eqn. (2.39) to
(2.40), we obtain the desired result:

^ - = Pi (2.54)
OXi

Since our derivation is based on general principles, and does not in any
way depend on the special conditions, number of degrees of freedom,
or choice of coordinates in our particular problem, it is clear that eqn.
(2.54) holds, with full generality, for any pair (phqt) of canonically con-
jugate momentum and position coordinates.
 CHAPTER THREE

Kant

The specter of determinism and its implications for moral responsibil-

ity acted as a powerful motive on Kant's "critical" investigation of the
structure of human reason and the limits of human knowledge. He was
convinced that mathematical physics was on the right path and con-
stituted an example that all natural sciences ought to follow. "I assert"
- he wrote in 1786 - "that each special discipline concerning nature
(besondere Naturlehre) can contain only so much genuine science as it
contains mathematics."1 But he stoutly opposed the facile opinion that
modern physics can yield metaphysical conclusions concerning the
subjects of greatest interest for mankind: God, freedom, and immor-
tality. On such matters he "found it necessary to deny knowledge in
order to make room for faith" (1787, p. xxx). Kant's faith was a dis-
tillation of Christianity. He understood it, however, not as a supernat-
ural gift, but as the natural response of our "theoretical reason" to the
living fact of "practical reason", the cognitive echo of the voice of duty,
so to speak.
Pious Christians had voiced qualms about modern natural philoso-
phy since its inception. So Blaise Pascal, after making splendid con-

Preface to Metaphysical Principles of Natural Science (Ak. IV, 470). Kant clarifies
the meaning of this statement by means of the following example: "As long, there-
fore, as no constructible concept has been found for the chemical actions of matters
on one another, that is, no law of the approximation and removal of its parts by
which, say, in proportion to their densities and the like, their motion and its conse-
quences can be made a priori intuitive and represented in space [...], chemistry will
be no more than a systematic craft [Kunst] or experimental study, but never a genuine
science [Wissenschaft], for its principles are merely empirical [...], and therefore do
not make the possibility of chemical phenomena in the least understandable, because
they do not admit the application of mathematics." (Ak. IV, 470; my italics).

97
98 Kant

tributions to geometry and physics, wrote c. 1660 about Cartesian

mechanicism: "One ought to say in general: 'It happens by figure and
motion'; for that is true. But to say which, and to compose the machine,
is ridiculous, for it is useless and uncertain and wearisome. And if it
were true, we do not think that all philosophy is worth an hour of
trouble" (Pensees, no. 192; ed. Chevalier). Christian writers heaped
abuse on sweet Spinoza, who, long before mathematical physics had
anything substantial to show in favor of determinism, had, in his
mock-geometric Ethics (1677), proclaimed its universal rule. Religious
worries may also have prompted some ideas of Leibniz, which set
imprecise but unsurpassable limits to physics, and they certainly
inspired Berkeley's invention of positivism. I shall deal briefly with
Leibniz and Berkeley in §3.1. After that, leaving aside the question of
motives, I shall dwell at some length on Kant's conception of the
sources and scope of Newton's conceptual frame, for it was the first
full-blown philosophy of physics and remains to this day the most
significant.

3.1 Leibniz and Berkeley on the Scope of Mathematical Physics

3.1.1 The Identity of Indiscernibles

Leibniz was persuaded that every creature of God is an irreplaceable
individual, not merely a particular realization of a common blueprint.
The omniscient Creator has, of course, full knowledge of each even
before creating it, for He certainly knows what He is doing when he
decides to create it. Moreover, He has the power to annihilate any one
of them - or many, or all but one - irrespective of the rest. These few
and - for a Christian - fairly obvious theological assumptions account
for the main traits of Leibniz's philosophy. There is a complete indi-
vidual concept of each individual creature, containing all its properties
and its entire - eternal - history. All true statements are either (i) ana-
lytic statements in Kant's sense, that is, statements of the form 'S is P\
where P is a predicate contained in the complete individual concept of
the subject S, or (ii) analytic statements in a sense closer to Frege's, viz.,
logical consequences of statements of type (i) and the laws of logic.
Two creatures must differ in some respect and cannot share the same
complete concept. Thus God can tell between any two of His creatures
merely by surveying their respective concepts. This is Leibniz's Princi-
 3.1 Leibniz and Berkeley on Physics 99

pie of the Identity of Indiscernibles.2 In reply to an objection by Clarke,

Leibniz stressed that this principle is not a logical truth but a conse-
quence of "divine wisdom"3 inasmuch as there is no reason why there
should exist several creatures with exactly the same properties. And yet
one wonders how God could even conceive the plan of creating two
or more indiscernible creatures to reject it as unwise. Of course, our
human concepts are usually common to many entities that we never-
theless succeed in distinguishing; but that is because we supplement the
clarity of the shared concept with confused and obscure ideas of sense,
which a perfect intellect is able to analyze into clear and distinct sets
of notions, a different one for each entity.
Be that as it may, what matters here are not so much Leibniz's
grounds for upholding the Identity of Indiscernibles as the implications
that the principle has for physics. Leibniz talked uninhibitedly about
them in a letter to the Dutch physicist de Voider on 20 June 1703:
Things which are different must differ in something, that is, they must
have in themselves some specifiable diversity. It is surprising that this
most obvious axiom has not been applied by men, like so many others.
But the generality of men are content to satisfy their imaginations and
do not care about reasons [...]. Thus they commonly use only incom-
plete and abstract (or mathematical) concepts, which thought supports
but which nature does not know in their bare form; such notions as that
of time, also of space or of what is only mathematically extended, of
merely passive mass, of motion considered mathematically, etc. Such
concepts men can easily fancy to be diverse without diversity - for
example, two equal parts of a straight line, since the straight line is some-
thing incomplete and abstract, which is worth considering only for the
sake of theory. But in nature any straight line is distinguished from any
other by its contents. Hence it cannot happen in nature that two bodies
are at once perfectly similar and equal. Also things which differ in posi-
tion must express their position, that is, their surroundings, and so must

2
Some twentieth-century philosophers have tried to render the Identity of
Indiscernibles innocuous by treating Hs identical with S' as one of the predicates con-
tained in the complete concept of any given individual S. They argue that, if P stands
for the conjunction of all the other predicates of S, there could well be another subject
S* such that S* is P and yet differs from S, inasmuch as S is P and identical with S,
while S* is P and identical with S*. Leibniz was too clever and too busy to indulge
in such ploys.
3
Mr. Leibnitz's Fifth Paper, §25, in Alexander (1956, p. 62).
100 Kant

not be distinguished only by their location or by an extrinsic denomi-

nation, as these are commonly understood. So in nature there can be no
bodies as they are commonly conceived, like the atoms of the Dem-
ocriteans and the perfect globules of the Cartesians, and these are
nothing but the incomplete thoughts of philosophers who do not suffi-
ciently look into the natures of things.
(Leibniz GP, II, 249-50)4
It follows at once that modern matter, as conceived in the seventeenth
century (§1.3), is just a fiction, serviceable for the intellectual endeav-
ors of finite minds who cannot have adequate thoughts about anything
real. And the same holds, of course, for more recent views, in which
nature is made to consist ultimately of a few irreducible homogeneous
"elements". Since mathematical structures can only specify their real-
izations up to isomorphism and experiments can teach nothing unless
they are repeatable, it is clear that in a Leibnizian world the
mathematico-experimental physics of Galileo and Newton and of
Leibniz himself cannot claim to properly know things as they actually
are. The Identity of Indiscernibles accounts in part for Leibniz's
repeated statement that physics busies itself with "well-founded
appearances" or "true phenomena". He wrote the following to
Arnauld on 9 October 1687:
Matter, considered as the mass in itself, is just a sheer phenomenon or
well-founded appearance, as are space and time also. It does not even
have the precise and definite qualities which could make it pass for a
determined being [. . .] because in nature even the figure which is essen-
tial to a limited extended mass is never, strictly speaking, exact or deter-
mined, due to the actual division of the parts of matter to the infinite.
There is never a globe without irregularities, or a straight line without
intermingled curvings, or a curve of a particular finite nature which is

Compare the following passage from First Truths (c. 1680-84):

Perfect similarity occurs only in incomplete and abstract concepts, where
matters are conceived, not in their totality, but according to a certain single
viewpoint, as when we consider only figures and neglect the figured matter. So
geometry is right in studying similar triangles, even though two perfectly
similar material triangles are never found. And although gold or some other
metal, or salt, and many liquids, may be taken for homogeneous bodies, this
can be admitted only as concerns the senses and not as if it were true in an
exact sense.
(Leibniz OFI, p. 519)
 3.1 Leibniz and Berkeley on Physics 101

not mixed with some other, and this in its small parts as in the large
ones; so that far from being constitutive of the body, figure is not even
an entirely real and determinate quality outside of thought. One can
never assign a definite and precise surface to any body, as could be
done if there were atoms. I can say the same thing about magnitude and
motion, namely, that these qualities or predicates are phenomenic, like
colors and sounds, and though they contain more distinct knowledge,
they can no more sustain a final analysis. As a consequence, extended
mass [.. .] which consists only in these qualities is [.. .] a mere phe-
nomenon like the rainbow.
(Leibniz GP, II, 118-19)5

The Identity of Indiscernibles does not impose any precise boundaries

on physical inquiry, but it certainly precludes it from reaching meta-
physical conclusions.

3.1.2 Mentalism and Positivism

George Berkeley is best known for his contention that bodies do not
exist without the minds that perceive them. He argued that a body's
primary qualities of shape, extension, and impenetrability are incon-
ceivable apart from its so-called secondary qualities of color, hardness
or softness, warmth or coldness, and so on. Since the latter were admit-
tedly mind-dependent, the same was true of the former. Modern matter

' Cf. the following passage from First Truths (cf. note 4). The paragraph in brackets
was crossed out by Leibniz.
There is no actual determinate figure in things, for none can satisfy infinitely
many impressions. So neither a circle nor an ellipse nor any other line defin-
able by us exists except in our understanding, or if you will, before the lines
are drawn or their parts separated.
{Space, time, extension and motion are not things but modes of consider-
ation which have some ground [modi considerandi fundamentum habentes]}.
Extension, motion, and bodies themselves, insofar as they consist in exten-
sion and motion alone, are not substances but true phenomena, like rainbows
and parhelia.

The two passages I have quoted to illustrate Leibniz's phenomenalism are compara-
tively early, but he continued to express this sentiment until the last years of his life;
thus, on 11 February 1715 he wrote to Remond: "Matter itself is nothing but a phe-
nomenon, though a well-grounded one, resulting from monads" (GP, II, 636; monad
- i.e., 'unit' - is Leibniz's term for an individual being).
102 Kant

- colorless, odorless, insipid - is sheer fiction, like Locke's abstract

idea of a triangle that is neither equilateral, nor isosceles, nor scalene.
Philosophers are prey to such fancies because they fail to understand
the power of signs, by means of which one is able to refer separately
to an aspect of a concrete object while forgetting all the other aspects
that go with it. (Thus the word 'triangle' refers to the triangularity of
any triangular figure, and not to the comparative size of its respective
sides, which, however, is well defined for each.) Berkeley was emphatic
that he was arguing only against certain aberrations of philosophers,
and had no quarrel at all with the commonsense beliefs of ordinary
people, "not yet debauched by learning" (1710, §123).

I do not argue againt the existence of any one thing that we can appre-
hend either by sense or reflection. That the things I see with my eyes and
touch with my hands do exist, really exist, I make not the least ques-
tion. The only thing whose existence we deny is that which philosophers
call matter or corporeal substance. And in doing of this there is no
damage done to the rest of mankind, who, I dare say, will never miss it.
(Berkeley 1710, §35)

Berkeley's denial of matter motivates his views on the scope and the
purpose of physics. This is not to discover the nature of things, but
only to ascertain the regularities of phenomena:

There are certain general laws that run through the whole chain of
natural effects; these are learned by the observation and study of nature
and are by men applied as well to the framing artificial things for the
use and ornament of life as to the explaining the various phenomena -
which explication consists only in showing the conformity any particu-
lar phenomenon has to the general laws of nature or, which is the same
thing, in discovering the uniformity there is in the production of natural
effects.
(Berkeley 1710, §62)

Thus, for Berkeley a particular phenomenon is scientifically explained

by conceiving it as an instance of a type that in turn regularly plays a
specific role in a typical series of phenomena. This is not to be con-
fused with causal explanation, which consists, of course, in finding the
cause of the phenomenon, that is, the agent that produced it. Accord-
ing to Berkeley, only minds can act as causes, and the vast majority of
natural phenomena should be ascribed directly to the activity of God.
Physics does not look for causal explanations.
 3.1 Leibniz and Berkeley on Physics 103

Though it be supposed the chief business of a natural philosopher to

trace out causes from the effects, yet this is to be understood not of
agents but of principles, that is, of component parts, in one sense, or of
law or rules, in another. In strict truth, all agents are incorporeal, and
as such are not properly of physical consideration.

There is a certain analogy, constancy, and uniformity in the phenomena

or appearances of nature, which are a foundation for general rules: and
these are a grammar for the understanding of nature, or that series of
effects in the visible world whereby we are enabled to foresee what will
come to pass in the natural course of things.
(Berkeley 1744, §§247, 252)

As I noted at the end of §2.5.1, a full century after Berkeley the French
philosopher Auguste Comte proclaimed that explanation by laws, not
causes, was the distinctive feature of mature, "positive" science.
This is not the only way in which Berkeley anticipated the concep-
tion of science of latter-day positivism. In the dialogues against free-
thinkers he published anonymously in 1732, Berkeley stresses that
words and other signs need not evoke ideas to be useful in scientific
discourse. For example, "in casting up a sum, where the figures stand
for pounds, shillings, and pence," it is obviously unnecessary to form
in each step, "throughout the whole progress of the operation," ideas
of pounds, shillings, and pence; "it will suffice if in the conclusion those
figures direct our action with respect to things" (1752, VII, §5; my
italics). Likewise, future and past events can be calculated from present
data through long chains of reasoning consisting for the most part of
words to which no ideas are attached. After recalling the disagreement
and confusion surrounding the term 'force' in the scientific literature
of that time, Berkeley's spokesman, Euphranor, continues:

And yet, I presume, you allow there are very evident propositions or the-
orems relating to force, which contain useful truths: for instance, that a
body with conjunct forces describes the diagonal of a parallelogram in
the same time that it would the sides with separate. Is not this a princi-
ple of very extensive use? Doth not the doctrine of the composition and
resolution of forces depend upon it, and, in consequence thereof, num-
berless rules and theorems directing men how to act, and explaining phe-
nomena throughout the Mechanics and mathematical philosophy? And
if, by considering this doctrine of force, men arrive at the knowledge of
many inventions in Mechanics, and are taught to frame engines, by
104 Kant

means of which things difficult and otherwise impossible may be per-

formed; and if the same doctrine which is so beneficial here below serveth
also as a key to discover the nature of the celestial motions; shall we
deny that it is of use, either in practice or speculation, because we have
no distinct idea of force?
(Berkeley 1752, VII, §7)

As the reader may recall, twentieth-century positivists - after vainly

struggling to secure the thoroughgoing "empirical meaning" of scien-
tific language - finally reached a position akin to Berkeley's on the sci-
entific utility of terms without referent, which they euphemistically
dubbed "theoretical terms" (cf. Carnap 1956; see also §§7.1, 7.2).

3.2 Kant's Road to Critical Philosophy

Kant's interest in the physical sciences is clear from his first two pub-
lications, the M.A. dissertation on the problem of live forces (1746)
and The Natural History and Universal Theory of the Sky (1754). In
the former he deals with the dispute between Cartesians and Leib-
nizians on whether the physical quantity conserved in elastic collisions
is proportional to mv or to mv1 (see §1.5.2). In the latter he put
forward a Newtonian hypothesis concerning the formation and
evolution of the solar system, and introduced the now current view
that most of the nebulae apparently scattered among the stars are in
fact gigantic star systems, separated from the Milky Way by enormous
distances.
In §2.5.2 I referred to Kant's Monadologia physica (1756), where
he presents the world as a collection of simple substances (monads),
which are interactive centers of force and also centers of perception.
The actual physical interaction of monads was of course denied by
Leibniz and his more faithful followers, but it was countenanced by
Martin Knutzen, with whom Kant studied philosophy in Konigsberg.
Knutzen presumably taught him also to view sense perception in the
orthodox Leibnizian way, that is, as an unclear and indistinct form of
intellection. The rejection of this view and the recognition of sensibil-
ity as a peculiar and independent source of knowledge was probably
the "great light" that Kant said had dawned on him in 1769, precipi-
tating the development of his mature philosophy (Ak. XVIII, 69). The
autarchy and even the primacy of feeling and sensation were asserted
in the eighteenth century, with increasing self-assurance, from Shaftes-
 3.2 Kant's Road to Critical Philosophy 105

bury to Sade, and Kant himself, a fervent admirer of Rousseau,6 was

certainly touched by contemporary thinking on morality and aes-
thethics. But his explicit acknowledgement of sensibility as an epistemic
faculty with its own principles occurred in connection with a problem
that is native to modern physics and its philosophy, namely, the debate
on the nature and ontological status of space.
On this issue, Kant initially held that the geometric system of spatial
relations is grounded on and abstracted from the actual interaction of
things. In particular, the fact that space has just three dimensions is a
consequence of the inverse square law of gravitation (1746, §10).7 This
relationist view should not be equated with Leibniz's, for whom things
did not interact at all. On the other hand, it somehow anticipates
Riemann's view, which in turn strongly influenced Einstein's (see §§4.1,
5.3). But Kant himself discarded it. He was probably impressed by
Euler's defense of Newton's absolute space as a prerequisite for a sat-
isfactory description of the phenomena of motion,8 but he went much
further. In a short paper published in 1768 in a local magazine he
argued that bodies depend for their very essence on their relation to
space.
Kant's argument is best explained through an example. Consider two
screws, one of them of the regular kind that will enter into a wall if
driven with a screwdriver rotating clockwise, the other an exact mirror
image of the former, which therefore advances only if the screwdriver is
rotated counterclockwise. Suppose that you sell such screws and receive
a written order in which one of them is described in terms of the mutual
distances and relative positions of its parts. No matter how detailed the
description, you cannot know what sort of screw is being requested,
unless you are told how it is set - or, as Kant said, oriented - in the sur-
rounding space. The ambiguity is not removed by referral to a room, or

6
In a private annotation, reproduced in Ak. XX, 58f., Kant draws a remarkable par-
allel between Newton and Rousseau. The rational order in nature and in human
society depends on the laws discovered, respectively, by each of them. "After Newton
and Rousseau, God is vindicated (gerechtfertigt)".
7
Obviously the inverse square law can only be stated as such with respect to a previ-
ously given geometry. So we ought to understand Kant as meaning something like
this: The "order of coexistence" of things that interact according to Newton's Law
of Gravity must possess the structure of a three-dimensional Euclidian space in which
the gravitational force between any two mass points is inversely proportional to the
square of their distance.
8
See Euler (1748; 1765, ch. II, §§78-94; 1768-74).
106 Kant

a city, or even a galaxy, for such relative spaces can in turn be oriented
one way or the other. So the character of each screw rests in the end on
its peculiar relation to infinite space. The same holds of course for the
many species of plants and animals - notably molluscs - which display
spirals oriented in a definite, hereditary sense. Kant concludes that space
is required for the full determination and hence for the existence of
bodies and so cannot be derived from them by abstraction.
In Kant's judgment, this conclusion forced him either to regard space
as a self-subsisting infinite entity indistinguishable from God himself
(as in Spinoza's Ethics) or to produce a wholly new way of conceiving
it. He chose the latter alternative. He claimed that differences in spatial
orientation - that is, between right and left, or between clockwise and
counterclockwise - cannot be intellectually grasped through general
concepts but are directly sensed through the positioning of our own
body. He saw this as an indication that sensibility is an independent
source of knowledge, irreducible to intellect, and that space is inti-
mately related to it. The same holds for time, which Kant - like
Newton, Leibniz, and Euler - unhesitatingly assimilated to space. For
a while he thought that by keeping the understanding {intellectus, Ver-
stand) separate from sensibility - in effect, by purging it from any
notions involving space or time - we would finally succeed in estab-
lishing metaphysics as a solid science of God and the soul. But further
research dispelled this illusion, and Kant's doctrine of sensibility, space,
and time became the foundation of his outright denial of the very pos-
sibility of such a science.
The first work published by Kant after the "great light" of 1769 was
the inaugural dissertation he submitted to the Faculty of Philosophy in
Konigsberg when he took possession of the Chair of Logic and Meta-
physics (1770). Time and space are described there as the "forms" of
the world of sense (mundus sensibilis), in a sense of 'form' that I shall
now explain. Kant characterizes a 'world' as a whole that is not in turn
a part (1770, §1). The matter of a world consists of its parts, "which
here we assume to be substances" (§2 I), while its form consists in "the
coordination of substances". This he conceives as something "real and
objective", indeed as "the principle of possible [mutual] influences of
the substances constituting the world".

For the identity of a whole is not secured by the identity of its parts but
requires the identity of its characteristic composition. Above all, this rests
on a real ground. For the nature of the world, which is the internal first
 3.2 Kant's Road to Critical Philosophy 107

principle of any variable determinations pertaining to its state, cannot be

opposed to itself and is therefore naturally, i.e. of itself, unchangeable.
Hence in any world there is a constant, invariable form attributable to
its nature, which is the perennial principle of any contingent and transi-
tory form that belongs to the world's state. Those who disdain this reflec-
tion are defeated by the concepts of space and time, as self-given original
conditions, by virtue of which many actuals relate to one another as parts
belonging together [uti compartes] and constitute a whole.
(Kant 1770, §2 II)

Besides matter and form, Kant's elucidation of 'world' includes one

more item, which he calls universitas and explains as "the absolute
totality of parts belonging together". Though seemingly easy, this idea
is "a crux for the philosopher", for "it is hard to conceive how the never
ending series of states of the universe eternally succeeding each other
can be reduced to a whole comprising absolutely all changes [...]. For
nothing succeeds the whole series, and yet, given a series of successive
items, only the last is succeeded by none: so there must be a last item
in eternity, which is absurd." The difficulty arises also in the case of
simultaneous infinity. "For simultaneous infinity offers eternity an
unexhaustible material for successively progressing to infinity through
its countless parts, which series would yet be actually given, complete
in all numbers, in simultaneous infinity, so that a series which can never
be completed by successive addition could nevertheless be given in its
entirety." According to Kant there is, however, a clear way out of this
"thorny question" since neither the successive nor the simultaneous
coordination of a multitude "belongs to the intellectual concept of a
whole, but only to the conditions of sense intuition [intuitus sensitivi]"
for both modes of coordination "rest upon concepts of time" (§2 III).
The meaning and weight of this remark become clear in the light of
Kant's revival of the ancient distinction between the intelligible and the
sensible world. The former he conceived as the totality of things as they
are "in themselves", which - he believed in 1770 - is accessible to our
intellect. The latter is the gathering of all appearances (apparentia,
phaenomena) displayed by things through our sensibility. He describes
sensibility as the receptivity of the mind, through which its state of
awareness can be affected in a definite way by the presence of an object
(§3). Things do not strike our senses with their form; so, to gather the
manifold of sense affections into a whole "an internal principle of the
mind is required by which that manifold acquires an aspect according
to stable and innate laws" (§4). This principle is the form of the sensi-
108 Kant

ble world. One expects that such a principle of unity should be unitary,
and indeed Kant often describes it in the singular, for example, as "a
certain law inherent in the mind for coordinating among themselves the
sensa that issue from the object's presence" (§4), or as "a definite law
of the mind, by which it is necessary that everything which (through its
qualities) can be an object of the senses be seen as necessarily belong-
ing to the same Whole". It turns out, however, that there are, accord-
ing to Kant, two "absolutely primary universal formal principles of
the phenomenal world, which are like the schemata and conditions of
everything else that is sensual [sensitivi] in human knowledge", namely,
time and space. In the Critique of Pure Reason (1781,1787), the duality
of these "primary principles" acquires a semblance of justification from
Kant's doctrine that time is the form of the "inner sense" through which
I appear to myself, while space is the form of "outer sense" through
which I get to know the presence of other things. But in the inaugural
dissertation I find no trace of this - in my view disastrous - teaching.
On the other hand, at least two passages of the work make a bold sug-
gestion that, if carried through, would secure the oneness of Kant's
"formal principle of the phenomenal world". The first of them was
quoted at the end of the foregoing paragraph; in it Kant takes for
granted that "simultaneous coordination" - which is displayed as
spatial relations - is based on a time concept (1770, §2 III). The second
passage is the footnote to §14.5, which again takes up the subject of
simultaneity and which, among other interesting things, says the fol-
lowing: "If time is represented by a straight line extended to infinity
and simultaneous things at each point of time by lines serially applied
to the former, the surface which is thus generated will represent the phe-
nomenal world [. . .]." Obviously, this representation would be much
more adequate if we attach a full copy of space to each point of the line
representing time. The suggestion clearly is that we treat three-
dimensional space as an aspect or, more precisely, a substructure of the
four-dimensional world (or "spacetime"). Kant, however, did not
pursue this suggestion. Instead, he kept space separate from time, even-
tually associating each with a different "side" of sense.9

9
Kant had no use for the fusion of space and time in a single four-dimensional con-
tinuum. He had long been familiar with the idea of w-dimensional space for n > 3
(cf. Kant 1746, §10); but he presumably conceived it as a metric space governed by
the ^-dimensional analogue of Pythagoras's Theorem. Now, within Newtonian
 3.2 Kant's Road to Critical Philosophy 109

Kant's arguments for his conception of space and time stress three
main points: (i) our ideas of space and time are not intellectual - for
they do not relate to particular spaces and times like a concept to its
instances, but like a whole to its parts -; nevertheless, (ii) they are not
acquired through the senses - for our knowledge of space geometry
and the linear order of time is not liable to correction and improve-
ment through experience -; and (iii) they involve requirements -
expressed in the said incorrigible knowledge - that every object of the
senses must comply with.10 Kant concludes that time and space origi-
nate in our own sensibility. Therefore, although things as known
through our senses are thoroughly impregnated with spatial and tem-
poral properties and relations, these need not belong to things as they
are in themselves. In Kant's view, this solves the paradoxes of com-
pleted infinity (see refs. to Kant 1770, §2 III). The paradoxes depend
on the assumption that existing things are fully determined in every
respect.11 But appearances are not thus determined, so their potential
infinity need not be actually completed. Of course, the solution works
only if things in themselves are neither spatial nor temporal. Now, a
moment's reflection shows that if we eliminate all explicitly or implic-
itly spatial and temporal properties and relations from our description
of bodies and events, there will be nothing left to say about them. So
Kant's new conception of space and time and his solution of the para-
doxes of infinity imply that our senses do not teach us anything at all
about things as they are in themselves.
In 1770, Kant thought that this result was very promising for meta-
physics. By keeping our understanding clean of any ideas stemming
from our sensibility - and this includes, of course, every reference to
space or time - we would soon be able to know the core of things. It
is baffling that Kant should ever have entertained such hope, for he
had equated sensibility with our capacity to receive information from
objects. Hence the purified understanding could only know what it

mechanics there is no room for such a spacetime metric. Indeed, the application of
Pythagoras's Theorem to Minkowski's relativistic spacetime involves the rather untidy
trick of using time coordinates that are imaginary numbers.
10
The arguments are found in their original form in Kant (1770, §§13-14). Kant (1787,
§§3-7) contains the final version.
11
This assumption probably has to do with the Christian belief that, apart from God
himself, who lacks nothing, all existing things are God's creatures, and therefore their
totality has everything it was meant to have.
110 Kant

drew out of itself. The first clear evidence that he realized this difficulty
and changed his program for metaphysics accordingly is his letter to
Marcus Hertz of 21 February 1772. Kant asks, what is the ground of
the relation of an idea to its object? Let me call this 'the question of
1772'. Kant observes - naively yet plausibly - that if the idea contains
the way in which the subject is affected by the object, the idea is related
to its object as an effect to its cause. "But our understanding is not the
cause of the object through its ideas [...], nor is the object the cause
(in a real sense) of the ideas of the understanding. The pure concepts
of the understanding cannot therefore be abstracted from sensations or
express the receptivity of ideas through the senses, but they must have
their sources in the nature of the soul [. . . ] " (Ak. X, 130). Kant recalls
that in the inaugural dissertation of 1770 he had said that "the ideas
of sense present things as they appear, and intellectual ideas present
them as they are". But, "if such intellectual ideas rest on our own inter-
nal activity, wherefrom comes the agreement that they are supposed to
have with objects which, however, are not produced by them?" (Ak.
X, 131). The question of 1772 is thus closely connected with the epis-
temological problem that lies at the center of the Critique of Pure
Reason, viz., how can we possess information about objects that is not
supplied by the action of those objects on our senses?12 Kant did not
doubt that such information was contained in arithmetic, geometry,
and the more general statements of physics (e.g., the principle that the
quantity of matter remains constant in all natural processes). More-
over, the metaphysical textbooks used in German universities by him
and his colleagues purported to give information about things in
general (general metaphysics or ontology, the science of being qua
being) and about God, the human soul, and the totality of creatures
(special metaphysics, comprising natural theology, rational psychology
and philosophical cosmology), which, by its very nature, cannot be
based on sense experience. However, the erratic performance of these
supposedly scientific disciplines made Kant wary of them.

12
In Kant's parlance, an informative statement is said to be synthetic, and a statement
not based on sense data is said to be a priori. Hence, Kant's epistemological problem
can be concisely formulated as follows: How are synthetic a priori statements possi-
ble} The possibility of analytic statements - i.e., statements that do not convey any
information about the objects that they mention, but merely speak about what is
implicit in the meaning of their terms - does not constitute a problem for Kant.
(Analytic statements, of course, are not - and need not be - supported by sense expe-
rience and are consequently all a priori.)
 3.2 Kant's Road to Critical Philosophy 111

The Critique of Pure Reason delivers Kant's solution to the ques-

tion of 1772 in the so-called transcendental deduction of the pure con-
cepts of the understanding (1781, pp. 95-130; 1787, pp. 129-69).13 I
cannot reproduce here Kant's convoluted argument, but I shall try to
summarize its main steps. A manifold of successive sense impressions
can be apprehended only if it is retained, reproduced in (short-term)
memory, and recognized as such. All this involves a mental activity of
unifying and binding together (Kant's term is "synthesis"). To recog-
nize parts of a manifold as remaining the same and belonging together
the active mind must of course be aware of its own identity. "Thus all
manifold of intuition has a necessary relation to the 'I think' in the
same subject in which this manifold is found." This relation requires,
however, that "all my ideas (even if I am not conscious of them as such)
meet the condition under which alone they can stand together in one
universal self-consciousness" (1787, p. 132). This condition amounts
to the possibility of being combined into a single interconnected system,
governed by certain principles. The human understanding is charac-
terized by Kant as the power to effect such combinations. So the said
principles must reflect the rules under which the understanding oper-
ates on the manifold of sense. Indeed, according to Kant, the primary
pure concepts of the understanding (or "categories") are just abstract
representations of those rules and "contain nothing more than the
unity of reflection about phenomena, insofar as they must necessarily
belong to a possible empirical consciousness" (1787, p. 367).14

13
The text of 1787 is quite different from that of 1781, which, however, according to
Kant, should still be read as a complement of the new version (cf. 1787, p. xlii). Note
that 'deduction' in this context simply means 'justification' or 'proof of legitimacy'
(1781, p. 84). The adjective 'transcendental' applies in Kant's parlance to knowledge
that "busies itself not with objects, but with our way of knowing objects, insofar as
this ought to be possible a priori" (1787, p. 25). I discussed both versions of the tran-
scendental deduction in Torretti (1967, pp. 262-385) (in Spanish). For a detailed
commentary in English, see Paton (1936, I, 313-585).
14
Kant took special pride in his alleged discovery of the complete list of the categories.
Although his views on this matter have almost never been shared by others (see
however Reich 1932), I shall explain them briefly in this note. Kant contended that
traditional formal logic knew enough about the operations of the human under-
standing to provide a full classification of the acts of judgment by which different
ideas are brought under a single concept. According to him, "the same function which
gives unity to the different ideas in a judgment also gives unity to the bare synthesis
of different ideas in an intuition" (1781, p. 78f.; 1787, p. 105). Since the categories
express in the abstract the several modes in which this unifying function is exercised
112 Kant

Thus the original and necessary consciousness of one's own identity is

at the same time consciousness of an equally necessary unity of the syn-
thesis of all phenomena according to concepts, that is, according to rules,
which not only make phenomena necessarily reproducible but also in so
doing determine an object for the intuition of them, i.e. the concept of
something wherein they are necessarily interconnected.
(Kant 1781, p. 108; my italics)

Therefore, in Kant's mature thinking, even sense impressions owe to

the understanding their reference to objects. The agreement between
such objects and the pure concepts of the understanding is no longer
puzzling, for these concepts express the rules by which the under-
standing organizes the manifold of sense into a system of objects. Such
agreement, however, is confined to the objects of sense. As Kant puts
it: "The categories [. . .] do not afford us any knowledge of things
except through their possible application to empirical intuition; that is,
they serve only for the possibility of empirical knowledge" or "expe-
rience" (1787, p. 147).15 Consequently, there can be no metaphysical

on the manifold of sense, their full list should match exactly the logical classification
of the ways of exercising the same function in judgment. To secure this match, Kant
distinguished two kinds of judgment of the form '5 is P' besides those admitted in
traditional logic, viz., the singular judgments, in which the predicate P is arbitrary
but the subject S is the proper name of an individual, and the indefinite judgments,
in which the subject 5 is arbitrary but the predicate P expresses a privation. (They
allegedly correspond to the categories of unity and of limitation, respectively.) But
even with Kant's doctoring the match remains doubtful. Thus, in spite of Kant's
lengthy explanations, few have been privileged to see how the category of commu-
nity or reciprocity between agent and patient matches the class of disjunctive judg-
ments (of the form '5 is either Pt or P2 o r . . . Pw'). More plausible and yet, in my
opinion, quite perverse is Kant's view that the intellectual operation yielding a causal
statement of the form 'A causes £' exercises precisely the same function as hypo-
thetical judgments of the form 'If p, then q\
15
This important conclusion is further elucidated in the following text:
Space and time, as conditions of the possibility that objects be given to us, are
valid no further than for objects of the senses, and therefore only for objects
of experience. Beyond these limits they represent nothing, for they are only in
the senses and have no existence outside them. The pure concepts of the under-
standing are free from this limitation, and extend to objects of intuition in
general, be it like ours or not, provided that it be sensible and not intellectual.
But this further extension of concepts beyond our sensible intuition is of no
help to us in anything. For as such they are empty concepts of objects, which
do not even enable us to judge whether those objects are possible or not. They
are mere forms of thought without objective reality, because we have no intu-
 3.3 Kant on Geometry, Space, and Quantity 113

science of theology, psychology, or cosmology, and our interest in God,

immortality and freedom, should seek satisfaction from sources other
than our cognitive powers. On the other hand, a science of being in
general or ontology is indeed possible, but only insofar as it deals with
the conditions that all phenomenal objects must meet to become inte-
grated in our experience. Such a study ultimately provides a meta-
physical foundation for mathematical physics. In the next three sections
I shall discuss some aspects of Kant's work in this area.

3.3 Kant on Geometry, Space, and Quantity

Since its Greek beginnings, geometry was a paradigm of secure knowl-
edge. Plato stressed its independence from sense experience; indeed, he
even held that sense appearances were inherently unable to satisfy geo-
metric relations with perfect accuracy. According to him, the certainty
and precision of geometry show that our human souls are rooted in
another world; they give us intimations of immortality and warrant
our capacity to achieve no less reliable knowledge of right and wrong.
Modern thinkers conceived geometry as the science of space, which
they either identified with matter (Descartes) or regarded as an eternal
precondition for matter's existence (Newton), so they had no doubts
as to the exact realization of geometric truths in the physical world.
But they continued to see geometry as a purely intellectual achieve-
ment, and even the arch-empiricist Locke dreamt of a deductive science
of morality on the analogy of mathematics.
Kant's new conception of space radically changed this state of affairs.
Ethics now has to go it alone and can draw no comfort from the success
of geometry. Geometry, on the other hand, is now bound to physics
more than ever, inasmuch as its truth rests wholly on the fact that it
spells out conditions of the possibility of our knowledge of physical
objects. (Of course, since such objects are just phenomena, i.e., objects
for us, the conditions of the possibility of our knowing them are also
conditions of the possibility of the physical objects themselves.) Our
notion of space is pure - for it is presupposed by our external percep-

ition at hand to which the synthetic unity of apperception - which constitutes

the whole content of these forms - could be applied so that they might deter-
mine an object. Only our sensible and empirical intuition can procure them
meaning and reference.
(Kant 1787, pp. 148-49)
114 Kant

tions and therefore cannot be extracted from them (1770, §15A) - and
intuitive - for it is the notion of a unique object, of which particular
spaces are parts, not a common concept of which particular spaces are
instances. According to Kant, the axioms of geometry bear witness to
this pure intuition. "That space has no more than three dimensions,
that between two points there is single straight line, that about a given
point on a plane a circle can be described with a given radius, are not
conclusions inferred from some universal notion of space, but can only
be discerned, so to speak, concretely in space itself" (1770, §15C).
Indeed Kant maintained that even geometric demonstrations do not
proceed by sheer logical deduction but must resort to intuition at every
significant step.16 Moreover, according to him, those who try to develop
a nonstandard geometry labor in vain, for they are forced to use the
standard - intuitive - notion of space in support of their fictions.17
Summing up, for Kant "space is not something objective and real,
not a substance, not an attribute, not a relation, but as it were a scheme
for coordinating together absolutely everything that is externally
sensed, which is subjective and ideal and issues from the nature of the

16
Kant (1770, §15C, last sentence); cf. (1781, pp. 716f., 734f., 782f.). Modern
axiomatics belies Kant's claim, but in his time elementary geometric demonstrations
did in fact resort to intuitive facts that were not expressed in the axioms and postu-
lates. According to Jaakko Hintikka (1967), when Kant spoke of the incessant appeal
to intuition in geometric demonstrations he referred to the use of existential instan-
tiation in proper geometric reasoning ("consider this particular - yet arbitrarily
chosen - triangle . . . " ) , not to the use of figures to fill in the gaps of reasoning. Hin-
tikka's interpretation clears Kant from the - to my mind, venial - sin of not foresee-
ing the achievements of Pasch (1882) or Hilbert (1899), but I am not persuaded by
it: Proof by existential instantiation works for predicates and relations displayed by
the particular individual brought to mind only if they are also made explicit in the
general premise that is being instantiated, and are thus seen to lie within one's con-
ceptual grasp.
17
Kant (1770, §15 E). Kant's friend, Johann Heinrich Lambert, was one of the fore-
runners of non-Euclidian geometry. In his posthumously published Theory of
Parallels (1786) he suggests that trigonometric relations on a sphere with radius
V—1 provide a model for a two-dimensional geometry in which Euclid's Postulate V
is false (see §4.1.1). Lambert's tract was written c. 1766, but there is no evidence that
Kant ever saw the manuscript. Nor does the extant correspondence between both
men contain any reference to nonstandard geometries, but Kant may have learned
about them in conversation. Strictly speaking, Lambert's sphere with an imaginary
radius is not a figure of standard geometry. Still, conservative philosophers of a later
age were wont to see the mathematicians' appeal to such para-Euclidian models as a
sign that non-Euclidian geometries cannot stand on their own feet.
 3.3 Kant on Geometry, Space, and Quantity 115

mind according to a stable law" (1770, §15D). Although with respect

to things in themselves space is something "imaginary,. . . with respect
to anything sensible not only is it most true, but it is the foundation
of all truth in external sensibility. For things cannot appear to our
senses under any aspect except through the mind's power of coordi-
nating all sensations according to a stable law inherent in its nature"
(1770, §15E). Therefore, nature is exactly subject to the precepts of
geometry "not according to some fabricated hypothesis, but by virtue
of an intuitively given subjective condition of all phenomena which
nature can ever manifest to our senses" (Ibid.). Because of it, geome-
try is the paradigm and the vehicle of all scientific evidence, for, "as
geometry studies the relations of space, whose notion contains the very
form of all sense intuition, nothing can be clear and perspicuous in
what is perceived externally except through the same intuition whose
contemplation is the business of geometry" (1770, §15C).
In Kant's later writings this doctrine of space - and the parallel doc-
trine of time - is on the whole preserved. Indeed, in 1791 he described
it as one of the two hinges on which true philosophy turns, the other
one being "the reality of the concept of freedom" (Ak. XX, 311).
However, his innovative views on the role of the understanding in the
constitution of the subject matter of physics inevitably produced a shift
in his thinking about intuition. The shift shows up in a small change
introduced by Kant in the second edition of the Critique of Pure
Reason. In both the first and the second editions space is said to be
"nothing but the form of all phenomena of outer sense" (1781, p. 26;
1787, p. 42), but the notion of the form of a phenomenon is explained
first as "that which makes that the manifold of the phenomenon is intu-
ited [as] ordered in certain relations" (1781, p. 20),18 and later as "that
which makes that the manifold of the phenomenon can be ordered in
certain relations" (1787, p. 34; cf. Ak. XVII, 639). This change is nec-
essary because all coordination and hence all relational order in a man-
ifold are now considered to be the work of the understanding.19

18
I insert the word 'as' to meet the requirements of English syntax; but its German
equivalent 'als' does not occur in Kant's text. The following rendering is perhaps more
accurate: "that which makes that the manifold of the phenomenon is intuited in a
certain relational order". However, it makes the changes in the wording of the second
edition look greater than they really are.
19
"The binding together [Verbindung] of a manifold in general can never come to us
through the senses and cannot therefore be already contained in the pure form of
sensible intuition. For it is an act [Actas] of the spontaneity of the cognitive faculty,
116 Kant

Enlightened by this view, Kant distinguishes between (i) the form of

intuition, that is, the condition inherent in our sensibility that con-
strains us to perceive everything as embedded in the structures of space
and time, and (ii) the cognitive grasp of those structures by what he
now terms formal intuition.
Space, represented as an object (as we actually need to do in geometry)
contains more than the mere form of intuition, for it contains a grasp-
ing together [Zusammenfassung] in one intuitive representation of the
manifold given according to the form of sensibility. Thus the form of
intuition gives just a manifold, but formal intuition gives unity of rep-
resentation. This unity [.. .] presupposes a synthesis, which does not
belong to the senses but through which all concepts of space and time
first become possible. [. . .] Through it - as the understanding determines
sensibility - space and time are first given as intuitions.
(Kant 1787, pp. 160-61n.)
The active contribution of the understanding to the geometric order-
ing of things is not properly highlighted in the Critique and is usually
ignored by commentators, but §38 of Prolegomena leaves no doubt
about it.20 This §38 is meant to illustrate Kant's contention that "the
understanding does not draw forth from nature its (a priori) laws, hut
prescribes them to her" (1783, §36, last sentence). As an example of
such a law he proposes the following: If two chords AB and CD meet
inside a circle at P, then, no matter how the chords are chosen, the
rectangles formed from the segments of either chord are always equal
(AP x PB = CP x PD). Kant asks himself:

Does this law lie in the circle or in the understanding? That is, does this
figure, independently of the understanding, contain in itself the ground
of the law; or does the understanding, having constructed the figure
according to concepts of its own (namely, of the equality of the radii),
introduce thereby into the figure this law of chords intersecting each
other in geometrical proportion? When we follow the proofs of this law
we soon see that it can only be derived from the condition on which the
understanding based the construction of the figure, namely, the equality
of the radii.
(Kant, Ak. IV, 321)

and since this [spontaneity], to distinguish it from sensibility, must be called under-
standing, so is all binding together [. . .] an action [Handlung] of the understanding"
(Kant 1787, pp. 129-30).
20
On Prolegomena, §38, see Friedman (1992, Ch. 4).
 3.3 Kant on Geometry, Space, and Quantity 117

Kant proceeds to examine a generalization of his example and then

asks again:
Do these laws of nature lie in space, and does the understanding learn
them merely by exploring the rich store of meaning which resides in
space? Or do they lie in the understanding, and in the way in which it
determines space according to the conditions of the synthetic unity to
which all its concepts point? Space is something so uniform and so inde-
terminate with regard to all particular properties, that we should cer-
tainly not seek in it a treasury of laws of nature. Instead, what determines
space to assume circular shape, or the figures of a cone and a sphere, is
the understanding, insofar as it contains the ground of the unity of their
constructions. The mere universal form of intuition, called space, is thus
indeed the substrate of all intuitions which can be determined [as refer-
ring] to particular objects, and the condition of the possibility and the
variety of such intuitions certainly lies in it. But the unity of objects is
entirely determined by the understanding, and indeed according to con-
ditions which lie in its own nature.
(Kant, Ak. IV, 321-22)
Kant's strong statement about the indeterminacy of space "with
regard to all particular properties" must, however, be taken with a
pinch of salt, for he continued to think that the understanding is
required by "the mere universal form of intuition, called space" to
bestow on it a Euclidian structure. This can be seen in the section on
"Axioms of intuition" in the Critique of Pure Reason (1787, pp.
202-207). It concerns the principle governing the application of the
categories of quantity to the manifolds displayed in sense awareness.
The principle is: "All intuitions are extensive magnitudes"'. Since "all
phenomena contain, as to their form, an intuition in space and time,"
they can only be grasped "by the synthesis of the manifold through
which the representations of a determinate space or time are generated,
i.e. by putting together the homogeneous manifold and becoming con-
scious of its synthetic unity" (1787, pp. 202-203). Thus, "I cannot rep-
resent to myself a line, however small, without drawing it in thought,
that is, generating from a point all its parts one after the other. [. . .]
Similarly with all times, however small: I think only of the successive
advance from one instant to another, whereby through all parts of time
and their addition a determinate time-magnitude is finally generated"
(1787, p. 203). An extensive magnitude is one "in which the repre-
sentation of the parts makes possible the representation of the whole"
(Ibid.), so that a line and a time interval can only be conceived as exten-
118 Kant

sive magnitudes, and the same evidently holds for any more complex
spatio-temporal configuration. This conclusion is fairly weak and
seemingly uncontentious. But Kant jumps from it to a much stronger
one, viz., that "this successive synthesis of productive imagination in
the generation of figures" is the foundation of geometry and its axioms,
for the latter "express the conditions of sense intuition a priori under
which alone" the pure concepts of quantity can be applied to an exter-
nal phenomenon (1787, p. 204). To illustrate this assertion, Kant judi-
ciously picks two axioms that are not exclusively Euclidian, viz. (i) Two
points are joined by a single straight line, and (ii) Two straight lines
never enclose an area. But there can be little doubt that, under inter-
rogation, he would have placed on a par with them all the axioms
required for proving Pythagoras's Theorem, and hence for character-
izing Euclidian distance. (After all, without a distance function, a geom-
etry can hardly be seen as dealing with extensive magnitudes.) How,
despite the uniformity and indeterminacy of space, the understanding
is constrained by "the conditions of sense intuition a priori" to per-
forming only Euclidian constructions is, I am afraid, anything but
perspicuous.

Appendix
Kant does not view geometry as the only application of the categories
of quantity. The following remarks deal briefly with two further aspects
of his thinking on this matter.
Some followers of Kant - among them the great Irish mathemati-
cian W. R. Hamilton - fell into the temptation of conceiving arithmetic,
on the analogy of geometry, as based on the a priori intuition of time.
But Kant astutely resisted it.21 On 25 November 1788, he wrote to
Pastor Schultz that "time, as you correctly note, has no influence on
the properties of numbers [. . .] and the science of number is - despite
the succession required by every construction of magnitude - a purely
intellectual synthesis, which we represent to ourselves in thought" (Ak.
X, 557). Of course, the existence of such a "purely intellectual syn-
thesis" is hard to reconcile with Kant's contention that no cognitive

21
Despite his baffling characterization of number as "nothing else but the unity of the
synthesis of the manifold of an homogeneous intuition in general by means of my
generating time itself in the apprehension of the intuition" (1781, pp. 142-43).
 3.3 Kant on Geometry, Space, and Quantity 119

synthesis is purely intellectual, and that in any case it must involve

either time or space. Thus, it turns out that - contrary to the conven-
tional wisdom - Frege's attempt to establish arithmetic as a chapter of
quantificational logic (i.e., the theory of singular, existential, and uni-
versal propositions - the formal logical counterpart of Kant's categories
of quantity) came in fact to Kant's rescue in this matter. For if the truths
of arithmetic are actually truths of logic, they are, in Kant's terms, a
priori but not synthetic, and therefore need not rest on one of the forms
of intuition.
According to Kant only extensive magnitudes, such as volumes or
durations, in which the whole is grasped as compounded from multi-
ple parts, are conceived under the categories of quantity, viz., One,
Many, and All; while intensive magnitudes or degrees belong under
the categories of quality, viz., Reality, Negation, and Limitation. He
explains this as follows:

What corresponds in empirical intuition to sensation is reality (realitas

phaenomenon); what corresponds to the lack of it is negation = 0. Now,
every sensation is capable of diminution, so it can decrease and gradu-
ally vanish. Thus, between phenomenal reality [Realitdt in der Erschei-
nung] and negation there is a continuous connection through many
possible intermediate sensations, the difference between which is always
less than the difference between the given one and zero or complete
negation.
(Kant 1787, p. 209)

Consider two actual sensations, a and fe, such that a is equal to one of
the possible sensations intermediate between b and its complete nega-
tion. We say then that the realities corresponding to a and b sport the
same quality to a different degree, and that the degree disclosed by a
is less than that disclosed by b.
As a description of ordinary usage and its underlying rationale the
above is fairly accurate. But Kant goes further. The statement that "in
all phenomena, the real which is an object of sensation, has an inten-
sive magnitude, i.e. a degree" (1787, p. 207) expresses, according to
him, a condition of the possibility of experience, without which it is
not possible to grasp an object in space and time. He rests this claim
on the allegation that it is always possible for an empirical conscious-
ness to change gradually "so that the real in it vanishes completely and
there remains a merely formal consciousness (a priori) of the manifold
in space and time". Whence, conversely, it is possible to synthesize "the
120 Kant

generation of the magnitude of a sensation",22 from its beginning, the

pure intuition = 0, to an arbitrarily chosen intensity of it (1787, p. 208).
This is Kant's principle of "anticipations of perception". It accounts
for the fact - first noted by Hume (THN, I.Li, p. 6) - that we some-
times can anticipate qualities we have never sensed, for example, a
nuance of blue intermediate between two given samples of paint.

3.4 The Web of Nature

In Kant's thinking, the constitution of our experience of objects in space
and time depends chiefly on the use of the categories of relation, viz.,
substance-and-attribute, cause-and-effect, and community of interac-
tion. Kant states three principles - called by him "Analogies of Expe-
rience" -, which govern, respectively, the application of each category,
plus a general principle of the Analogies, which sets the stage for the
others. He offers proofs of all four. These principles raise different
problems that still engross the philosophy of physics. I shall discuss
them separately in the next four subsections.

3.4.1 Necessary Connections

In 1781, pp. 176-77, the general principle of the Analogies read thus:
"All phenomena are subject a priori, with regard to their existence, to
rules concerning the determination of their mutual relations in one
time." In 1787, p. 218, it had been changed to: "Experience is possi-
ble only through the representation of a necessary connection of per-
ceptions." The earlier version expresses in a general way the function
of the three special principles. On the other hand, Kant's argument for
the principle - added in 1787 - resulted in the second version.
The argument runs as follows. In experience, perceptions come
together casually, and do not by themselves disclose a necessary con-
nection between them. "Perceptual grasp [Apprehension] is just a
putting together of the manifold of empirical intuition", and it does
not include "a representation of the necessity of the connected exis-
tence in space and time of the phenomena it puts together". However,
because experience is knowledge of objects through perceptions, rela-

22
Kant says that "eine Synthesis der Grossenerzeugung einer Empfindung" - literally,
"a synthesis of the magnitude-generation of a sensation" - is possible. My paraphrase
involves a deliberate - and probably failed - attempt to make Kant's meaning clearer.
 3.4 The Web of Nature 121

tions touching the existence of the manifold must be represented in

experience not as the manifold is merely put together in time, but as
it is in time objectively. Since time itself cannot be perceived, "the deter-
mination of the existence of objects in time can only occur through
their being combined in time in general, and therefore only through a
priori connecting concepts". Such concepts "always carry with them
necessity, so that experience is only possible through a representation
of the necessary connection of phenomena" (1787, p. 219).
The purport of this argument will, I hope, become clearer in the light
of Kant's treatment of causality (§3.4.3). But we need not wait for it
to inquire what sort of necessity must, by Kant's argument, be present
in experience, or, more precisely, what are the source and the scope of
such necessity. A careful look at Kant's reasoning prompts us to dis-
tinguish three sorts of necessity, (i) By the transcendental deduction, it
is necessary^ that the manifold of sense be unifiable by the under-
standing in a single, coherent, self-conscious spatio-temporal experi-
ence, (ii) If Kant's table of categories is final and complete, then it is
necessary2 that the synthesis of the manifold of sense which is thus nec-
essarily possible should yield instances of his categories or of concepts
derived from them, (iii) If any entity - be it an object, an event, or a
system of objects or events - is an instance of a specific concept, it is
necessary?, that the said entity shall possess all the attributes entailed
by this concept. In particular, if the entity e is manifested through a
variety of phenomena /"1?. . ., fm any connection Xhk between, say, fh
and fk that is required for e to be an instance of a concept C is of course
a necessary connection if e actually is such an instance.
Note that all concepts carry with them necessity of this third kind,
whether or not they derive from Kant's table of categories. Thus, if
something is correctly diagnosed as a healthy mammal, it must contain
a healthy heart; if an event is correctly described within current parti-
cle physics as a collision between a negative and a positive electron, it
must be a source of radiation. Necessity?, is often scorned as merely
"verbal" necessity, in part no doubt because it is so obvious and per-
vasive, but also because, being conditional on the right application of
concepts, it does not protect us against uncertainty. Commonsense con-
cepts are often fuzzy and have a feeble grip on their instances. A
mammal's health can fail suddenly, for example, if its heart stops.
Science tries to work with concepts that are well-defined and stable. If
two objects are in fact what is known in current physics as two dif-
ferently charged electrons, it is inevitable that, upon meeting, they fuse
122 Kant

into a spurt of radiation. Yet even in this case uncertainty persists, for
the conceptual systems of physics are subject to seemingly endless
revision.
I expect that after finishing this book the reader will be persuaded
that natural necessity, as understood in modern physics, is simply
necessity3, and that the appearance of its being something stronger and
harder to conceive is due to the richness of physical concepts, the com-
plexity of their instances, and the firm grip that the former get on the
latter through experiment and measurement. Kant, however, was not
of this persuasion. As we shall see below, he argues in detail that the
constitution of experience in time must be presided by the three cate-
gories of relation. In this way he claims necessityi not for concepts in
general, but for these concepts specifically. As members of Kant's list,
these categories possess moreover necessity2, and admit no substitutes.
The necessity that they carry with them, and which is displayed in expe-
rience by their instances, is of course necessity3, but Kant apparently
thought that it would be not be securely grounded unless the concepts
from which it stems are specifically required for the possibility of expe-
rience (are necessary!) and belong to the small permanent inventory of
primary concepts of the human understanding (are necessary2).

3.4.2 Conservation of Matter

The First Analogy or "Principle of the Permanence of Substance" reads
thus: "In all change of phenomena, substance persists, and its quantum
does not increase or decrease in nature" (1787, p. 224).23 Kant's proof
is quite remarkable: "All phenomena are in time", in which alone we
can represent their coexistence and succession. Time therefore remains
unchanged, while succession and coexistence "can only be represented
as determinations of it". Time, however, "cannot be perceived by
itself". Hence, "in the objects of perception, i.e. in phenomena, there
must be found the substrate which represents time in general, and in
which all change or coexistence can be perceived through the relation
of phenomena to it" (p. 225). Kant explains that "the substrate of all
that is real, i.e. belonging to the existence of things, is the substance".

23
The statement of the First Analogy in the first edition made no mention of a fixed
quantum: "All phenomena contain the permanent (Substance) as the object itself, and
the changeable as its mere determination, i.e., as a way in which the object exists"
(1781, p. 182).
 3.4 The Web of Nature 123

"Consequently, the permanent, in relation to which alone all time rela-

tions of phenomena can be determined, is the substance in the phe-
nomenon, i.e. what is real in it, which as substrate of all change remains
always the same. As this cannot change in existence, its quantum in
nature cannot increase or decrease" (p. 225).
This argument raises a problem for Kant interpreters. This is not the
place to discuss it, but allow me to state it. Throughout the Critique,
Kant maintains the existence of things-in-themselves, independent of
human experience, whose true nature cannot be known from their
appearances in space and time. "In fact, when we consider the objects
of sense, rightly, as mere phenomena, we grant thereby that a thing in
itself lies at their foundation, although we know not how it is consti-
tuted in itself, but only [know] its appearance, that is, the way in which
our senses are affected by this unknown something" (1783, §32; my
italics). But Kant's proof of the First Analogy neatly disposes of this
claim: To be "read" as experience, the changing sense appearances
must be "spelled" as manifestations of an underlying substrate;
however, such a substrate is just the conceptual representative of time
- the universal form of our sensibility -, and is nothing apart from its
role as a steady referent "in relation to which alone all time relations
of phenomena can be determined".24
I turn now to other questions that are more directly relevant to
physics. Suppose that the constitution of experience in time does indeed
require that we tie the kaleidoscopic flow of phenomena to a perma-
nent substance. Still, one might ask, why should its permanence show
up as a conserved quantity?25 The only answer I can find lies in Kant's
identification of "the substance in the phenomenon" with "what is real
in it". That "the real" should have an intensive magnitude in all phe-
nomena was for Kant a condition of the possibility of experience

24
For the "spell/read" metaphor, see Kant (1781, p. 314; 1783, §30). Kant's proof of
the First Analogy of Experience agrees well with - and further clarifies - Locke's well-
known remark on our "notion of pure substance in general": Whoever examines it
"will find he has no other idea of it at all, but only a supposition of he knows not
what support of such qualities which are capable of producing simple ideas in us"
(1690, Il.xxiii.l). Obviously, such an idea cannot stem from sense impressions.
25
As I said in note 23, a quantum of substance was not mentioned in the 1781 for-
mulation of the First Analogy. However, already in 1781 (p. 185) Kant illustrated the
First Analogy with the story of the chemist who figured out the weight of smoke by
subtracting the weight of the ashes left over from the weight of the wood burnt. (This
story was already documented in late Antiquity; see Lucian, Demonax, 39.)
124 Kant

(1787, p. 207). This answer, however, raises other questions. First, the
First Analogy asserts the conservation of a single physical quantity,
which is cautiously described as "the quantum of substance", but
which, as I intend to show in the next paragraph, can be none other
than the quantity of matter, as measured by Newtonian mass; yet in
Kant's discussion of intensive magnitude, sense qualities of different
sorts are mentioned as having degrees, which take their places along
different scales, so there must be more than just one physical quantity
in this sense (viz. temperature, sound pitch, color saturation, etc.). One
can reply that these are degrees of sensation, not of reality. The latter
is "that which corresponds to sensation" (1787, p. 182), and Kant
repeatedly calls it "matter" (1787, pp. 34, 182, 609f.; 1786, in Ak. IV,
481). Indeed, the aim of mechanistic natural philosophy was to pin all
variety and variation in sensation on the distribution and redistribu-
tion of matter alone; and Kant, in his critical writings, certainly appears
to have shared this aim.26 Still, one may wonder, secondly, whether the
argument for intensive magnitudes is not undermined by the First
Analogy's conservation principle. For that argument - summarized in
the Appendix to §3.3 - rested on the gradual variability of empirical
consciousness and hence of the real that it grasps. So if the real is nec-
essarily fixed, the reason given for assigning it a quantity does not seem
to hold. This difficulty is solved, I think, by distinguishing between (i)
the degree to which matter is present on a particular occasion at a point
in space, and (ii) the quantity of matter in {a) a finite region V, or (b)
in nature as a whole. We met this distinction already at the beginning
of §2.1, in the definition of Newtonian mass: (i) is density, conceived
as a time-dependent scalar field (Chapter Two, note 4), whose value
can vary from point to point and may grow with time at a given point
from 0 to any finite value; (ii) is mass, conceived (a) as the integral
Jvpd V of the density p over the finite region V, or even (b) as the inte-

26
However - in stark opposition to seventeenth-century mechanicism - he conceived
matter as being solely a source of force, and he thought that there were natural forces
of several kinds. He says in 1786: "All that is real (alles Reale) in the objects of the
external senses, which is not a mere determination of space (place, extension and
shape), must be regarded as motive force; so that the so-called solidity or absolute
impenetrability is expelled from physics as a vacuous concept, and is replaced with
repulsive force, while on the other hand the true immediate attraction is defended
against the sophistries of a metaphysics which misunderstands itself, and is declared
necessary, as fundamental force, for the very possibility of the concept of matter."
(Ak. IV, 523).
 3.4 The Web of Nature 125

gral j%pd V of the density p over the whole of Newtonian space %. There
is no question that both (iia) and (iib) can retain a constant value while
the density p varies in time and space. The First Analogy requires the
constancy of (iib). It is doubtful, however, that it makes any sense to
speak of the constancy of j^pdV unless the integral converges, which
it can only do under stringent additional conditions. In physics, of
course, 'conservation of mass' normally means something else, namely,
that the value of JvpdV over any finite region V at the end of a time
interval x is always equal to its value at the beginning of x plus the net
inflow of mass across the boundaries of the region during x. I am not
sure that the principle thus understood would meet Kant's
requirements.
I still have to justify my assertion that the quantum of substance
that, according to Kant, can neither increase nor decrease in nature,
is precisely Newton's quantitas materiae. Although this is not made
explicit in the proof of the First Analogy or in the comments that follow
it, Kant does refer to chemical calculations based on the conservation
of mass as a telling example of the First Analogy's use (1781, p. 185;
see note 25). Then, in 1786, after defining "the quantity of matter" or
"mass" as "the aggregate of what is moveable in a definite space (die
Menge des Beweglichen in einem bestimmten Raum)" (Ak. IV, 538),
he derives the following "law", which he links directly to the First
Analogy:
First Law of Mechanics. In all changes of corporeal nature, the quantity
of matter in the whole remains the same, unincreased and undiminished.
(Kant, Ak. IV, 541)
Why, then, does Kant studiously avoid the term 'matter' in his discus-
sion of the First Analogy? Kant neatly distinguishes (i) "transcenden-
tal philosophy", which receives no information from the senses, and
deals, independently of any particular object of experience, with "the
laws which make possible the concept of nature in general", from (ii)
the "metaphysics of corporeal nature", which takes "the empirical
concept of matter" for granted and investigates the extent to which its
object can be known a priori (1786, in Ak. IV, 469f.). The First
Analogy belongs to transcendental philosophy; the "First Law of
Mechanics" is its application to the metaphysics of bodies. So Kant
would be right in leaving 'matter' out of the former. But is the abstract
concept of matter involved in these considerations a truly empirical
concept? By Kant's definition, matter can only be known through sen-
126 Kant

sation, that is, empirically; but this does not imply that the concept
'matter' is empirical. Indeed, there is an almost perfect match between
Kant's definition of the matter of phenomena as "that which corre-
sponds to sensation" (1787, p. 34) and his characterization of reality
in phenomena as "that which, in empirical intuition, corresponds to
sensation" (1787, p. 209). And, as we saw above, the proof of the First
Analogy turns on the identification of the permanent with "what is real
in [the phenomenon] {das Reale derselben), which as substrate of all
change remains always the same" (1787, p. 225). So his avoidance of
the term 'matter' may be somewhat disingenuous after all. Still, one
may wish to distinguish between the fully abstract concept of matter
in the definition just quoted and the narrower concept that the "meta-
physics of corporeal nature" takes as its starting point. Matter is
characterized there as "that which in external intuition is an object of
sensation" and is specified further as "the moveable" that "fills space"
and "qua moveable, possesses motive force" (1786, in Ak. IV, 481,
480, 496, 536). Perhaps this concept of matter as the object of outer
sense is in effect empirical. Still, according to Kant, only matter thus
understood can afford the permanent reality demanded by the First
Analogy. For, as he argues in the "refutation of idealism",

All determination of time presupposes something permanent in percep-

tion. But this permanent cannot be an intuition in me. For all grounds
of determination of my existence which can be found in me are repre-
sentations and require, as such, a permanent [something] distinct from
them, in relation to which their change, and so my existence in the time
wherein they change, may be determined.
(Kant 1787, p. 276, as corrected on p. xxxix)

So "we have nothing permanent which we can place as intuition under

the concept of a substance, except only matter" (1787, p. 278). This
is to be understood, of course, as the object of outer sense.
Besides the principle of the conservation of mass, mathematical
physics has countenanced, since its inception, several other conserva-
tion principles. Due to their global scope, empirical evidence for them
has often seemed inadequate, so there has been a tendency to vindicate
them by reason alone. We have seen Descartes derive the conservation
of motion from the immutability of God (§1.3), and Leibniz argue for
the conservation of "force" (i.e., energy) on the ground that one cannot
get something for nothing (§1.5.2). By the 1780s three conservation
principles were well entrenched in mechanics, regarding (a) linear and
 3.4 The Web of Nature 127

(P) angular momentum, as well as (y) mechanical energy. By a remark-

able mathematical theorem due to Emmy Noether (1918), these prin-
ciples can be shown to follow respectively from the symmetries of
Euclidian space (a,(3) and Newtonian time (y).27 From an orthodox
Kantian standpoint Noether's theorem yields a perfect a priori justifi-
cation for these three principles. Nowadays, however, one would rather
argue, by means of it, for the adoption of this or that particular sym-
metry group, from whatever empirical evidence can be mustered for
the corresponding conservation principles.
Although Kant's thoughts on the constitution of human experience
were chiefly directed toward the rational foundation of physics, they
were meant to apply also to prescientific experience as described in
ordinary discourse. Now, in everyday talk - at any rate in English and
other European languages - we do refer the quickly varying aspects of
sense experience to more or less stable things. But these are many, not
just one, as Kant's argument for the First Analogy would seem to
suggest.28 And, although they usually last a good deal more than their
fugitive states, they are certainly not everlasting. Thus, ordinary usage

The conservation of linear momentum is a consequence of the homogeneity of space

(invariance of spatial relations under translation in any direction), the conservation
of angular momentum is a consequence of the isotropy of space (invariance of spatial
relations under rotation about any point), and the conservation of mechanical energy
- in a conservative mechanical system - is a consequence of the homogeneity of time
(invariance of temporal relations under translation). See Landau and Lifschitz (1960,
Ch. II). For a very readable proof of Noether's theorem, see Lovelock and Rund
(1975, pp. 201-206).
Throughout most of the section on the First Analogy, Kant uses the word 'substance'
(Substanz) in the singular. It occurs in the plural (Substanzen) only toward the end,
in the following two sentences: "Alteration can therefore be perceived only in sub-
stances", and "Substances (in the phenomenon) are the substrate of all determina-
tions of time" (both in 1787, p. 231). The next occurrence of 'substance', at the
beginning of the proof of the Second Analogy, is strongly monistic: "All phenomena
in the succession of time are one and all only alterations, i.e. a successive being and
not-being of the attributes of the substance, which abides" (1787, p. 232). But in the
section on the Third Analogy, Kant persistently speaks of interacting substances (in
the plural). In fact, his proof of the First Analogy is quite compatible with a plural-
ity of substances if the permanent ultimately consists of ingenerate and indestructible
indivisible corpuscles or atoms. But in his maturity Kant rejected all forms of
atomism, including the dynamic monadology of his youth. He conceived matter as a
continuum whose presence at each point of space is represented by a variable density
scalar. Such a view of matter can no doubt be expressed in pluralistic terms, but a
monistic formulation sounds much more natural.
128 Kant

supports the philosophical category of substance-and-attribute, but it

does not agree with Kant's handling of it. He went so far as to write
that if some substances are born and others perish, this "would remove
the sole condition of the empirical unity of time, and phenomena would
then relate to two different times, in which existence would flow in
parallel streams, which is absurd" (1787, pp. 231f.). I find it hard to
think of a more preposterous philosophical claim. Durable things can
stand proxy, as in Kant's argument, for the one permanent time we
cannot perceive, even if they are not eternal. It is enough that they
mutually interact within partially overlapping sets that, so to speak,
take turns at guarding the identity and continuity of time. I do not
expect my watch to last forever; but the time it keeps need not break
down with it; it can be kept further by my next watch, provided that
there is a third one that coexists for a while with each, was seen to
agree with the former, and is used for setting the latter.

3.4.3 Causality
Even if we do not subscribe to Kant's theory of the human under-
standing, we may readily agree that the category substance-and-
attribute is deeply entrenched in our everyday language - where it
shows up in the subject-predicate structure of simple sentences -; that
it is also found, with little or no modification, in the classical physi-
cist's conception of matter whose quantity is conserved through its
manifold changes of state; and that Kant's contention that substance
functions in the constitution of experience as a tangible representative
of time is, if not true, at least reasonable. When we come to the cate-
gory cause-and-effect, things are less straightforward. It is clear that
we continually use some such category in the analysis and interpreta-
tion of ordinary events, for describing which we have a rich stock
of causal verbs.29 On the other hand, we often hear of a 'principle
of causality' that ruled over mathematical physics until Quantum
Mechanics allegedly dethroned it, but which might be restored to
its former glory if the de Broglie-Bohm "causal interpretation" of
Quantum Mechanics is successful. However, as we shall see, between
the ordinary concept of a causal relation and the applications of the

29
The following "small selection" is due to Anscombe: scrape, push, wet, carry, eat,
burn, knock over, keep off, squash, make (e.g., noises, paper boats), hurt (1971,
p. 9).
 3.4 The Web of Nature 129

said principle of physics there are incongruities that prompt one to

think that 'causality' in the name of the latter is a misnomer (and a
source of confusion, given that physics, in its experimental practice,
cannot do without the ordinary concept). Due in part to these incon-
gruities, the Second Analogy of Experience, conceived by Kant as a
common foundation of the use of both the physical principle and the
ordinary concept of causality, in the end accounts for neither. To justify
these assertions, I shall first go over Kant's argument for the Second
Analogy, and then discuss its relations to the ordinary notion of
cause-and-effect and to the so-called principle of causality of classical
physics.
In the second edition of the Critique of Pure Reason, the Second
Analogy bears the title of "Principle of Succession in Time in Accor-
dance with the Law of Causality" and is stated as follows: "All alter-
ations happen in accordance with the law of the connection of cause
and effect" (1787, p. 232). In the first edition it was called "Principle
of Generation (Erzeugung)" and read thus: "Everything that happens
(begins to be) presupposes something which it follows in accordance
with a rule" (1781, p. 189). Despite the enormous change in wording,
Kant apparently believed that both texts were semantically equivalent,
for he did not rewrite the first edition's discussion of the Analogy, but
merely placed in front of it two new paragraphs, headed by the word
"Proof". In these paragraphs he recalls that, as a consequence of the
First Analogy, all phenomena that succeed each other in time are merely
alterations, "i.e., a successive being and not-being of the attributes" of
a permanent substance. He then argues for the Second Analogy
(version of 1787) as follows: When I perceive that two phenomena A
and B succeed each other I connect together two perceptions in time.
This connection is not a gift of sense intuition, but "the product of a
synthetic faculty of imagination" (1787, p. 233). This can connect the
phenomena so that A precedes B or so that B precedes A. Since time
itself is not perceived, it is not possible to ascertain which of the two
phenomena came first "in the object" by comparing them both with
time. Since the objective order of succession does not depend on the
way the phenomena are given to the senses, and it cannot be derived
from their relation to time, it must lie in our conceptual grasp of it:
"the relation between the two states must be so thought that it is
thereby determined as necessary which of them must be placed before
and which after, instead of the other way around" (p. 234; my italics).
The rest of the proof merely applies the argument for the general prin-
130 Kant

ciple of the Analogies (§3.4.1) to the special case of succession in time:

"The concept which carries with it a necessity of synthetic unity can
only be a pure concept of the understanding, which does not lie in per-
ception, and here it is the concept of the relation of cause and effect,
the former of which determines the latter in time, as its consequence"
(p. 234).
This understanding of causality as the relation between two events
(i.e., two transient phenomena or "alterations", to use Kant's term),
one of which necessarily follows the other in accordance with a rule,
comes straight out of Hume. Hume had argued that, since the idea of
necessary connection is not obtained from sense impressions, it can
only reflect our compulsive tendency to think of the cause in the pres-
ence of the effect and to think of the effect in the presence of the cause
due to the habit of perceiving them together. To counter this degrada-
tion of physical to psychological necessity Kant resorted to his doctrine
concerning the constitution, by the human understanding, of a web of
objective time relations in nature, but he accepted Hume's analysis of
the causal relation almost unmodified.30 Yet that analysis has little to
do with our ordinary concept of cause-and-effect. Humean causation
is a relation between two events, but our causal verbs normally take
persons or things as subjects.31 In ordinary causation, the cause usually
exists before the effect, but it can very well come into being together
with it (like the newborn child and his first act of breathing, accord-
ing to some definitions of birth). Everyday causal inquiries usually ask
(a) who - or what - is to blame for some, usually unwelcome, change
in our environment, or (b) which effects a person must cause to achieve
some desired state of affairs. Inquiries of type (a) appeal to the regu-

30
We should recall, however, that according to Hume the proximate cause must be con-
tiguous to its effect. Kant wholly disregards this requirement, presumably because it
rules out instantaneous action-at-a-distance.
31
See note 29. Our word 'cause' is derived from the Latin causa, which primarily meant
'legal case' - whence also 'plea', 'excuse', 'pretext', 'motive', 'purpose', and 'reason'
- but which occurs in philosophical literature as the standard Latin equivalent of the
Greek word aixta. Now, aixia was Aristotle's word for 'cause' in a sense so broad
that some modern translators render aixia (in Aristotle) as 'explanation'; but in ordi-
nary Greek aixia primarily meant 'responsibility', mostly in a bad sense, that is,
'blame' (but also 'merit' - Aeschylus, Septem, 4). The noun aixia was closely related
to the adjective aixux;, 'responsible', 'culpable', and the verb aixidojiai, 'accuse',
'censure', or 'lay to one's charge', 'impute', aixia is first attested in Pindar and
Herodotus, but aixio<; and aixidoum are old, common words, that were already
found in Homer.
 3.4 The Web of Nature 131

larities of nature (e.g., paternity inquiries resort to the laws of genet-

ics), but only for guidance. A person X can be held responsible for a
change Y even if effects such as Y do not follow from actions such as
X's necessarily or even regularly. U is guilty of V's death if V died from
one bullet shot at her by 17, even if four other bullets missed the target
or caused only minor wounds and a sixth bullet, shot at V's guest,
failed to kill him because he reacted better to medical care. Inquiries
of type (b) do, of course, depend decisively on the regularity of phe-
nomena, but for which they would rarely be of any use. And yet even
in this case necessity is not an essential feature of causation. Consider
an agent U who, in the light of such an inquiry, causes V1?. . ., Vn to
achieve W. Surely the connection between U and the V's is not a nec-
essary one, or else he would not have had to inquire about them. As
for the connection between the V's and W, U surely would like it to
be necessary, but ordinarily this will not be the case. Typically, there
might be several, partially overlapping, sets of means available, Y1 -
{V1?. . ., V/,}, T 2 , • • • ? yn a n d U will choose to execute one of them
after estimating their respective cost and probability of success. If we
never settled for less than a necessary connection between means and
ends, there is precious little we could get done.
So much for the similarity between Kant's Humean category of
cause-and-effect and the homonymous concept one meets in life. But
perhaps what Kant - and Hume - had in mind was not the anthropo-
morphic concept of causation we inherited from our prehistoric ances-
tors, but the scientific concept involved in the "principle of causality"
that is said to govern classical physics (and which allegedly broke down
with the advent of Quantum Mechanics). As we shall now see, this is
admirably suited for the role assigned by Kant to cause-and-effect in
the construction of an objective time order. Still, due perhaps to Kant's
desire to keep his categories in touch with folk concepts,32 his "law of
the connection of cause and effect" does not quite agree with the physi-
cists' "principle of causality". The latter is in fact only a short - but
high-sounding - way of referring to the following characteristic of clas-
sical physical systems: Their evolution in time is governed by a system
of differential equations that, by virtue of its mathematical properties,
normally has unique solutions, in which case, if the value of certain

32
Folk concepts form the "massive central core of human thinking which has no history
- or none recorded in histories of thought" (Strawson 1959, p. 10). They give firmer
support to Kant's view of changeless reason than the fickle notions of science.
132 Kant

physical quantities at a particular instant is given exactly, the state of

the system is fixed for all times (see §2.5.3, after eqns. (2.43)). Thus,
the physicists' "principle of causality" is in effect a principle of deter-
minism,33 and therefore, according to our discussion in the foregoing
paragraph, it is quite foreign to ordinary causal thinking. Of course,
scientists are wont to use words in different, often quite disparate,
senses (think of 'vector' as used in physics and in epidemiology).
However, since commonsense causal thinking is a central feature of
laboratory life, it is convenient to avoid any confusion between 'cau-
sation' and 'determinism' - at any rate, a physicist's interventions in
his experimental equipment must surely be up to him, and not fixed
since time immemorial by the evolution of a physical system that
embraces them both.
There are other discrepancies between the concept of cause-and-
effect, in both the ordinary and the Kantian acceptance, and the idea
of a physical system's deterministic evolution under differential equa-
tions with unique solutions. Suppose we try to describe the evolution
of such a system S in causal terms. To do so we must consider causal-
ity, with Kant and Hume, as a relation between events, the relata being
in this case the states of S at different times. The state st of S at any
given time tx certainly "follows in accordance with a rule" on its state
s0 at some earlier time to. But could one say, without sounding artifi-
cial, that s0 is the cause of Si? Should I say, for example, that the present
angular momentum of the earth and its position relative to the fixed
stars (from which I see the sun right over my head, moving westward
at 15° per hour) cause the angular momentum and the position that
the earth will have 18 hours from now (from which I shall see the sun
rise in the east)? One may feel tempted to see this as a case of indirect
causation, today's state causing tomorrow's through all the states that
the system will have in between. However, in ordinary conversation we
would never assert that A indirectly causes B unless we believe that
there is some effect C that A causes directly and which in turn brings
about B. Thus, if we say that the Luddite terrorist Mr. U in Sacramento
indirectly caused the death of Ms. V, a communications engineer in
New York, with a mailbomb, we imply that he intervened personally
at a definite point in the process leading to the explosion of the bomb

33
For example, Heisenberg (1927, p. 197), gives a "sharp formulation of the Law of
causality (Kausalgesetz)" as follows: "When we know the present exactly we can cal-
culate the future". See also Frank (1932, pp. 3Off.) and Hopf (1948, pp. 1-2).
 3.4 The Web of Nature 133

that killed her, either by packaging it and mailing it, or by ordering his
assistant to do so, or in some other way. But in the evolution under
differential equations of our system S there is no state st that is caused
directly by an earlier state. Between a given state Si and any earlier
state s0 there must be an uncountable infinity of states, or else their
succession could not be governed by differential equations. Since the
same is true again of s0 and each state between it and su none of these
states can be singled out as a direct effect of s0. The contrast between
the discreteness of causal chains, as ordinarily understood, and the con-
tinuity of evolution under differential equations moved J. R. Lucas to
present the latter not as a mere application, but rather as a "general-
ization" - in fact, a creative extension - of commonsense causal think-
ing (1984, chapter X). Still, this approach does not take care of the
biggest discrepancy. If a closed system S evolves under differential equa-
tions with unique solutions, the state s of S at any particular time t
determines every other state s' of S, no matter whether s' follows or
precedes s. And, of course, s' also determines s. In other words, the
binary relation 'x determines y\ where x and y are different states in
the evolution of a physical system subject to differential equations, is
a symmetric relation; whereas 'x causes y' is antisymmetric: If x causes
y, it is certainly false that y causes x.34
On the other hand, the physicist's understanding of successive phe-
nomena as states of a system that evolves under a set of differential
equations with unique solutions bestows necessity on the temporal
relations between those phenomena in a manner that is unrivaled by
any other form of thought. This is necessity3 in the sense of §3.4.1:
State Si necessarily follows state s0 after time tx - t0 (if t0 < tx) or is fol-
lowed by it after time to-ti(iito>ti), because both states lie on the
same solution of the said set of equations and correspond respectively
to times tx and t0. If we represent the solutions of our set of equations
- as indeed we may - by curves in a space of sufficiently large dimen-
sion number, we see that Si follows or precedes s0, at the stated time
intervals, with the same kind of necessity that constrains two straight
lines on a Euclidian plane II to meet at some point of II, unless there
34
Antisymmetry holds for both common sense and Humean causality. The former must
be antisymmetric because the relata are heterogeneous: If x is to blame for y, y cannot
be to blame for anything, and x is not the sort of thing that something else could be
blamed for. Humean causation is a binary relation between entities of the same (most
general) kind, viz., events, but the effect must succeed the cause in time, and tempo-
ral succession is of course antisymmetric.
134 Kant

is a third straight line that meets them both at right angles, or that
makes the distances from the foci of an ellipse to any point on it add
up to the length of the diameter through the foci. If sa did not stand
in precisely that temporal relation to s0, s0 and Si could not be what
the physicist takes them to be, viz., just those states of just that system.
The strength and the scope of this kind of necessity are more readily
acknowledged in the case of simple geometric figures in three-
dimensional space because the mathematical concept of a physical
system is so much more complex and it is so much harder to ascertain
whether and in what terms a given set of phenomena should be brought
under it. Anyway, the necessity3 that the classical concepts of physical
systems carry with them serves the demand for an objective ordering
of phenomena in time (put forward by Kant as necessary^ much better
than the presumptive necessity2 of the category of cause-and-effect. For
the latter is not only doubtful or liable to exceptions, as we have seen;
but even if this category were applicable in Kant's sense, it would con-
stitute only an arbitrary and unintelligible connection between its
relata. Kant concedes as much when - commenting on his Analogies
of Experience - he compares the meaning of 'analogy' in mathematics
and in philosophy. A mathematical analogy states the equality of two
quantitative relations (ratios), so that when three of the four quanti-
ties involved are given the fourth can be "constructed" (if x:a::b:c, then
x - able). But a philosophical analogy asserts the equality of two qual-
itative relations, so that "from three given terms I can only know the
relation to a fourth one, but not this fourth term itself; yet I have a
rule for seeking it in experience, and a mark by which to find it there"
(1787, p. 222). This is a good deal less than what mathematical physics
can do for us. Just think of astronomers who, after a few sightings of
a newly discovered comet, merely by conceiving it as part of the Solar
System (regarded as a practically closed Newtonian gravitational
system), are able to construct its trajectory for the next six months and
to understand why it cannot be otherwise.

3.4.4 Interaction
The Third Analogy governs the application of the category of interac-
tion. Since this category is Kant's own creature, we do not have to con-
sider whether his treatment of it agrees with ordinary usage. Our
attention will go to the category's prototype in Newtonian physics and
to the role that Kant assigned to it in the constitution of experience.
 3.4 The Web of Nature 135

The concept of interaction is meant to fit the relation between two

bodies that attract each other in accordance with Newton's Law of
Universal Gravitation. This relation is perfectly symmetric, although,
if the bodies have unequal masses, it has different effects on each.35
Kant's words suggest at times that interaction is just two-way causa-
tion.36 Can this be right? Causation is antisymmetric: If x causes y, then
y does not cause x. Thus, 'x interacts with y' cannot, under pain of
self-contradiction, mean the same as 'x causes y and y causes x\ Of
course, if x and y stand for bodies exerting a Newtonian gravitational
pull on one another, 'x interacts with y' is not intended to mean that
'x causes y and y causes x\ but rather that 'x causes a state (viz. of
acceleration) or a change (viz. of velocity) in y while y causes a (similar)
state or change in x\ This explication, however, will not work if 'x
causes y' is given the Humean sense of a relation between events, for
an event is not the sort of thing that suffers changes or abides in states.
So, if Kant's concept of causation agrees on this point with Hume's, it
cannot be used in this way for explicating his concept of interaction.
It is preferable to take seriously his description of interaction as a cat-
egory, that is, as a basic, irreducible concept, that cannot be under-
stood solely in terms of other concepts.37
Throughout the section devoted to the Third Analogy, Kant speaks
of interaction as a relation between substances. This agrees well with
what I have just said, but it might seem to raise a problem with Kant's
proof of the Analogy. The Third Analogy, like the first two, was rewrit-
ten for the Critique of 1787, but in this case the proof added in 1787
suits the text of 1781 quite well. The earlier version was: "All sub-
stances, insofar as they exist simultaneously (zugleich), stand in thor-
oughgoing community (i.e. mutual interaction)" (1781, p. 211). This
was replaced by: "All substances, insofar as they can be perceived in

35
A crumb of bread and the planet earth attract each other with forces of exactly the
same magnitude, but only the former experiences a significant acceleration as a result
of this, because F = ma, and the earth's m is so very much larger than the crumb's.
Interaction in Kant's sense obviously covers Coulomb attraction or repulsion between
electric charges and also momentum exchange in elastic collisions. Indeed, the
Third Analogy translates, in the Metaphysical Principles of Natural Science, into the
following version of Newton's Third Law of Motion: "Third Law of Mechanics.
In all communication of motion, action and reaction are always equal" (1786 in
Ak. IV, 544).
36
For instance, in the paragraph beginning o n p . 2 6 1 of Kant (1787).
37
Cf. Kant's letter t o J o h a n n Schultz of 17 February 1784 (Ak. X , 3 6 6 - 6 8 ) .
136 Kant

space as simultaneous, are in thoroughgoing interaction" (1787, p.

256). The proof recalls that two things are said to be simultaneous
when the perception of one of them can follow that of the other, and
vice versa. "Thus I can direct my perception first to the Moon and
then to the Earth, or, conversely, first to the Earth and then to the
Moon; and because the perceptions of these objects can alternatively
follow each other, I say that they exist simultaneously" (1787, p. 257).
However, we cannot perceive time and learn that two things exist
simultaneously by observing their placement in it. All we can gather
from our temporal grasp of things is that each perception is present in
the subject when the other one is not, and vice versa, but not that the
objects are simultaneous, that is, that when one exists the other one
also exists at the same time, and that this is necessary in order that the
perceptions can alternate as they do.

Consequently, in order to say that the alternating sequence of percep-

tions is grounded in the object and thereby to represent simultaneous
existence as objective, we require a pure concept of the alternating
sequence of the properties of these things existing simultaneously outside
each other. Now, the relation between substances one of which has prop-
erties whose ground is contained in the other is the relation of influence,
and when each, reciprocally, contains the ground of properties in the
other, this is the relation of community or interaction. Therefore, the
simultaneous existence of substances in space cannot become known in
experience except on the assumption of their mutual interaction. Con-
sequently, this is also the condition of the possibility of the things them-
selves as objects of experience.
(Kant 1787, pp. 257i.)

The problem with this argument is that, in the light of Kant's dis-
cussion of the First Analogy, if there is more than one substance, they
must all be coeval, for the birth of some while others perish would
destroy the unity of time (§3.4.2, last paragraph). Hence, all substances
exist simultaneously all the time, whether they interact or not. The First
Analogy thus undermines Kant's argument for the Third and makes the
latter, as formulated in 1781, completely idle. However, the text of
1787 - "All substances, insofar as they can be perceived in space as
simultaneous, are in thoroughgoing interaction" - might avoid this
reproach if, with some hermeneutic good will, one seeks the key to its
meaning in the words I have italicized. We perceive substances only
through their transient states, and so - presumably - we can perceive
 3.4 The Web of Nature 137

them as existing at the same time only if the states through which we
perceive them occur simultaneously. From this perspective, Kant's argu-
ment for the Third Analogy may be reconstructed as follows: The
simultaneous occurrence of state a of substance A and state b of sub-
stance B will be established as a matter of objective fact only by
grounding ^ on B's being b and b on A's being a; therefore, unless all
the substances are in thoroughgoing interaction, the simultaneous
occurrence of their states cannot be known and, consequently, the
simultaneous existence of the substances themselves cannot be
perceived.
If this argument is valid, the simultaneity of distant events presup-
poses instantaneous distant interaction. In this way, by dint of Kant's
philosophical ingenuity, the most objectionable feature of Newtonian
gravity is turned into a precondition of the empirical knowledge of
spatial objects in time. Indeed, Kant has nothing practical to say about
the synchronization of distant events, which certainly could not be
carried out in his time - or even now - by observing gravitational inter-
actions. However, by mentioning the need for a physical foundation of
objective simultaneity he probably contributed to motivate Einstein's
more fruitful handling of this question in 1905 (§5.1).38
The Third Analogy rounds off the Kantian construction of nature
as a field of human experience. In Kant's own eloquent words:
By nature (in the empirical sense) we understand the connection of phe-
nomena, as regards their existence, according to necessary rules, that is,
according to laws. There are certain laws, indeed a priori laws, which
first make nature possible. Empirical laws can operate and be discovered
only through experience, and indeed in consequence of those original
laws through which experience itself first becomes possible. Our analo-
gies therefore properly represent the unity of nature in the connection of
all phenomena under certain characters which only express the relation
of time (insofar as time comprises all existence) to the unity of apper-
ception, which can only be achieved in synthesis according to rules. So,
taken together, the analogies say that all phenomena lie, and must lie, in

38
Kant expressly mentions light as a vehicle for the propagation of simultaneity,
although he must have known that it travels with finite speed: "From our experiences
one may easily gather that only the continuous influences in all points of space can
lead our senses from one object to another; that light, playing between our eye and
the heavenly bodies, effects a mediate community between us and them, and thereby
demonstrates the simultaneous existence of the latter" (Kant 1787, p. 260).
138 Kant

one nature, because without this a priori unity no unity of experience,

and therefore no determination of objects in it, would be possible.
(Kant 1787, p. 263)

3.5 The Ideas of Reason and the Advancement of Science

As the final item in our selection of Kantian themes, I propose to deal
briefly with the Ideas of reason and their significance for the philoso-
phy of physics.
Having shown that all purported knowledge of God, freedom, or
immortality is illusory, Kant did not dismiss the pseudoscience of meta-
physics as a manifestation of man's silliness and conceit, but ascribed
its origin to a natural illusion of reason, a "transcendental mirage" that
will not vanish merely because the critique of reason has exposed its
vacuity ("e.g. the mirage in the statement: 'the world must have a
beginning in time'" - Kant 1787, p. 353). This explanation of meta-
physics as a necessary evil is closely related to Kant's partition of the
intellectual powers of man among two "faculties", viz., understanding
(Vet'stand), which constitutes experience by articulating sense appear-
ances as objective phenomena, and reason proper (Vernunft), which
guides the understanding in this process and therefore may be said to
regulate experience. Reason performs this function by setting certain
unattainable goals that will keep the human understanding busy
forever. Every such goal is represented in thought by what Kant calls
an Idee, that is, a "necessary concept of reason, such that no object
congruent with it can be given to the senses" (1787, p. 384; cf. 1783,
§40; 1790, §57, Anm. I). I render this Kantian term as 'Idea' with a
capital T to distinguish it from the ordinary English word 'idea' (which
is closer to Kant's Vorstellung). Although Ideas cannot determine any
object, "they can serve the understanding as a canon for its extended
and consistent employment; the understanding does not hereby get to
know any more objects than it would by its own concepts, but is better
and further conducted in this knowledge" (1787, p. 385).
The conception of reason as a guide of life can be traced back to the
ancient Stoics, who called it to hegemonikon, 'the guiding [principle]'.
Kant's originality lies in considering it as a guide of science, and indeed
in the thought that science can use a guide. Such a thought was out of
the question while 'science' designated the repository of all truth,
shining forever in God's mind, of which our human science was a tiny
- yet otherwise unadulterated - portion. What was needed then was a
 3.5 The Ideas of Reason 139

guide of ignorance, that is, a method for progressively getting rid of it

and increasing one's share in divine science. But Kant, as we have seen,
radically separated human from divine science, the former being con-
fined to phenomena while the latter presumably embraces things-in-
themselves. In fact, "intellectual intuition" - the sort of knowledge that
God, if He exists, has of everything - is introduced by Kant only as a
negative Idea: a paradigm of what human knowledge cannot be. By con-
trast, human science is essentially an enterprise, forever unfinished and
thoroughly drenched with ignorance, so it is no wonder that it should
need guidance. Specifically, the understanding, in all areas of experi-
ence, faces conditioned aspects of phenomena, whose conditions it must
determine, for instance, by locating them in spatial and temporal sur-
roundings, or by analyzing them into parts, or by finding their causes.
Typically, such conditions are conditioned in turn: Locations have their
own surroundings, parts are analyzable wholes, causes are events
effected by other causes. In each line of inquiry, reason prescribes the
search for the totality of conditions of every conditioned feature of
things. Such totality, of course, will never be given in experience and
therefore can only be represented by an Idea. Thus, the Ideas

of totality in the synthesis of conditions are necessary - and grounded

in the nature of human reason - at any rate as tasks for carrying through
the unity of the understanding, where possible, up to the unconditioned;
even though these transcendental concepts otherwise lack any suitable
application in concreto, so that their sole utility lies in setting the under-
standing on such a course that its employment is both extended to the
uttermost and made thoroughly consistent with itself.
(Kant 1787, p. 380)

Kant is emphatic that reason is never directly concerned with an

object, but only with the understanding. "It does not, therefore, create
any concepts (of objects), but only orders them, and gives them that
unity which they can have in their widest possible extension, that is,
with respect to the totality of series" (1787, p. 671). Reason
directs the understanding towards a certain goal, with a view to which
the guiding lines of all the latter's rules converge to a point. This is indeed
only an Idea (focus imaginarius), that is, a point from which the con-
cepts of the understanding do not actually proceed, for it lies wholly
outside the limits of possible experience; but it serves however to procure
them maximal unity together with maximal scope. From this we get the
illusion that the lines issue from an object outside the field of empirically
140 Kant

possible knowledge (just as we see objects behind the surface of a

mirror). Though we can hinder this illusion from deceiving us, it is
nonetheless inevitably necessary if besides the objects in front of our eyes
we also want to see those which lie far behind our backs, i.e., in our
case, if we want to direct the understanding beyond every given experi-
ence (a part of the whole possible experience), and so also towards the
furthest and greatest possible enlargement.
(Kant 1787, pp. 672-73)
Different forms of this illusion generate the three branches of meta-
physica specialis as delineated by Christian Wolff, viz., natural theol-
ogy, rational psychology, and philosophical cosmology. The first two
rest, according to Kant, on fallacious inferences, which he explains and
refutes in some of the more readable portions of the Critique of Pure
Reason (1781, pp. 341-406, 571-642; 1787, pp. 399-432, 600-670).
In the remainder of this chapter I shall deal only with the contradic-
tions of philosophical cosmology, better known as Kant's antinomies.39
The four cosmological questions leading to the antinomies concern,
(I) the temporal origin and spatial boundary of the world, (II) the divis-
ibility of bodies, (III) the existence of an uncaused initial cause in every
causal series, and (IV) the thoroughgoing necessity or utter contingency
of physical events. They are age-old problems. In the Monadologia
physica (1756) Kant valiantly attempted to solve the second one (as I
noted in §2.5.2). And in his inaugural dissertation he presented the first
as an apparently insuperable obstacle to the very notion of world as
an "absolute totality of parts belonging together" (see §3.2, after the
indented quotation from Kant 1770, §2 II). In 1772 or 1773, while
working on the Critique of Pure Reason, he lighted on the idea that
we had here a conflict of reason with itself - neatly articulated in four
theses and antitheses to match the fourfold table of categories - which
can be overcome only by admitting that bodies and processes in space
and time are mere phenomena and not things-in-themselves.
The thesis of the First Antinomy is that the world (a) has a begin-

39
The title of the relevant chapter is "The Antinomy of Pure Reason", in the singular,
meaning the state of opposition (anti) to its own law (nomos) in which reason finds
itself by virtue of the cosmological contradictions. However, Kant subsequently refers
to the contradictions themselves - and/or the purportedly valid arguments that lead
to them - as "the Antinomies" (first through fourth). This impropriety has been uni-
versally adopted in the literature and is probably the source of the misnomer 'antin-
omy' applied to all sorts of contradictions and paradoxes.
 3.5 The Ideas of Reason 141

ning in time and (b) is limited in space. To prove it, Kant argues that,
if it were not so, (a) the present would be at the end of an eternal series
of events, yet the notion of such a completed infinity is absurd; and (b)
an infinite time would be required for synthesizing a spatially infinite
world, which, by the same token, is absurd as well. The antithesis
denies both parts of the thesis. If such denial were false, Kant argues,
then (a) there would be a first instant in the history of the world, pre-
ceded by empty time, but nothing can be born in an empty time;40 and
(b) since the whole world can be spatially limited only by empty space,
there would be a relation between the world, that is, the absolute total-
ity of spatial objects, and the absolute absence of them, "but such a
relation, and consequently the limitation of the world by empty space,
is nothing" (1787, p. 457).
In the Second Antinomy the existence of ultimate indivisible parts
of matter faces the infinite divisibility of bodies. The latter follows, of
course, from the infinite divisibility of space: The smallest body can be
divided, at least in thought, into two smaller bodies (each, say, with
one-half its volume). But it runs against the following difficulty: If every
body is composed of other smaller bodies, without end, then nothing
remains when the relation of composition is removed; and yet a com-
posite can only subsist on the strength of the reality of its parts. So
some parts must be indivisible, even in thought. But this clashes with
the divisibility of space.
According to Kant, the First and the Second Antinomies arise from
the assumption that physical objects are fully determinate, so that, in
either case, the argument proving the falsehood of the thesis necessi-
tates the truth of the antithesis, and vice versa. He believes that this
assumption holds for things-in-themselves.41 However, if physical
objects are phenomena, which are being gradually determined in the

40
"Because no part of such a time possesses, as compared with any other, a distin-
guishing condition of existence rather than of non-existence, and this applies whether
the thing is supposed to arise of itself or through some other cause" (Kant 1787, p.
456; Kemp Smith translation).
41
This belief was common among modern Christian metaphysicians, who apparently
thought that nature would not measure up to the Creator if any determinables
remained undetermined. But it was certainly not shared by Aristotle, who maintained
that bodies, although indefinitely divisible, are not therefore infinitely divided (Phys.
263a29). Why Kant, notwithstanding his devastating critique of Leibniz-Wolfflan
metaphysics, continued to uphold the thoroughgoing determinacy of things-in-them-
selves is a mystery that I am unable to unravel.
142 Kant

course of the progressive construction of experience by the human

understanding, the two theses and the two antitheses can all be false
at once, and therefore the contradictions vanish.
The Third and Fourth Antinomies can be seen as variants of the
problem of the groundless ground, which is aptly illustrated by an
ancient myth. To the question "Why doesn't the earth fall?" the myth
replies "Because it stands on top of an elephant". The elephant, in turn,
stands on the back of a tortoise, which stands on top of another ele-
phant, and so on. Evidently, the piled-up creatures will not provide the
required support unless there is one at the bottom that is able to float
freely. But if something can possess this property, why not the earth
itself? In Kant's book the problem unfolds in two antinomies, corre-
sponding, respectively, to the relational category of cause-and-effect
and to the modal category of necessity. This ensures the correspon-
dence between the system of antinomies and the table of categories,
but it also allows Kant to deal separately with two different principles
adduced by metaphysicians as groundless grounds - uncaused causes
- of events in the world, namely, God and human freedom. The proof
of the thesis of the Third Antinomy is a cosmological argument for the
existence of freedom; the proof of the thesis of the fourth foreshadows
the cosmological proof of the existence of God. Note that, despite the
allegedly modal character of the Fourth Antinomy, the concept of cause
occurs prominently in the formulation and the proofs of its thesis and
antithesis.
We need not go further into the Fourth Antinomy, which does not
mean much for the philosophy of physics (and does not show Kant at
his best). On the other hand, there are two points in the discussion of
the Third that deserve our attention.
(i) The thesis is: "Causality in accordance with the laws of nature
is not the only one from which the phenomena of the world can all be
derived; to explain them it is necessary to assume also causality through
freedom" (1787, p. 472). In the Observation following its proof, Kant
countenances a conception of nature as woven from causal chains,
some of which begin with causeless acts of freedom.
If, for instance, I at this moment arise from my chair, in complete
freedom, without being necessarily determined thereto by the influence
of natural causes, a new series, with all its natural consequences in infini-
tum, has its absolute beginning in this event, although as regards time
this event is only the continuation of a preceding series. For this resolu-
tion and act of mine do not form part of the succession of purely natural
 3.5 The Ideas of Reason 143

effects, and are not a mere continuation of them. In respect of its hap-
pening, natural causes exercise over it no determining influence whatso-
ever. It does indeed follow upon them, but without arising out of them;
and accordingly, in respect of causality though not of time, must be enti-
tled an absolutely first beginning of a series of appearances.
(Kant 1787, p. 478)

This is not a conception that many scientifically minded philosophers

are prepared to endorse, and Kant himself emphatically rejects it - in
the Observation on the antithesis - because, if allowed, "the connec-
tion of phenomena determining one another with necessity according
to universal laws, which we call nature, and with it the criterion of
empirical truth, whereby experience is distinguished from dreaming,
would for the most part disappear" (1787, p. 479). And yet the exis-
tence of deterministic developments with free beginnings - surely the
most shocking feature of the said conception - is taken for granted in
the daily practice of laboratory physics. The experimenter sets up and
sets going, one assumes, of her own free will an almost closed physi-
cal system whose evolution is governed to a good approximation by
this or that system of differential equations, and lets it run until she
intervenes, again freely, to practice some measurement or to willfully
alter its course. The scientist's freedom is no less essential than the
system's determinism - or quasideterminism - to the epistemic purpose
of this exercise. As Hawking and Ellis pointedly note, "the whole of
our philosophy of science is based on the assumption that one is free
to perform any experiment" (1973, p. 189; quoted in context in §7.1).
(ii) The antithesis is: "There is no freedom, and everything in the
world happens solely in accordance with the laws of nature" (1787, p.
473). In the Observation following its proof, Kant assumes that a thor-
oughgoing causal concatenation of events is essential for the objective
articulation of experience. This claim, already adumbrated in his dis-
cussion of the Second and Third Analogies, is put forward again in
subsequent sections concerning the solution of this antinomy. In them
Kant describes the "principle of the thoroughgoing connection of all
events in the sensible world according to unchangeable natural laws"
(1787, p. 564) as "a law of the understanding from which no devia-
tion is allowed and no phenomenon can be exempted, on any pretext
whatsoever", for it is only by virtue of it that "phenomena can con-
stitute a nature and yield objects of experience" (1787, p. 570). We
must bear in mind that these strong statements refer to phenomena in
144 Kant

space and time. As we saw above, Kant's solution of the first two antin-
omies rests on the fact that phenomena are not fully determined:
Before, after, beyond, and within any known collection of phenomena
one must expect to find others, not yet known, preceding, succeeding,
surrounding, or articulating the former. Consequently, any definite
set of events must sport vacant slots for connection "according to
unchangeable natural laws" with still other events.42 No system of
lawful connections among given events can presently be thoroughgo-
ing. Thus, Kant's claim is purely programmatic: Faced with any phe-
nomenon, the scientist should try to ascertain all its causal slots and
to fill them by making out both the phenomena on which the given
phenomenon depends and those that depend on it. This is a task
without end, for every time a slot is filled new vacant ones are exposed
(in the phenomena adduced to fill the former). So what shall we make
of Kant's admonition that no phenomenon can be exempted "on any
pretext whatsoever" from the principle of thoroughgoing connection?
Evidently this is an exhortation not to yield in the quest for causal
links, no matter how difficult it is to find them. But what if a particu-
lar phenomenon resists all our efforts to connect it through and
through to the universal web of nature? Surely this will not, pace Kant,
bring about the breakdown of experience. For experience is not right
now - and never has been - a thoroughly connected system of phe-
nomena, but very much rather a thoroughly fragmented one, despite
our success in capturing vast segments of it in a few - different, not
always mutually compatible - conceptual nets. Thoroughgoing con-
nection may be - perhaps - the goal of experience, but, contrary to
Kant's suggestion, it certainly is not its prerequisite.
The Third Antinomy has brought us back to the theme with which
this chapter began. The thesis is made to stand for moral freedom, the
antithesis for universal determinism. Kant embraces the latter without
exception for phenomena in space and time, while insisting that free
initiatives can conceivably be attributed to things-in-themselves. I do
not know whether this solution is sufficient to rescue morality, but it
is surely worthless for the physicist who expects to be free to intervene

42
Just as the development of a tumor, observable with the naked eye, turns out to
depend on cellular changes, visible under the microscope, the latter, too, probably
depend in turn on processes still unknown occurring at the atomic and nuclear levels.
Our understanding of the cellular processes must leave an opening, so to speak, for
determination by such other processes, or it is bound to be wrong.
 3.5 The Ideas of Reason 145

in her experiments here and now, in the light of incoming results, and
not just in the timeless realm where her moral character is supposedly
chosen once and for all.

I cannot go into Kant's Metaphysical Principles of Natural Science

(1786), and his posthumous manuscript on The Transition from
the Metaphysical Principles of Natural Science to Physics (Ak. XXI -
XXII), lest this chapter should grow out of proportion. Friedman
(1992, Chapters 3 and 5) provides excellent guidance on both. I
shall only mention one significant addition that Kant made in 1786
to his teachings on space, and which is an apt and persuasive
application of his notion of Idea. Kant emphasizes that absolute space
"cannot be an object of experience, for space without matter is not an
object of perception, and yet it is a necessary concept of reason, and
so nothing more than a mere Idea" (Ak. IV, p. 558). He elaborates
further:

In order that motion be given, even if only as phenomenon, an empiri-

cal representation of space is required with respect to which the move-
able is to change its relation. However, the space, which must be
perceived, has to be material and so, pursuant to the concept of matter
in general, itself moveable. To think of it as moving one need only to
conceive it as contained in a broader space and to assume that the latter
is at rest. But the same consideration can be applied to such a broader
space, and so on and on, without ever attaining through experience a
motionless (immaterial) space with respect to which one could absolutely
attribute motion or rest to any matter [ . . . ] . Whence it is clear: First,
that all motion and rest is merely relative and that none can be absolute,
i.e. that matter can be conceived as being in motion or at rest merely in
relation to matter, never with respect to sheer space, so that absolute
motion, i.e. motion conceived without any reference of one matter to
another, is absolutely impossible. Secondly, that for this very reason a
concept of motion or rest in relative space which is valid for every phe-
nomenon is not possible, but one must conceive a space in which this
[relative space] can be thought to be in motion and which does not again
depend on another empirical space and so is not conditioned in turn;
i.e., an absolute space to which all relative motions can be referred, so
that everything empirical is moveable in it [ . . . ] . Absolute space is there-
fore necessary not as the concept of an actual object, but as an idea
which ought to serve as a rule for regarding every motion in it as purely
146 Kant

relative; and all motion and rest must be referred (reduziert) to absolute
space, if their phenomena are to be transformed into a definite concept
of experience (which unifies all phenomena).
(Kant 1786, in Ak. IV, pp. 558-60)
At first sight, this seems to contradict Kant's doctrine about the "formal
intuition" (1787, p. 160n.; quoted in §3.3), by which space "is repre-
sented as an infinite given magnitude" (1787, p. 39; Kant's italics). One
should, however, bear in mind that we are dealing here with kinemat-
ics, not geometry. According to Kant (1768; 1770, §15.C; 1783, §13),
in geometry the intuitive presence of absolute space shows up in the
distinction between incongruous counterparts (such as the two differ-
ently oriented screws mentioned in §3.2). But to make space available
as a frame of reference for motion there must be some way of effec-
tively identifying four of its points, not all on one plane, throughout
the duration of the motion. As a matter of fact, this can be done only
by marking those points on a rigid body that is assumed to be at rest.
Whence Kant's conclusion "that matter can be conceived as being in
motion or at rest merely in relation to matter, never with respect to
sheer space", and that absolute space, understood as the ultimate kine-
matic frame of reference, is just a regulative idea.
 CHAPTER FOUR

The Rich Nineteenth Century

In the North Atlantic countries that have been the main stage of our
story, the nineteenth century was a time of enormous production, not
only of manufactured goods but also of art and literature, science, and
philosophy. From the great wealth of new ideas in nineteenth-century
mathematical physics we can consider only a very small part, chosen
mainly for their impact on twentieth-century physics and philosophy.
I begin with the new geometries (§4.1), whose emergence is often
treated as a chapter in the history of mathematics and its philosophy
but which in fact attracted some of the mathematicians who developed
them - notably Riemann - chiefly for their potential significance for
physics (which Einstein subsequently made good in a wholly unex-
pected way). The next two sections deal with the concept offield,espe-
cially in electrodynamics (§4.2), and with the introduction of chance
into thermal physics (§4.3). Finally, we shall take a glance at some
nineteenth-century philosophies, which set the tone of twentieth-
century debates (§4.4)

4.1 Geometries

4.1.1 Euclid's Fifth Postulate and Lobachevskian Geometry

I can still recall my frustration when, early in my first term of high
school geometry, the teacher "proved" - such was his word - Euclid's
Theorem 1.29 "by parallel transport", in effect by sliding a wooden
square along a steel ruler pressed on the blackboard. I had been fasci-
nated by his hortative talk about geometry's breadthless lines and the
power of deductive proof and saw the tottering advance of the square
from one precarious chalk band to another as one more example of

147
148 The Rich Nineteenth Century

Figure 10

the seemingly unlimited capacity of adults to fail their promises. I

learned later that Euclid inferred that theorem from his Postulate V, an
unproven assumption that can be paraphrased as follows (see Fig. 10):

Any two coplanar straight lines which are intersected by a transversal

straight line meet on that side of the transversal on which the internal
angles they form with the transversal add up to less than two right
angles.1

'Postulate' is the standard translation of aitema, which literally means

a 'demand' or 'request' that the audience must grant before geometry
can get going. Of course, a request is anything but self-evident,2 so
many mathematicians sought to prove Postulate V. Some succeeded in

1
On each side of a straight line lies a region of the plane, which has no points in
common with the region on the other side. The internal angles formed by the trans-
versal with one of the lines are the two angles that lie on the same side of this line
as the intersection of the transversal with the remaining line. On each side of the
transversal there are two internal angles. Obviously, the pair on one side adds up to
less than two right angles if and only if the pair on the other side adds up to more
than two right angles. If each pair adds up to two right angles, then, by Euclid's Pos-
tulate V, the straight lines do not meet, i.e., they are parallel lines. Thus, Postulate V
entails (and is also entailed by) the following proposition, known as Playfair's Axiom:
Given a straight line X and a point P not on X, there is on the plane (A,,P) one and
only one straight line through P that is parallel to X.
2
This is nicely illustrated by Euclid's Postulate III, "To draw a circle with any center
and any radius", which is not only not self-evident but also downright impossible if,
as most contemporaries of Euclid believed, the world is wholly contained within a
finite surface.
 4.1 Geometries 149

deriving it from other, equally strong propositions, which were also in

need of proof. Saccheri (1733), who tried to prove it by reductio, that
is, by inferring a contradiction from its negation, obtained from the
latter a surfeit of surprising but not inconsistent theorems, but finally
stopped their flow and forced a contradiction by sleight of hand. Sac-
cheri's theorems were later independently rediscovered by Gauss,
Lobachevsky, and Bolyai, who, unbeknownst to each other, treated
them as propositions of a new geometry. Given that Lobachevsky was
the first to publish, the system of geometry that negates Postulate V
but agrees with Euclid on every question which does not depend on
that postulate is properly called Lobachevskian geometry?
The key difference between Euclidian and Lobachevskian geometry
is that in the latter two figures can have the same shape only if they
are equal in size. To be specific, consider two convex polygons AXA2
. . . An and BiB 2 . . . Bm and designate the angle at A, by a, and the
angle at B{ by (3,; (1 < / < n). If a, = p, for each index /, then both poly-
gons are similar and there is a constant factor k such that AnAx = kBnBi
and, for every / < w, A,A/+i = kBtBt+1. In Euclidian geometry k can be
any real number, but in Lobachevskian geometry similarity can hold
only if k = 1, that is, if the polygons are congruent. This implies that
on a Lobachevskian plane there are no rectangles, and of course no
squares; nor are there any cubes or rectangular parallellepipeds in
Lobachevskian 3-space. The sum a of the internal angles of a
Lobachevskian triangle is always less than two right angles: c = n - 8,
the defect 8 being proportional to the area of the triangle. Since 0 < a,
8 < n and the area of a Lobachevskian triangle has an upper bound
equal to n times the constant ratio between the area and the defect.
Consider a straight line m and a point P outside it. P and m deter-
mine a plane (m,P). If (ra,P) is Lobachevskian, there is more than one
straight line on it that goes through P and never meets m. Let h be the
distance from P to m. Let the perpendicular from P meet m at M. By

3
Gauss, born in 1777, was already convinced by 1799 that Postulate V could never
be proved from more perspicuous premises; c. 1813 he began work on what he ini-
tially called "anti-Euclidian geometry", but published nothing, for fear of "the outcry
of Boeotians", as he later said. Lobachevsky, born in 1793, announced his findings
in French in 1826 in a public lecture at the University of Kazan and first published
them in 1829-30, in Russian, in four installments carried by a local journal. Bolyai,
born in 1802, cryptically alluded to his wonderful discoveries in a letter to his father
of 23 November 1823, but he published them only in 1832, in a 26-page appendix
to his father's Tentamen in elementa matheseos.
150 The Rich Nineteenth Century

definition, PM = h. Let n be a straight line through P on the plane

(ra,P). Let an be the smallest angle that n forms with PM. There is an
angle Tl(h), - the angle of parallellism for h - such that m meets n if
and only if an < H(h). If an = H(h), we say that n is parallel to ra.4 In
Euclidian geometry, Tl(h) is a right angle, no matter what the value of
h. In Lobachevskian geometry, Tl(h) is an acute angle, which decreases
as h increases.5 In each instance of Lobachevskian space there is a def-
inite distance K, such that II(K) = n/4 (half a right angle). If PM = K,
the two straight lines through P that make with PM an angle equal to
7i/4 are parallel to m and perpendicular to one another. The constant
K can be taken as an absolute unit of length, characteristic of the
instance in question.
There is a legend that Gauss tried to test the physical truth of
Lobachevskian geometry by measuring the defect of a large triangle
formed by the tops of three mountains in Germany.6 However, as early
as 1819 Gauss had told a correspondent that "in the light of our astro-
nomical experience, the constant [K] must be enormously larger than
the radius of the earth" (WW VIII, 182). To someone acquainted with
this fact, the legendary attempt should have seemed preposterous. On
the other hand, Lobachevsky did try to evaluate the constant K by mea-
suring the defect of the triangle formed by three well-known stars.
According to his calculations, the defect amounted to 3.7-millionths of
a second of arc, well within the range of observational error. So
Lobachevsky concluded that "all lines subject to our measurements,
even the distances between heavenly bodies, are too small in compari-
son with the line which plays the role of a unit in our theory, so that
the usual equations of plane trigonometry must still be viewed as
correct, having no noticeable error" (ZGA, I, 22).
More interestingly perhaps, Lobachevsky contemplated the use of

4
Under this definition of parallellism, which was independently adopted by Gauss,
Lobachevsky, and Bolyai, there are at most two parallels to a given straight line
through a given point outside it. It differs from the standard definition: The straight
lines m and n are parallel to each other if they lie on the same plane and do not meet.
(See the next note.)
5
Continuing with the preceding note, we can now see that on a Lobachevskian plane
there are, under the new definition of parallellism, exactly two parallels to a given
straight line through a given point outside it but infinitely many straight lines through
that same point that never meet that line.
6
Cf. Torretti (1978, p. 381, n. 40), for a brief comment on the origin of this legend
and references to A. I. Miller, who first exposed it.
 4.1 Geometries 151

two or more incompatible geometries in physics. He maintains that in

nature we only know movement, "without which sense impressions are
impossible"; and that "all other concepts, e.g. geometrical concepts,
are generated artificially by our understanding, which derives them
from the properties of movement" (ZGA, I, 76). "In nature there are
neither straight nor curved lines, neither plane nor curved surfaces,
[but] only bodies; so that all the rest is created by our imagination and
exists solely in the realm of theory" (p. 82). Since our geometry con-
structs its concept of space from an experience of bodily motion due
to physical forces, we might well make allowance for more than one
geometry, corresponding to different kinds of natural forces.

To explain this idea, we assume that [. . .] attractive forces decrease

because their effect is diffused upon a spherical surface. In ordinary
geometry the area of a spherical surface of radius r is equal to 47ir2, so
that the force must be inversely proportional to the square of the dis-
tance. I have found that in imaginary [i.e., Lobachevskian - R. X] geom-
etry the surface of a sphere is equal to n(er — e~r); such a geometry could
possibly govern molecular forces, whose variations would then entirely
depend on the very large number e.
(Lobachevsky, ZGA, I, 76)

Lobachevsky also brought up the question of the consistency of the

new geometry, which Gauss and Bolyai apparently took for granted.
He derived the fundamental equations of trigonometry, which, as in
the Euclidian case, suffice to determine all metric relations in his
system. He noted that any contradiction that might eventually emerge
in a theorem of the new geometry must therefore be implicit in the said
equations. But "these equations become [the familiar] equations of
spherical trigonometry as soon as we substitute ^API, fcV^i, W^i for
sides a, b, c" (ZGA, I, 65; recall Lambert's suggestion cited in Chapter
Three, n. 17). Consequently, if any contradiction can be derived in
Lobachevskian geometry, a matching contradiction must occur in stan-
dard geometry. This is a proof of relative consistency: Lobachevskian
geometry is consistent unless ordinary spherical trigonometry is incon-
sistent. The proof rests on the purely formal agreement between two
sets of equations, so it does not require, like the better known proofs
of Beltrami and Poincare, a contrived interpretation of curved surfaces
and lines in one system as planes and straights of the other. Moreover,
it proves the consistency of Lobachevskian geometry, including the
negation of Postulate V, relative to a part of Euclidian geometry that
152 The Rich Nineteenth Century

does not depend on Postulate V. Thus, any contradiction that might

emerge in the new geometry will not stem from the negation of Pos-
tulate V but must be contained already in the assumptions shared by
both systems.

4.1.2 The Proliferation of Geometries and

Klein's Erlangen Program
Lobachevskian geometry did not become a subject of public debate
until the 1860s, when Gauss's favorable view of it became known
through the publication of his correspondence with Schumacher. On
the other hand, a much more radical revision and extension of stan-
dard geometry was being actively pursued by mathematicians since the
publication of Poncelet's Traite des proprietes projectives des figures
(1822). The new geometry, known as protective geometry, enriches
each Euclidian straight line m with an ideal point at infinity where m
meets its parallels (i.e., every straight coplanar with m that does not
meet m at an ordinary point). By this seemingly innocuous trick many
proofs are made easier and many theorems simpler. But, while the
relations between neighboring ordinary points remain unaltered, the
addition of the points at infinity completely disrupts the global neigh-
borhood structure of Euclidian space. Every neighborhood of the point
at infinity on a straight m contains ordinary points from both extremes
of ra; thus, the projective straight has the neighborhood structure of a
Euclidian circle! Readers who are acquainted with the rudiments of
topology will appreciate the deep difference between Euclidian and
projective space if I say that the latter is compact while the former is
not.7
The success of projective geometry led to a multiplication of geo-
metric systems. Felix Klein sought to make sense of their diversity from
a single unifying point of view in the booklet, generally known as the
Erlangen Program, that he published as he joined the faculty at the
University of Erlangen (1872). Klein's proposal turns on the concept
of a group of transformations, 'Group' is taken here in the sense famil-
iar to students of algebra (and to all those who still remember Rubik's
cube). As this fairly simple notion is central to much of recent physics,
the reader who is not acquainted with it should make a resolute effort
to understand the following explanation. Consider an arbitrary set G

7
Some elementary information about topological spaces is given in Supplement III. 1.
 4.1 Geometries 153

(its elements can be numbers, or rotations of a cube, or anything else).

Assume that there is a "multiplication table" that assigns a definite
element of G to each pair of elements of G. (The element assigned to
the pair (a,b) is called the product of a and b and is denoted by ab;
note that ab need not be the same as ba.) Assume that "multiplication"
is associative, in other words, that for any three elements a, b, and c
of G, a(bc) = (ab)c. Assume that there is an element e of G such that,
for every element a of G, ae = ea - a. We call e the zero or neutral
element of G. Assume, finally, that for every element a of G there is
an element a* of G such that aa* = a*a = e. a* is called the inverse of
a. If these four assumptions are fulfilled, the set G possesses the struc-
tural properties that characterize it as a group. Let me give two simple
examples:
(i) A permutation is a one-one mapping of a set onto itself. There
are six possible permutations of the set {1,2,3}, formed by the first three
positive integers. For reference, I shall designate them with the first six
letters of the alphabet. I write beneath each integer the integer assigned
to it by each permutation:
\ , (123^ (\2?>\

'H312J H j KJ
The product xy of permutations x and y is the permutation resulting
from carrying out first permutation y, and then permutation x. Thus,
ab(l) = a(b(l)) = a(3) = 2; ab(2) = 3 and ab(3) = 1, so that ab = f. The
reader should compute the rest of the multiplication table and verify
that it defines a group, that the neutral element is e, and that the
inverses of the other five elements are, respectively, a* = a, b* = b, c*
= c, d* = /i and f* = d. Note that the three permutations e, d, and f
form a group by themselves, a subgroup of the larger group, formed
by a, by c9 d, e, and f.
(ii) Take the set Z of all integers and view the sum of any two of
them as their product (in the sense introduced above). Addition of inte-
gers is, of course, associative. For any integer a, a + 0 = 0 + a = a.
Finally, for every integer a there is an integer -<a, such that a + (-a) =
(-a) + a = 0. So (Z,+,0) is a group.
A group's structure is fully specified by its multiplication table,
regardless of the particular nature of its elements. For example, the
group of permutations of {1,2,3} is structurally identical with the group
154 The Rich Nineteenth Century

formed by the following operations on an equilateral triangle with ver-

tices A, B, and C: flippings about the triangle's three heights and clock-
wise rotations about its center of gravity by 120°, 240° and 360°. To
see this, denote the three rotations by d, /", and e, respectively; by a the
flipping about the height through A, and so on. Evidently, the three
rotations constitute a subgroup, with the multiplication table
d e f
d f d e
e d e f
f e f d
To explain what Klein means by a group of transformations, I note
that 'transformation' is just another word for permutation, which
mathematicians prefer to use when the set that is being mapped onto
itself is what they call a space - that is, a structured set at least remotely
resembling Euclidian space. If (p and \\f are two transformations of a
space X, their product cp\|/ is the transformation of X that results from
carrying out \|/ first, followed by (p. This operation is clearly associa-
tive. The identity mapping I2: *•-»#, which maps each element x of X
to itself, is of course a transformation. Moreover, for every transfor-
mation cp of X there is a transformation cp"1, such that (p^cp = qxp"1 = Js
((p"1 - the inverse of cp - assigns to each element x of X the element to
which x was assigned by cp).8 Consider now the set of all transforma-
tions of X that satisfy some specific condition. This set constitutes a
group of transformations of X if and only if it contains the identity
mapping 7Z, the inverse of every member of the set and the product
of any two members of the set. Take for example the isometries of
Euclidian space, that is, the transformations that map every segment
onto a segment of the same length. This means that, if cp is an isome-
try and p and q are any two points in space, the distance 8(p,q) between
these points equals 8((p(/?),cp(g)). Obviously, the identity mapping is an
isometry, and so is the inverse of any isometry. And, of course, if cp and
\\f are isometries, 8((p\|/(/?),(p\|/(g)) = 8(\|/(p),\|/(g)) = 5(p,g), so that <p\|/ is
an isometry too. Thus the isometries form a group.
The last example also illustrates the crucial concept of invariance.
Let R be an w-ary relation on X (in other words, 'R' is a predicate that

!
We have that, for each x in X, there is one and only one y in Z such that x —
Thus, y = (fHx).
 4.1 Geometries 155

can be meaningfully applied to lists of n objects in X).9 We say that a

transformation cp of Z preserves R and that R is invariant under cp if R
is true of the list ((p(xi),... , (p(*«)) whenever it is true of (xu . • • ? #«).
If R is invariant under every transformation of a group G of transfor-
mations of X, we say that G preserves R and that R is G-invariant.
Thus the distance between point pairs is preserved by the group of
isometries, and so are the size of angles, the area of surfaces, and the
volume of spatial figures.
Klein perceived that the relations studied by different geometries
were in effect the invariants of different transformation groups and
proposed the following formulation for the most general problem of
geometry:
Let there be given a manifold and a group of transformations in it. To
investigate the configurations belonging to the manifold with respect to
such properties as remain invariant under the transformations of the
group.
(Klein 1893, p. 67)10
Klein adds that "this is the universal problem which spans not only
ordinary geometry but also, in particular, the new geometric methods
to be mentioned later and the different treatments of manifolds with
arbitrarily many dimensions". Since the transformation group adjoined
to the manifold can be arbitrarily chosen, every way of dealing with
the general problem - that is, each of the new forms of geometry men-
tioned in Klein's booklet, and many others still awaiting development
- is equally justified (Ibid.).

To avoid verbal complications I speak of 'objects in £', meaning not just the points
of 2 but also the sets, sets of sets, and so on, that can be constructed from those
points. A mapping cp: Z —> X, that assigns points to points, induces mappings on the
domains formed by such sets, sets of sets, etc. These mappings are customarily des-
ignated by the same name (p. Thus, if AABC is the triangle with vertices A, B, and C,
cp(AABC) is the triangle with vertices cp(A), <p(B), and <p(C).
The word 'manifold' translates the German 'Mannigfaltigkeit', which was used by
Klein's contemporary Georg Cantor to designate what he later called 'Menge', i.e., a
wholly arbitrary, unstructured set. However, the word was also used in a narrower
sense, for structured sets of real or complex numbers or of lists (ordered w-tuples) of
such numbers, with the standard topologies inherited from the real number field U
or the complex number field C (for more on algebraic fields, see Supplement 1.2).
Klein's sense is probably somewhat wider than this narrower sense but essentially
akin to it. On this point, cf. Torretti (1978, pp. 137f.). For the current use of 'man-
ifold' in mathematics, see Supplement III.
156 The Rich Nineteenth Century

This last remark was probably motivated by an important discov-

ery concerning Lobachevskian and Euclidian geometry that Klein pub-
lished at about the same time as the Erlangen Program and that, he
thought, resolved the disputed question as to "the true geometry"
(Klein 1871, 1873, 1874; cf. Klein 1890). I cannot give here an accu-
rate description of it, but the following rough indications should be
sufficient for our purposes.11 To make things easier I consider only the
two-dimensional case. By a collineation I mean a transformation of the
projective plane that maps each triple of collinear points onto a triple
of collinear points. Collineations form a group, and plane projective
geometry studies precisely the invariants of this group. One such invari-
ant is the function known as the cross-ratio, which assigns a real
number to every list of four collinear points. An interesting family of
figures in projective - as in ordinary - geometry are the curves called
conies. A conic on the projective plane can be defined analytically - as
in ordinary geometry - as the locus of points whose coordinates satisfy
some definite quadratic equation or geometrically by its behavior under
certain transformations. Following an idea of Cayley, Klein considered
the cross-ratio of point quadruples (P1,P2,P3,P4) such that P3 and P4 lie
on a given conic £. Since P3 and P4 must be the points where the straight
through Pi and P2 meets £, the said cross-ratio may be regarded as
depending only on Vx and P2, that is, as a function of point pairs. The
collineations that map a given conic onto itself form a group, and
the said function is clearly an invariant of this group. Let d^ denote the
principal value of the natural logarithm of this function. Klein showed
that the restriction of d^ to a well-chosen region 91^ of the projective
plane - with 2ft^ bounded by or in some other way dependent on £ -
behaved like an ordinary distance function. Depending on the nature
of the conic £, the structure ffi^dt) - that is, the region 2/^, endowed
with the distance function d^ - satisfies all theorems in the plane
geometries of Euclid or Lobachevsky or in a third kind of geometry
discovered by Klein, which he termed 'elliptic' (his names for the other
two systems were 'parabolic' and 'hyperbolic', respectively). Thus,
depending on whether £ belongs to one or the other of three types of
conic, the group of collineations that map £ onto itself is structurally
identical with one of the three groups of Lobachevskian, Euclidian,
or elliptic isometries. Similar results hold for the three-dimensional

11
For a passable sketch, see Torretti (1978, pp. 125-32) (pp. 110-25 contain prelimi-
nary explanations). For a detailed modern exposition, see Redei (1968).
 4.1 Geometries 157

case, with £ a quadric surface instead of a conic. In the light of them

one sees better why Klein thought that the different geometries had
"equal rights" or "equal justification" (gleiche Berechtigung - 1893,
p. 67).

4.1.3 Riemann on the Foundations of Geometry

In a lecture delivered to the Faculty of Philosophy at Gottingen in 1854,
Bernhard Riemann put forward a general view of geometry that is
broader and deeper than Klein's.12 He noted that geometry has hith-
erto taken for granted the notion of space and the basic concepts for
constructions in space. These assumptions of geometry are spelled out
in nominal definitions and axioms that throw no light on their mutual
relations, so we cannot see whether and to what extent their combi-
nation is necessary or even possible. This is due to the fact that "the
general concept of multiply extended magnitudes", of which space is
an instance, has not been worked out. He proposes therefore to con-
struct such a concept "from general concepts of magnitude":
It will ensue that a multiply extended quantity admits diverse metric rela-
tions, and that space is therefore only a special case of a triply extended
quantity. A necessary consequence of this is that the theorems of geom-
etry cannot be inferred from general concepts of quantity, but that
those properties which distinguish space from other conceivable triply
extended quantities can only be obtained from experience. Thus arises
the task of inquiring after the simplest facts from which the metric rela-
tions of space can be specified; a task which by its very nature is not
completely determined, for several systems of simple facts can be pro-
posed which suffice to determine the metric relations of space - the most
important one for our present purpose being that laid down by Euclid.
Like all facts, these facts are not necessary but have only empirical cer-
tainty: they are hypotheses. One can therefore study their probability -
which within the bounds of observation is anyway very large - and there-
upon judge the admissibility of extending them beyond the bounds of
observation, both in the direction of the immeasurably large as in that
of the immeasurably small.
(Riemann 1867, pp. 133-34)

12
Although Riemann's lecture, "On the Hypotheses that lie at the Foundation of Geom-
etry" preceded Klein's program by 18 years, it was not printed until 1867 (after
Riemann's death). This may help to explain why Klein paid so little attention to
Riemann's conception and came up with a program that fails to cover it.
158 The Rich Nineteenth Century

An n-iold extended magnitude (n = 1, 2, 3 , . . .) as conceived by

Riemann is substantially the same as what we now call a real n-
dimensional smooth manifold.13 For brevity, I shall say '^-manifold'.
Since contemporary cosmology and celestial mechanics conceive the
world as a 4-manifold of uncertain shape, it is important to get the gist
of Riemann's idea. I shall try to explain it as intuitively as I can. Then
I shall consider his proposals concerning metric relations on an n-
manifold.
Any smooth surface, such as a sphere, a Mobius strip,14 or the
surface of a sculpture by Henry Moore, illustrates the concept of a 2-
manifold. Note that in these three examples a neighborhood of each
point can be mapped continuously one-to-one onto a flat piece of paper
- say, a page of an atlas - but no such mapping is possible for the
surface as a whole. Analogously, a neighborhood of each point can be
mapped continuously one-to-one on an open set of R2 (the set of all
ordered pairs of real numbers, with the standard topology generated
from open rectangles).15 Such a mapping is called a coordinate system
or chart. A collection of charts of parts of a given surface if constitutes
an atlas for if if (i) each point of if lies in the domain of at least one
chart, and (ii) given any two charts g and h in the collection, the com-
posite mappings g°h~l and hog'1 are differentiable wherever they are
defined. The mappings g°h~l and hog'1 are coordinate transforma-

13
Riemann's "w-fold extended magnitude" is the direct source of our notion of a
smooth, or differentiable, manifold. We make a distinction between real and complex
manifolds, depending on the nature of the coordinates employed for identifying their
points, and between manifolds of finite or infinite dimension, depending on the
number of coordinates assigned to each point. Riemann contemplated manifolds of
infinite dimension (1867, last two sentences of 1.3) but went on to discuss only n-
manifolds with real-valued coordinates. Still, the subsequent introduction of complex
manifolds was true to his spirit.
14
To form a Mobius strip, take a rectangular strip of paper, twist it, and paste together
the two short sides so that the upper left-hand corner coincides with the lower right-
hand corner and the lower left-hand corner coincides with the upper right-hand
corner. Note that the resulting surface can be said to have only one side, in the fol-
lowing intuitive sense: Any point on the surface that can be reached from another by
piercing the surface can also be reached by sliding a pencil over the surface, without
ever touching its edges.
15
'Open set' and 'topology generated from .. .' are defined in Supplement III. 1. An open
rectangle in (R2 is a set of ordered pairs (x,y), such that x and y are real numbers sat-
isfying the inequalities a < x < b, c < y < d, for any given a, b, c, and d with —°o < a,
c < oo and —oo < b, d < <».
 4.1 Geometries 159

tions; they map number pairs to number pairs, so it makes good sense
to say that they are differentiate.16 The surface if endowed with an
atlas si is a 2-manifold. An ^-manifold is defined similarly, substitut-
ing Un for R2 in the definition of 'chart'. If $Ri,<s£i) is an w-manifold,
(TI2M2) is an ra-manifold, and f is a mapping of HRi into 3M2, we say
that f is differentiable at a point p of Wlx if there is a chart b defined
at p and a chart g defined at f(p) such that go fob'1 is differentiable at
h(p). fis differentiable tout court if it is differentiable at every point of
SRi. As one readily sees, every w-manifold is like Rn locally, on a neigh-
borhood of each point, but it can differ widely from Un globally.
Atlases provide the means of subjecting manifolds to the power of
mathematical analysis while also keeping track of their variegated
overall shapes.17
I shall now define the tangent space at a point of an w-manifold.
This concept does not occur in Riemann's lecture. Yet, combined with
two more ideas that I shall mention in the next paragraph (which were
even further from his thought), it has turned out to be indispensable
for understanding his proposal on metric relations as well as the math-
ematical treatment of physical fields (§4.2). Let us concentrate once
more on 2-manifolds. It is intuitively obvious that, if p is any point of
our smooth surface if, there is a plane tangent to if at p. We seek a
way of extending this structure to less tangible manifolds. Consider a
smooth curve on if. We think of it as being drawn by a point that
moves over if during a period of time represented by an open - finite
or infinite - real interval J5. So we identify our curve with a mapping
y: $ —> if, which assigns to each instant t in $ the position y(t) of the
moving point at that instant. The collection of these positions - the
range of y - we call a patb; the variable t that, so to speak, regulates
the deployment of y along its path is known as the curve's parameter.
To simplify my exposition I shall hereafter assume that all curves are
injective, that is, that each point of the path corresponds to one and
only one value of the parameter. A given path can be the range of many

16
By 'differentiable' I mean differentiable to every order.
17
I ought perhaps to mention that any atlas si for a manifold W determines a unique
maximal atlas «s4max- It is the collection of every conceivable chart x such that, for
every y in si, the coordinate transformations x o y'1 and y o x~x are differentiable wher-
ever they are defined. simzx obviously contains si. The topology generated from the
domains of the charts in sim3x is the manifold topology of M (see Supplement III.2).
Henceforth every manifold mentioned is endowed with its manifold topology, unless
otherwise stated.
160 The Rich Nineteenth Century

curves, defined on different intervals and drawn at different paces; such

curves are said to be reparametrizations of each other. If we regard if
as a part of ordinary space, with its familiar metric relations, and we
assume that the point's motion is smooth, with no brusque changes of
speed or direction, we can assign to the curve y, at a given instant £0?
a definite velocity, which can be represented by a vector tangent to if
at the corresponding point p = y{t0). The vectors representing the veloc-
ity at p of each curve through p generate a two-dimensional vector
space18 that is tangent to if at p. This is the prototype for the concept
of tangent space that I shall define. However, since ^-manifolds are not
endowed from the outset with metric relations, I must proceed more
deviously. The remainder of this paragraph will probably be too tough
for some readers and yet, I hope, quite useful for others. Consider the
collection 2F(/?) of all differentiable real-valued functions defined on
some neighborhood of point p. With the ordinary operations of func-
tion addition and multiplication by a constant, ^(p) has the structure
of a vector space.19 Each function cp in ?F(p) varies with £, along the
path of y, in some neighborhood of p. Its rate of variation at p = y(t0)
is properly expressed by the derivative d(q)o y)/d£ at t = t0. As (p ranges
over 3F(p), the value of d(($oy)/dt at t = to is apt to vary in R. So we
have here a mapping of $F(p) into R, which I denote by yp (also by y(u),
if p = y(u)). It is in fact a linear function20 and therefore a vector in the
dual space 2P(p) of real-valued linear functions on 2F(/?).21 The vectors

18
Supplement I provides elementary information about vector spaces.
19
If cp and V|/ are any two functions in $F(p), their sum ((p + \j/) assigns the value
\\f(q) to each point q in the intersection of their respective domains. If (p is in
and a is any real number, their product tfcp assigns the value ay(q) to any q in the
domain of (p.
20
Linear functions on vector spaces are defined in Supplement 1.5. It is clear that, for
any real numbers a and b, and any two functions (p and \\f in $F(p)9 %(a<Q + by) =
a%(<$>) + b%(\\f).
21
The foregoing explanation throws light on a piece of notation that the reader may
have seen and which I shall use in the sequel. Consider a chart x defined on a neigh-
borhood Ux of our point p that assigns to each point q in Ux the coordinates x1(q)
and x2(q). There is a path through p along which x1 changes while the other coordi-
nate remains fixed. The curve that assigns each point on that path to the successive
values of coordinate x1 is the parametric curve of that coordinate through p. Con-
sider its tangent vector at p. According to our definition, it is a linear mapping that
assigns to each function cp in &>(p) its rate of variation at p as x1 changes while x2
remains fixed. In other words, it assigns to cp the partial derivative dq/dx1^. It is there-
fore customary to denote the tangent vector to the parametric curve of coordinate x1
 4.1 Geometries 161

yp corresponding to all the curves y whose paths contain the point p

span a two-dimensional subspace of <F*(p). This subspace is, by defi-
nition, the tangent space of if at p. The tangent space TqWl at a point
q of an w-manifold 3K is defined in exactly the same way, substituting
n for 2 in all occurrences of the dimension number.
Two further complications that Riemann presumably never had in
mind have contributed in the twentieth century to the final clarifica-
tion of his philosophy of geometry. First, the tangent spaces at all points
of an n-manifold Wl can be bundled together into a single 2^-manifold
T3W in a wholly natural way. The projection mapping n: TM —> Wl
assigns to each tangent vector v in TSft the point n(v) at which it is
tangent to Wl. The structure (T3W,3K,7i) is the tangent bundle over Wl.22
Second, any vector space V is automatically associated with other
vector spaces, such as the dual space T * of linear functions on T, the
spaces of so-called covariant multilinear functions on T, "contravari-
ant" multilinear functions on T*, and "mixed" multilinear functions
on both.23 This holds, of course, for each tangent space of a manifold
3K. Moreover, there is a natural way of bundling together into a k-
manifold (for some suitable integer k) all the vector spaces of a defi-
nite type associated with the tangent spaces of Wl (e.g., all the spaces
of bilinear functions on TpWft for every p in 9K).
Armed with such post-Riemannian notions we can now turn to
Riemann's revolutionary views on metric relations in physical space.
He observes that concepts of magnitude or quantity (Grossenbegriffe)
can only be applied to a general concept that admits different modes.
Depending on whether the transition from one mode to another is con-
tinuous or not, we have a continuous or a discrete variety. In the latter

at p by dldx\. Note, by the way, that d/dx\ and d/dx2\p span the tangent space at p.
The mapping p •-> d/dxk\p is a differentiable mapping of Ux into the tangent bundle
TUX (see the next paragraph and note 22). The vector it assigns to each point p
belongs precisely to the tangent space at p. A mapping of this sort, defined on a man-
ifold Ti, constitutes what is known as a vector field on Wl. These notions and nota-
tions are readily extended, mutatis mutandis, to charts in ^-manifolds.
22
The natural manifold structure of TWl can be readily seen if we recall that each
tangent space is an n-dimensional real vector space and therefore isomorphic to the
vector space U". So one can pick, for each p in W, an isomorphism hp: Tp$Jl —> IR".
Given a chart g of Wl defined on a suitable set U of M we can define a chart h on
n~l(U) c TWl as follows: For each v in iCl(U), h(v) = (g(nv),hnv(v)). Clearly, h(v) is a
list of 2M real numbers.
23
Multilinear functions are defined in Supplement 1.5.
162 The Rich Nineteenth Century

case, the size of different parts can be compared by counting the modes
that they contain. But in the continuous case this can only be done by
measurement. As described by Riemann, measurement is effected by
superposition of the quantities to be compared and therefore "requires
a means of transporting one quantity to be used as standard for the
other" (1867, p. 135). In the case of physical space, the behavior of
standards of measurement under transport depends of course on the
forces of nature and can only be learned by experience. At the human
scale, metric relations in space agree well with the theorems of Euclid-
ian geometry. However, since all measurements are approximate, such
agreement does not warrant the validity of Euclidian geometry in the
very large and in the very small. Indeed, "the empirical concepts
on which the metric determinations of space are grounded, viz., the
concept of a solid body and that of a light ray, would seem to fail in
the infinitely small; so it is quite conceivable that the metric relations
of space in the infinitely small do not satisfy the hypotheses of geom-
etry, and in fact one should assume this if one can thereby explain the
phenomena in a simpler way" (p. 149). To decide such matters one
ought to start from the well-corroborated conception of phenomena
laid down by Newton and gradually rework it under the pressure of
facts that cannot be explained by it. However, to ensure "that this work
is not hindered by the narrowness of concepts and that the progress of
knowledge of the connection of things is not obstructed by traditional
prejudices", a purely mathematical investigation of the conceivable
alternatives will be helpful (p. 150). With this purpose in mind,
Riemann undertakes such an investigation in Part II of his lecture.
The main obstacle obstructing the progress of geometry in
Riemann's intellectual environment was the prejudice that spatial mea-
surements must be carried out with rigid bodies (cf. Ueberweg 1851;
Helmholtz 1866, 1868).24 This is possible only in a space in which
every figure can be moved about undeformed. As we shall see, Riemann
spelled out the condition under which this requirement is met and
described a spectrum of geometries that meet it, neatly defining their
place in his own scheme of things. But he took a much broader view
of geometry. He agreed that "metric determinations require that mag-

24
For a short report in English about these works, see Torretti (1978, pp. 260-64,
155-71). The said prejudice motivated the rejection of Riemannian geometry by Poin-
care (1891, p. 773) and Russell (1897, §§23, 143-57; see his mea culpa in Russell
1959, pp. 39f.).
 4.1 Geometries 163

nitudes be independent of place," but pointed out that "this can

happen in more than one way" (1867, p. 138). The one he chooses for
further elaboration in his lecture rests on the assumption that the length
of lines is independent of the way they lie in space, so that every line
is measurable by every other. Under this assumption, spatial measure-
ments require inextensible cords rather than rigid rods. Riemann
understood that the tailor's tape is more versatile than the clothier's
yardstick. If the said assumption holds for an /t-manifold 9W, the length
of a path in Tt must be an intrinsic property of the path, belonging to
it as a 1-manifold embedded in 3W, regardless of its relation to the
points outside of it.
Traditionally, the length of an arbitrary line is figured out by inscrib-
ing in it a series of polygonal lines of ever shorter and more numerous
sides and calculating the limit to which the series of their total lengths
converges as the number of sides grows beyond all bounds. To be spe-
cific, if y is a smooth curve in Euclidian space % - that is, if y is a dif-
ferentiable mapping of a real interval (a9b) into % - and x = (x1,*2,*3)
is a Cartesian coordinate system defined on %, the length of y^ path is
given by the integral

f | ^ | (4.2)
Here the integrand is, of course, the limit to which the length of a chord
drawn from y(u) to a neighboring point y(u + h) converges as h goes to
0. Thus conceived, the length of a contorted line does not stand on its
own feet, but depends on the availability of straight-sided polygons.
However, the concept of a tangent space explained above furnishes the
means for overcoming this limitation. Let the curve y be a differentiable
mapping of the real interval (a,b) into an w-manifold Wl. Suppose that,
for each p in 3K, there is a real-valued mapping on the tangent space
TP3K that assigns a length \ip(v) to each vector v in Tp-Wl. Suppose, more-
over, that the correspondence p «-» \xp is a differentiable mapping of W
into some suitable bundle over 3K. One can then take the length
V^i{u){y(u)) of the vector y(«), tangent to the path of y at point y(u) in 9W,
as an index of the curve's advance as it goes through that point. The
full length of y*s path can then be defined as the value of the integral

(4.3)

As required above, this definition is intrinsic to the path of y because

(i) it can be shown that the length thus defined is the same for all repa-
164 The Rich Nineteenth Century

rametrizations of y? and (ii) for each u e (a,b), the vector y(u) spans the
one-dimensional tangent space at y(u) of the one-manifold constituted
by the path itself, so the integral does not in any sense depend on the
way this 1-manifold lies in Wl.
A mapping that assigns lengths to the vectors of a vector space V is
called a norm in T. One naturally expects that
Nl. the length of the 0-vector is 0;
N2. the length of any other vector is a positive real number;
N3. if two vectors add up to the 0-vector, their lengths are equal; and
N4. the length of the sum of any two vectors is equal to or less than
the sum of their lengths.
These conditions, however, can be met by a wide variety of norms.
Riemann pursues one method of defining them, which he calls the "sim-
plest case" (1867, p. 140). It is especially worth looking at - given the
practical success of Euclidian geometry at the human scale - because
metric relations determined on a 3-manifold by this method agree
well25 with the familiar Euclidian ones on a small neighborhood of each
point. Riemann's simplest case is based on Gauss's formula for com-
puting the length of arcs on any curved surface in Euclidian space in
terms of the surface's own intrinsic geometry (i.e., without paying
attention to the surrounding space). Riemann applies Gauss's approach
to ^-manifolds. This is how he states the gist of it: In the integral Jds,
which is equal to the length of a given curve, the integrand "ds = the
square root of an everywhere positive homogeneous function of the
second degree in the [coordinate differentials], in which the coefficients
are continuous functions of the [coordinates]" (p. 140). Therefore, in
a region U of an w-manifold W charted by the coordinate functions x1,

d ]^dhdxk (l<h,k<n) (4.4)

where the coefficients ghk vary continuously with the coordinates. Let
the curve in question be given - as in the discussion preceding eqn.
(4.3) - by a mapping u •-> y{u) of a real interval (a,b) into U. Then,
according to eqn. (4.4), its length is given by this integral:
dxh dxk t

By using the post-Riemannian ideas introduced above, this can be

elucidated as follows. Let Wl be an w-manifold and g a differentiable
25
To first order in the coordinate differentials.
 4.1 Geometries 165

mapping of Wl into the bundle of bilinear functions on its tangent

spaces, such that g assigns to each point p in 30? a bilinear function gp
on TP3W. A mapping of this sort is called a tensor field, g is a Rie-
mannian metric on Wl if, for each p e-Ul, the function gp is (i) sym-
metric, that is, gp(v,w) = gp(w,v) for all vectors v and w in TP9K; (ii)
positive definite, that is, gp(v,v) > 0 unless */ is the 0-vector; and (iii)
nondegenerate, that is, gp(v9w) = 0 for all vectors w in TP9K if and only
if v is the 0-vector. The mapping v *-> ^gp(v,v) is clearly a norm in TpWl.
This is the norm we employ for defining the length of curves in agree-
ment with Riemann's "simplest" method. A manifold in which lengths
- and with them all metric relations - are defined in this way is called
Riemannian. Thus, if the curve y in eqn. (4.3) lies in a Riemannian n-
manifold (3ft,g), the integral giving its length can be rewritten as:
cb . —
^gy(u)(yM,y(u))du (4.6)
To recover eqn. (4.5) from (4.6) we must confine the range of y to the
domain U of coordinate functions x1,. . . , xn. As is indicated in note
21, for each coordinate function xk there is a vector field d/dxk on U.
In the light of the preceding explanation it is clear that, for each pair
(h,k) satisfying the inequality in (4.4), the mapping

lxh p'dx
k

is a differentiable real-valued function on U. Let us denote it by

The tangent space Ty(M)9K is spanned by the vectors
d
1 y n
dx :'" dx
so the vector j(u) is equal to a linear combination of them, namely,
dx^ dx"
(4.7) 26
du 1
.a*
By substituting from eqn. (4.7) into (4.6) and writing ghk for the value
of ghk at y{u), we obtain the integral (4.5). The definition of the length

26
To see that this equality holds, try the following exercise. Imagine that y is the para-
metric curve through y(u) of a coordinate function belonging to some suitable chart.
(This fantasy is permissible because we have assumed that y is injective.) Obviously
y= dldu. Consider a differentiable function cp defined on a neighborhood of y(u) con-
tained in the domain of the said chart and in that of chart x. Clearly
dxl\ chq (4.8)
a*1 dx" ) du \
166 The Rich Nineteenth Century

of a curve leads at once to the notion of geodesic (or straightest) curves,

which are characterized by the fact that their length is extremal, that
is, either the greatest or the shortest among all the curves that follow
neighboring paths between the same two points.
In his study of curved surfaces in Euclidian space, Gauss introduced
a real-valued function, the Gaussian curvature, which measures a
surface's local deviation from flatness in terms of the surface's intrin-
sic geometry, without regard to the way it is embedded in space. From
this point of view surfaces such as a cone or a cylinder, which can be
flattened without stretching or tearing, are no less flat than the plane,
and their Gaussian curvature is everywhere equal to 0. On the other
hand, the surface of an egg is nowhere flat, and it evidently deviates
most from flatness at one of its tips; its Gaussian curvature is every-
where positive, and greatest at this point. Consider, finally, the surface
of a saddle; some points on it are the intersection of some curves curled
upward and other curves curled downward; the Gaussian curvature
is so defined that it takes a negative value at such points. Riemann
extended this concept of curvature to Riemannian w-manifolds. He
observed that each geodesic through a point in such a manifold is fully
determined by its tangent vector at that point. Consider a point p in a
Riemannian ^-manifold (%Jl,g) and two linearly independent vectors v
and w in TpWl. The geodesies determined by all linear combinations of
v and w form a 2-manifold about p, with a definite Gaussian curva-
ture kp(vyw) at p. The real number kp(v,w) measures the curvature of
Wl at p "in the surface direction" given by v and w (Riemann 1867, p.
144). Riemann (1861) conceived a global mapping on 9K, depending
on the metric g, that yields the said values kp(v,w) on suitable argu-
ments (p,v,w). This object - or rather its twentieth-century rational
reconstruction - is now known as the Riemann tensor. In post-
Riemannian jargon, we describe it as a differentiable mapping that
assigns to each p in a Riemannian w-manifold Tl a 4-linear function
on the tangent space TpWt.27 It is a tensor field like the metric g

27
The requirement of differentiability makes sense because the Riemann tensor maps
the manifold W into the bundle of 4-linear functions on its tangent spaces. Let the
Riemann tensor be p •-»• Rp; then

For readers who have some inkling of the subject I will note further that in a Rie-
mannian manifold (2ft,g) there is a canonic isomorphism between each tangent space
 4.1 Geometries 167

(although one of rank 4, while the metric is of rank 2). Given the above
definition of kp(v,w), it is clear that, if n = 2, the Riemann tensor
reduces to the Gaussian curvature function.
By using his extended concept of curvature, Riemann was able
to characterize with great elegance the metric manifolds in which
all figures can be freely moved around without changing their size
and shape. These are the manifolds in which measurements can be
performed with rigid rods and they are, of course, a very peculiar
subspecies of Riemann's "simplest" case. They are the Riemannian
manifolds of constant curvature, in which kp(v,w) has exactly the same
numerical value at each point p for every local pair of directions v and
w. "For obviously the figures in [such manifolds] could not be arbi-
trarily displaced and rotated if the curvature was not the same in every
direction at every point" (1867, p. 145). This idea can be nicely com-
bined with Klein's classification of metric geometries. Regarded as
Riemannian 3-manifolds, Euclidian or parabolic space has constant
zero curvature, Lobachevskian or hyperbolic space has constant nega-
tive curvature, and elliptic space has constant positive curvature. A dif-
ferent kind of space of constant positive curvature occurs in Einstein's
earliest cosmological solution of his gravitational field equations. In
this "Einstein world" each maximal set of simultaneous events consti-
tutes a 3-manifold - with the neighborhood structure of the 3-sphere
S3 - in which the global spacetime metric induces a Riemannian metric
of constant curvature > 0.28 The spatial geometry thus defined has

TpWl and its dual, the cotangent space (TpWl)*9 that assigns to each v in TPW the
linear function v* in (TpTl)* defined by v*(w) - gp(v,w) for any w in TPW. v and v*
are usually regarded as the "contravariant" and the "covariant" form of one and the
same geometric object. By using the canonic isomorphism we can readily match the
"covariant" or (0,4) form of the Riemann tensor, as characterized above, with other
forms of it found in the literature. For example, one defines its (l,3)-form, con-
travariant in the first argument and covariant in the other three, by stipulating that
if Rp stands for its value at /?, for each p in Wl, while Rp denotes the value at p of
the said covariant form, then, for every quadruple (a,b,c,d) of vectors in TpWl

Rp(a,b,c,d) = R'p(a*,b,c,d) (4.9)

S3 is the topological space formed by the set of all real number quadruples (x,y,z,w)
that satisfy the equation x2 + y2 + z2 + w2 = 1, with the subset topology induced by
the standard topology of U4. The spatial metric of constant positive curvature in the
said Einstein world agrees with the metric induced in S3 by the usual flat metric of
the 4-manifold 1R4 in which it is embedded.
168 The Rich Nineteenth Century

sometimes been called "Riemannian" in a narrow sense, but it is prefer-

able to call it spheric. Pursuant to the Erlangen Program, each of these
geometries of constant curvature is characterized by its own group of
isometries. But Klein's conception is too narrow to embrace all Rie-
mannian geometries. Indeed, in the general case, the group of isome-
tries of a Riemannian w-manifold is the trivial group consisting of the
identity alone, and therefore it utterly fails to capture the peculiarity
of the respective geometry.

4.2 Fields
The physico-mathematical concept of afieldcan be traced back to eigh-
teenth-century work in fluid dynamics by Euler and others. A fluid is
represented as a continuum characterized by physical quantities that
may vary smoothly from point to point and from moment to moment.
Take, for instance, the distribution of matter in the fluid. This is rep-
resented by the density function p, which, at any given moment, assigns
to each point in the continuum a definite real number of mass units
per unit of volume. The total mass m of the fluid is given by the inte-
gral JypdV taken over the fluid's total volume V. In the case of water
and other such liquids, p is usually regarded - with mild idealization
- as constant in time and uniform in space; therefore m = pjvd V = p V,
and V does not increase or decrease. But this idea of an incompress-
ible fluid of uniform density is obviously inadequate in the case of
gases, in which p must be thought to change gradually over space and
time. Still, in either case the density function is a constant or smoothly
varying assignment of real numbers to the points of a continuum, that
is, a so-called scalar field.29 Suppose now that the fluid is in motion,
relatively to an appropriate rigid reference frame F. Let x1, x2, x3 be
Cartesian coordinates affixed to F. If p is a particular point of the fluid,

29
As is explained in Supplement 1.3, a scalar is an object that, when "multiplied" by a
vector, changes its size but not its direction, and so rescales it. The most common
vector spaces take real numbers as scalars, whence the use of 'scalar field' for a
smooth assignment of real numbers to points and 'vector field' for a smooth assign-
ment of vectors. In these expressions, the English word 'field' translates the French
'champ' and the German 'Feld'. Unfortunately, the algebraic structure from which
the scalars of a vector space are drawn is also called a 'field' in English ('corps' in
French and 'Korper' in German). So a scalar field on a space 5 is a smooth mapping
of 5 into a field of scalars - the word 'field' being used in completely disparate mean-
ings in its two occurrences in this sentence.
 4.2 Fields 169

its coordinate xl(p) (i = 1,2,3) is changing in any given instant t at the

rate expressed by the derivative dxl(p)/dt\t. The assignment to each p
in the fluid of velocity components
dx\p) dxHp) dx\p)
dt ; dt
suitably describes the motion of the fluid at time t, relative to the Carte-
sian system (x 1 ,* 2 ,* 3 ). Of course, the motion is a relation between the
fluid and the frame F, and it is not affected by the particular coordi-
nate system affixed to F that we choose for describing it. In view of
this, the motion of our fluid at time t is represented - since the late
nineteenth century - by the assignment of a coordinate-free - or geo-
metric - object to each /?, viz., the velocity vector \p(t). Both methods
of description are directly related, as follows: Let e, be a unit vector at
/?, pointing in the direction in which the coordinate xl increases while
the other two remain fixed (i = 1,2,3); then
dx\p)
e3 (4.10)
dt dt
The assignment p >-> vp(t) - where p ranges over the points of a fluid
and vp(t) is the momentary velocity of p in the chosen frame of refer-
ence - is, presumably, the earliest example of a vector field.30 This
method of representation can be naturally extended to the forces
exerted on each point and the resulting accelerations.
Early in the nineteenth century Laplace and Poisson described the
action-at-a-distance of central gravitational and electrostatic forces by
means of what we now would say are vector fields. Given a static dis-
tribution of electric charge in empty space, we consider its action under
Coulomb's Law 31 on a test charge, that is, on a point charge so small
that it does not significantly contribute to the effect of the charge dis-
tribution. If the test charge stands motionless at point /?, we denote by
Ep the force per unit charge exerted on it by the charge distribution.
Relative to a Cartesian coordinate system (x 1 ,* 2 ,* 3 ) affixed to the dis-

30
One ought to ask oneself, to what vector space does vp belong? The only answer that
makes sense to me is that it is a vector in the tangent space at the point momentar-
ily occupied by p in the space defined by the reference frame. But this answer became
possible only much later. While the underlying space was Euclidian, nobody thought
of distinguishing it from its tangent spaces.
31
See page 82.
170 The Rich Nineteenth Century

tribution's rest frame, Ep can - like vp above - be analyzed into com-

ponents. We write:
E, =E1(p)t1 + £2(p)e2 +£ 3 (p)e 3 (4.11)
with ei, e2, e3 as in eqn. (4.10). The mapping p •-» Ep is the electric field
generated by the given charge distribution. I deliberately say 'is', not
'represents', for, at this stage, the electric field itself was just a conve-
nient mathematical representation of a physical reality consisting of the
charges. Indeed the vector notation, which so eloquently wraps up the
three coordinate-dependent components £i, £ 2 , £3 into the one geo-
metric object E, was not invented until much later, after Maxwell had
conceived the electric and magnetic fields as the actual physical support
of electric, magnetic, and optical phenomena. But before turning to
him, I shall refer to an important mathematical idea that came up
together with the introduction of central force fields.
To introduce a slightly different perspective, I shall link it to the
gravitational field generated by an arbitrary distribution of point
masses. To get a proper characterization of it, one need only substitute
in the preceding paragraph 'mass' for 'electric charge' and 'charge', G,
G for E, £, and 'Newton' for 'Coulomb'. Recall Leibniz's discussion
of the conservation of "live force" (§1.5.2). In the light of it, it is clear
that - in an ideal situation in which only gravitational forces count and
friction may safely be ignored - the mechanical work done to move a
test particle against the gravitational field forces is fully recovered by
letting the particle fall back to the starting point. This perfect balance
does not depend on the shape of the upward and downward paths nor
on the time spent on them. In vector notation,
-ds = O (4.12a)
no matter what the closed path along which this integral is evaluated.
In component notation, relative to a Cartesian system (x1,:*;2,*3) affixed
to an inertial frame, eqn. (4.12a) can be rewritten:
JG1dx1 +G2dx2 +G3dx3 =0 (4.12b)
Equation (4.12b) obtains if and only if there is a scalar field p •-»
such that

or - again in vector notation - if and only if

 4.2 Fields 171

G = -VO (4.13a)32
In other words, due to the fact - expressed in eqn. (4.12) - that grav-
itational forces are conservative (see §2.5.3, after eqn. (2.14)), the
information content of the vector field G is also encoded more simply
in the scalar field O, which is known as the gravitational potential. The
electrostatic potential cp is defined in the same way: Substitute E and E
for G and G in the foregoing discussion.33
Maxwell's theory of electricity and magnetism was first adumbrated
in his paper "On Faraday's lines of force" (1855/56), in which he
sought to work out Faraday's conception of magnetic lines of force in
a mathematically precise and mechanically viable manner. I mentioned
in §2.5.2 the discoveries of Oersted and Ampere, as well as the latter's
purported explanation of them by means of attractive and repulsive
central forces. Faraday did not go along with Ampere's account. He
felt no sympathy for instantaneous action at a distance, and his own
discoveries led him to see the true protagonist of electromagnetic
action, not in the matter-borne electric charges, but in the - possibly
empty - space between them.34 The reader has probably observed how
iron filings strewn among magnets spontaneously align themselves
along smooth curves that join the magnetic poles and appear to con-
tinue inside the magnets. Faraday took this as an indication that such
curves - which he termed magnetic lines of force - are the site of natural
powers that are ready to act on a susceptible material as soon as it is
placed on them. Because these lines are curved^ Faraday felt that he
could not conceive them "without the conditions of a physical exis-
tence in that intermediate space".35 He became convinced of the phys-

32
The minus sign in eqns. (4.12) is just a convenient - and well-entrenched - conven-
tion. Obviously, if there is a scalar field <j> satisfying these equations, there necessar-
ily is another one for which the equations hold with the minus sign deleted, viz., -O.
33
Need I mention that the potentials O and (p are defined in each case only up to an
arbitrary constant? Since dk/dx1 = 0 for any constant k, if O is a solution of eqns.
(4.12b), <1> + k is another. This is not important in the present context, but it gives
an inkling of the conventional element in physico-mathematical representations.
34
In November 1845, Faraday introduced the word 'field' as a name for just such a site
of potential electromagnetic action (Diary, §7979; cf. EREM, §2247). In 1850, he
dropped the word after defining it in terms of lines of force (EREM, §2806), but it
had already gained the favor of younger physicists like Kelvin. See references in
Gooding (1981, footnote 35), and in OED, s.v. 'field', 17a.
35
Faraday (EREM, §3258); cf. §3263: "To acknowledge the action in curved lines,
seems to me to imply at once that the lines have a physical existence. It may be a
172 The Rich Nineteenth Century

ical reality of magnetic lines of force through his work on the effects
of magnetism on light (after 1845); he could, however, find support-
ing evidence in the phenomena of electromagnetic induction, which
was perhaps his most striking - and historically significant - discovery
(c. 1831). Faraday had shown that an electric current appears in a
closed wire loop in three situations that, using the idea of lines of force,
can be described as follows: (i) The loop moves across magnetic lines
of force (issuing from a magnet or surrounding a current-carrying wire,
as in Oersted's experiment); (ii) the loop rests and a magnet or current-
carrying wire is moved in its neighborhood in such a way that the lines
of force issuing from and presumably dragged by the latter successively
cut the loop; and (iii) an electric current is started (interrupted) near
the loop, so that, while that current is increasing (decreasing) the loop
is swept by an expanding (collapsing) system of lines of force. Faraday
devised a way of quantifying the lines of force that issue from a magnet
or a current of known intensity and showed that the electromotive
force developed in a loop in the said three cases is always proportional
to the number of lines per second intersected by the loop. Faraday also
extended this idea of lines of force to electrostatics and gravitation but
would grant that gravitational lines are imaginary, because he thought
that they are always straight.36
Maxwell (1855/56) treated Faraday's electric (and magnetic) lines of
force as a simple geometrical consequence of the disposition, present at
each point p in the surroundings of an electrified conductor (or, respec-
tively, of a magnetic body), to urge in a definite direction a small charged
body (or one of the poles of a small magnet) that is placed at p.
When a body is electrified in any manner, a small body charged with
positive electricity, and placed in any given position, will experience a
force urging it in a certain direction. If the small body be now negatively

vibration of the hypothetical aether or a state or tension of that aether equivalent to

either a dynamic or a static condition; or it may be some other state, which though
difficult to conceive, may be equally distinct from the supposed non-existence of the
lines of gravitational force [.. .]."
See the quotation in note 35. Maxwell, the mathematician, disabused him in a letter
that he wrote to Faraday in 1857: "The lines of Force from the Sun spread out from
him and when they come near a planet curve out from it so that every planet diverts
a number depending on its mass from their course and substitutes a system of its own
so as to become something like a comet, if lines of force were visible" (Campbell and
Garnett 1884, p. 203; quoted in Harman 1982b, p. 88).
 4.2 Fields 173

electrified, it will be urged by an equal force in a direction exactly

opposite.
The same relations hold between a magnetic body and the north or
south poles of a small magnet. If the north pole is urged in one direc-
tion, the south pole is urged in the opposite direction.
In this way we might find a line passing through any point of space,
such that it represents the direction of the force acting on a positively
electrified particle, or on an elementary north pole, and the reverse direc-
tion of the force on a negatively electrified particle or an elementary
south pole. Since at every point of space such a direction may be found,
if we commence at any point and draw a line so that, as we go along it,
its direction at any point shall always coincide with that of the resultant
force at that point, this curve will indicate the direction of that force for
every point through which it passes, and might be called on that account
a line of force. We might in the same way draw other lines of force, till
we had filled all space with curves indicating by their direction that of
the force at any assigned point.
(Maxwell 1890, vol. I, p. 158)

Thus, in effect, Maxwell conceived the electric or magnetic force that

would be exerted on a test object at a given point as a vector at that
point, and Faraday's lines of force as the paths in space to which such
vectors are tangent. Maxwell remarks that Faraday's method of repre-
senting a force field by lines of force will "tell us the direction of the
force, but we should still require some method of indicating the inten-
sity of the force at any point" (Ibid.). He proposes to consider Faraday's
lines "as fine tubes of variable section carrying an incompressible
fluid". Then, "since the velocity of the fluid is inversely as the section
of the tube, we may make the velocity vary according to any given law,
by regulating the section of the tube, and in this way we might repre-
sent the intensity of the force as well as its direction by the motion of
the fluid in these tubes" (Maxwell 1890, vol. I, pp. 158-59). Maxwell
emphasizes that he does not mean to postulate a hypothetical fluid to
explain actual phenomena. The fluid in question here "is merely a col-
lection of imaginary properties which may be employed for establish-
ing certain theorems in pure mathematics in a way more intelligible to
many minds and more applicable to physical problems than that in
which algebraic symbols alone are used" (p. 160). Indeed, it is not
"assumed to possess any of the properties of ordinary fluids" except
those required for the present exercise, viz., "freedom of motion and
resistance to compression". If I understand him well, Maxwell is saying
174 The Rich Nineteenth Century

that he resorts to his fictitious fluid only as a more familiar alternative

realization of the mathematical structure that he submits is also em-
bodied in electromagnetic phenomena. The algebraic symbolism
adopted by mathematicians for describing this structure would, strictly
speaking, also be sufficient for the physicist's purpose, which is to grasp
electromagnetism by means of it; but the hydrodynamic analogy is nev-
ertheless introduced because, in Maxwell's judgment, it is both didac-
tically enticing and heuristically fruitful.
In his next paper on the subject, Maxwell (1861/62) produced a
mechanical model of electromagnetic action of a very different char-
acter. While in the earlier paper he used "mechanical illustrations to
assist the imagination, but not to account for the phenomena", he now
proposed "to examine magnetic phenomena from a mechanical point
of view, and to determine what tensions in, or motions of, a medium
are capable of producing the mechanical phenomena observed" (1890,
vol. I, p. 452). Maxwell invites the reader to
suppose that the phenomena of magnetism depend on the existence of a
tension in the direction of the lines of force, combined with a hydrosta-
tic pressure; or in other words, a pressure greater in the equatorial than
in the axial direction: the next question is, what mechanical explanation
can we give of this inequality of pressures in a fluid or mobile medium?
The explanation which most readily occurs to the mind is that the excess
of pressure in the equatorial direction arises from the centrifugal force
of vortices or eddies in the medium having their axes in directions par-
allel to the lines of force.
(Maxwell 1890, vol. I, p. 455)
To ensure that neighboring vortices can rotate in the same sense,
Maxwell inserts between them rows of little idle wheels. I cannot go
further into this quaint and yet admirable construction. Suffice it to say
that by pursuing it far enough Maxwell reached the conclusion that
light is an electromagnetic process. His judicious choice of mechanical
analogues for known electromagnetic quantities enabled him to calcu-
late - "from the electro-magnetic experiments of MM. Kohlrausch and
Weber" (p. 500) - the speed of propagation of transverse waves in his
hypothetical medium by a straightforward application of a formula
from continuum mechanics.37 He found it to be 310,740 kilometers per
second "in air or vacuum" (p. 499), a value so close to the velocity of

37
Viz. V = Vra/p, where V is the said speed, m is the coefficient of transverse elasticity
and p is the density.
 4.2 Fields 175

light in air as determined by Fizeau (314,858 kilometers per second),

"that we can scarcely avoid the inference that light consists in the trans-
verse undulations of the same medium which is the cause of electric
and magnetic phenomena" (p. 500; Maxwell's italics). This identifica-
tion led Maxwell henceforth to designate this medium by the same
name 'aether' employed for it in the wave theory of light.38
Maxwell did not abide by this "rotatory theory of magnetism", in
which he perceived a tendency "towards the to me inconceivable and
.*. no doubt to the misty".39 In his Treatise on Electricity and Mag-
netism (1873), he noted that this theory should be taken only for "a
demonstration that mechanism may be imagined capable of producing
a connexion mechanically equivalent to the actual connexion of the
parts of the electromagnetic field". However, "the problem of deter-
mining the mechanism required to establish a given species of connex-
ion between the motions of the parts of a system always admits of an
infinite number of solutions",40 and he had no evidence that the par-
ticular solution described in the rotatory theory was the true one.
Therefore Maxwell, although he continued to believe that electromag-
netic phenomena are the manifestation of a mechanical process,
governed by Newton's Laws of Motion, did not again put forward a
hypothetical account of its operation. The third and last of his great
papers on electrodynamics, "A Dynamic Theory of the Electromagnetic
Field" (1864), turns on the idea
that there is an aethereal medium pervading all bodies, and modified
only in degree by their presence; that the parts of this medium are
capable of being set in motion by electric currents and magnets; that this
motion is communicated from one part of the medium to another by
forces arising from the connexions of those parts; that under the action
of these forces there is a certain yielding depending on the elasticity of
these connexions; and that therefore energy41 in two different forms may

38
The word 'aether' is the Latin transcription of Greek ai6fip = 'heaven'. Aristotle used
the word for his fifth element, of which the heavens are made. Descartes used it for
the more subtle form of matter, filling the interstices between grosser, ponderable
particles. In this acceptance, 'aether' was an obvious candidate for the name of the
medium whose vibrations constitute light according to the wave theories of Huygens,
Young, Fresnel, and others.
39
Maxwell to William Thomson, 15 October 1864; Glasgow University Library, Kelvin
Papers M 17 (quoted in Harman 1987, p. 277).
40
Maxwell (1873, vol. II, p. 417; 1891, vol. II, p. 470).
41
As noted in Chapter One, note 30, 'energy' - meaning 'capacity to do work' - gained
currency in the 1850s. Maxwell has the following to say on his use of it here:
176 The Rich Nineteenth Century

exist in the medium, the one form being the actual energy of motion of
its parts, and the other being the potential energy stored up in the con-
nexions, in virtue of their elasticity.
(Maxwell 1890, vol. I, pp. 532-33)
We are thus led "to the conception of a complicated mechanism
capable of a vast variety of motion, but at the same time so connected
that the motion of one part depends, according to definite relations,
on the motion of other parts, these motions being communicated by
forces arising from the relative displacement of the connected parts, in
virtue of their elasticity. Such a mechanism must be subject to the
general laws of dynamics" (p. 533; my italics). But Maxwell no longer
fancies a mechanical contraption that would account for electromag-
netic phenomena; he resorts instead to Lagrange's methods of analyt-
ical mechanics, which - as I noted in §2.5.3 - were designed to facilitate
the description of a mechanical system whose inner workings are
unknown. Early in the paper, Maxwell introduces the term 'electro-
magnetic field' for "the space in the neighbourhood of the electric or
magnetic bodies" (p. 527). The connexion between an electric current
and the magnetic forces excited by it in the field is discussed in abstract
dynamical terms, leading to the notions of (a) the electromotive force,
which "is not ordinary mechanical force, at least we are not as yet able
to measure it as common force", and which does not move "merely
the electricity in the conductor, but something outside the conductor,
and capable of being affected by other conductors in the neighbour-
hood carrying currents" (p. 539); and (b) the electromagnetic momen-
tum, "every change of which involves the action of an electromotive
force, just as change of momentum involves the action of mechanical

In speaking of the Energy of the field [. . .] I wish to be understood literally.

All energy is the same as mechanical energy, whether it exists in the form of
motion or in that of elasticity, or in any other form. The energy in electro-
magnetic phenomena is mechanical energy. The only question is, Where does
it reside? On the old theories it resides in the electrified bodies, conducting cir-
cuits, and magnets, in the form of an unknown quality called potential energy,
or the power of producing certain effects at a distance. On our theory it resides
in the electromagnetic field, in the space surrounding the electrified and mag-
netic bodies, as well as in those bodies themselves, and is in two different
forms, which may be described without hypothesis as magnetic polarization
and electric polarization, or, according to a very probable hypothesis, as the
motion and the strain of one and the same medium.
(Maxwell 1890, vol. I, p. 564)
 4.2 Fields 177

force" (p. 542). According to Maxwell, the latter is the same quantity
that Faraday called "the electrotonic state"; Maxwell later dubbed it
"the vector potential" (1873, §405; cf. 1891, vol. II, pp. 29, 187, 232).
Like their mechanical prototypes, both quantities are vectors varying
smoothly in size and direction from point to point of the field.
After inviting us to pick three orthogonal directions in space as the
axes of Cartesian coordinates and "to let all quantities having direc-
tion be expressed by their components in these three directions",
Maxwell introduces the "General Equations of the Electromagnetic
Field", 20 partial differential equations in 20 unknowns, which link
the electromagnetic momentum and the electromagnetic force with
other electromagnetic quantities (pp. 554-62; the unidentified quota-
tions in this paragraph and in note 42 are all from these pages). Trans-
lated into the vector notation that was subsequently developed by his
friend Tait and his follower Heaviside, and which so naturally fits
Maxwell's thinking, the 20 equations reduce to 8, involving 6 vector-
ial and 2 scalar variables. Besides the two vectors already named, there
are two that were already familiar in theories of electricity, viz., (c) the
"Current due to true Conduction", that is, the quantity of electricity
transmitted from one part of a body to another in unit time per unit
area; and (d) the "Magnetic Intensity", or "force acting on a unit mag-
netic pole placed at the given point"; plus two more that were first
introduced by Maxwell, viz., (e) "Electric Displacement", which he
described as "the opposite electrification of the sides of a molecule or
particle of a body which may or may not be accompanied with trans-
mission through the body"; and (f) "Total Current (including varia-
tion of displacement)", which is defined by Maxwell's equations (A) as
the vector sum of the conduction current and the time derivative of the
electric displacement. The scalars are (g) "the quantity of free positive
electricity" per unit of volume, and (h) the electrostatic potential.
A discussion of Maxwell's equations would be quite out of place
here.42 They contain or entail the laws of electricity that were formerly

42
Some readers will wish to know how the 20 equations of Maxwell 1864 relate to the
four streamlined vector equations that go under Maxwell's name in college textbooks
such as Feynman et al. (1964; see Table 18-1 in vol. II). The answer is that the latter
can be obtained from the former with some rewriting (in modern vector notation),
some reinterpretation, and - in the case of two of them - some straightforward math-
ematical reasoning. I denote electromagnetic momentum by A, magnetic intensity by
H, electromotive force by F, conduction current by j , electric displacement by D,
178 The Rich Nineteenth Century

secured by Coulomb, Ampere, and Faraday but go well beyond them.

Maxwell's innovations, notably the inclusion of "variation of dis-
placement" in the total current, drew support mainly from the fact that
the theory could, by Lagrange's methods, be shown to be dynamically
viable. The theory's most striking novelty concerns the existence of
electromagnetic waves: The equations admit solutions consisting of
mutually induced electric and magnetic undulatory disturbances prop-
agating in free space with the speed of light. But, apart from Maxwell's
bold conjecture that such disturbances are the stuff that light itself is
made of, there was not a shred of evidence for them. The theory was
resisted on the continent, particularly by Helmholtz, and was looked
askance in Britain by Maxwell's older friend, Lord Kelvin (born

quantity of free electricity by e, and electrostatic potential by *F (e and *F are used

by Maxwell in this sense). I use the nabla operator

\dx'dy'dzl
Let me also put E for the electromotive force on a conductor at rest, as defined in
Maxwell's eqns. (35):

E = -dA/dt-Vx¥ (35)

Maxwell's eqn. (G) is

e + VD = 0 (G)

which - restricted to free space and other regions where D = £0E and e equals the
charge density p - agrees, up to a conventional choice of sign, with Feynman's
"Maxwell equation I":

V • E = p/e 0 (I)

Putting, as is now usual, B = uH, where ji is "the ratio of the magnetic induction in
a given medium to that in air under an equal magnetizing force", eqns. (B) in Maxwell
(1864) can be written:

B = -VA (B)

Together with eqn. (35) and the mathematical theorem "V x Vo = 0 for every scalar
a", this yields "Maxwell equation II", viz.,

V x E = -dB/d£ (II)

Since, for any vector V, V • (V x V) = 0, eqn. (B) also yields Feynman's "Maxwell
equation III":

-VB = 0 (III)
 4.2 Fields 179

William Thomson), but it won acceptance among the younger electri-

cians in the United Kingdom - Lodge, Heaviside, FitzGerald - who
made it the groundwork of their research. The Maxwell equations
became the core of electrical science only after Heinrich Hertz - a
former assistant of Helmholtz - showed in 1888 that he could gener-
ate and receive electromagnetic waves with electrical laboratory equip-
ment. However, the program of conceiving a working mechanical
model of the electromagnetic aether never took root on the continent,43

Maxwell's eqns. (C) can be written as

/ (C)

which, under the restriction stipulated above for eqn. (G), translates, in SI units, into
Feynman's "Maxwell equation IV":

c 2 VxB = jeo1+dE/d? (IV)

Maxwell's "complete equations of electromotive force on a moving conductor" -

labeled (D) in his paper - are also worth looking at. By writing v for the conductor's
velocity, they become, in my notation,

F = (vxB) + E (D)

which reads, of course, just like the standard formula for the so-called Lorentz force
- i.e., the "ordinary mechanical force" - exerted by the magnetic field B and the elec-
tric field E on a body carrying a unit of charge and moving with velocity v. See eqn.
(7.3) and the text surrounding it on page 427.
The program, however, lived on in the British Isles. We hear, for instance, of one
W. M. Hicks, at St. John's College, Cambridge, who "since graduating 7th wrangler
in the Cambridge Mathematical Tripos of 1873 [and until the early years of the twen-
tieth century] devoted the greater part of his research effort to the mathematical
analysis of vortex motion in hydrodynamical ether" (Warwick 1995, p. 308). Hunt
(1987) gives an enthralling report on G. F. FitzGerald's aether models and their heuris-
tic utility. FitzGerald warned against mistaking his models - "made up of wheels and
india-rubber bands, [or] of paddle-wheels, with connecting canals" - for a likeness
of reality, but he still believed that "what physicists ought to look for is such a mode
of motion in space as will confer upon it the properties required in order that it
may exhibit electromagnetic phenomena" (FitzGerald 1885, p. 162; quoted in Hunt
1987). As late as 1904, Larmor, reviewing Kelvin's modeling exercise in his Baltimore
Lectures (delivered in 1884 but first published that year), wrote: "We are at the
parting of the ways. Is it incumbent on us to treat the aether as strictly akin to the
material bodies around us? or may we assign to it a constitution of its own, to be
tested by its success in comprehending the complex of known relations of physical
systems?" (Larmor 1904, quoted in Wise and Smith 1987, p. 323). Larmor's ques-
tions were rhetorical, for he had de facto answered 'No' to the first and 'Yes' to the
second in his book, Aether and Matter (1900).
180 The Rich Nineteenth Century

where indeed the very idea that electromagnetic phenomena rest on

some form of mechanical interaction - subject to Newton's Laws of
Motion - between ponderable matter and the aether fell into abeyance.
H. A. Lorentz, surely the most significant contributor to electrody-
namics at the turn of the century, postulated a completely motionless
aether.44 So with him the electromagnetic waves could not be mechan-
ical vibrations - involving elastic displacements, however small - but
only oscillatory changes of length and direction of the electric and mag-
netic field vectors at each point. From here to the modern view of
geometric object fields pinned directly on space (or on spacetime),
without a ghostly material medium to carry them, it was but a small
step (though one that only a truly free spirit like Einstein could have
taken).45

4.3 Heat and Chance

Maxwell is best known for his equations of the electromagnetic field
(§4.2). But a no less decisive contribution to physics was his use of the
probability calculus in the theory of heat. Through the study of thermal
phenomena in gases he sought definitive evidence for the thesis that
heat is motion - specifically, that the energy contents of steam and other
hot gases consists, for the most part, in the kinetic energy of its wildly
agitated particles. To cope with the immense number of molecules in
any observable volume of gas Maxwell took his cue from Quetelet, a
former astronomer turned social scientist who successfully employed
the probability calculus to handle demographical statistics. In this
section I shall try to ascertain the meaning of physical probability as
it was understood by Maxwell and his successors. But first let us
deal briefly - in §§4.3.1 and 4.3.2 - with the development of thermal
physics leading to his move.

44
Contrary to what one sometimes reads, Lorentz's motionless aether does not in any
way presuppose the conception of an absolute space. "When I say for brevity's sake
that the aether is at rest, I mean only that no part of this medium is displaced with
respect to its other parts and that all perceptible motions of the heavenly bodies are
motions relative to the aether" (Lorentz 1895, §1, in CP, vol. 5, p. 4).
45
For a concise and illuminating sketch of the development of the field concept from
Faraday and Maxwell, through Lorentz, to Einstein, see Nersessian (1984, Part II).
The standard reference on the history of electrodynamics in the late nineteenth
century is Whittaker (1951/53, vol. I, Ch. IX-XIII). For a richer and deeper view, see
Buchwald (1985, 1994).
 4.3 Heat and Chance 181

4.3.1 Heat as Motion

At the beginning of the nineteenth century the received scientific view
was that there exists a weightless, elastic, fluid substance - known as
caloric - responsible for all thermal phenomena.46 Parts of caloric were
supposed to repel each other and to be attracted by other material sub-
stances. The absorption of caloric furnished a seemingly obvious expla-
nation of the expansion of bodies with heat. Temperature gradients
reflected the tendency of caloric to flow from one place to another. Local
temperature, however, depended not only on the local abundance of
caloric but also on the capacity of local matter for storing it. (It was
well known that to warm up a pound of water by one degree one must
burn more fuel - and so, presumably, release more caloric - than to do
the same to a pound of mercury.) The conception of heat as motion
favored by Bacon and Boyle in the seventeenth century was generally
regarded as obsolete. Indeed, Rumford did his best to gather empirical
support for it as he supervised the manufacture of ordnance for the
Bavarian army. He observed that the rate at which heat is produced
while boring a cannon remains steady and does not gradually diminish
as it should if heat were a substance stored in and extracted from the
whole metal body. Nor could the heat released proceed exclusively from
the thin layer of metal being reduced to chips, for its total quantity was
out of all proportion with the mass of the chips obtained.47 So he con-
cluded that the heat observed was not a substance previously stored in
the cannon but only a display of the metal's growing internal agitation.
The scientific establishment, however, remained unconvinced48 until the

46
Caloric, like aether, was the subject of much respectable scientific work before being
pronounced nonexistent. My short remarks cannot do justice to caloric physics, its
diversity, and its applications. For a detailed report, see Fox (1971).
47
While the borer detached 54 g of "metallic dust, or, rather, scaly matter" from the
bottom of the cylinder, the heat developed in the cannon - apart from any heat loss
to the environment - was sufficient to bring to the boil some 2,300 g of ice-cold water.
Rumford not only judged it improbable that such a small metal mass should contain
that much caloric, but, by "experiments, made for the express purpose of ascertain-
ing that fact", he showed that "the capacity for Heat of the metal of which great
guns are cast is not sensibly changed by being reduced to the form of metallic chips
in the operation of boring cannon" (Rumford 1798, in CW, I, 10).
48
One notable exception was young Humphry Davy, who, shortly after Rumford com-
municated his results to the Royal Society of London, published his own friction
experiments with blocks of ice, which - he believed - showed that "heat cannot be
182 The Rich Nineteenth Century

1840s, when Joule measured the mechanical equivalent of heat and

Helmholtz and others took this as proof that the capacity for doing
mechanical work - or energy, as it would soon be called - is conserved
in nature and that heat is just one of its transitory guises.
In about 1838 Joule began experimenting on what we would now
call energy conversion, with the aim of improving the electric motor
and eventually replacing with it the fairly inefficient steam engines of
the time. He was soon disappointed in his hopes, but he had shown
by the way that, since an electric current would produce heat and
mechanical work in fixed amounts depending on the current's inten-
sity, the mechanical and heating powers of a current stand in a fixed
ratio, which he reckoned as 838 foot-pounds per BTU.49 Emboldened
by the fact that "the magnetic engine [i.e. the electric generator] enables
us to convert mechanical power into heat by means of the electric cur-
rents which are induced by it", he ventured to think "that by inter-
posing an electro-magnetic engine [i.e., an electric motor] in the circuit
of a battery a diminution of heat evolved per equivalent of chemical
change would be the consequence, and this in proportion to the
mechanical power obtained" (Joule 1843, in Joule 1884, p. 120). This
was hard to prove, for heat loss to the environment can easily escape
detection,50 but Joule was able to show by a clever experiment that
heat literally converts into mechanical work. He compressed air to 22
atmospheres inside a vessel placed in a water tank and released the air
into a second vessel in the same tank. If the second vessel was kept at
atmospheric pressure, the inflowing air had to do work against it and
the water in the tank grew colder, but no temperature change was
observed if the second vessel was exhausted and the air entered it
without overcoming any resistance. As Joule noted, these results are
baffling if heat is a substance, but they "are such as might have been

considered as matter" but must be regarded as a "peculiar motion, probably a vibra-

tion of the corpuscles of bodies" (quoted in Wolf 1939, p. 198). However, the effects
observed by Davy were probably due to the conduction of heat from the environ-
ment to the ice.
One foot-pound is the quantity of work required to raise one pound to the height of
one foot (in London, I presume). One BTU (or British Thermal Unit) is currently
defined as the quantity of heat required to increase the temperature of one pound of
water from 63 °F to 64°F; Joule's unit was slightly different; see the quotation in the
main text, at the end of the paragraph.
In 1862 it was shown experimentally by Him that "when heat does work in an engine
a portion of the heat disappears" (Maxwell 1883, p. 147).
 4.3 Heat and Chance 183

deduced a priori from any theory in which heat is regarded as a state

of motion among the constituent particles of bodies" (Joule 1844, in
Joule 1884, p. 186). Joule developed diverse methods for measuring
the mechanical equivalent of heat with increasing precision. In 1849
(Joule 1884, p. 328), he announced that "the quantity of heat capable
of increasing the temperature of a pound of water (weighed in vacuo,
and taken at between 55° and 60°) by 1° Fahr, requires for its evolu-
tion the expenditure of a mechanical force represented by the fall of
772 lbs through the space of one foot", a result that is remarkably
close to the modern value of 778.51
Joule's results did not meet immediate acceptance. However,
Helmholtz cited them as evidence for the conservation of energy in his
pioneering paper of 1847. By that time, the idea of energy conserva-
tion was cropping up in independent, often obscure, publications -
Kuhn names "twelve men who, within a short period of time, grasped
for themselves essential parts of the concept of energy and its conser-
vation" (1959, p. 321) - but Helmholtz was the first one to state it
with full clarity and precision as a fundamental principle of physics.52
He puts it forward as "a physical presupposition (Voraussetzung)" the
implications of which he develops and compares with "the empirical
laws of natural phenomena in the different branches of physics" (1847,
p. 1). This development and comparison fill most of the paper, but here
we need only consider the general remarks contained in the Introduc-
tion. Helmholtz notes that his theses can be reached from two propo-
sitions that, he contends, are logically equivalent (identisch), namely,
(i) "the statement that it is not possible to obtain limitless energy
(Arbeitskraft, i.e., 'force for doing work' - R. T.) from the effects of
any combination of natural bodies on one another", and (ii) "the
assumption that all effects in nature must be referred to attractive and

51
The said value was obtained by Joule through friction experiments in water. In the
same paper, he reported a value of 776.045, obtained through friction against cast
iron. However, in Joule's opinion it was "highly probable that the equivalent from
cast iron was somewhat increased by the abrasion of particles of the metal during
friction, which could not occur without the absorption of a certain quantity of force
in overcoming the attraction of cohesion" (1884, p. 328).
52
Helmholtz (1847) speaks of the conservation of Kraft, which literally means force.
In his text, Kraft sometimes designates Newtonian force, but more often the capac-
ity for doing mechanical work. Of course, it is only in the latter sense that Kraft is
a conserved quantity according to Helmholtz. In a footnote added in 1881, Helmholtz
uses 'Energie' for 'Kraft' in this sense (WA, I, 29).
184 The Rich Nineteenth Century

repulsive forces (Krdfte) whose intensity depends only on the distance

of the interacting points" (1847, p. 1). Proposition (i) amounts to
the impossibility of perpetual motion, which we already saw Leibniz
invoke against Descartes (§1.5.2). Helmholtz has no difficulty in
showing that proposition (i) follows from (ii), that is, that physical
systems under the exclusive rule of central forces are conservative (in
the sense explained in §2.5.3). But he also argues that proposition (ii)
follows from (i), so that a perpetual motion should be feasible in a
physical system that is governed by forces other than central:
If the natural bodies also exhibit forces (Krdfte) which depend on time
and speed or act in directions other than the straight lines joining each
pair of acting material points - e.g. rotatory ones - , then systems of
such bodies would be possible in which energy (Kraft) is either lost or
gained ad inftnitum.
(Helmholtz 1847, pp. 19f.)

Helmholtz retracted these words in 1881, noting that they hold only
if Newton's Third Law of Motion is generally valid (WA, I, 71; cf. the
footnotes to WA, I, 20, 21). Even this restriction may, however, be
insufficient to rescue Helmholtz's claim, for the equality of action and
reaction can hold - in a sense - together with energy conservation in
a classical electromagnetic system in which the force on a moving
charge depends on the latter's velocity, provided that energy and
momentum can be stored in the field.
The Introduction to Helmholtz's paper of 1847 also contains a
philosophical argument for energy conservation that deserves our
attention. While the "experimental part of our sciences" seeks for the
laws that will enable one to bring particular processes in nature under
general rules (Helmholtz's examples are the law of the refraction and
reflexion of light, and Mariotte and Gay-Lussac's law regarding the
volumes of gases), the theoretical part endeavors "to find the unknown
causes of the processes from their visible effects". This endeavor is
justified and indeed necessitated by "the principle that every change
in nature must have a sufficient cause". Some of the causes we
discover are changeable, so we must look for their causes in turn.
Thus, "the final goal of the theoretical natural sciences is to discover
the ultimate unchangeable causes of natural processes" (p. 2). In
its quest science uses two "abstractions", namely, Materie ('matter'),
that is, the sheer existence of things apart from their effect on
other things and on our sense organs, and Kraft ('force'? 'energy'?),
 4.3 Heat and Chance 185

that is, the capacity of things to produce effects. The problem of

referring natural phenomena to ultimate unchanging causes therefore
takes this form: "as ultimate causes [. . .] one ought to find unchang-
ing Krdfte" (p. 4).53 Helmholtz takes for granted the analysis of
nature into chemical elements, that is, "matters with unchanging
Krdfte (undestructible qualities)". Under this analysis, the only possi-
ble changes in the universe are spatial, that is, motions; "so the forces
can only be moving forces, dependent in their action upon spatial rela-
tions" (Ibid.). A few additional reflections lead to the concept of central
forces and to the following conclusion:

The problem of the physical sciences is finally determined thus: To refer

natural phenomena to unchanging attractive and repulsive forces, whose
intensity depends on distance. The solvability of this problem is at the
same time the condition of the complete intelligibility of nature.
(Helmholtz 1847, p. 6)

In 1881 Helmholtz reconsidered his distinction between the experi-

mental search for laws and the theoretical search for causes. He says
that the philosophical discussion I have just summarized was more
strongly influenced by Kant than he would still judge proper. Remark-
ably, Helmholtz's emended view comes even closer to Kant's philoso-
phy (cf. §3.4):

Only later have I understood that in fact the principle of causality is

nothing but the assumption that all natural phenomena accord with law.
Law acknowledged as objective power we call Kraft. By dint of its ety-
mological meaning, cause (Ursache)54 is the unalterably permanent or
existent behind the changes of phenomena, namely Matter (Stoff), and
the law of its action is Kraft.
(Helmholtz WA, I, 68)

Soon after the publication of Helmholtz's paper the English-speak-

ing physicists who shared his views began using 'energy' as a general
term for his conserved Kraft (cf. Chapter One, note 30). According to
this school of thought - to which Kelvin, Rankine, Maxwell, and, of

53
'Krafte' is the plural of 'Kraft'. The reader may decide when to render it as 'forces'
and when as 'energies'.
54
'Ursache', the German word for 'cause', literally means 'primal thing' or 'thing-
at-the-origin'.
186 The Rich Nineteenth Century

course, Helmholtz himself belonged - energy is present in bodies in

two main forms only: (i) as energy of motion or kinetic energy', equal
to one-half the body's mass times its squared velocity, and (ii) as energy
of position or potential energy, equal to the work that must be
done against ambient forces to carry the body from a position of
(conventionally) zero potential energy to its present place. The received
inverse-square laws of gravity, electrostatics, and magnetostatics
naturally led to this concept of energy of position, and our physicists
looked forward to explaining chemical and electrodynamic energy in
some such way. Latent heat - that is, the heat absorbed without
temperature change by melting solids and boiling liquids - was
readily understood in similar terms, as the work done to break a solid's
rigidity or a liquid's cohesion, which could later be recovered by
freezing or condensation. But manifest heat - showing up through
temperature increases - was regarded by these physicists as kinetic
energy. In his textbook on the theory of heat, Maxwell concedes that
"the evidence for a state of motion, the velocity of which must far
surpass that of a railway train, existing in bodies which we can place
under the strongest microscope, and in which we can detect nothing
but the most perfect repose, must be of a very cogent nature before we
can admit that heat is essentially motion" (1883, pp. 302f.). The alter-
native is "that the energy of a hot body is potential energy". Now, this
form of energy "depends essentially on the relative position of the parts
of the system in which it exists", so that motion is necessarily involved
"in every transformation of potential energy" (p. 303). Therefore heat
transfer, such as is bound to occur wherever there is a temperature gra-
dient, must involve motion, even though none is visible. This proves
that every hot body contains invisibly moving parts and that at least
part of its thermal energy arises from their motion.55 The study of gases
- says Maxwell - provides good evidence that "a very considerable
part of the energy of a hot body" (p. 304) is in fact the kinetic energy
of its submicroscopic parts.

55
The proof, of course, depends on the premise that all nonkinetic energy is energy of
position. Physicists like Mach, Ostwald, and Duhem, who rejected the analysis of
matter into molecules as an unnecessary metaphysical hypothesis, did not subscribe
to that premise but simply viewed energy as a fundamental physical quantity that is
transformed in fixed ratios from one to another of its many guises: mechanical,
thermal, electromagnetic, chemical, and so on. This school of thought presumably
inspired Sigmund Freud with his notion of psychic energy, although he never both-
ered to propose an experiment for ascertaining its mechanical equivalent.
 4.3 Heat and Chance 187

4.3.2 The Concept of Entropy

If heat is energy and energy is conserved, why cannot we satisfy our
energy requirements through a scheme like the following? (i) Extract
heat from the environment and convert it into kinetic energy; (ii) use
the kinetic energy thus obtained to run the transportation system and
to generate electricity; and (iii) as wheels chafe the ground and resis-
tors warm up in the electric network, let the heat return to the envi-
ronment and recycle it. Sadi Carnot (1824) explained within the theory
of caloric why this ecophile's dream is unfeasible. Just as the substance
water yields mechanical work as it falls from a higher to a lower level,
so the substance caloric yields mechanical work as it falls from a higher
to a lower temperature. Therefore, to obtain useful work from a hot
source heat must be transferred to a colder sink. Moreover, Carnot
showed by a masterly piece of reasoning that the efficiency of a heat
engine - that is, the amount of work that it can obtain from a unit of
heat as this falls to a lower temperature - has an upper bound that
depends not on the machine's design, nor on the working substance
employed in it (e.g., steam, gasified petrol, etc.), but only on the tem-
peratures between which the engine operates.
In view of Carnot's achievement and its well-corroborated predic-
tions and applications, physicists like Kelvin were reluctant to give up
the caloric theory.56 When the kinetic theory prevailed c. 1850, the
new science of heat founded by Kelvin and Clausius in the wake of
Helmholtz and Joule rested on two principles, viz., the Conservation
of Energy and the so-called Second Principle of Thermodynamics,
embodying the gist of Carnot's idea without the water-level metaphor.
Kelvin and Clausius gave two different versions of this principle, which,
in fact, entail each other. Let me quote them in Enrico Fermi's words:

56
This may have been partly due to careless reading, for Carnot did not conceal his
doubts with regard to the prevailing theory of heat (1824, p. 37nl). He also asked
rhetorically: "Can one conceive the phenomena of heat and electricity as due to any-
thing else than certain motions of bodies?" (p. 21nl). Indeed, as Maxwell (1883, pp.
146f.) showed, Carnot's analysis is inconsistent with the caloric theory. To see this,
suppose that an amount of kinetic energy W is obtained while heat is transferred
from a warmer to a colder body. Let Q be the quantity of heat drawn from the source.
W can in turn be converted into a quantity of heat q, for example, by causing it to
move a water paddle. So, unless a part of Q is destroyed to compensate for the gen-
eration of W - which is of course impossible if heat is caloric - the process will create
a net increase of heat q - which again is impossible if heat is caloric.
188 The Rich Nineteenth Century

A transformation whose only final result is to transform into work heat

extracted from a source which is at the same temperature throughout is
impossible (Postulate of Lord Kelvin).
A transformation whose only final result is to transfer heat from a body
at a given temperature to a body at a higher temperature is impossible
(Postulate of Clausius).
(Fermi 1937, p. 30)
Building on Carnot's work, Kelvin developed the absolute scale of
temperature and Clausius the concept of entropy. Entropy soon became
entangled in philosophically significant debates, so I must stop to
explain it. The long-winded road leading to its definition is in itself of
interest for philosophers, as an egregious instance of creative concept
formation in mathematical physics.57 Ultimately, I do not think that
entropy is any more contrived than force or momentum, or the New-
tonian concepts of space and time. But these have managed, despite
their novelty and artificiality, to pass for mere refinements of familiar
ideas, whereas entropy - perhaps because it was carefully thought out
from the beginning and was given a newly coined name58 - has always
appeared to be foreign to common experience. In this it anticipates
many quantities in twentieth-century physics, such as spin, strangeness,
and spacetime curvature, for which no folk prototype can be cited.
I begin with Carnot's theorem on the efficiency of heat engines. Such
an engine typically produces work W by taking a quantity of heat Qx
from a body B1 at temperature T1 and surrendering a quantity of heat
Q2 to a body B2 at temperature T2 < T\. If all forms of energy are
expressed in the same unit (say, joules), obviously W = Qi - Q2. The
engine's efficiency is naturally expressed by

,-£-.-£
57
I confine my exposition to the original definition in the 1850s. For a philosophical
discussion of the subsequent development, generalization and reinterpretation of
'entropy' in physics, see Bartels (1994, pp. 135-219).
58
By Clausius. Rankine called it "thermodynamic function". Structurally, the Greek
word evxpoTCia stands to xpOTifj ('turn, turning') in the same relation as evepTEioc
stands to Epyov ('work'). 'EvepYEict, meaning 'activity', 'actuality', 'vigor of style', was
used abundantly by Greek prose writers since Aristotle. On the other hand, evtpOTtia
is extremely rare in extant literature; it occurs once in the Homeric Hymn to Hermes
(245), meaning 'trick, wile', and once in Hippocrates (De decente habitu, 2), meaning
'modesty'.
 4.3 Heat and Chance 189

To prove his theorem, Carnot constructed the concept of an ideal

engine, known as a Carnot engine, that runs cyclically - periodically
returning the working substance to its initial state59 - and reversibly -
so that it can also run backwards, converting heat Q2 at temperature
T2 into heat Qi at temperature 7\ with a net expenditure W of work
supplied by external sources.60 Because they are reversible, all Carnot
engines working between a given pair of temperatures are equally effi-
cient, and no beat engine working cyclically between the same tem-
peratures can be more efficient than them. Let © denote a Carnot
engine that produces work W by taking a quantity of heat Qx from a
source at temperature Ti and surrendering a quantity of heat Q2 to a
sink at the lower temperature T2, and let &* be any other engine -
reversible or not - that obtains work W* by extracting heat Q* at tem-
perature Ti and surrendering heat Q* - W* at the lower temperature
T2. Without significant loss of generality we may assume that Qi/Q*
is rational, so that there are integers m and w, relatively prime, such
that Qi/Q* = m/n.61 We can imagine a heat engine @, each cycle of
which consists of n cycles of (£ running backwards followed by m
forward cycles of ©*. If ©* were more efficient than S , that is, if
(W*/Q*) - (W/Qi) = AT] > 0, each cycle of ® would obtain the quan-
tity of work AW = mQ *Ar| > 0 by taking heat nQ2 from a source at
temperature T2 and returning to it heat m(Q* - W*) at the same tem-
perature, without causing any other net change in the world, m viola-
tion of Kelvin's postulate. (And AW could, of course, be used for
heating up any body already at a temperature higher than T2, in vio-
lation of Clausius's postulate.) Therefore, © is at least as efficient as
©*. In particular, if ©* is another Carnot engine, it must be exactly as
efficient as (£, for we can then imagine a heat engine 2) *, each cycle of
which consists of m cycles of (£* running backwards followed by n
forward cycles of ©, and show by the above argument that ®* would
violate Kelvin's postulate if © were more efficient than ©*.
If all Carnot engines operating between two given temperatures have
the same efficiency, the ratio Q2/Qi between the heat Q 2 delivered by

Thanks to this feature, the working substance in a Carnot engine contains the same
amount of energy at the beginning and at the end of each cycle and one need not
inquire how much energy is needed to bring it from its initial to its final state.
For a description of a Carnot engine, see the appendix at the end of §4.3.
If Qi/Q* is irrational, then, for every positive real number e, no matter how small,
there is a pair of integers mz and «e, relatively prime, such that |«EQi - meQ*\ < £.
190 The Rich Nineteenth Century

such an engine to a sink at temperature T2 and the heat Qi it draws

from a source at temperature T2 can depend only on those tempera-
tures (cf. eqn. (4.14)). We can therefore put

77- = q>(Ti,T2) (4.15)

where cp is a real-valued function of temperature pairs, independent of

engine design. The function (p has the following property:62 For any
three temperatures To, Tl5 and T2,

So a function of temperature 0 can be defined, relative to an arbitrary

temperature To, by the condition
e(T) = fe<pCTo,T) (4.17)

where k is a positive, real-valued constant. By substituting from eqn.

(4.15) into (4.17) we have that, for any choice of £,

(Note that k is the only conventional element in the definition of 0, for

if we define 0' relative to To' as in eqn. (4.17), 0' satisfies eqn. (4.18)
just like 0, so 0 and 0' differ at most by a constant factor.)
Ti > T2 entails that 0(Ti) > 0(T2). Otherwise, a Carnot engine
running in reverse between these temperatures could draw an amount
of heat Q 2 at temperature T2 and deliver an equal or smaller amount
of heat Qx at the higher temperature Tx without adding any energy
from external sources, in violation of Clausius's postulate. So there is
a one-one correspondence between the values of 0 and temperature
states, as labeled by any scale, and it is possible to define a scale of
temperature such that Q(T) = T throughout. This is Kelvin's absolute
scale. In contrast with traditional scales, which are based on the

To prove it, consider two Carnot engines, (^ and (£2, operating between temperatures
Ti and To, and T2 and To, respectively. Let (£i absorb a quantity of heat Qi at temper-
ature T1 and deliver Qo at To, while (£2 absorbs a quantity of heat Q2 at temperature
T2 and also delivers Qo at To. The engine (£ whose cycle consists of one forward cycle
of (£2 followed by one backward cycle of (S^ is also a Carnot engine, for a sequence of
two reversible cycles is itself reversible. (£ absorbs Qi at temperature T1 and delivers
Q2 at T2. Obviously, <p(Ti,T2) = QJQ, = (Q2/Qo)/(Qi/Qo) = cpfTo,
 4.3 Heat and Chance 191

thermal properties of this or that thermometric substance, the Kelvin

scale rests on the universal properties of Carnot engines. To set up a
temperature scale one normally picks a unit and a zero. In this case,
however, the latter choice is not up to us. By substituting from eqn.
(4.18) into (4.14) we see that the efficiency rj12 of a Carnot engine oper-
ating between temperatures 7\ and T2 is given by
6(T2) 6(70-9(7,)
1112 1 (4J9)
- -6(7,)- 6(7,)
Therefore, the temperature represented by 0 = 0 must be such that every
Carnot engine approaches 100 percent efficiency as its sink approaches
that temperature. This limit temperature is the same for all Carnot
engines, regardless of the temperature of their heat sources.63 On the
other hand, we are free to define the unit of the Kelvin scale (known
as 1 kelvin, or IK). This amounts in effect to fixing the multiplicative
constant k in eqn. (4.17). To make the kelvin as close as possible to
the familiar Celsius degree, we agree that there will be a difference of
100K between the freezing and the boiling points of water at atmos-
pheric pressure. The efficiency r| of a Carnot engine operating between
100°C and 0°C must then satisfy the relation
(* + 100)K-*K_ 100K
1
(x + 100)K (x + 100)K
Equation (4.19) could be solved for x if one could accurately measure
r|. This would give us the Kelvin temperature of the freezing point of
water (at atmospheric pressure). Metrologists, however, have found it
more expedient to assign 273.16 K to the triple point of water, an easily
reproducible state in which water, ice, and vapor coexist without
noticeable changes.
We are finally in a position to define entropy. Henceforth all tem-
peratures are assumed to be given in the Kelvin scale. By using eqns.
(4.14) and (4.19) we can write

63
Suppose that, for an arbitrarily small positive real number e, a Carnot engine
drawing heat from a source at temperature T1? has the efficiency (0(T\) —
and a Carnot engine ©25 drawing heat from a source at temperature T2, has the effi-
ciency (6(T2) - e)/0(T2). The sink of both machines must then be at the same tem-
perature 6"!(£).
192 The Rich Nineteenth Century

for the amount of work that a Carnot engine operating between tem-
peratures Ti (high) and T2 (low) extracts from a quantity of heat Qi.
As we know, W is an upper bound on the work produced by real heat
engines, operating irreversibly in the said conditions. By using eqn.
(4.18) we verify that, in the case of Carnot engines, the factor Q/Ti
equals Q2/T2. In other words, in a reversible thermodynamic process
the ratio Q/T between the quantity of heat exchanged at a fixed tem-
perature at any point of the process and the temperature at which the
exchange takes place is conserved. A physical quantity with this prop-
erty is worth looking into.
Following Clausius's lead, we consider a cyclic process in the course
of which a system ® exchanges heat with n bodies at temperatures T1?
. . . , Tn. Let Qk denote the quantity of heat exchanged at temperature
Tk. We take Qk positive if it is heat absorbed by © and negative if it
is delivered by (L We shall prove that

I§^0 (4.22)
the equality sign holding if and only if the process is reversible. To do
so, we introduce an additional body at an arbitrary temperature To >
0 and n Carnot engines ©i,. . . , ©„, such that G^ operates between Tk
and To exchanging at Tk the quantity of heat -Q&.64 So, by eqn. (4.18),
the quantity of heat Qo exchanged by ©^ at To must be

Qo=Qk^~ (4.23)

Consider now the complex process consisting of one cycle of © fol-

lowed by one cycle of each of the Carnot engines © l5 . . . , ©„. The net
exchange of heat at temperature Tk amounts to Qk - Qk = 0 (1 < k <
«), while the net exchange of heat at temperature To is equal to

l
k=l k=l k

Now, if Qo > 0, our complex process will absorb a positive amount of

heat at temperature To while surrendering no heat at all at tempera-

64
In other words, ©^ surrenders |Q*| at Tk if 3 absorbs |Q*| at Tk, and &k absorbs \Qk
at Tk if (5 surrenders \Qk\ at Tk. Therefore, &k absorbs (surrenders) {TQ/Tk)\Qk\ at To
if © absorbs (surrenders) |Q^| at Tk; this accounts for the identity of sign in both sides
of eqn. (4.22). For different values of k, the reversible engine &k will produce work
or consume it, as the circumstances require.
 4.3 Heat and Chance 193

tures Ti,. . . , Tn. So the only net effect of the process is to convert heat
Qo into work. Since Qo is obtained from a source at uniform temper-
ature, this violates Kelvin's postulate. Consequently, Q o < 0. Since To
> 0, inequality (4.22) follows. If the cycle performed by © is reversible,
our complex process can be run in the opposite sense, in which case
the heat quantities Q 1 ? . . . , Qn change signs. By the same argument
that has just led us to inequality (4.22) we conclude that, in this case,

Z-7p<0 (4.25)

So, if the process performed by © is reversible, relations (4.22) and

(4.25) must hold simultaneously. This can happen if and only if both
are equalities. Q. E. D.
We now change the conditions on ©. Instead of exchanging finite
amounts of heat with a finite number of bodies, © will exchange infin-
itesimal quantities of heat with a continuous spread of sources. Let dQ
denote the heat exchanged by © with a source at temperature T.65 In
the general case,

l^r^O (4.26)
the integral being taken over a complete cycle of ©. If the cycle is
reversible,
- An*
0 (4.27)

Equation (4.27) entails that for all reversible processes that take our
system from a state A to a state B, the integral \BAdQIT has the same
value (for if one such process follows the reverse of another, they jointly
complete a cycle, the integral over which is equal to 0), dependent only
on the initial and final states, not on the course of the process. There-
fore, given a conventional reference state O, we can define a property
of the thermal state A by
-A dO
-~ (4.28)

65
We take dQ to be positive if © absorbs heat from the source and negative if the
source absorbs heat from (5. Hence, if dQ > 0, the temperature of © is not greater
than T, nor is it smaller than T if dQ < 0. Therefore, if the process is reversible, the
temperature of (5 in each heat exchange must be equal to that of the relevant source.
194 The Rich Nineteenth Century

Clausius called S(A) the entropy of A. Evidently,

Consequently, for any state A, the entropy S'(A) defined as in eqn.

(4.28) with respect to another reference state O' differs from S(A) only
by an additive constant (equal to S(O')). This relativity in the concept
of entropy was overcome in the twentieth century by Walther Nernst,
who showed that the entropy of a system at 0 K is independent of every
macroscopic particularity of the system and so should be regarded as
a universal constant that one can put equal to 0.
The definition of entropy as a property of state in eqn. (4.28) pre-
supposes that the integral on the right-hand side depends only on O
and A. The integral must therefore be taken over a continuous succes-
sion of reversible heat exchanges. The same condition holds for the
integral J|dQ/T on the left-hand side of eqn. (4.29). But once entropy
has been defined as a property of thermodynamic systems, one may
well compare the entropy difference on the right-hand side with the
said integral taken over a continuum of arbitrary heat exchanges. Con-
sider a cycle formed by an arbitrary process from A to B combined
with a reversible one from B to A. Then, by eqn. (4.26), 0 > <JWdQ/T
= \ldQIT + J^dQ/T; hence, by eqn. (4.29), 0 > UQIT + S(A) - S(B),
so that, in the general case,

f^ (4.30)
If the process under consideration occurs in a completely isolated
system, dQ = 0; therefore, the entropy of the final state B is always
equal to or greater than that of the initial state A. Thus, in a closed
thermodynamic system the entropy can never decrease. This proposi-
tion, which we have derived from the postulates of Kelvin and Clau-
sius, also entails them and is therefore often employed to state the
Second Principle of Thermodynamics in a concise and pleasantly eso-
teric way.
Philosophers became interested in entropy because it was assumed
that the universe is a closed system possessing a definite energy and
entropy. Since the energy cannot change and the entropy can only
increase, the conclusion seemed inevitable that the universe would one
day reach the maximum entropy compatible with its energy contents,
after which it could no longer change in any way. Many mature
 4.3 Heat and Chance 195

thinkers, who had at most a few decades to live, were deeply impressed
- and depressed - by the anyway distant prospect of this so-called heat
death of the universe. The subject was debated in the context of the
mechanical conception of heat that we shall deal with in the next two
subsections, so I shall come back to it in §4.3.4.

4.3.3 Molecular Chances

In his paper "On the Kind of Motion We Call Heat" (1857), Clausius
combined the kinetic conception of heat with a molecular view of
matter to form what became known as the kinetic theory of gases.66
He assumes that every body consists of very small particles or mole-
cules^ roughly similar in size, which are bound to one another in solids
(firmly) and liquids (pliably) but which in the gaseous state move freely
in straight lines and rebound under the action of repulsive forces when
they run against other molecules. Clausius says that he was moved to
publish his ideas by Kronig's "Groundlines of a Theory of Gases"
(1856), which anticipated some of them. Both Kronig and Clausius
appeal to probability at some point in their arguments. Kronig con-
siders a gas consisting of equal, perfectly elastic spheres that move,
without interacting, inside a cubic container; each sphere travels per-
pendicularly to two of the six container walls and rebounds elastically
in the opposite direction when it collides with either of them. Kronig
admits that, at the scale of gas molecules, even the smoothest wall is
very rough.
The path of each gas atom must therefore be very irregular, so that it
eludes calculation. However, according to the laws of the probability cal-
culus one may assume, instead of this perfect irregularity, a perfect reg-
ularity.
(Kronig 1856, p. 316; quoted by Schneider 1988, p. 300; my
italics)
Clausius (1857) bestowed more freedom on his molecules but retained
the mutual independence of molecule motions, which entails, of course,

The theory can be traced back to Daniel Bernoulli's Hydrodynamica, Sect. 10 (1738).
Versions of it had been argued for, with little success, by Herapath (1821) and by
Waterston (1846). The name 'kinetic theory of gases' is not quite adequate, for a
theory that conceived gases as continuous fluids could - and surely would - be kinetic
too. 'Molecular-kinetic' is a better name.
196 The Rich Nineteenth Century

that they never meet each other.67 Regarding the collisions with the
container walls, Clausius notes that if the latter are perfectly smooth
and the bouncing is perfectly elastic, the speed and angle of incidence
are equal to the speed and angle of reflexion, but that this need not be
so in a more general case.
Still, by the rules of probability', one can assume that there are just as
many molecules whose angles of reflexion lie within a certain interval,
e.g. between 60° and 61°, as there are whose angles of incidence lie in
that interval, and also that, on the whole, the speed of the molecules is
not changed by the wall. Thus it will make no difference in the final
result if one assumes that for each molecule the angle and speed of reflex-
ion are equal to those of incidence.
(Clausius 1857, §15; my italics)
Neither Kronig nor Clausius - in 1857 - carried their use of prob-
ability any further, but in the paper of 1858 mentioned in note 67 Clau-
sius figured out the mean free path of a molecule by straightforward
probability calculations. A partial look at these calculations will help
us fix the sense in which the concept of probability made its first
appearance in thermal physics (and see whether it retained this same
sense later). Clausius bids us consider a space - which I shall denote
by if - containing a great many molecules at rest, distributed without
any regular order but with uniform density. Each molecule has a
"sphere of action" of radius p, inside which other molecules are
repelled. There is, on average, one molecule in each cube of volume ^ 3 .
Clausius calculates the probability W that a molecule of the same sort,
moving in a straight line through if, will travel a distance x undeterred
by the repulsive forces of the static molecules. The calculation consists
of two parts: (i) a formal calculation leading to the result

W = exp(-ox) (4.31)68

and (ii) a calculation of the value of a from the above physical assump-
tions. Part (i) was redone much more elegantly and concisely by

67
If this condition were even approximately realistic, gases would mix together much
faster than they actually do, and one would rarely get to see puffs of smoke. Reminded
of this fact by Buijs-Ballot (1858), Clausius (1858) made allowance for intermolecu-
lar collisions and calculated the mean free path of a gas molecule, i.e., the average
distance it travels between two such collisions. The figure he got was largely com-
patible with phenomena.
68
I write exp(x) for e? (e the basis of natural logarithms). I reserve e for the electron
charge.
 4.3 Heat and Chance 197

Maxwell (1860, Prop. X), whom I shall follow. Anyway, only part (ii)
involves Clausius's understanding of physical probabilities. Maxwell
writes ocdx for the probability of a particle striking other particles while
traveling among them through a distance dx. In other words, if N such
particles traveling through if independently - perhaps on separate
occasions - reach a distance x, Nadx of them would be stopped before
getting to the distance x + dx. In other words,
dN
— = -Na or N = Cexp(-ox) (4.32)
dx
"Putting N = 1 when x = 0, we find exp(-ocx) for the probability of a
particle not striking another before it reaches a distance x" (Maxwell
1890, vol. I, p. 386; my notation). This is precisely the relation
expressed by eqn. (4.31). To calculate the value of a we reason - after
Clausius - as follows. We divide the space if into slabs perpendicular
to the direction of motion of our moving molecule. Let an average slab
of thickness X contain n static molecules. The slab's area is therefore
nX2. The section of each molecule's sphere of action is 7ip2, so the total
area across which passage is blocked is nnp2. The probability that the
moving molecule will be stopped as it traverses a typical slab of thick-
ness X is the ratio between the blocked area and the total section of
the slab: np2/X2. To get the probability ocdx that the moving molecule
will strike a static molecule as it travels the distance dx, we multiply
this ratio by dx/X. So

^ d x (4.33)

whence, from eqn. (4.31)

W = exp(-ox) = exp(-xjcp2^3) (4.34)
Clausius's argument evidently treats the molecule-studded space if as
the three-dimensional field of a new-fangled pinball game, into which
the moving molecule is thrown in a particular direction at random, that
is, in such a way that it has precisely the same chance of taking any of
the parallel paths available in that direction. There is no essential dif-
ference between the derivation of eqn. (4.33) and the standard calcu-
lation of the probability of outcomes, say, in a game of roulette. Here
we divide the number of holes corresponding to the outcome in ques-
tion (e.g., Red, or Third dozen) by the total number of holes available.
The result obtained is correct if the ball has precisely the same chance
198 The Rich Nineteenth Century

of falling into any hole (which, one assumes, is the case if the initial
angular velocity of the wheel and the initial position and velocity of
the ball are picked at random from among all the viable alternatives).69
Of course, in Clausius's calculation one cannot compare the number
of blocked paths with the total number of paths because there are
uncountably many of them. So we take a measure over each set of
paths. This is naturally given by the area of a section of 6f, each point
of which is perpendicularly intersected by one of the paths in question.
We may therefore conclude that the concept of probability was first
introduced into thermal physics with the same sense it possessed on its
emergence in the late medieval and early modern discussions of games
of chance: It quantifies the comparative ease with which a definite phys-
ical chance setup produces this or that outcome of interest.70
To calculate the mean free path of a molecule, Clausius assumes that
the molecules in if are no longer static but move in straight lines in
every direction with the same constant speed. This assumption is still
too unrealistic and was dropped by Maxwell in his first paper on the
subject (1860), in which, among other things, he calculated the likeli-
est distribution of velocities among the molecules of a (streamlined)
gas. Prompted by Clausius's work, Maxwell sought here to "lay the
foundation" of the kinetic theory of gases "on strict mechanical prin-
ciples" by demonstrating "the laws of motion of an indefinite number
of small, hard, and perfectly elastic spheres acting on one another only
during impact" (Maxwell 1890, vol. I, p. 377).
If the properties of such a system of bodies are found to correspond to
those of gases, an important physical analogy will be established, which
may lead to more accurate knowledge of the properties of matter. If
experiments on gases are inconsistent with the hypothesis of these propo-
sitions, then our theory, though consistent with itself, is proved to be
incapable of explaining the phenomena of gases. In either case it is nec-
essary to follow out the consequences of the hypothesis.
(Maxwell 1890, vol. I, p. 378; my italics)

69
Suppose that the roulette game is governed by classical mechanics, so that any given
set of initial conditions I determines one and only one outcome O. We conceive I as
a point in the system's phase space S. If the roulette game is fair, there will be, within
every neighborhood of J in S, no matter how small, sets of initial conditions leading
to every possible outcome. Thus, the slightest variation in the conditions is apt to
radically change the outcome.
70
On chance setups see Hacking (1965), or my own presentation in Torretti (1990,
§4.3).
 4.3 Heat and Chance 199

Thus, for Maxwell, the aim of a physico-mathematical theory like the

one he proposes in this paper is heuristic, not dogmatic. If the theory
is successful, that is, if the logical consequences derived from its ground
assumptions roughly agree with the phenomena it is meant to explain,
it may be serviceable for scientific inquiry by raising questions, moti-
vating experiments and suggesting theoretical refinements whose con-
sequences agree even better with phenomena. It would be foolish,
however, to assert the hypothesis of a successful theory as a dogma
concerning the nature of things, for its agreement with phenomena can
only be approximate and there might always be another, very differ-
ent hypothesis that agrees with them equally well.
Our concern here is not with the physics or the mathematics of
Maxwell's theory of gases, but with the probability arguments that
Maxwell puts forward at some critical junctures. I presume that the
probability assumptions on which those arguments rest also fall within
the scope of the preceding remarks. The first such assumption (Prop. II)
concerns the motion after impact of two spheres that strike each other
while moving in opposite directions with velocities inversely as their
masses. Maxwell has just shown that each ball moves with the same
speed before and after impact, and that the directions before and after
impact lie in the same plane with the line of centers and make equal
angles with it (Prop. I). He goes on to ask for the probability that the
direction after impact lies between given limits. He answers as follows:
In order that a collision may take place, the line of motion of one of the
balls must pass the centre of the other at a distance less than the sum of
their radii; that is, it must pass through a circle whose centre is that of
the other ball, and radius (s) the sum of the radii of the balls. Within
this circle every position is equally probable, and therefore the proba-
bility of the distance from the centre being between r and r + dr is
2rs~2dr. Now let (() be the angle [. . .] between the original direction and
the direction after impact, then [the angle between the original direction
and the line of centres] = (j)/2, and r = ssin((|)/2), and the probability
becomes jsin §d§. The area of a spherical zone between the angles of
polar distance 0 and $ + dty is 2n sintydty;therefore if co be any small area
on the surface of a sphere, radius unity, the probability of the direction
of rebound passing through this area is CO/4TC; SO that the probability is
independent of (|), that is, all directions of rebound are equally likely.
(Maxwell 1890, vol. I, p. 379; my italics)
The assumption I have italicized comes out of the blue. There is no
thought here of a physical process that might be regarded as a chance
200 The Rich Nineteenth Century

setup apt to yield with the same probability every position within the
said circle. What we have is a general concept - viz., 'to be found within
the said circle' - coupled with a probability distribution over the set of
its alternative specifications. In contrast with earlier uses of the concept
of probability, this distribution does not express the relative ease with
which one or the other alternative is produced by a definite physical
process whose outcomes are instances of the general concept. I cannot
see that such a distribution has any meaning apart from the statistics
by which it would be confirmed.
To establish a semantic connection between probability statements
and statistical data was surely the main purpose of the interpretation
of probability in terms of relative frequency introduced in the 1840s
by several authors, including the Cambridge mathematician R. L. Ellis
(1849). In its more sophisticated versions, the probability of finding a
feature A among objects of a class B is defined as the limit of the rel-
ative frequency of A's in a random sequence of B's.71 Although the
probability calculus was originally developed around the concept of
relative ease of occurrence, it admits this interpretation because some
of its theorems provide a close link between probability and relative
frequency. Thus, the so-called Strong Law of Large Numbers implies
that if (i) B is the class of outcomes of a chance setup, (ii) p is the prob-
ability of obtaining an A in a single trial, and (iii) P is the probability
that the relative frequency of A's in an infinite sequence of independent
trials72 converges to the limit /?, then P = I.73 This is, of course, triv-
ially true if probability is defined, as above, by the frequency limit.
Note, however, that the tie between frequency and probability that
flows from the probability calculus is itself probabilistic, not factual.

By 'random sequence' I mean an infinite sequence such that, given any positive integer
n, the list of the first n terms in the sequence provides no hint whatsoever concern-
ing the (n + l)-th term.
Two or more events are said to be (statistically) independent if the probability of each
is in no way affected by the occurrence or nonoccurrence of the others. Thus, a
sequence of independent trials on a chance setup is always random in the sense of
note 71. However, when we resort to the concept of a random sequence in the defi-
nition of probability we cannot use independence as a criterion of randomness, for,
as we have just seen, independence is itself defined in terms of probability.
Compare the so-called Weak Law of Large Numbers: Take B and p as above, and let
Pk denote the probability that the proportion of A's falls within Ilk of p in a run of
k independent trials: Then the sequence Pu . .. , Pn,... converges to the limit 1 as n
increases beyond all bounds (Jakob Bernoulli 1713, pp. 225f.).
 4.3 Heat and Chance 201

Moreover, by the very nature of infinite sequences, any frequency limit

is compatible with every finite set of statistics (for let a be an infinite
sequence of B's in which the relative frequency of A's converges to the
limit p; let x be a finite list of JB'S - no matter how long - in which the
relative frequency of A's is q ^ p; let a* denote the sequence formed
by x followed by a; then the relative frequency of A's also converges
to p i n a*).74
Maxwell resorts again to probabilistic ideas in the derivation of the
distribution of velocities among the molecules in his gas model. The
task is "to find the average number of particles whose velocities lie
between given limits, after a great number of collisions among a great
number of equal particles" (1860, Prop. IV). Maxwell analyzes the
velocity v of a typical particle into three mutually perpendicular com-
ponents that I shall denote by vx, vy, and vz.75 Following him, I write
Nf(Vi)dvt for the number of particles whose zth velocity component lies
between v{ and vt + dvt (i = x, y, z)9 where N is the total number of
particles and f is a function to be determined (on U). Maxwell says that
each velocity component does not in any way affect the other two,
"since these are all at right angles to each other and independent"
(1890, vol. I, p. 380). So, he concludes, the number of particles whose
velocity lies (i) between vx and vx + dvx, and also (ii) between vy and
vy + dvy, and also (iii) between vz and vz + dvz, is
Nf{vx)f{vy)f{vz)dvxdvydvz (4.35)
To understand the argument leading to eqn. (4.35) we should note first
that if Nf(Vi)dvt has the above meaning, then f(vx)dvx, f(vy)dvy and

74
According to the great frequentist statistician, Richard von Mises, this obvious fact
may be safely ignored because the hypothetical random sequences into which statis-
tical data are embedded are built on the "silent assumption" that "in certain known
fields of application of probability theory (games of chance, physics, biology, insur-
ance, etc.) the frequency limits are approached comparatively rapidly (the rate of
approach being different for different problems)" (1964, p. 108). Therefore, the
assigned probabilities should lie close to the relative frequencies observed in long,
though finite, series of observations. "This assumption has nothing to do with the
axioms of the probability calculus [. . .] and is not explained by any results of theo-
retical statistics," for the latter, in fact, depend upon it (Ibid.). However, "the whole
body of our experience in applications of probability theory seems to prove that rapid
convergence indeed prevails - at least in such domains - as a physical fact, confirmed
by an enormous number of observations. [. . .] In such domains, and only in them,
statistics can be used as a tool of research" (pp. 109, 110).
75
Maxwell's notation is x for vx, y for vy9 and z for vz.
202 The Rich Nineteenth Century

f(vz)dvz stand, respectively, for the relative frequency of the particles

meeting conditions (i), (ii), and (iii) among N particles. If conditions
(i), (ii) and (iii) are statistically independent (in the sense explained in
note 72), the probability that one particle satisfies them all is calcu-
lated by multiplying the probability assigned to one condition by the
probabilities assigned to the others. So, if one equates probability with
relative frequency, as one can do with little risk of error if N is very
large, f(vx)f(vy)f(vz)dvxdvydvz is the probability that a single particle
meets conditions (i), (ii), and (iii) and therefore the relative frequency
of such particles among the total N. Result (4.35) follows at once, and
the function f is readily determined from it.76
Maxwell (1866) dealt again with the collective behavior of gas mol-
ecules, considered this time "not as elastic spheres of definite radius,
but as small bodies or groups of smaller molecules repelling one
another with a force whose direction always passes very nearly through
the centres of gravity of the molecules, and whose magnitude is repre-
sented very nearly by some function of the distance of the centres of
gravity" (1890, vol. II, p. 29). This change of model was motivated,
he says, by his experiments on the viscosity of air at different temper-
atures. When tackling the distribution of velocities he did not assume
the mutual independence of a particle's orthogonal velocity compo-
nents for "this assumption may appear precarious" (p. 43). So he
derived "the Final Distribution of Velocity among the Molecules of
Two Systems acting on one another according to any Law of Force"
from a consideration of velocity changes in molecular collisions. Mol-

76
Maxwell reasons thus: If we suppose the N particles to start from the origin at the
same instant, eqn. (4.34) is the number in the element of volume dvxdvydvz after unit
time, so the number referred to unit volume is Nf(vx)f{vy)f(vz). "But the directions
of the coordinates are perfectly arbitrary, and therefore this number must depend on
the distance from the origin alone" (1890, vol. I, p. 381), that is, f(vx)f(vy)f(vz) =
(p(vx2 + vy2 + vz2). Solving this functional equation, we find
f(vx) = Cexp(Av2x) and cp(r2) = C2 exp(Ar2)
If A were positive, the number of particles would increase with the velocity, and their
total number would be infinite. So we make A negative and equal to — OC2. Then the
number of particles whose first velocity component lies between vx and vx + dvx is
NC exp (-vx2cf2)dvx. Integrating over 1R we find the total number of particles N =
NCcrJn; so C = (cn/ft)"1. Therefore
 4.3 Heat and Chance 203

ecule velocities are represented "in direction and magnitude" by lines

drawn from a common origin O (i.e., by vectors). "The extremities of
these lines will be distributed over space in such a way that if an
element of volume dV be taken anywhere, the number of such lines
which will terminate within dV will be f(r)dV, where r is the distance
of dV from O" (p. 43). Let dV(X) stand for all the velocities whose
representative lines terminate inside a sphere of volume d V centered at
X. Consider the set °€ of pairwise colliding molecules such that the
initial velocity of a molecule in each pair lies in dV(A) and its final
velocity lies in d V(A'), while the initial and final velocities of the other
molecule lie, respectively, in dV(B) and in dV(B'). Let <€* be defined by
reversing all the velocities of the molecules in %. Thus <€* is the set of
pairwise colliding molecules whose initial velocities lie in dV(A') and
in dV(B')? while the final velocities lie, respectively, in dV(A) and in
d V(B). When the number of molecules in ^ equals the number of mol-
ecules in c€*, "then the final distribution of velocity will be obtained,
which will not be altered by subsequent exchanges". This remark yields
at once "a possible form of the final distribution of velocities" (p. 45).77
According to Maxwell, it is the only form, for if there were another,
the exchange between velocities represented by dV(A) and dV(A')
would not be equal. Thus, Maxwell's result rests on the unspoken
proposition that in a molecule population governed by the laws of clas-
sical mechanics any particular configuration % of colliding pairs is, in
the long run, exactly as probable as the configuration defined by revers-
ing the velocity of each molecule in C6. The final distribution will, of
course, be reached only after a great number of collisions, that is, in a
well-shuffled gas. However, "the great rapidity with which the encoun-
ters succeed each other is such that in all motions and changes of the
gaseous system except the most violent, the form of the distribution of
velocity is only slightly changed" (p. 46).
Another way of understanding and using probability in physics,

Let a lowercase a (b, etc.) stand for the distance from O to A (B, etc.) and therefore
also for the magnitude of the vector OA (OB, etc.). Sf and Sf* contain the same
number of molecules if and only if f(a)f(b) = f(a')f(b'), where / is the function defined
in the above quotation from p. 43. Maxwell assumes that all molecules in Sf n dV(A)
have the same mass Mi and that all molecules in Sf n dV(B) have mass M2. Energy
conservation entails that M^2 + M2b2 = M^d1 + M2b2. So f(a) = Ciexp(-<32/oc2) and
f(t>) = C2exp(-b2/p2), where M ^ 2 = M2p2. The constants Q and C2 are readily com-
puted as in 1860 (see note 76).
204 The Rich Nineteenth Century

which is central in the work of Gibbs (1902) and Einstein (1902, 1903,
1904), was adumbrated in Maxwell's derivation of the mean free path
(1860, Prop. X). I shall explain it briefly in its mature form and then
return to Maxwell's argument. We consider a classical mechanical
system formed by N dimensionless particles, so that its mechanical state
is fully characterized by three position and three momentum coordi-
nates. Obviously, if N is very large, one can never know the 6N coor-
dinates that specify the state of the system. Typically, one can only
know a few of its properties, which are then compatible with a large
family of alternative states, corresponding to many different sets of
additional properties. So we represent every conceivable state of our
system by a different point in a 6N-dimensional Euclidian space, the
system's phase space (§2.5.3, after eqn. (2.37)). The observable or
macroscopic state of the system, specified by the few properties we
know, is then represented by a proper part of the phase space, formed
by all the points whose 6N coordinates are compatible with that
macroscopic state. The classical physicist assumes, of course, that the
actual - microphysical - state of the system is represented by a single
point in phase space. However, he does not know which one it is. He
is therefore content to study an ensemble (Gibbs's term) of possible
systems, any one of which might be the real one, for they all share with
it the said macroscopic state. This enables him to estimate the proba-
bility that the system under consideration possesses this or that hidden
property of interest and will therefore evolve in this or that way. To
do so he assumes that the real system has an equal chance of being any
of the possible systems and calculates what fraction of the possible
systems shares the property of interest. This fraction is the probability
that the real system has the said property. Probability is thus equated
with relative frequency but, mark you, with frequency in an imaginary
ensemble, not among actual things or events. Just as Galileo and Pascal
calculated the probability of making a certain number with a pair of
dice by counting all the alternative equipossible dice throws and com-
puting what fraction of them yields that number, so the statistical physi-
cist estimates the probability that a gas will take a particular course by
measuring the set of points representing its alternative equipossible
microphysical states and ascertaining what fraction of the set repre-
sents states that put the gas on that course. In either case probability
is quantified possibility. There is one big difference, however, that dis-
sociates Gibbs and Einstein from Galileo and Pascal: In the ensemble
approach probability is in no way tied to a physical chance setup of
 4.3 Heat and Chance 205

which the actual dynamical state of the real system is an outcome. This
agrees, of course, with what I said earlier about Maxwell's handling of
the direction in which a molecule moves after striking another one
(1860, Prop. II). It is fairly well known that, 30 years before Gibbs and
Einstein, the ensemble approach had been used by Boltzmann (1871a,
1872 in WA, I, 261f., 401f.). I submit that the idea was already at work
in Maxwell's argument for eqns. (4.32): The N particles that are sup-
posed to travel independently through the given space if do not con-
stitute a physical system - for what would then prevent them from
occasionally interacting with one another? - but an ensemble of uni-
molecular systems.

4.3.4 Time-Reversible Laws for Time-Directed Phenomena?

Maxwell's molecular-kinetic theory of gases was extended and per-
fected by Ludwig Boltzmann, who made the subject his life work. The
theory successfully accounted for many phenomena, at least approxi-
mately, but often gave the wrong numbers for specific heats. It won
acceptance in the British Isles but was resisted on the continent by Ernst
Mach and other prominent physicists (including Max Planck). They
criticized it on methodological grounds, as a purely speculative hypoth-
esis that was not justified by experience and was unnecessary for the
advancement of science. But they also adduced the theory's wrong pre-
dictions78 and its conceptual difficulties. Among the latter there is one
of considerable philosophical interest that I shall now discuss.
To see the difficulty one must bear in mind that the equations of

78
The molecular-kinetic hypothesis, combined with Newtonian mechanics, also entails
the absurd Rayleigh-Jeans formula for black-body radiation, according to which the
total energy of a hot body that absorbs and emits radiant energy in all frequencies is
given by

—-— \) d\) = oo
c Jo
where x> stands for the frequency, T stands for the temperature, k is Boltzmann's con-
stant, and c is the speed of light. Einstein, who independently discovered this formula
(1905i), cited its inconsistency as evidence for the emergent quantum theory. That
same year Einstein also gave a molecular-kinetic explanation of Brownian motion
(1905k) that was subsequently confirmed by Perrin's admirable experimental work
and finally persuaded the Continental scientific establishment of the existence of
atoms and molecules (see Perrin 1913; for a brilliant philosophical analysis of Perrin's
statistical arguments, see Mayo 1996, pp. 220-50).
206 The Rich Nineteenth Century

classical mechanics for a conservative dynamical system are invariant

under time reversal In other words, if you replace, say, in the Lagrange
equations (2.33) every occurrence of the time variable t with -t, the
changes cancel out in each term, so you end up with exactly the same
set of equations.79 Therefore, for every solution of the Lagrange equa-
tions there is another solution like the former in which -t is substituted
for t. In a trivial sense, such a substitution can be taken as a mere rela-
beling of successive instants; as time goes by, the time coordinate
decreases instead of growing. But there is another, physically more sig-
nificant interpretation: If we read the time coordinate in the standard
way (as monotonically increasing with the lapse of time), the substitu-
tion of -t for t amounts to a general reversal of velocities. Consider a
mechanical system if moving according to the Lagrange equations
during the interval (-1,1). Let q(t) = (qi(t),. . . , qn(t)) be the list of gen-
eralized coordinates representing £f at a particular time t e ( - l , l ) . The
mapping q:(-l,l) —> Un by t *-* q(t) is a curve in the configuration space
of if that traces the evolution of if during (-1,1). The generalized veloc-
ities of if at t are represented by a vector in the tangent space at q(£),
viz., q(t) = (qu . . . , qn). The pair (q(t), q(t)) fully represents the state
of if at t. Consider now the mapping r:(-l,l) —> R" by t >-> r(t) =
q(—t). r traces the evolution of a system if* whose state at a particular
time t e ( - l , l ) is represented by (r(f), r(*)> = (q(-t)9 -q(-t)). if'' plays
back, so to speak, in reverse time order the evolution of if. Time-
reversal in variance entails that the motion of if* also satisfies the
Lagrange equations.
The molecular-kinetic theory purportedly accounts for the macro-
scopic behavior of gases by conceiving it as the manifestation of mol-
ecular motions governed by the Lagrange equations. But familiar
instances of gas behavior are not matched in nature by similar series
of events in reverse time order. Thus, the pressurized gas in a punc-
tured balloon spurts out of it until the inside and outside pressures are

79
The Lagrangian L equals T - V. The potential V depends on the positions but not
on t, and the kinetic energy T is proportional to the squared velocities and therefore
retains its value when t is multiplied by - 1 . So time reversal does not affect the second
term dL/dq, in eqns. (2.33). Nor does it affect the first, d/dt(dL/dqj), for, as I explain
in the text, if -t is substituted for t, cjj must be replaced with -q, and, plainly,

dfdL^ d ( dL
 4.3 Heat and Chance 207

equalized; but we have yet to see air enter of itself into a flat tire until
it reaches the pressure prescribed by the manufacturer. If you open the
door connecting a cold room with a warm one, the air in both will
mix; but the air in a room at even temperature will not separate spon-
taneously into a hot part that goes to one corner and a cold one that
goes to another. The notorious time-directedness of thermal phenom-
ena finds expression in Fourier's equation of heat conduction, which is
not invariant under time reversal, and, of course, in the Second Prin-
ciple of Thermodynamics as stated in eqn. (4.30). The molecular-
kinetic theory therefore had to face the seemingly impossible task of
deriving such laws, and the irreversible natural processes that suggested
them, from the time-reversal invariant laws of mechanics.
Boltzmann tackled this task head on. In 1872 he introduced a quan-
tity, which (in 1896)80 he designated by H and defined by
H= f flog/dco
•'CO
(4.36)

Here (i) log is the natural logarithm function; (ii) /dco abbreviates
f(vx,vy,vz,t)dvxdvydvz, which stands for the number of particles whose
velocity components relative to the Cartesian coordinates x, y, and z
lie, at time t, between vx and vx + dvx, between vy and vy + dvy, and
between vz and vz + dfz, respectively; and (iii) it is assumed that the
velocity cell dco "can be infinitesimal yet still contain many molecules"
(Boltzmann 1896a, §15; LGT, p. 111). Boltzmann showed that in a
closed system dHldt < 0, with equality holding if and only if /"happens
to be Boltzmann's improved version of Maxwell's velocity distribution

In 1872 Boltzmann defined the quantity of interest in terms of the distribution of

kinetic energies among the molecules of the gas and denoted it by £ (for entropy).
The original definition, which can be found in Boltzmann (1872, eqn. 17, in WA, I,
p. 335), reads as follows:

Here f does not stand for the velocity distribution function as in eqn. (4.35), but it
is defined thus: Let r be a region of the space occupied by the gas, of arbitrary shape
and unit volume, large enough to contain many molecules; f(x,t)dx denotes the
number of molecules in r whose kinetic energy at time t lies between x and x + dx
(WA, I, 321). The letter H, introduced by Burbury (1894), is in all likelihood not the
eighth letter of the English alphabet, but a Greek eta (Gibbs used a lowercase eta as
a symbol for entropy).
208 The Rich Nineteenth Century

function. In other words, H can only decrease, unless the distribution

of velocities has reached its final equilibrium form, in which case H is
constant. This proposition, known as Boltzmann's H-Theorem, evokes
at once the law (4.30), by which entropy can only increase, except in
the limiting case of reversible processes, in which it remains unchanged.
And indeed Boltzmann shows that in all cases in which the entropy can
be defined as in §4.3.2, it turns out to be, in the molecular-kinetic inter-
pretation, equal to -H (up to a positive constant factor).81
Boltzmann (1872) openly suggested that his H-Theorem follows
from the laws of classical mechanics.82 This would imply that these
laws, though invariant under time reversal, impose a time-directed evo-
lution on large molecular populations. The absurdity of this implica-
tion was perceived by Kelvin (1874), and also by Loschmidt (1876),
himself a noted contributor to the molecular-kinetic theory.83 They
made the point that I explained above, namely, that for every dynam-
ically possible evolution % of a classical system of molecules in motion,
the reverse evolution %* is equally possible. Specifically, suppose that
% comprises a continuum of states, a chronologically ordered finite
subset of which we number from 0 to 2m (m > 0). Consider now the
dynamical evolution % * of a copy of the former system, such that in a

81
Of course H also has a meaning in situations for which the Clausius entropy is not
defined. Boltzmann therefore boasted that he had generalized the entropy principle
"in that we have been able to define the entropy in a gas that is not in a stationary
state" (1896, §8; in Boltzmann LGT, p. 75).
82
After deriving the form of f (in the sense of note 77) at the terminal stage of evolu-
tion, Boltzmann writes: "Thus it is rigorously proved that, no matter what the dis-
tribution of kinetic energy at the beginning of time, after a very long time it must
always necessarily approach the distribution discovered by Maxwell" (1872 in WA,
I, 345). Cf. the title of Boltzmann (1871b): "Analytical proof of the Second Princi-
ple of the Mechanical Theory of Heat from the Theorems on the Equilibrium of
Kinetic Energy" (WA, I, 288). On the other hand, the introductory section of Boltz-
mann (1872) stresses the fundamental role of probability in the molecular-kinetic
theory and warns the reader against confusing "an incompletely proven theorem,
the correctness of which is therefore problematic, with a completely demonstrated
theorem of the probability calculus: the latter - like the result of any other calculus
- is a necessary consequence of certain premises, and if these premises are correct, it
is confirmed by experience if sufficiently numerous cases are subject to observation,
as is always the case in the theory of heat, given the enormous number of molecules
in a body" (WA, I, 317).
83
Indeed Boltzmann (1905, p. 236) gave his friend and senior colleague Loschmidt
credit for setting him on the path of molecular-kinetic research.
 4.3 Heat and Chance 209

definite halfway state m* the molecules hold the same relative posi-
tions as in state m o f ? , but have their respective velocities reversed.
Since %* before and after m* merely repeats % in the reverse time order,
it must successively go through 2m + 1 states, which we shall label 0*
to 2ra *, such that - if the integer k ranges from -m to m - the mole-
cules hold the same relative positions in state (m - k)* as in state (m
+ k) of %, but have their respective velocities reversed. Evidently, the
distribution function /", which depends on the squared velocities (cf.
note 76), is the same at(m + k) and at(m-k)*; so, if H, and H{* denote
the value of H for either system in the ith and /*th states, respectively,
it is clear that

Him+k) = H{m_k)* (-m<k< m) (4.37)

Therefore, if in agreement with Boltzmann's H-Theorem the values of
H for the said stages of % satisfy the inequality
H0>H2>...>Hw>Hw+1>...>H2w (4.38)
the values of H for the said stages of % * must satisfy the inequality
Ho* < H2* < . . . < Hm* < H(w+1)* < . . . < H2w* (4.39)

This clashes openly with the H-Theorem, except in situations in which

the Maxwell-Boltzmann velocity distribution holds (so H is constant).
Countering Loschmidt's criticism Boltzmann stressed that the H-
Theorem does not follow from the laws of mechanics alone, but from
these laws together with the probability calculus and suitable probabil-
ity assumptions. He formulated those assumptions in several not alto-
gether clear ways. Applied to the above example the gist of them may
be stated thus: In the arbitrarily chosen dynamical evolution % the initial
molecular positions and velocities should be totally uncorrelated. Since
% * is expressly defined in terms of a set of previously specified positions
and velocities, that are reached in % after numerous collisions, the initial
positions and velocities in % * are not uncorrelated, so it is no wonder
that the H-Theorem does not apply to it. Therefore, the H-Theorem is
not meant to be universally true, but only overwhelmingly probable. The
same holds, according to the molecular-kinetic theory, for the Second
Principle of Thermodynamics. In this theory it is not impossible that gin
and tonic spontaneously separate after being thoroughly mixed in a
glass, but only extremely improbable - more so, indeed, than that
every inhabitant of a large country commits suicide, purely by accident,
on the same day, or that every building burns down at the same
210 The Rich Nineteenth Century

time.84 Still, Loschmidt's objection should also be understood statisti-

cally. It is not a matter of choosing at random a typical evolution % and
then, after it is picked, reversing the molecular velocities in one of its
states. Think rather of the ensemble of dynamically possible evolutions
of any given collection of molecules. For every member % of the ensem-
ble in which H never increases, there must be a matching member % * in
which H never decreases. Boltzmann's mature reading of the H-
Theorem, developed in answer to criticism by Zermelo, implicitly vin-
dicates Loschmidt's objection in this ensemble form.
To fight the molecular-kinetic theory, Planck's assistant, Ernst
Zermelo, recalled that, by a theorem proved by Poincare, every closed
mechanical system must eventually return to a state as similar as you
wish to its present state.85 "Therefore, in such a system irreversible
processes are impossible, for (apart from singular initial states) no
single-valued continuous function of the state variables, such as
entropy, can increase continuously; if there is a finite increase, there
must be a corresponding decrease when the initial state recurs"
(Zermelo 1896, p. 485). Boltzmann (1896b, 1897) did not deny
Zermelo's claim, but was content to show that the (microphysical)
states of a gas in which H is at or very near its absolute minimum are
overwhelmingly more probable than those in which it is far above it.
From this he concluded that:
(I) the gas spends most of the time in or near the equilibrium condi-
tion characterized by the Maxwell-Boltzmann velocity distribu-
tion, in which H is as small as it can be and most likely to remain
constant, and
(II) if the gas departs from that condition by any chance, then, at any
particular moment in which H is not minimal, H will most prob-
ably decrease, indeed all the more probably the greater the depar-
ture from minimality.86
84
The last two examples are given by Boltzmann, who wryly remarks: "Yet the insur-
ance companies get along quite well by ignoring the possibility of such events" (LGT,
p. 444).
85
Represent the different states of the system by points in its phase space. Let P repre-
sent the system's initial state. Take any neighborhood N(P) of P. Then, by Poincare's
theorem, the system's trajectory in phase space will enter N(P) again and again in the
course of everlasting time.
86
Paul and Tatiana Ehrenfest (1906) constructed a beautiful example of a function of
time that is always more likely to decrease than to increase - unless it stands at its
absolute minimum - although it depends on a process that is invariant under time
reversal.
 4.3 Heat and Chance 211

Thus, the long-term evolution of a closed molecular system certainly

includes periods of increasing H in which the Second Principle of Ther-
modynamics is transgressed; but most transgressions are small and
short-lived, so it is no wonder that we do not usually perceive them.87
Still, they can be seen under the microscope in the guise of Brownian
motion, the ceaseless agitation of particles suspended in fluids, about
one micron in diameter, which was first accurately described by the
botanist Robert Brown (1828), and which Einstein (1905k, 1906b)
successfully explained as a consequence of the frequent collisions of
such particles with the much smaller molecules that surround them.88
The Brownian particles perform uninterrupted mechanical work at the
expense of heat extracted from the fluid even if the latter is at the same
temperature throughout. As Perrin eloquently puts it:

We have only to follow, in water in thermal equilibrium, a particle denser

than water, to notice that at certain instants it rises spontaneously, thus
transforming a part of the heat of the medium into work. If we were no
bigger than bacteria, we should be able at such moments to fix the dust
particle at the level reached in this way, without going to the trouble of
lifting it and to build a house, for instance, without having to pay for
the raising of the materials.
(Perrin, Atoms, p. 87)

87
Of course, if the universe is a Boltzmann molecular-kinetic system spending most of
its time near equilibrium, it must have gone through an enormous entropy decrease
when our sun and the other stars were formed. Boltzmann's theory does not preclude
such a stroke of luck, without which, in fact, Boltzmann himself could not have been
born. Moreover, if the universe were infinite and eternal — as Boltzmann apparently
thought - , the occurrence at some time and place of a low-entropy bubble, less than
a trillion light-years in diameter, and returning to heat death in a few dozen billion
years, would be practically certain. Cf. the long quotation from Boltzmann (1898,
§90) given in the main text, near the end of this section.
88
A molecular-kinetic explanation of Brownian motion had been suggested by Gouy
(1888), among others. But it ran against the following difficulty: the observed speed
of Brownian particles was very much less than the speed calculated from the molec-
ular-kinetic theory. It took Einstein's peerless imagination to realize that we can at
no time properly observe the momentary velocity of a particle that moves to and fro
propelled by some 1022 collisions per second, and that the particle's trajectory from
one second to the next is a polygonal line that is very much richer in turnabouts and
therefore very much longer than what it looks to us. Therefore, the particle's true
speed is a good deal greater than what it seems. To come anywhere near perceiving
the Brownian motion's actual complexity we would need, besides the microscope that
enables us to discern particles about 10~18 cubic meters in size, a device enabling us
to discern events lasting, say, less than 10~18 second.
212 The Rich Nineteenth Century

In reaching conclusion (II), Boltzmann took for granted that a phys-

ical system is more likely to evolve from a less probable to a more prob-
able state than vice versa. To infer conclusion (I) he had to assume that
the time average of a physical quantity throughout the evolution of a
closed physical system equals its so-called phase average, that is, its
present average over an ensemble of similar closed systems. Both
assumptions are questionable, but this is not the place to discuss
them.89 Before changing the subject I must, however, refer to a remark
by Boltzmann that, although irrelevant to his physics and unworthy of
his genius, has had a lasting effect on lesser minds. It is contained in
one of the final sections of his Lectures on Gas Theory, entitled "Appli-
cation to the universe". I reproduce it in context:

One can think of the world as a mechanical system of an enormously

large number of constituents, and of an immensely long period of time,
so that the dimensions of that part containing our own "fixed stars" are
minute compared to the extension of the universe; and times that we call
eons are likewise minute compared to such a period. Then in the uni-
verse, which is in thermal equilibrium throughout and therefore dead,
there will occur here and there relatively small regions of the same size
as our galaxy (we call them single worlds) which, during the relative
short time of eons, fluctuate noticeably from thermal equilibrium, and
indeed the state probability in such cases will be equally likely to increase
or decrease. For the universe, the two directions of time are indistin-
guishable, just as in space there is no up or down. However, just as at a
particular place on the earth's surface we call "down" the direction
toward the center of the earth, so will a living being in a particular time
interval of such a single world distinguish the direction of time toward
the less probable state from the opposite direction (the former toward
the past, the latter toward the future). By virtue of this terminology, such
small isolated regions of the universe will always find themselves "ini-
tially" in an improbable state. This method seems to me the only way
in which one can understand the second law - the heat death of each
single world - without a unidirectional change of the entire universe
from a definite initial state to a final state.
(Boltzmann 1898, §90; LGT, p. 447; my italics)
The whole passage is interesting for the light it throws on Boltz-
mann's final reading of his H-Theorem. Clearly, Loschmidt's objection

89
The reader who wishes to pursue the fascinating subject of heat and chance will find
in Sklar (1993) a trustworthy and illuminating guide, with abundant references to
the scientific and philosophical literature.
 4.3 Heat and Chance 213

has been fully assimilated. I have set the remark in question in italics.
By virtue of it, Boltzmann's controversial probabilistic assumption
about the transition from states of lesser to states of greater probabil-
ity is upgraded to a truth of grammar. No scientist could ever observe
a global entropy decrease in his world region, because advance toward
a more probable state, that is, one of maximal entropy, is by definition
the mark of the time direction from past to future. Boltzmann's maneu-
ver is of a type that - although common in philosophy - is quite
unusual in science. He needs, of course, to retain the everyday meaning
of 'past' and 'future' and of the temporal direction from one to the
other, for otherwise his gambit would be no more than wordplay. Yet
he produces a definition of those terms that is not warranted by ordi-
nary usage. I have just seen two swallows fly past my window, head
forward, from left to right. In perceiving their flight in this time order
I took no cognizance of the global thermal state of my "single world"
- extending to the furthest quasars - , let alone of its progress toward
a more probable condition. Would I see the birds' heads trail their
bodies if I lived in a "single world" in which entropy is decreasing? To
be specific, imagine that the system of observable galaxies, which is
currently expanding, begins to recontract. (Such a turnabout would be
bound to occur if - as seems likely - General Relativity is approxi-
mately right and - contrary to current data - the average density of
nongravitational energy were greater than the so-called critical density.)
If the concept of entropy is applicable to the system of galaxies, their
coming together would involve a sustained, very large entropy
decrease. Would this affect our perception of temporal order? Would
we see birds fly backwards? Or would birds living in a contracting
"single world" fly tail first, so that people in that "single world" - who
identify the future with the time direction of increasing entropy - see
them fly in the same direction as we do? What about the galaxies?
Shouldn't they too look as if they moved backwards, so that even after
they are closing in on us they still seem to move away from us? If Boltz-
mann's criterion of time order makes any sense, "the direction of time"
depends on the global evolution of entropy over our entire region of
the universe,90 and it must therefore be reversed in our minds as soon
as the distant galaxies begin to approach us. Yet at that moment the

90
If the sense of time direction depended on local entropy change, then, by Boltzmann's
criterion, the observer who follows the Brownian motion of a particle would expe-
rience the future direction toggle back and forth.
214 The Rich Nineteenth Century

light signals we are receiving from them are still old signals, issued
many millions of years earlier, while their distance from us was
growing. Will that light still appear redshifted after the sources turn
around? Or will it appear blueshifted? And how will the new light
appear when it reaches us from the nearest, now incoming galaxies?
Does the change in the direction of time change the way the galactic
spectra are shifted? Or do scientists who live in a "single world" in
which entropy is decreasing, prompted by their particular experience
of time order, replace the Doppler shift formula with one in which
blueshift indicates recession? Such questions are silly, indeed, but no
more so than the remark that motivates them.91

Appendix
A Carnot engine can be built from the following pieces: a piston P in
a cylinder that contains the working substance G; a stand S; and two
bodies Bx and B2, at temperatures 7\ and T2, respectively, with 7\ > T2
(see Fig. 11). It is assumed that the piston, the cylinder walls, and the
stand are perfect nonconductors of heat, but that the bottom of the
cylinder is a perfect conductor whose capacity for storing heat is
negligible. On the other hand, B2 and Bx have such a large heat capac-
ity that their temperature does not vary significantly when they
exchange heat with G in the manner described below. A cycle of the
Carnot engine consists of four stages:
I. With G at temperature T2, the cylinder is placed on the stand S.
No heat can go in or out. The piston P is forced down very slowly
until the temperature of G rises to TV Let Wi denote the amount
of work done by P on G in this stage.
II. With G at temperature Tu the cylinder is placed on Bu which, as
we may recall, is at that same temperature. The piston P is allowed
to rise. Since the bottom of the cylinder is a perfect conductor, heat
from Bx flows into G and prevents it from being cooled by the
expansion. Temperature remains constant. Let Q1 be the quantity
of heat that G absorbs from B\ and Wn the amount of work done
by G on P in this stage.

91
It may be objected that such questions only look silly to me because I do not take an
Archimedean standpoint outside time from which to view them (cf. Price 1996). Such
a feat, however, is beyond my capacity. Indeed, in my irrevocably timebound situa-
tion, the very idea of standing outside time only compounds the silliness.
 4.4 Philosophers 215

n
Figure 11

III. With G still at temperature Tl5 the cylinder is once again placed
on S. P is allowed to rise. Since no heat can go in or out of the
cylinder, the temperature of G falls as it expands. The process is
stopped when the falling temperature is back at T2. Let Wm be the
amount of work done by G on P in this stage.
IV. With G at temperature T2, the cylinder is placed on B2, which, as
we may recall, is at that same temperature. The piston P is forced
down very slowly until it returns to the position it had at the begin-
ning of stage I. This time, compression does not cause a rise in tem-
perature, for the bottom of the cylinder is a perfect conductor and
the extra heat flows to B2 while the temperature remains constant.
Let Q2 be the quantity of heat that G surrenders to B2 and WiV the
amount of work done by P on G in this stage.
At the end of stage IV the working substance G has recovered the
temperature and the volume, and thus also the pressure, it had at the
beginning of stage I and the engine is ready to begin a new cycle. The
net amount of work extracted during one cycle is W = - Wi + Wn + Wm

4.4 Philosophers
Nineteenth-century philosophy is overwhelming in variety, density, and
volume. Much of it concerns the natural sciences, which were increas-
ingly regarded as the most rewarding form of intellectual endeavor.
I can deal here only with a very small sample of nineteenth-century
216 The Rich Nineteenth Century

philosophers of science. The four I have chosen were not the ones most
revered by their contemporaries - that prize goes to the positivists
Comte and J. S. Mill - but they have had a great impact on my gen-
eration, in the second half of the twentieth century, three of them -
Whewell, Peirce, and Duhem - as powerful sources of inspiration, and
the fourth one - Mach - as the major source of the late form of posi-
tivism that we have had to fight (he was, however, much more open-
minded and wrote more stimulating and instructive books than his
epigons). Reluctantly I leave out the two great philosopher-scientists
Helmholtz and Poincare; however, in the space available I could only
have summarized what I have said about them elsewhere.92

4.4.1 William Whewell (1794-1866)

To a casual reader it may seem that Whewell's philosophy of human
knowledge is only a watered-down version of Kant's. Like Kant,
Whewell insists on the active role of the knowing mind, through which
alone knowledge achieves universality and necessity. However, dis-
regarding Kant's meticulous classification of the mental ingredients of
knowledge into "forms of sense", "categories of the understanding",
and "ideas of reason", he brings every contribution of the mind under
the single term 'ideas', which, besides space, time, and cause, also
covers number, resemblance, force, polarity, medium (for the trans-
mission of signals), and so on. Whewell maintains that ideas - in his
sense - are in one way or another conditions of the possibility of expe-
rience, and repeats Kant's "transcendental" vindication of space as the
foundation of geometric truths (1847, I, 85f.). But he does not try to
show - as Kant did in the "proofs" we examined in §3.4 - that the
ideas of cause, etc., are prerequisites of objectivity and therefore justi-
fied in their ordinary scientific use. Now, Kant's conclusions concern-
ing the limits of a priori knowledge depend on his precise grounding
of it. By cavalierly forgoing this grounding, Whewell is able to avoid
those conclusions and cheerfully lists "natural theology" among the
sciences, right after linguistics and ethnography (1847, II, 117).
To a closer look, however, Whewell's writings offer some new and

92
On Poincare, see Torretti (1978, pp. 320-58; 1983, pp. 83-87). On pages 155-71 of
the earlier book, I comment on Helmholtz's philosophy of geometry. The discussion
of extensive quantities in Torretti (1990, pp. 58-65), is ultimately based on Helmholtz
(1887).
 4.4 Philosophers 217

illuminating thoughts on the natural sciences, which I shall epitomize

and illustrate by a series of quotations.93 Let me first note that
Whewell's portmanteau description of the mental principles of knowl-
edge as 'ideas' is anything but careless. Such looseness is required to
make room for the historical development of knowledge. Although
Whewellian ideas are supposed to express the permanent nature of the
human mind, they do so progressively, as human experience evolves.
For Whewell
Ideas are not Objects of Thought, but rather Laws of Thought. Ideas are
not synonymous with Notions; they are Principles which give to our
Notions whatever they contain of truth.
(Whewell 1847, I, 29)
And yet, according to him, ideas are modified by their employment in
experience.
Ideas cannot exist where Sensation has not been [...]. Hence, at what-
ever period we consider our Ideas, we must consider them as having been
already engaged in connecting our Sensations, and as having been mod-
ified by this employment. By being so employed, our Ideas are unfolded
and defined; and such development and definition cannot be separated
from the Ideas themselves.
(Whewell 1847, I, 43-44)
Therefore, the philosopher may not claim to have now - or at any other
time - a definite and definitive awareness of the mind's contribution to
experience.94 This contribution does not consist of innate factors whose
inventory and mutual relations one could establish once and for all.
Our fundamental ideas are necessary conditions of knowledge, universal
forms of intuition, inherent types of mental development; they may even
be termed, if any one chooses, results of connate intellectual tendencies;
but we cannot term them innate ideas [...]. For innate ideas were con-
sidered as capable of composition, but by no means of simplification: as
most perfect in their original condition; as to be found, if any where, in

93
Whewell's main books, the History of the Inductive Sciences (1837) and the Philos-
ophy of the Inductive Sciences (1840, 2nd much revised ed. 1847), are very readable,
but most students will probably be put off by their size. They can profitably use the
excellent selection in Butts (1968).
94
Kant failed to see that this follows from his assertion that all knowledge begins with
experience and none precedes it (1787, p. 1), and went on to prolixly set forth his
tripartite architecture of reason.
218 The Rich Nineteenth Century

the most uneducated and most uncultivated minds; as the same in all ages,
nations, and stages of intellectual culture; as capable of being referred to
at once, and made the basis of our reasonings, without any special acute-
ness or effort: in all which circumstances the Fundamental Ideas of which
we have spoken, are opposed to Innate ideas so understood.
(Whewell 1851, in Whewell 1860, pp. 530f.)
Such "fundamental ideas" are not necessarily ultimate elements of
knowledge; "they are the results of our analysis so far as we have
yet prosecuted it; but they may themselves subsequently be analysed"
(p. 531; my italics).
Indeed, for Whewell, 'ideas and sensations' form just one of several
pairs of terms by which English speakers denote the "fundamental
antithesis of philosophy". Its "simplest and most idiomatic expression"
is the opposition between 'thoughts and things'; others are 'necessary
and experiential truths', 'subjective and objective', 'form and matter',
and 'theories and facts'.
The fundamental antithesis of philosophy is an antithesis of inseparable
elements. Not only cannot these elements be separately exhibited, but
they cannot be separately conceived and described. The description of
them must always imply their relation; and the names by which they are
denoted will consequently always bear a relative significance. And thus
the terms which denote the fundamental antithesis of philosophy cannot
be applied absolutely and exclusively in any case.
(Whewell 1844, in Whewell 1847, II, 651-52)
The revolution of the stars about the poles, the rotation and transla-
tion of the earth, and the mutual attraction of the sun and planets are
described by some as theories and as facts by others.
In these cases we cannot apply absolutely and exclusively either of the
terms, Fact or Theory. Theory and Fact are the elements which corre-
spond to our Ideas and our Senses. The Facts are Facts so far as the
Ideas have been combined with the sensations and absorbed in them: the
Theories are Theories so far as the Ideas are kept distinct from the
sensations, and so far as it is considered as still a question whether they
can be made to agree with them. A true Theory is a fact, a Fact is a
familiar theory.
(Whewell 1844, in Whewell 1847, II, 652)

In a Fact, the Ideas are applied so readily and familiarly, and incor-
porated with the sensations so entirely, that we do not see them, we see
 4.4 Philosophers 219

through them. A person who carefully notes the motion of a star all
night, sees the circle which it describes, as he sees the star, though the
circle is, in fact, a result of his own Ideas.
(Whewell 1847, I, 40)
In Whewell's vocabulary, 'ideas' are "certain comprehensive forms
of thought"; while "the special modifications of these ideas which are
exemplified in particular facts" are termed 'conceptions' (1847, II,
5-6). The construction of science comprises two main processes, viz.,
the Explication of Conceptions, by which conceptions are "carefully
unfolded" and "made more clear in themselves", and the Colligation
of Facts, "by which the conceptions more strictly bind together the
facts" (II, 5). These two processes must go hand in hand. The "right
definition of a Term may be a useful step in the explication of our con-
ceptions", but only when we contemplate "some Proposition in which
the Term is employed" (II, 12). For "that which alone makes it worth
while" to clarify a conception "is the opportunity of using it in the
expression of Truth" (II, 13). So "the question really is, how the Con-
ception shall be understood and defined in order that the Proposition
may be true". Now, as Whewell emphatically recalls, "the establish-
ment of a Proposition requires an attention to observed Facts, and can
never be rightly derived from our Conceptions alone" (II, 12).
However, "facts cannot be observed as Facts, except in virtue of the
Conceptions which the observer himself unconsciously supplies" (II,
23). Therefore, the two processes by which science is constructed, "the
Explication of the Conceptions of our own minds, and the Colligation
of observed Facts by the aid of such Conceptions", are inseparably con-
nected with each other (II, 46). When jointly employed in collecting
knowledge "they constitute the mental process of Induction-, which is
usually and justly spoken of as the genuine source of all our real general
knowledge respecting the external world" (II, 46-47). The vulgar
conceive of induction as a process "by which we collect a General
Proposition from a number of Particular Cases: and it appears to be
frequently imagined that the general proposition results from a mere
juxta-position of the cases, or at most, from merely conjoining or
extending them" (II, 48). Not so, says Whewell.

In each inference made by Induction, there is introduced some General

Conception, which is given, not by the phenomena, but by the mind.
The conclusion is not contained in the premises, but includes them by
the introduction of a New Generality. In order to obtain our inference,
220 The Rich Nineteenth Century

we travel beyond the cases which we have before us; we consider them
as mere exemplifications of some Ideal Case in which the relations are
complete and intelligible. We take a Standard, and measure the facts by
it; and this Standard is constructed by us, not offered by Nature. We
assert, for example, that a body left to itself will move on with unaltered
velocity; not because our sense ever disclosed to us a body doing this,
but because (taking this as our Ideal Case) we find that all actual cases
are intelligible and explicable by means of the Conception of Forces,
causing change and motion, and exerted by surrounding bodies.
(Whewell 1847, II, 49)
So in every induction "there is some Conception superinduced upon
the facts", and Whewell proposes that we henceforth regard this as
"the peculiar import of the term Induction" (II, 50). This decisive step
in the formation of science, "the Invention of a new Conception in
every inductive inference", had rarely been noticed by Whewell's pre-
decessors. The reason for this is not hard to see: Such acts of inven-
tion soon slip out of notice.
Although we bind together facts by superinducing upon them a new
Conception, this Conception, once introduced and applied, is looked
upon as inseparably connected with the facts, and necessarily implied in
them. Having once had the phenomena bound together in their minds
in virtue of the Conception, men can no longer easily restore them back
to the detached and incoherent condition in which they were before they
were thus combined. [. . .] As soon as the leading term of a new theory
has been pronounced and understood, all the phenomena change their
aspect. There is a standard to which we cannot help referring them.
(Whewell 1847, II, 51-53)
Although Whewell occasionally refers to a "logic of induction" -
perhaps as a concession to his contemporaries - and devotes several
chapters to "methods of induction" (II, 395-425), he saw and said
clearly - a full century before Feyerabend (1970) - that scientific dis-
covery is not subject to any rules.
Scientific discovery must ever depend upon some happy thought, of which
we cannot trace the origin; - some fortunate cast of intellect, rising above
all rules. No maxims can be given which inevitably lead to discovery.
(Whewell 1847, II, 20f.)

The Conceptions by which Facts are bound together, are suggested by

the sagacity of discoverers. This sagacity cannot be taught. It commonly
 4.4 Philosophers 221

succeeds by guessing; and this success seems to consist in framing

several tentative hypotheses and selecting the right one. But a supply of
appropriate hypotheses cannot be constructed by rule, nor without
inventive talent.
(Whewell 1847, II, 467f.)

Talent can certainly benefit from instruction. This must be sought,

however, not in abstract methodological prescriptions, but in actual
instances of scientific thinking and scientific practice. So Whewell filled
page after page of his philosophical books with historical examples and
wrote the three volumes of his History (1837).
Twentieth-century positivists would maintain, of course, that the
rules of inductive logic are not meant to preside over the process of dis-
covery, but to control the validity of itsfindings.Whewell expressly asks
how we are to find what new conception succeeds in binding together
the facts. How is the trial to be made? What is meant here by 'success'?
(II, 45). Compared with the convoluted contraptions of modern confir-
mation theory, his approach to the testing of hypotheses may well seem
casual (see II, 60-95). He repeats the usual bland generalities about the
comparison of hypotheses with the facts and stresses the force of con-
viction issuing from the prediction of novel phenomena (as opposed to
the mere explanation of those already familiar). But he makes no move
to quantify degrees of confirmation or the weight and relevance of evi-
dence. In view of the dismal failure of all attempts to achieve cogency
in this matter, his relaxed informality seems wise. Only in cases of a pecu-
liar and indeed quite extraordinary sort did he claim finality for the con-
clusions of induction. He coined for them the term 'consilience of
inductions' (II, 65). This "takes place when an Induction, obtained from
one class of facts, coincides with an Induction, obtained from another
different class" (II, 469). Consilience - says Whewell - "is a test of the
truth of the Theory in which it occurs" (Ibid.).

The evidence in favor of our induction is of a much higher and more

forcible character when it enables us to explain and determine cases of
a kind different from those which were contemplated in the formation
of our hypothesis. The instances in which this has occurred, indeed,
impress us with a conviction that the truth of our hypothesis is certain.
No accident could give rise to such an extraordinary coincidence. No
false supposition could, after being adjusted to one class of phenomena,
exactly represent a different class, when the agreement was unforeseen
and uncontemplated. That rules springing from remote and unconnected
222 The Rich Nineteenth Century

quarters should thus leap to the same point, can only arise from that
being the point where truth resides.
(Whewell 1847, II, 65)
Whewell's paramount example of consilience is Newton's discovery
that the three Keplerian laws of planetary motion follow from his Law
of Gravity although no link between them had been noticeable before.
Moreover, Newton's Law also accounted for such dissimilar phenom-
ena as the tides and the precession of the equinoxes (II, 66). Now,
despite its enormous predictive fruitfulness, Newtonian instantaneous
attraction at a distance is no longer thought to be the best, let alone
the true, explanation of the phenomena that we nevertheless continue
to call 'gravitational' (see §5.4). So Newton's theory is also a preemi-
nent instance of inconclusive consilience, and Whewell has been rightly
taken to task for thinking otherwise. Still, there is no question that con-
silience, when it occurs, impresses us with a conviction of certainty.
Moreover, there is one respect in which the consilience of inductions
seems indeed to be definitive: The unifications that it brings about
appear to be irreversible, at least within a given scientific tradition.
Thus, no one today would dream of placing different subsets of the
so-called gravitational phenomena under disparate conceptions. We
naturally expect every purported successor to Newton's theory of
gravity to account for the fall of heavy bodies and the circulation of
the planets, not forgetting the tides and the precession of the equinoxes,
just as any substitute for the wave theory of light - Whewell's second
example of consilience - must countenance some sort of waves, be it
only waves of probability.

4.4.2 Charles Sanders Peirce (1839-1914)

Peirce once characterized philosophy as "the attempt to form a general
informed conception of the A//" (CP 7.579). Since we cannot reason-
ably hope to attain individually "the ultimate philosophy which we
pursue", it must be a goal "for the community of philosophers" (CP
5.265), a public work, "meant for the whole people and [. . .] erected
by the exertions of an army representative of the whole people", like
a cathedral (CP 1.176; cf. 6.315). Early "a passionate devotee of
Kant",95 Peirce retained a lifelong preference for triads (CP 4.3,

95
"I believed more implicitly in the two tables of the Functions of Judgment and the
Categories than if they had been brought down from Sinai" (CP 4.2).
 4.4 Philosophers 223

6.321ff., 1.369, 6.202, 1.471ff., etc.; cf. 1.568) and fondly cited Kant's
teaching about the "architectonic" construction of philosophy (CP 6.9,
5.5, 1.176; cf. Kant 1781, pp. 831ff.). And yet, were it not for these
reminders, you would hardly think of philosophical system building as
you navigate the ocean of his Collected Papers. Gathered by "pitch-
forking" Peirce's published articles and massive Nachlass (CP 1.179),
they glaringly flaunt their variegated origin, despite the editors' Pro-
crustean endeavor to fit them into a systematic plan.96 Moreover, it
would seem that Peirce, a profoundly original thinker of incisive intel-
ligence, had a hyperkinetic mind, which made it difficult for him to
keep a steady course even within a single paper or a connected series
of papers. So his writings resemble anything but a masterpiece of archi-
tecture. In a way this is all for the best. As we zigzag among Peirce's
startling insights and clever arguments, the Collected Papers come out
as immensely more attractive and instructive than, say, the fossil trea-
tises of Schopenhauer or Spencer.
One can, indeed, by dexterous hermeneutics elicit one or more
systems of philosophy97 from Peirce's "odds and ends of commentary"
(CP 6.150). Maybe this is the proper way to handle the heritage of a
man who relished coining "ism" words to bear on his standard. But I
see no gain in corralling a thinker I admire into a philosophical "posi-
tion" that anyway is bound to prove untenable. So, instead of trying to
weave Peirce's thoughts on science into the systematic context to which
they presumably belong, I shall consider only three topics relevant to
physics on which he threw much light. This approach, I think, comes
closer than would a rational reconstruction of Peirce's system or systems
to the "ism" for which he is best remembered, of which he said:
Pragmatism is not a Weltanschauung but is a method of reflexion having
for its purpose to render ideas clear.98
The topics I have chosen for consideration are: (i) the maxim of prag-
matism, (ii) inference, and (iii) determinism and chance.

96
A new edition of Peirce's writings in chronological order and satisfying modern stan-
dards of scholarship is being published by Indiana University Press since 1982. As I
write this, only five volumes have appeared, covering the period from 1857 to 1886.
Peirce remained a very prolific writer until two or three years before his death in
1914.
97
M. G. Murphey distinguishes four such systems in his article on Peirce in Edwards's
Encyclopedia of Philosophy.
98
CP 5.13nl; c. 1902. From Peirce's personal interleaved copy of the Century
Dictionary.
224 The Rich Nineteenth Century

(i) In the first of six papers on "the logic of science" published in

1877-1878, Peirce defines 'inquiry' as the struggle to overcome the irri-
tation of doubt by reaching a state of belief (CP 5.374). "Hence" - he
concluded - "the sole object of inquiry is the settlement of opinion"
(5.375). He considers four methods of fixing belief. The first, the
method of tenacity', consists in "taking any answer to a question which
we may fancy, and constantly reiterating it to ourselves" (5.377). The
second, the method of authority', confers this function on an institu-
tion whose sole purpose is "to keep correct doctrines before the atten-
tion of the people, to reiterate them perpetually, and to teach them to
the young; having at the same time power to prevent contrary doc-
trines from being taught, advocated, or expressed" (5.379). The fail-
ings of tenacity and authority have led to the development of a third
method, which Peirce describes as follows: "Let the action of natural
preferences be unimpeded, then, and under their influence let men,
conversing together and regarding matters in different lights, gradually
develop beliefs in harmony with natural causes" (5.382). However, this
method, well exemplified by the history of metaphysical philosophy,
has also failed to produce a general agreement and to stabilize belief.
"To satisfy our doubts, therefore, it is necessary that a method should
be found by which our beliefs may be caused by nothing human, but
by some external permanency - by something upon which our think-
ing has no effect" (5.384). This must be something "which affects, or
might affect, every man. And, though these affections are necessarily
as various as are individual conditions, yet the method must be such
that the ultimate conclusion of every man shall be the same. Such is
the method of science" (Ibid.).
The next paper in the series examines the first step of the scientific
method, namely, how to make our ideas clear. Peirce assumes that the
sole aim of thought is to produce belief (5.396)," and that all belief

Did Peirce abide by this view of thought? Twenty years later, in the Cambridge lec-
tures of 1898 on "Reasoning and the Logic of Things", he asserted that "what is
properly and usually called belief, that is, the adoption of a proposition as a KTfjjLia
tq ocia [a possession forever] has no place in science at all" (CP 1.635). However, in
the 1878 paper, Peirce did not equate belief with something everlasting. "Belief" - he
wrote then - "is a rule for action, the application of which involves further doubt
and further thought", so, "at the same time that it is a stopping-place, it is also a
new starting-place for thought" (5.397). Thus, Peirce's views of 1878 and 1898 might
 4.4 Philosophers 225

"involves the establishment in our nature of a rule of action, or, say

for short, a habit" (5.397). If two beliefs "appease the same doubt by
producing the same rule of action, then no mere differences in the
manner of consciousness of them can make them different beliefs"
(5.398). Hence the root of every real distinction of thought, no matter
how subtle, lies in what is tangible and conceivably practical: "There
is no distinction of meaning so fine as to consist in anything but a pos-
sible difference of practice" (5.400). Therefore, to make our ideas clear
we must follow this maxim:
Consider what effects, that might conceivably have practical bearings,
we conceive the object of our conception to have. Then, our conception
of these effects is the whole of our conception of the object.
(Peirce CP 5.402)
In Baldwin's Dictionary (1902), Peirce defined 'pragmatism' as "the
opinion that metaphysics is to be largely cleared up by the application
of the [above] maxim", which he quoted verbatim (CP 5.2). He added
that, "after many years of trial", he still thought the maxim to be "of
great utility"; but he explicitly distanced himself from William James,
who in his Will to Believe (1897) and elsewhere had "pushed this
method to such extremes as must tend to give us pause" (5.3). Taken
literally the maxim would indeed sweep away the difference between
rational and irrational numbers, for actual measurements can only
yield rational values. This was a consequence that Peirce, as a trained
mathematical physicist, could not stomach (cf. 5.32-33). He mentioned
it in the said Dictionary article, and went on to add that the doctrine
of pragmatism "appears to assume that the end of man is action - a
stoical axiom which to the present writer at the age of sixty, does not
recommend itself so forcibly as it did at thirty" (5.3). If action wants
an end and that end must be something of a general description, then
"the spirit of the maxim itself, which is that we must look to the upshot
of our concepts in order rightly to apprehend them, would direct us

not be as irreconcilable as they seem. Still, one notices at least a change of emphasis
from his early claim that the "sole motive, idea and function" of thought is "to
produce belief" (5.396), to the subsequent one that "there is . . . no proposition at all
in science which answers to the conception of belief" (1.635). Such change probably
has to do with his later view that the end of man is not simply "action" - as he may
have thought at age 30 - but "the development of concrete reasonableness" (5.3; cf.
5.433, 1.615).
226 The Rich Nineteenth Century

towards something different from practical facts, namely, to general

ideas, as the true interpreters of our thought" (Ibid.).
(ii) Peirce distinguished three kinds of inference: deduction, induc-
tion, and abduction.
Deduction comprises every inference in which the premises are so
related to the conclusion that it is impossible that the latter be false
while the former are true (CP 2.778). It is worth remembering that in
the 1870s Peirce himself and Gottlob Frege, unbeknownst to one
another, revolutionized the theory of deductive inference (which - as
Kant 1787, p. viii, noted - had not progressed since Antiquity). They
founded what Peirce called 'the logic of relatives', involving predicates
assertible of two or more objects. They also introduced the now famil-
iar analysis of general statements as quantified compound state-
ments.100 Thanks to their momentous work they were both able to
identify deduction with mathematical proof. "Every science has a
mathematical part, a branch of work that the mathematician is called
to do. We say: 'Here, mathematician, suppose such and such to be the
case. Never you mind whether it is really so or not; but tell us, sup-
posing it be so, what will be the consequence.'" (CP 1.133). This the
mathematician does by deduction, which, of course, "does not lead to
any positive knowledge at all, but only traces out the ideal con-
sequences of hypotheses" (7.207).101 Peirce's work on deduction per-
suaded him that every deductive inference can be represented by a
diagram from which one intuitively gathers that the conclusion neces-
sarily follows from the premises. Deduction - he wrote to Calderoni
c. 1905 - "consists in constructing an image or diagram in accordance

100
Readers unacquainted with modern formal logic may benefit from the following.
Denote arbitrary objects by italic lowercase letters. The expression 'is greater than'
- as in 'x is greater than y' - is a binary predicate; 'gives' - as in 'x gives y to z' -
is a ternary one. Aristotle classified general statements into universal statements of
the form 'All Ps are Qs' and particular statements of the form 'Some Ps are Qs'. For
Peirce and Frege, 'All Ps are Qs' is just a short way of saying that 'For any object
x (in the universe of discourse), if x is P, then x is Q', whereas 'Some Ps are Qs'
says that 'In the universe of discourse there is at least one object x such that x is P
and x is Q'. The generalizing operators 'For any x' and 'There is an x' are called
quantifiers.
101
"It is impossible to reason necessarily concerning anything else than a pure hypoth-
esis. Of course, I do not mean that if such pure hypothesis happened to be true of
an actual state of things, the reasoning would thereby cease to be necessary. Only,
it never would be known apodictically to be true of an actual state of things" (Peirce
CP 4.232).
 4.4 Philosophers 227

with a general precept, in observing in that image certain relations of

parts not explicitly laid down in the precept, and in convincing oneself
that the same relations will always occur when that precept is followed
out" (8.209; cf. 1.66).
In 1878 Peirce defined induction as synonymous with statistical infer-
ence, viz., "the inference that a previously designated character has
nearly the same frequency of occurrence in the whole of a class that it
has in a sample drawn at random out of that class" (6.409). He gave a
similar definition in 1883 (2.702), but in an important manuscript, "On
the Logic of Drawing History from Ancient Documents Especially from
Testimonies", written perhaps in 1901 (CP 7.163-255), he tentatively
proposed a distinction between three kinds of induction, such that the
definition of 1878 applies only to the first.102 All three kinds share the
property that Peirce regards as "essential" and "intrinsic" (5.579) to this
method of inference, viz., that it is self-correcting:

Induction is the experimental testing of a theory. The justification of it

is that, although the conclusion at any stage of the investigation may be
more or less erroneous, yet the further application of the same method
must correct the error. The only thing that induction accomplishes is to
determine the value of a quantity. It sets out with a theory and it
measures the degree of concordance of that theory with fact.
(Peirce CP 5.145)

The process of testing it will consist, not in examining the facts, in order
to see how well they accord with the hypothesis, but on the contrary in
examining such of the probable consequences of the hypothesis as would
be capable of direct verification, especially those consequences which
would be very unlikely or surprising in case the hypothesis were not true.
(CP 7.231)

102
Strictly speaking, to a subspecies of the first. But this is because in the paper of c.
1901 Peirce gives a definition of randomness by virtue of which no random samples
can be drawn from infinite collections. Consequently, the definition of 1878 does
not cover the case in which we sample a countable collection "in order to ascertain
the proportionate frequency with which its members have a certain character desig-
nated in advance of the examination" (7.212), nor that in which we find a count-
able series "in an objective order of succession, and wish to know what the law of
occurrence of a certain character among its members is, without at the outset so
much as knowing whether it has any definite frequency in the long run or not"
(7.213). These are, then, the other subspecies of the first kind of induction.
228 The Rich Nineteenth Century

When we adopt a certain hypothesis, it is not alone because it will

explain the observed facts, but also because the contrary hypothesis
would probably lead to results contrary to those observed. So, when we
make an induction, it is drawn not only because it explains the distri-
bution of characters in the sample, but also because a different rule
would probably have led to the sample being other than it is.
(CP 2.628)
Readers acquainted with modern statistics will perceive that Peirce
anticipates here the approach subsequently developed by Neyman and
Pearson.103 He explicitly renounced the opposite statistical school, pur-
portedly based on Bayes's Theorem, long before it became fashionable:
The theory here proposed does not assign any probability to the induc-
tive or hypothetic conclusion, in the sense of undertaking to say how
frequently that conclusion would be found true. It does not propose to
look through all the possible universes, and say in what proportion of
them a certain uniformity occurs; such a proceeding, were it possible,
would be quite idle. The theory here presented only says how frequently,
in this universe, the special form of induction or hypothesis would lead
us right. The probability given by this theory is in every way different -
in meaning, numerical value, and form - from that of those who would
apply to ampliative inference the doctrine of inverse chances.
(Peirce CP 2.748)
Throughout his life, Peirce always insisted in the existence of a third
type of inference besides deduction and induction. He called it by dif-
ferent names - viz., 'abduction', 'retroduction', 'hypothesis', 'pre-
sumption' - and described it in mildly different ways while consistently
regarding it as the mainstay of scientific thinking, the sole source of
new ideas in science (5.145) and indeed of new truths (7.219). Here
is how Peirce characterized abduction in Baldwin's Dictionary (s.v.
'Reasoning'):
Upon finding himself confronted with a phenomenon unlike what he
would have expected under the circumstances, [the reasoner] looks over
its features and notices some remarkable character or relation among
them, which he at once recognizes as being characteristic of some
conception with which his mind is already stored, so that a theory is

103
Also Deborah Mayo's "error statistics", as she points out in her illuminating ana-
lysis of Peircean error correction (Mayo 1996, pp. 412-41). I owe to Mayo the four
Peirce extracts on induction.
 4.4 Philosophers 229

suggested which would explain (that is, render necessary) that which is
surprising in the phenomena.
(Peirce CP 2.776)

There is no warrant for this inference. The hypothesis inferred is often

"utterly wrong" and "even the method need not ever lead to the truth;
for it may be that the features of the phenomena which it aims to explain
have no rational explanation at all". Nevertheless, abduction is justified
insofar as it is "the only way in which there can be any hope of attain-
ing a rational explanation" (2.777). Or, as Peirce concisely put it in his
1903 lectures on pragmatism, "Abduction consists in studying facts and
devising a theory to explain them. Its only justification is that if we are
ever to understand things at all, it must be in that way." (5.145).
Peirce once proposed some trivial rules by following which "the
process of making an hypothesis should lead to a probable result"
(2.634). Clearly he was paying lip service to the contemporary preju-
dice in favor of rules of method; for in this matter of theory building
- as Paul Feyerabend (1970) said after Cole Porter - "anything goes".
Peirce also comments on "a rule of abduction much insisted upon by
Auguste Comte, to the effect that metaphysical hypotheses should be
excluded" (7.203). By 'metaphysical' Comte means a hypothesis that
has no experiential consequences. Peirce agrees that

an explanatory hypothesis, that is to say, a conception which does not

limit its purpose to enabling the mind to grasp into one a variety of facts,
but which seeks to connect those facts with our general conceptions of
the universe, ought, in one sense to be verifiable, that is to say, it ought
to be little more than a ligament of numberless possible predictions con-
cerning future experience, so that if they fail, it fails.
(Peirce CP 5.597)

However, "Comte's own notion of a verifiable hypothesis was that it

must not suppose anything that you are not able directly to observe",
and he ought therefore "to forbid us to suppose that a fossil skeleton
had ever belonged to a living ichtyosaurus" and indeed "to believe in
our memory of what happened at dinnertime today" (5.597). Thus, in
Comte's own narrow sense of 'verifiable', his rule is perverse. On the
other hand, in Peirce's sense, the rule is sound but superfluous, for a
hypothesis that is unverifiable in this wider sense cannot even be con-
structed. A proposition cannot "refer to anything with which experi-
ence does not connect us". Indeed, "the entire meaning of a hypothesis
230 The Rich Nineteenth Century

lies in its conditional experiential predictions" and "if all its predic-
tions are true, the hypothesis is wholly true" (7.203).
Although "abduction is, after all, nothing but guessing" (7.219), it
is apparently guided by some kind of instinct (1.630, 5.604) or natural
affinity for truth (7.220), for otherwise it would be well-nigh impos-
sible for us ever to guess right. "Think of what trillions of trillions of
hypotheses might be made of which one only is true; and yet after two
or three or at the very most a dozen guesses, the physicist hits pretty
nearly on the correct hypothesis. By chance he would not have been
likely to do so in the whole time that has elapsed since the earth was
solidified" (5.172). Peirce repeatedly identified this human talent with
the "natural light" {lume naturale) to which Galileo appealed "at the
most critical stages of his reasoning" (1.80; cf. 6.477, 1.630, 5.604).
He was fond of equating it with the innate abilities of animals and to
trace it somehow to natural selection:
If man had not had the gift, which every other animal has, of a mind
adapted to his requirements, he not only could not have acquired any
knowledge, but he could not have maintained his existence for a single
generation.
(Peirce CP 5.603)
Yet he must have realized that the instinctive understanding of natural
relations that is required to preserve our species could well be confined
to the human scale and fail utterly at the frontiers of physical inquiry,
for he referred to correct abductive guessing as a hope implicit in
scientific practice (much in the way that Kant talked of life after death
as a hope alive in ordinary moral practice):
We are therefore bound to hope that, although the possible explanations
of our facts may be strictly innumerable, yet our mind will be able, in
somefinitenumber of guesses, to guess the sole true explanation of them.
That we are bound to assume, independently of any evidence that it is
true. Animated by that hope, we are to proceed to the construction of a
hypothesis.
(Peirce, CP 7.219)
(iii) Brilliantly anticipating twentieth-century physics, Peirce (1892)
rejected "the doctrine of necessity", that is, the thesis of thorough-
going physical determinism adopted by so many of his contemporaries,
and subjected it to damaging criticism. Instead of it, he professed
"tychism", that is, the doctrine that ultimately nature is the realm of
 4.4 Philosophers 231

chance. This, he said, is subsidiary to "synechism", the metaphysics of

continuity that he embraced in the 1890s.104
The thesis of determinism is "that the state of things existing at any
time, together with certain immutable laws, completely determine the
state of things at every other time" (CP 6.37). Some "thinking men"
who professed it told Peirce that it is a presupposition or "postulate"
of scientific reasoning. However, "this does not make it true, nor so
much as afford the slightest rational motive for yielding it any cre-
dence" (6.39). Moreover, the very idea that scientific reasoning, and in
particular induction, may require such a postulate should, according
to Peirce, be dismissed in the light of our present understanding of
inference. On the other hand, there is no hope that induction might
itself provide a ground for determinism. Every purported application
of this view requires "that certain continuous quantities have certain
exact values". To someone who is familiar with experimental work
"the idea of mathematical exactitude being demonstrated in the labo-
ratory will appear simply ridiculous" (6.44).
Those observations which are generally adduced in favor of mechanical
causation simply prove that there is an element of regularity in nature,
and have no bearing whatever upon the question of whether such regu-
larity is exact and universal or not. Nay, in regard to this exactitude, all
observation is directly opposed to it; and the most that can be said is
that a good deal of this observation can be explained away. Try to verify
any law of nature, and you will find that the more precise your obser-
vations, the more certain they will be to show irregular departures from
the law. We are accustomed to ascribe these, and I do not say wrongly,
to errors of observation; yet we cannot usually account for such errors
in any antecedently probable way. Trace their causes back far enough
and you will be forced to admit they are always due to arbitrary deter-
mination, or chance.
(Peirce CP 6.46)105

Peirce (CP 6.202). 'Tychism' is formed from the Greek Ti3%r| = 'luck, chance';
'synechism', from <yov£%rj(; = 'continuous'.
Here, Peirce, like almost everybody else, associates physical determinism with cau-
sation. But elsewhere he forcefully criticizes this association, much on the same
grounds I adduced in §3.4.3. For him, causation is a "confused notion" (6.600) that
men in different stages of scientific culture have conceived in entirely different and
inconsistent ways. In particular, the common view "that our conception of cause is
that of the Aristotelian efficient cause will hardly bear examination. The efficient
cause was, in the first place, generally a thing, not an event; then, something which
232 The Rich Nineteenth Century

For Peirce, "every throw of sixes with a pair of dice is a manifest

instance of chance" (6.53). He debates this issue with an imaginary
determinist who recalls that "each die moves under the influence of
precise mechanical laws". Yet, counters Peirce, "the laws act just the
same when other throws come up". True, the determinist concedes, but
"the diversity is due to the diverse circumstances under which the laws
act: the dice lie differently in the box, and the motion given to the box
is different"; the "number of numbers, which expresses the amount of
diversity of the system, remains the same at all times" (6.55-56). Peirce
clinches the issue thus:
You think all the arbitrary specifications of the universe were introduced
in one dose, in the beginning, if there was a beginning, and that the
variety and complication of nature has always been just as much as it is
now. But I, for my part, think that the diversification, the specification,
has been continually taking place.

By thus admitting pure spontaneity or life as a character of the universe,

acting always and everywhere though restrained within narrow bounds
by law, producing infinitesimal departures from law continually, and
great ones with infinite infrequency, I account for all the variety and
diversity of the universe, in the only sense in which the really sui generis
and new can be said to be accounted for.
(Peirce, CP 6.57, 6.59)
I am not sure that I understand how tychism follows from synechism
according to Peirce. It certainly has to do with his unorthodox con-

need not do anything; its mere existence might be sufficient. Neither did the effect
always necessarily follow. True when it did follow it was said to be compelled. But
it was not necessary in our modern sense. That is, it was not invariable" (6.66). After
poking fun at those who admire John Stuart Mill for regarding "the cause as the
aggregate of all the circumstances under which an event occurs", Peirce notes that
this - otherwise commonplace view - rests on a misconception. "So far as the con-
ception of cause has any validity - that is [...], in a limited domain - the cause and
its effect are two facts"; however, contrary to what Mill "thoughtlessly" assumed,
"the objective history of the universe for a short time, in its objective state of exis-
tence in itself" is not what a fact is; "a fact is an abstracted element of that; a fact
is so much of the reality as is represented in a single proposition" (6.67). Nor does
modern physical determinism illustrate Kant's law of causality (the Second Analogy
of Experience). "There is no mechanical truth in saying that the past determines the
future, rather than the future the past" (6.600). Moreover, "Kant's 'Analogy' ignores
that continuity which is the life blood of mathematical thought" (Ibid.).
 4.4 Philosophers 233

ception of the continuum. Peirce studied and partially embraced

Cantor's ideas about infinite sets and transfinite cardinalities, but he
firmly rejected Cantor's view that continua are equinumerous with 2°°,
that is, the set of all mappings from the set of natural numbers into
{0,1}. A collection that meets this condition must consist of "absolutely
distinct individual objects" and therefore should be called a pseudo-
continuum (6.176), for "continuity is fluidity, the merging of part into
part" (1.164). According to Peirce, mathematics calls for continuity in
this sense. To show it, he proves, after Cantor, that no collection of
distinct individuals "can have a multitude [i.e., cardinality - R. T.] as
great as that of the collection of possible collections of its individual
members" (Peirce RLT, p. 158). He then bids us consider "a collection
containing an individual for every individual of a collection of collec-
tions comprising a collection of every abnumeral [i.e., uncountable -
R. T.] multitude".
This collection shall consist of all finite multitudes together with all
possible collections of those multitudes, together with all possible
collections of collections of those multitudes, together with all possible
collections of collections of collections of those multitudes, and so on ad
infinitum. This collection is evidently of a multitude as great as that of
all possible collections of its members. But we have just seen that this
cannot be true of any collection whose individuals are distinct from one
another. We, therefore, find that we have now reached a multitude so
vast that the individuals of such a collection melt into one another and
lose their distinct identities. Such a collection is continuous.
(Peirce RLT, p. 159)106
In other words, a continuum is what Peirce elsewhere calls a "collec-
tion too great to be discrete" (CP 4.180). More exactly, "a true con-
tinuum is something whose possibilities of determination no multitude
of individuals can exhaust" (1.170). Synechism regards all distinct
actualities as issuing from some such primeval continuum of poten-
tialities. "It must be by a contraction of the vagueness of that poten-

106
In this way, if Peirce's conception of the continuum makes any sense, it provides an
instantaneous solution to Cantor's antinomy of the set of all sets. Peirce's argument
for the reality of such continua is no less original and worth looking into: "The argu-
ment which seems to me to prove, not only that there is such a conception of con-
tinuity as I contend for, but that it is realized in the universe, is that if it were not
so, nobody could have any memory. If time, as many have thought, consists of
discrete instants, all but the feeling of the present instant would be utterly non-
existent." (CP, 4.641; cf. 1.167ff.)
234 The Rich Nineteenth Century

tiality of everything in general, but of nothing in particular, that the

world of forms comes about" (6.196). So "the very first and most fun-
damental element that we have to assume is a Freedom, or Chance,
or Spontaneity, by virtue of which the general vague nothing-in-
particularness that preceded the chaos took a thousand definite quali-
ties" (6.200). At this point, I presume, tychism joins synechism.

4.4.3 Ernst Mach (1838-1916)

Mach saw himself as a physicist, not as a philosopher.107 However, it is
in the latter capacity that he is best known. In our eyes, his most valu-
able work is contained in his "historico-critical" studies on the energy
principle (1872), the science of mechanics (1883), the theory of heat
(1896), and physical optics (1921). His criticism of what he considered
superfluous metaphysical ingredients in Newton's conceptual frame
impressed young Einstein, encouraged his defiance of established
physics, and guided his first steps toward a new theory of gravity. Mach
is also remembered for his teachings about the aim of science and the
common ground of physics and psychology, mainly because of the influ-
ence they exercised on the Vienna Circle, and through it on twentieth-
century positivism. The following notes focus on these teachings -
beginning with the second point - and on his criticism of Newton.
(i) Mach felt that mind-body dualism, especially in the clear-cut
version inherited from Descartes, is a needless, unwarranted obstacle
to the unity of science. A smooth transit from physics (and physiology)
to psychology and vice versa can be secured if we acknowledge that
we have to do with two perspectives, two different ways of articulat-
ing and combining the same set of elements. Such elements are "colors,
sounds, temperatures, pressures, spaces, times, and so forth, [. . .] con-
nected with one another in manifold ways; with them are associated

107
Cf. Mach (AS, p. 47): "I am a scientist and not a philosopher" (Mach AS, p. 47).
Ibid., p. 368: "Once more, there is no such thing as 'the philosophy of Ernst Mach'."
See also pp. xxxvi, 30 and Mach (El, pp. 15-16, 143). Through his experimental
work, Mach made significant contributions to various fields of physics. Several quan-
tities are named after him in fluid dynamics; for example, the Mach number is the
ratio of the local velocity of flow to the velocity of sound in a compressible fluid.
 4.4 Philosophers 235

dispositions of mind, feelings, and volitions" (Mach AS, p. 2). Certain

"complexes of colors, sounds, pressures, and so forth, functionally con-
nected in time and space" are comparatively stable. They are called
bodies. Physics studies their properties and mutual relations. Among
such bodies there is one, whose elements Mach denotes by K L M. . . ,
that appears to be more intimately connected than the other bodies
(denoted by A B C. . .) with "the complex of volitions, memory-images,
and the rest", which he denotes by a P y. . . . "Usually, now, the
complex a (3 y . . . K L M. . . , as making up the ego, is opposed to the
complex A B C. . . , as making up the world of physical objects; some-
times also a (3 y . . . is viewed as ego, and KLM . . . ABC. . . a s world
of physical objects" (AS, p. 9). Psychology studies all elements in their
relation to and interdependence with the body KLM... and its asso-
ciated elements a (3 y. . . . Considered from its standpoint, the elements
A B C. . . K L M. . . are called sensations; while in the perspective of
physics they are physical objects.108

Any one who has in mind the gathering up of the sciences into a single
whole, has to look for a conception to which he can hold in every depart-
ment of science. Now if we resolve the whole material world into ele-
ments which at the same time are also elements of the psychical world
and, as such, are commonly called sensations; if, further, we regard it as
the sole task of science to inquire into the connexion and combination
of these elements, which are of the same nature in all departments, and
into their mutual dependence on one another; we may then reasonably
expect to build a unified monistic structure upon this conception, and
thus to get rid of the distressing confusions of dualism.
(Mach AS, p. 312)
At first blush Mach's solution may seem incredibly amateurish. It
will appear to be less so if one considers how broadly he uses the term
'sensation'. To prove that there exist definite, specific sensations of
time, he proposes the example of two bars of music, in which the
sequence of the notes is quite different but which, when played con-
secutively, will at once be recognized as rhythmically identical.109 He
describes them as "two tonal entities which, acoustically, are differ-

108 "Phenomena may be subdivided into elements, which, in so far as they are connected
with certain processes of bodies, and can be regarded as conditioned by these
processes, we call sensations" (Mach 1875, p. 54; quoted by him in AS, p. 17n.).
109
The notes in the first bar are, in this order, C = J, E = J>, F = J>, G = J, followed by i. The
notes in the second bar are, in this order, B = J, G = J\ F = J>, E = J, followed by L
236 The Rich Nineteenth Century

ently colored, but possess the same temporal form" (AS, p. 248). By
'temporal form' Mach means a certain relational system or structure
embodied in both sound sequences. Now, if the concept of 'sensation'
in Mach's parlance is wide enough to include structures, his approach
is a good deal more promising than if the term had its usual meaning
and referred only to the notorious sense-data - viz., sounds, colors,
odors - of the empiricist tradition. Mach's intention probably was to
count only very simple structures among "the elements of the world"
(AS, p. 32), perhaps only such as can be apprehended at one fell swoop.
But there is no way of drawing a limit here, and, of course, if none is
drawn, Mach's "elements A B C . . . K L M . . . " , the common build-
ing ground of the sciences, comprise not just the sensory contents sug-
gested by Mach's manner of speech, but every contentual and formal
aspect discernible in experience. If such is the case, few will dispute
Mach's contention that "every physical concept means nothing but a
certain definite kind of connexion of the [. . .] elements which I have
denoted by A B C . . . " (AS, p. 42).
My vindication of Mach is perhaps disingenuous. Thus, in the last
quotation I have brazenly substituted '[. . .]' for the word 'sensory'. I
have also disregarded Mach's more obnoxious utterances, for instance,
that "the whole inner and outer world are put together, in combina-
tions of varying evanescence and permanence, out of a small number
of homogeneous elements" (AS, p. 22). Indeed, the very term 'element'
- which I have not tried to conceal - connotes a degree of finality and
irreducibility that the referents of current physical concepts plainly
cannot lay claim to. However, one can find plausible answers to these
objections. Mach, for one, stressed that his elements are provisional
(vorldufig - twice in italics in El, p. 12), viz., they are that beyond which
we cannot go for the time being. Science is concerned with "the func-
tional dependence (in the mathematical sense) of these elements from
one another" (El, p. 11; cf. AS, p. 35); usually, indeed, with the smooth
variation of elements of one kind as the elements of some other kind -
or some ordered w-tuple of kinds - vary smoothly. This alone entails
that the elements cannot all be homogeneous, so that Mach's contrary
indication must be a slip of the pen. Moreover, nontrivial functional
dependence and, in particular, smooth dependence presuppose that the
elements in question belong to structures apart from which they are
nothing or, at any rate, not what the scientists take them to be.
(ii) Mach insisted that no science would be necessary, and none
would ever have arisen, "if every particular fact, every particular phe-
 4.4 Philosophers 237

nomenon were immediately accessible as soon as we require to know

it" (Mach 1872, in 1909, p. 31). The function of science is, therefore,
"to serve as a substitute for experience"; its task is "to represent the
facts as completely as possible with the least expenditure of thought"
(Mach M, p. 465). We should be content to know the distance s tra-
versed by a freely falling body in each time interval t. "But what enor-
mous memory would be necessary to keep in our heads the relevant
table of 5 and t. Instead of it, we retain the formula s = 1/lgt2, [. . .] by
which we find the s pertaining to a given £[...]. This formula, this 'law'
does not in the least possess a greater objective value than all the par-
ticular facts together. Its value lies merely in the ease with which it is
used. It has an economic value" (1909, p. 31; cf. M, p. 461, on SnelPs
law). Besides gathering and filing a maximum of facts in a manageable
shape, science has a second task, namely, to analyze the more complex
facts into a minimal number of maximally simple ones. "This we call
explaining. The simple facts to which we reduce the more complex ones
are in themselves always incomprehensible, i.e. not further analyzable,
e.g. that a mass bestows acceleration on another mass. It is again a ques-
tion of economy, on the one side, and of taste, on the other, at which
incomprehensibles one chooses to stop" (1909, p. 31; my italics). "No
basic fact is more understandable than another. The choice of basic facts
is a matter of comfort, of history and of habit" (p. 33).
The pursuit of such economic aims is naturally led by interest. "When
we portray facts in thoughts, we never portray the facts as a whole, but
only the aspect which is important to us; we have here a goal which has
issued mediately or immediately from a practical interest. Our por-
trayals are always abstractions" (M, p. 458). Thus, "when we speak of
cause and effect we arbitrarily highlight those aspects to whose con-
nection we ought to pay attention as we portray a fact in the respect
which is important to us. In nature there is no cause and no effect.
Nature is simply there, just once" (M, p. 459). Likewise, when we
bestow a name on a "thing", that is, a fairly stable complex of elements,
we look away from its surroundings and the many little changes that
such a complex is continually going through. "The thing is an abstrac-
tion, its name a symbol for a complex of elements, whose changes we
ignore" (Ibid.). "Bodies are nothing but bundles of reactions connected
by law. The same holds for processes of every sort which we classify
and name to satisfy our need for overview. Be it waterwaves [. . .] or
soundwaves in the air [. . .] or an electric current [...], that which is
constant is always the connection of reactions according to law, and
238 The Rich Nineteenth Century

this alone. This is the critically purified concept of substance which, sci-
entifically, ought to replace the vulgar one" (El, p. 148).
Summing up: "Our concepts arise from the sensations, by way of
their connections; the aim of concepts is to lead us in every given case
by the shortest and most comfortable ways to the sensory representa-
tions which best agree with the sensations" (El, p. 144).110 Of course,
the agreement need not be better than is required by the momentary
interests and circumstances under which it takes place. Since such inter-
ests and circumstances vary from case to case, the intellectual por-
trayals of phenomena do not exactly agree with each other. According
to Mach, "biological interest impels the mutual correction of different
results of portrayal, towards the best possible, most advantageous bal-
ancing of differences" (El, p. 164). Thus, our concepts need not only
adjust themselves to facts, but also to one another. "The adjustment of
thoughts to facts [. . .] we call observation^ the adjustment of thoughts
to each other we call theory. Observation and theory cannot be sharply
separated, for almost every observation is already influenced by theory
and, if sufficiently important, will react upon theory" (pp. 164f.).
Scientific knowledge is concerned with the connection of phenom-
ena. Whatever we might make out as standing behind phenomena
"exists only in our understanding and has for us just the value of an
aid to memory or formula, whose form, being arbitrary and indiffer-
ent, can very easily change with our cultural standpoint" (Mach 1872,
in 1909, pp. 25f.). We are therefore free to devise transphenomenal
objects in any way we think useful for connecting the phenomena. We
are under no constraint to impose on them the familiar conditions of
sensory experience. In particular, there is no need to think of them as
spatial, that is, as sporting the same kind of relations as the visible and
the tangible, "any more than it is necessary to think of them as having
a definite pitch of sound" (1909, p. 27). As we shall see in Chapters
Five and Six, early twentieth-century physicists took advantange of this
freedom to a wholly unprecedented degree.

110
The passage continues with this remarkable metaphor: "Our true mental workers
are the sensory representations, while concepts are their supervisors and regulators,
which put the multitude of the former into place and assign them their jobs. In simple
affairs the intellect deals immediately with the workers, but in larger enterprises it
has to do with the leading engineers, which, however, would be of no use to it if it
had not also furnished them with reliable workers."
 4.4 Philosophers 239

(iii) Mach's "critical purification" of Newtonian mechanics con-

centrates on the concept of mass and the principle of inertia (Newton's
First Law).
According to Mach, Newton's definition of mass as the product of
density and volume is useless, for density, in turn, can only be evalu-
ated in terms of mass. In the wake of Euler's Mechanica (1736), most
writers on the subject opted for defining the mass of a particle as the
ratio between the net external force acting on the particle and the par-
ticle's acceleration.111 This approach, however, could not satisfy Mach,
who sought to "demythologize" the notion of force by defining it as
the product of mass and acceleration. So he tackles the problem in quite
another way. He considers two isolated bodies A and B, interacting
with one another. Initially he assumes that A consists of m and B con-
sists of m' equal bodies a, where m and m' are integers. Let (p and cp'
denote the accelerations experienced by A and B, respectively, due to
their interaction. Then, says Mach, "if we take into consideration the
sign of the acceleration", we have that mini' = —cpVcp. If |(p| = |(p'|, we
say that A and B have the same mass. This linguistic convention would
be highly inconvenient if there existed three bodies, A, B, and C, such
that A and B had the same mass, and B and C had the same mass, but
C and A had different masses. Mach argues, however, that if this could
happen, it would lead to situations blatantly incompatible with the
experiences summarized by the principle of energy conservation. Let
A, B, and C be three perfectly elastic bodies constrained to move
around a fixed, circular, frictionless ring. Suppose that, under our con-
vention, A has the same mass as B and B has the same mass as C, but
that, in an interaction between A and C, A experiences a greater accel-
eration than C. Then, if the three bodies are initially at rest on the ring,
and we impress on A a speed v toward B, A will eventually transmit
this speed to B which will in turn transmit it to C; but when C reaches
A the latter will acquire a speed v' > v. The process will repeat itself,
and new kinetic energy will accrue to the three-body system with each
cycle, thus generating a perpetuum mobile. This thought experiment
shows, in the light of experience, that sameness of mass as defined

111
Paul Appell gave this definition in his Traite de mecanique rationelle: "The mass of
a particle is the constant ratio that exists between the intensity of a constant force
and the acceleration impressed by it on the particle" (1893, p. 87; quoted in Jammer
1961, p. 89).
240 The Rich Nineteenth Century

above is a transitive relation. Therefore a definite mass can be consis-

tently assigned to every body, at least ideally, by the following proce-
dure: Pick any body A as the standard of mass. Any body with the
same mass as A is assigned 1 unit of mass. The mass of any other body
B is put equal to |cp7(p| units, where cp' is the acceleration experienced
by B and cp is the acceleration experienced by a body C of mass 1,
when B and C interact in isolation. Note that |cp7(p| can be any real
number, not necessarily a rational one.
Mach discussed Newton's First Law in a course of lectures in the
summer of 1868. The gist of his comments was printed in a note at
the end of his booklet on the energy principle (1872; in Mach 1909,
pp. 46-50). A longer and better known version of his criticism
appeared in Die Mechanik (1883) and was amplified in later editions
(see M, pp. 216-71). Mach complains that the First Law - quoted in
§2.1 - is ambiguous. It speaks of "uniform motion straight ahead" but
says nothing of the bodies to which "the direction and speed of the
moving body" are referred. Newton, of course, thought that the direc-
tion should be taken in absolute space and the speed in absolute space
and time. Mach regarded this as sheer nonsense.
A motion can be uniform with respect to another motion. The question
whether a motion is uniform in itself has no sense at all. Likewise, we
cannot speak of an "absolute time" (independent of every change). This
absolute time cannot be measured against any motion, so it has no prac-
tical and also no scientific value; no one is entitled to say that he knows
something about it, it is an idle, "metaphysical" concept.
(Mach M, pp. 217)

No one can say anything about absolute space and absolute motion; they
are mere figments of thought (Gedankendinge) which cannot be pointed
out in experience. All our principles of mechanics [. . .] are experiences
about the relative positions and motions of bodies. Before testing them
it was neither possible nor permissible to adopt them for the regions in
which they are now considered valid. No one is entitled to extend these
principles beyond the limits of experience. Such an extension is indeed
senseless, for no one would know how to apply it.
(Mach M, pp. 222-23)
Mach was indeed aware of the experiment through which, according
to Newton, the presence of absolute space is made manifest: A bucket
is suspended from a long cord and turned about until the cord is
strongly twisted. The bucket is then filled with water and held at rest.
 4.4 Philosophers 241

If it is suddenly impelled in the opposite direction, and continues to

move for some time while the cord untwists, "the surface of the water
will at first be flat, as before the bucket began to move; but after that,
the bucket, by gradually communicating its motion to the water, will
make it begin sensibly to revolve, and recede little and little from the
center, and ascend to the sides of the bucket, forming itself into a
concave figure (as I have experienced), and the swifter the motion
becomes, the higher will the water rise, till at last, performing its rev-
olutions in the same times with the vessel, it becomes relatively at rest
in it" (Newton 1726, p. 10). According to Newton, the water's
endeavor to climb the bucket's walls and recede from its axis bears
witness to its real rotation in absolute space, for the water's surface
remained initially unchanged while it rotated only in appearance with
respect to the adjacent bucket. Mach proposes an altogether different
reading of these phenomena:
Newton's experiment with the rotating water bucket teaches us only that
the rotation of the water relative to the bucket walls does not stir any
noticeable centrifugal forces; these are prompted, however, by its rota-
tion relative to the mass of the earth and the other celestial bodies.
Nobody can say how the experiment would turn out, both quantitatively
and qualitatively, if the bucket walls became increasingly thicker and
more massive - eventually several miles thick.
(Mach M, p. 226)
According to this view the steady speed and direction of inertial
motion depends on and should be referred to, not the chimera of
absolute space, but the actual distribution of matter in the entire
world (pp. 227, 228). As we shall see in §5.4, Einstein believed for
some time that his General Theory of Relativity embodied this view of
inertia.
The following passage, highlighted by Mach, contains what he
describes as "the most important result" of the preceding reflections:
The seemingly simplest propositions of mechanics are quite complicated
by nature. They rest on experiences which are still incomplete and indeed
can never be fully completed. Practically, they are sufficiently assured to
furnish a basis for mathematical deduction with a view to the needed
stability of our environment; but they should by no means be regarded
as mathematically final truths, but rather as propositions which admit
and indeed require a continual experimental control.
(Mach M, p. 231)
242 The Rich Nineteenth Century

I do not doubt that words like these encouraged Einstein and his
younger contemporaries to proceed with their reform of physics.

4.4.4 Pierre Duhem (1861-1916)

Like Whewell and Mach, Duhem was a practicing scientist who
devoted an important part of his adult life to the history and philoso-
phy of physics. With his studies on the origins of statics (1905-1906),
on the mechanical tradition linked to Leonardo da Vinci (1906-1913),
and on cosmology from Plato to Copernicus (1913-1958), he founded
single-handed the history of medieval physics.112 His philosophy is con-
tained in La theorie physique: son objet, sa structure (1906), which
may well be, to this day, the best overall book on the subject. Its main
theses, although quite novel when first put forward, have in the mean-
time become commonplace, so I shall review them summarily without
detailed argument, just to associate them with his name. But first I
ought to say that neither in the first nor in the second (1914) edition
of his book did Duhem take into account - or even so much as mention
- the deep changes that were then taking place in physics. It is mainly
for this reason that I classify him as a nineteenth-century philoso-
pher.113 Still, the subsequent success and current entrenchment of
Duhem's ideas are due above all to their remarkable agreement with -
and to the light they throw on - the practice of mathematical physics
in the twentieth century.
In the first part of La theorie physique Duhem contrasts two opin-
ions concerning the aim of a physical theory. For some authors, it ought
to furnish "the explanation of a set of experimentally established laws",
while for others it is "an abstract system whose aim is to summarize
and logically classify a set of experimental laws, without pretending to
explain these laws" (Duhem 1914, p. 3). Duhem resolutely sides with

112
When I was young, Duhem's scholarship used to be invidiously compared with
Anneliese Maier's, whose admirable "Studies on the Natural Philosophy of Late
Scholastics" were being published just then. I suppose that, by now, one may
commend Duhem's pioneer work with equanimity.
113
In a review of his own work that he submitted to the Academie des Sciences when
he was proposed for membership, Duhem (1913) professed support for Energetics.
This was the nineteenth-century school that opposed the analysis of matter into
atoms and molecules (see note 55). By 1913, Einstein's theoretical work on Brown-
ian motion and Jean Perrin's experimental research had already persuaded the more
vocal energetists that their position was untenable.
 4.4 Philosophers 243

the latter. His rejection of the former rests on his understanding of

'explanation' ('explication' in French), which he expresses as follows:
"To explain, explicare,114 is to divest reality from the appearances which
enfold it like veils, in order to see that reality face to face" (pp. 3-4).
Authors in the first group expect from physics the true vision of things-
in-themselves that religious myth and philosophical speculation have
hitherto been unable to supply. Their expectation makes no sense unless
(i) there is, "beneath the sense appearances revealed to us by our per-
ceptions, [. . .] a reality different from these appearances" and (ii) we
know "the nature of the elements which constitute" that reality (p. 7).
Thus, physical theory cannot explain - in the stated sense - the laws
established by experiment unless it depends on metaphysics and thus
remains subject to the interminable disputes of metaphysicians. Worse
still, the teachings of no metaphysical school are sufficiently detailed
and precise to account for all the elements of a physical theory (p. 18).
Duhem instead assigns to physical theories a more modest but
autonomous and readily attainable aim:
A physical theory is not an explanation. It is a system of mathematical
propositions, derived from a small number of principles, whose purpose
is to represent a set of experimental laws as simply, as completely and
as exactly as possible.
(Duhem 1914, p. 24)

Duhem's book is a long gloss on this definition. At one point, it may

seem trite: Physics advances by abstraction and generalization, from
facts to laws and from laws to theories. First, an enormous variety of
particular, complex facts is analyzed to find what they have in common.
This is summarized "in a law, that is, a general proposition combin-
ing abstract notions". Then a whole set of laws is considered. "A very
small number of extremely general statements, about some very
abstract ideas" is substituted for them. The physicist "chooses these
primary properties, formulates these fundamental hypotheses in such
a way that every law belonging to the set under study can be drawn
from them by a possibly very long but very secure deduction. This
system of hypotheses and the consequences that flow from them [. . .]
constitute the physical theory as defined by us" (p. 77). However,
Duhem gives a peculiar twist to this seemingly facile view: Physical

114
In Latin in the original. From 'plico' = 'I fold'; so 'explico' = 'I bring out of its folds,
I unfold'.
244 The Rich Nineteenth Century

laws are experimental and physical theories are mathematical-, this

combination of mathematics and experiment can only be achieved by
putting theories to work in the design of experiments and the gather-
ing of facts under laws.
According to Duhem, there are three operations involved in the con-
stitution of any physical theory (pp. 197f.). In the first place, a few
properties are chosen among the many revealed by observation; they
are regarded as primary and are represented by algebraic or geometric
symbols. Certain relations are postulated between these symbols; they
are the principles or "fundamental hypotheses" of the theory. The third
operation is the mathematical development of the theory. "Its purpose
is to teach us that, by virtue of the fundamental hypotheses [...], the
combination of such and such circumstances will bring about such and
such consequences; to announce to us, for example, that by virtue of
the hypotheses of thermodynamics, if we subject an ice block to such-
and-such compression, the block will melt when the thermometer will
indicate such-and-such a degree" (p. 198).115 The mathematical devel-
opment must bond (se souder) with observable facts at both ends. This
can be achieved only by a double translation: "To introduce the cir-
cumstances of an experiment into the calculations, a translation must
replace the language of concrete observation with the language of
numbers; to check the outcome predicted by the theory for this exper-
iment, a numerical value must be transformed into an indication stated
in the language of experience" (p. 199). According to Duhem, both
translations involve a measure of indeterminacy (or, as he bluntly says,
of betrayal). He bases this on the presumption that algebraic variables
must stand for numbers (p. 158). Such a view of algebra is surely too
narrow, but it is not one that can be easily countered with examples
drawn from the literature of physics. Anyway, to vindicate Duhem's
allegation it is enough to consider that every term occurring in a math-
ematical argument is precise and unambiguous in a way in which "the
language of concrete observation" is not. "The facts of experience, in
all their native brutality, cannot be used in mathematical reasoning. To

115
Duhem's description of the mathematical development of a theory agrees fairly well
with what Hempel and Oppenheim (1948), Braithwaite (1953), and others call the
explanation of laws by theories. They chose to retain this prestigious name for the
aim of science, although they conceived it very much like Duhem. Their decision
gave rise to a vast and mostly tedious literature about the true meaning of 'expla-
nation'. Whoever has had to wade through it is bound to admire the wisdom of
Duhem, who gave up the term without fuss.
 4.4 Philosophers 245

feed such reasoning they must be transformed and put into symbolic
form" (p. 298).
The indeterminacy of translation follows at once from Duhem's dis-
tinction between "practical" and "theoretical" facts. A practical fact
is a concrete fact described in ordinary language just as one happens
to observe it. A "theoretical fact", instead, is the "set of mathematical
data which replaces a concrete fact in the reasonings and calculations
of the theoretician" (p. 199). In such a fact nothing is vague or unde-
cided. If it concerns the temperature distribution in a body, the latter
is geometrically defined, its edges are widthless lines, its points are
dimensionless, lengths and angles are exactly assigned, and each point
of the body has a definite temperature, a number that cannot be con-
founded with any other number. A practical fact is a far cry from this,
and it cannot be described "without attenuating by means of the words
more or less (a pen pres) whatever is too determined in each statement"
(p. 200). Because of it, "an infinity of different theoretical facts can be
taken as the translation of the same practical fact" (p. 201), and vice
versa (pp. 225, 229, 230).
Of course, Duhem uses 'translation' as a metaphor.116 While the
rules of genuine translation, for example, between Greek and Latin,
are not a part of either language but can be established by comparison
from the outside (say, by a German philologist), the rules of Duhemian
translation are an integral part of the physical theory to and from
which the translation is effected. This affects the very essence of phys-
ical experiments.
A physical experiment is the precise observation of a group of phenom-
ena together with the INTERPRETATION of these phenomena; this inter-
pretation substitutes, for the concrete data actually collected, abstract
and symbolic representations which correspond to them by virtue of the
theories accepted by the observer.
(Duhem 1914, p. 222; Duhem's emphasis)

One cannot use the instruments found in physical laboratories if one

does not substitute for the concrete objects which compose those instru-

116
The metaphor was popularized in the English-speaking world by Campbell (1920).
I am not sure that everyone who used it realized that it was only a metaphor, and
that, for instance, every sentence in a paper by Rutherford is written in one and the
same language, viz., English. (Indeed, the mathematical and chemical formulas are
ideograms that can be read in any civilized language, just as Chinese characters can
be read in Mandarin and Cantonese.)
246 The Rich Nineteenth Century

merits an abstract and schematic representation which provides a handle

for mathematical reasoning, if one does not subject this combination of
abstractions to deductions and calculations which involve the acceptance
of theories.
(p. 231)

When a physicist makes an experiment two quite different representa-

tions of the instrument on which he operates occupy his mind. One is
the image of the concrete instrument he actually handles; the other is a
schematic model {type) of that same instrument, constructed by means
of symbols furnished by the theories. And, mark you, it is about this
ideal and symbolic instrument that he reasons, applying the laws and
formulas of physics to it.
(p. 235)
Since the laws of physics are based on the results of physical exper-
iments, the character of the latter affects the nature and scope of the
former. "A physical law is a symbolic relation whose application to
concrete reality requires that one knows and accepts a whole collec-
tion of theories" (p. 254). Such a law is always provisional and rela-
tive. It is provisional, not because it is true today and will be false
tomorrow, but because it represents the facts to which it is applicable
with an approximation that physicists now consider satisfactory, yet
will eventually judge insufficient. It is relative, not because it is true for
one physicist and false for another, but because the approximation
involved in it is good enough for one physicist's use of it and not for
the other's (p. 260). Indeed, "one can see the same physical law simul-
taneously adopted and rejected by the same physicist in the course of
the same work" (p. 262).117
Yet physicists show, as a matter of fact, "an irresistible aspiration"
toward a physical theory that would represent all experimental laws
by means of a single, logically consistent theory (p. 449). This aspira-
tion has been present and has exerted a powerful influence throughout
the history of physics. According to Duhem, neither physical knowl-
edge nor the philosophical analysis of the structure of physical theory

117
An outstanding example of this procedure was provided soon thereafter by Einstein
(1915h) when he computed the anomalous perihelion advance of Mercury using his
new law of gravity, while at the same time accepting the value computed by Newton's
law of the - 12 times larger - perihelion advance attributable to the presence of the
other planets. See §§5.4, 5.5, and 7.2.
 4.4 Philosophers 247

can justify this aspiration; so the physicist who yields to it - and which
does not? - implicitly endorses a metaphysical creed:
Which is this metaphysical proposition that the physicist will affirm, as
if by force, despite the restraints imposed by the method he habitually
uses? He will aver that, beneath the sensible data solely accessible to his
methods of research, realities are hidden whose essence cannot be
grasped by those same methods; that these realities are disposed in a
certain order which physical theory cannot contemplate directly; but that
physical theory, by its successive perfectionments, tends to arrange its
experimental laws in an order which is ever more similar to the tran-
scendent order by which realities are classed; so that, by virtue of this,
physical theory gradually approaches its limiting form as a natural clas-
sification^finally,that logical unity is a character without which physi-
cal theory cannot claim the status of a natural classification.
(Duhem 1914, p. 450)
These are the central tenets of a "believer's physics",118 which no reason
can justify, but without which "it would be unreasonable to work in
the progress of physical theory" (1908, p. 19). It often happens that a
hitherto unsuspected physical law is derived from a physical theory
and subsequently corroborated by experiment (1914, p. 450). Events
like this "press the physicist to assert that physical theory, as it pro-
gresses, becomes more similar to a natural classification, which is its
ideal and its goal" (p. 452). However, Duhem's strong and disciplined
intellect would not let him claim that this assertion is proved by such
events.
At any rate, it is clear that physics cannot proceed by induction
alone. For, as we have seen, no generalization from experiment can be
of any use to physics "unless it undergoes an interpretation which
transforms it into a symbolic law; and this interpretation involves the
acceptance of a whole set of theories" (p. 303). Many "symbolic trans-
lations" being equally admissible, the physicist must choose among
them "the one which will provide a fruitful hypothesis to the theory,
without his choice being guided by experience in any way" (Ibid.).
Since "the execution and interpretation of any physical experiment
involve the acceptance of a whole ensemble of theoretical proposi-
tions", theoretical physics can only be tested as a corporate body.

118
"Physique de croyant" is the title of the article from which the last quotation is
taken; it appeared in Annales de Philosophie chretienne in the fall of 1905 and was
later included by Duhem in the second edition of La theorie physique.
248 The Rich Nineteenth Century

The physicist can never subject an isolated hypothesis to experimental

control, but only a whole ensemble of hypotheses; when experience dis-
agrees with his previsions it teaches him that at least one of the hypothe-
ses in this ensemble is unacceptable and should be modified; but it does
not indicate to him which one should be changed.

Physics is not a machine one can take apart; one cannot try each piece
in isolation and wait, to adjust it, until its solidity has been minutely
checked. Physical science is a system that must be taken as a whole. It
is an organism no part of which can be made to function without the
remotest parts coming into play, some more, some less, but all in some
degree.
(Duhem 1914, pp. 284-85)
In every epoch physicists agree that certain elements in their system
are beyond question, so that any modification required by experimen-
tal evidence must bear on other elements. However, the privileged
status of this hard core of physics does not rest on any logical neces-
sity. So a physicist can always refuse to reconcile the theoretical scheme
with the facts by invoking causes of error and introducing corrections;
instead, "by resolutely bringing reform to the propositions declared
intangible by common consent, [he or she might] fulfill the work of a
genius who opens theory to a new career" (pp. 321f.; cf. p. 328).
According to Duhem, the choice of physical hypotheses is subject only
to these restrictions: Each must be consistent with itself and with the
rest, and they must form a system such that, "from their ensemble',
mathematical deduction can draw consequences that represent, to a
sufficient approximation, the ensemble of experimental laws" (p. 335).
Otherwise, theoreticians are free to lay down the foundations of their
systems in any way they think fit. When Duhem died in 1916,
Einstein's dazzling display of freedom in physics had been going on
for over a decade.
 CHAPTER FIVE

Relativity

'Relativity theory' or simply 'Relativity' is the standard name of two

quite different, yet subtly related theories put forward by Albert Ein-
stein in 1905 and 1915, respectively. The first, Special Relativity (SR),
rescued the Maxwell equations from seemingly catastrophic experi-
mental results by making deep changes in the basic concepts and laws
of Newtonian mechanics. The second, General Relativity (GR), solved
the problem of reconciling SR with Newton's theory of gravity by
transcending them both.1 For those of us who cherish physico-
mathematical theories more for their inherent beauty than for their
transient accuracy, SR and GR remain unmatched. Moreover, to this
day, they have enjoyed tremendous empirical success. SR is corrobo-
rated daily in every high-energy lab. GR accounts for all the phenom-
ena Newton classified as gravitational just as well or even better than
his theory. Moreover, it provides an amazing gravitational explanation
of other phenomena - such as the systematic shift in the spectrum of
light from distant galaxies and the pervasive microwave background
radiation -, which nobody even suspected c. 1910 and which would
not easily fit in a Newtonian framework.
This is not the place to deal even superficially with the many fruit-
ful applications of SR and GR. Our attention must go to their con-

1
The difference and the relation between SR and GR will, I hope, be made clear in this
chapter. For the time being let me just remark that SR can hold in a world governed
by GR only if that world is completely lacking in gravitational sources. Still, in a world
like ours, GR agrees excellently with SR in a small, freely falling lab, over short periods
of time. It is therefore not altogether unjustified to describe SR as a special version
and GR as a general version of one and the same theory. On the other hand, this
description conceals the drastic change of meaning and scope that SR suffers when
embedded in GR.

249
250 Relativity

ceptual problems and their philosophical significance. The latter has

been judged differently by different authors. In my view, it lies chiefly
in the fact that both theories are exemplary cases of far-reaching con-
ceptual change in fundamental physics, firmly rooted in the tradition
they go beyond. As such, they illustrate with exceptional clarity the
way in which rupture and continuity are combined in the history of
physics. Conceptual innovation was also the main source of the so-
called philosophical problems of Relativity, which generally stem from
misunderstandings and were proposed by physicists and philosophers
who resisted Einstein's reckless assault on their encrusted modes of
thought.

5.1 Einstein's Physics of Principles

Let me first explain the problem to which Einstein responded with SR.
By virtue of Newton's principle of relativity, any inertial frame of ref-
erence may be looked on as a frame at rest to which all motions are
referred (see Chapter Two, note 17, and the text linked to it). The valid-
ity of this principle remained unquestioned while physics dealt with
interactions that cause acceleration and depend on the mutual distances
- but not on the relative velocities - of the interacting bodies, that is,
until the 1850s (see the quotations from Helmholtz 1847, p. 6, on page
185). This changed with the advent of Maxwell's electrodynamics. The
constant speed c with which light travels in vacuo takes pride of place
in the Maxwell equations. Moreover, the velocity v with which a
charged body moves through an electromagnetic field determines both
the direction and the magnitude of the force that the field exerts on it.
Now, if v is the momentary velocity of an object B relative to an iner-
tial frame 3% and 2F in turn moves with constant velocity w relative to
another frame *§, then - one presumed - the same object's velocity rel-
ative to ^ is surely v + w. So, if F stands for the force per unit charge
exerted on our object by the electric and magnetic fields E and B, the
expression F = (v x B) + E (eqn. (7.3)) must be referred to a particu-
lar such frame.2 One assumed, rather naturally, that this is the frame
in which the electromagnetic aether, the site of the stresses represented
by E and B, is permanently at rest. It was - one concluded - relative

2
Otherwise, an object of unit mass that experiences acceleration (v x B) + E relative to
frame SF must experience acceleration ((v + w) x B) + E relative to frame % which is
absurd.
 5.1 Einstein's Physics of Principles 251

Figure 12

to this frame, and to this frame only, that light travels in vacuo in any
direction with speed c.
Of course, the earth does not rest in the aether, at least not the whole
year round.3 So one should, in principle, be able to measure the speed
u of the earth in the aether by the observation of optical effects. None
were detected. It was shown, however, that all first-order effects of the
earth's motion - that is, effects that depend on multiples of the ratio
ulc, but not on (u/c)2, (u/c)3, or other higher powers of this ratio -
cancel out if one assumes with Fresnel that a small fraction of the aether
is dragged by matter. On the other hand, the means of detection avail-
able until 1880 were insensitive to second-order effects, depending on
the much smaller ratio (u/c)2. But Albert Michelson's interferometer,
which was first built about that time, was capable of disclosing such
effects. It consists of two perpendicular steel arms, one of which is
aligned with the motion of the earth (see Fig. 12). Light issuing from
a common source is divided into two rays that travel along each arm,
are reflected by mirrors at either end, and then are mixed together to
produce a pattern of interference fringes that are observable through
a microscope. The pattern should vary when the instrument is rotated
by 90° (so that the arm that was formerly aligned with the motion of
the earth is now perpendicular to it, and vice versa). However, neither
Michelson (1881) nor Michelson and Morley (1887) - using an
improved interferometer - could measure any significant change in the
interference pattern.

3
The motion of the earth in the aether follows at once from the Copernican hypothe-
sis. Empirical proof was provided by Bradley's discovery of the aberration of starlight
(1728).
252 Relativity

To explain the negative result of Michelson's and Morley's experi-

ment, G. F. FitzGerald (1889) and H. A. Lorentz (1892) independently
conjectured that solids that move through the aether with speed v con-
tract in the direction of motion by a factor of Vl - {vie)1. This factor
is evidently very close to 1 if c is much larger than v. The conjecture
drew on the plausible idea that solids are held together by electro-
magnetic forces and should therefore suffer some deformation when
they move across the aether. The expression for the Lorentz-FitzGer-
ald contraction factor was not derived from some specific theory of the
solid state, but was directly figured out to compensate for the differ-
ence in the speed with which light was supposed to travel along one
and the other arm of the interferometer. Lorentz realized later that, to
obliterate all the predictable - yet not forthcoming - effects of motion
in the aether, he would have to tamper not only with lengths but also
with durations. He introduced the enigmatic notion of a local time,
measured on moving objects and 1/Vl - {vie)1 times longer than the
real time elapsed between the same events. In this way, the discussion
of space and time coordinates and of coordinate substitution eventu-
ally took center stage in Lorentz's study of the motion of bodies in the
quiescent aether. It should be noted, though, that the coordinate sub-
stitutions found in Lorentz's papers as late as 1904 are not quite the
same thing as the exact transformations that now bear his name (cour-
tesy of Poincare 1906), but are more in the nature of coordinate adjust-
ments, made necessary by the physical alteration of moving clocks and
rods, and avowedly valid only to a specified approximation.4
The task of describing electromagnetic phenomena relative to
moving bodies was handled in a totally different spirit by Einstein
(1905r). His work on radiation (1905i; see §6.1) persuaded him that
contemporary views concerning the microphysical basis of electrody-
namic phenomena needed thorough revision. Time was not ripe yet for

4
Lorentz (1904, §4, eqns. (4) and (5); my notation) introduces the auxiliary substitu-
tion x' = kit,, y = h\, Z' = /£, f = hk'1 - klvfy-2, where k = 1/(1 - (vie)1) and "the
coefficient / . . . is to be considered as a function of v whose value is 1 for v = 0, and
which, for small values of v, differs from unity no more than by a quantity of the
second order". Suppose now that £, t|, £, and x, are related to x, y, z, and t by the
Galileian transformation £ = x - vt, r\ = y, £ = z, x = t. Then the coordinate systems
(x,y,z,t) and (x',y',z',f) will be related by a Lorentz transformation - eqns. (5.5) on
page 257 - if and only if / = const. = 1. Lorentz argued for this equality from physi-
cal considerations. Poincare showed that unless / = 1, the coordinate substitutions
introduced by Lorentz do not constitute a group.
 5.1 Einstein's Physics of Principles 253

a new construction founded on hypotheses about the structure of

matter. So Einstein took his cue from classical thermodynamics, which
accounts for a vast and ubiquitous class of phenomena, without
making any assumptions about their deep underlying structure, by
means of two universal principles (cf. §4.3.2). He also proposed two
universal principles, which can be paraphrased as follows:
Relativity Principle. The laws by which the states of physical systems
change do not depend on the particular inertial frame of reference to
which those changes of state are referred.
Light Principle. Every light signal travels in vacuo relatively to a given
inertial frame of reference with constant velocity c, no matter what the
state of motion of its source.
The meaning of these statements will be clarified below; but there is
no doubt that, prima facie, they agree with experience. Indeed, Lorentz
and others strained their imaginations to explain why they did so,
although, on purely conceptual grounds, they ought not to. Instead of
postulating new microphysical interactions to account for the unex-
pected agreement, Einstein embarked in conceptual criticism. He
showed that one can very well conceive that the same light signal
travels with the same speed relative to two different frames moving
with constant velocity relative to each other; and that the common pre-
sumption that this is logically impossible rests on the uncritical accep-
tance of a specific, not logically necessary, scheme for the description
of motion. As it turns out, this scheme is admissible if and only if
neither light nor any other physical disturbance propagates with the
same speed in all inertial frames; and so its adoption by classical
physics unwittingly begs the question at issue.
To simplify matters I shall compare two inertial frames 2F and 3*\
endowed with right-handed systems of Cartesian coordinates x, y, z
and x\ y\ z\ such that on a particular instant the three axes of
the primed Cartesian coordinate system lie, respectively, along the
homonymous axes of the unprimed system.5 The frames are also
endowed with universal time coordinates t and t' - on which I shall
have more to say -, which take the same value 0 at the origins of the

5
A system of Cartesian coordinates x, y, z is right-handed if one can place the right
hand at the origin in such a way that the thumb points along the x-axis in the direc-
tion of increasing x, the forefinger points along the y-axis in the direction of increas-
ing y, and the middle finger points along the 2-axis in the direction of increasing z.
254 Relativity

Cartesian systems when both origins coincide. All coordinates of the

same kind are expressed in the same measurement units (e.g, meters
and seconds, or light-years and years).6 I assume moreover that SF'
moves relatively to 3F parallel to the x-axis, with constant speed v, in
the direction of increasing x. Before Einstein, mathematical physicists
took for granted that, in such a case, the primed and unprimed coor-
dinates described above are related by the following so-called Galileian
transformation?

x' = x - vt

t' = t

If this is so, then, necessarily, a photon traveling in 3F along the x-axis,

with constant speed c = dx/dt, travels in 3F' with constant speed

dx' d(x-vt) dt dx .r ..

(For brevity's sake - here and in the rest of this chapter - I say 'photon'
instead of 'light signal in vacuo\s) Obviously, Einstein's two principles

6
The general case can be reduced to this special case by performing a few, presumably
unproblematic operations, viz., (i) transform whatever space coordinates each frame
is endowed with into right-handed Cartesian coordinates; (ii) choose an instant of time
and apply to one of the Cartesian systems a translation that takes its origin to the
point which lies at that instant on the origin of the other Cartesian system; (iii) apply
to one Cartesian system a rotation that brings its axes to point in the same directions
as the axes of the other; (iv) add or subtract a constant to each value of the time coor-
dinates so that both take the value 0 at the common origin of the Cartesian systems
on the chosen instant; and (v) convert all coordinates to common measurement units.
7
A Galileian transformation is any coordinate transformation that results from apply-
ing - in any order - one or more coordinate transformations chosen from the fol-
lowing sets: (i) the Galileian boosts, defined by eqns. (5.1); (ii) the space translations
such as the one mentioned in (ii) of note 6; (iii) the space rotations such as that men-
tioned in (iii) of note 6; and (iv) the time translations described in (iv) of note 6. The
Galileian transformations form a group, of which each one of the sets (i)—(iv) is a sub-
group. Of course, Galileo Galilei never wrote down a Galileian transformation, and
presumably did not even dream of them.
8
The idea that electromagnetic radiation is emitted and absorbed in discrete quanta of
energy - subsequently called 'photons' - is due to Einstein (1905i); but he did not
 5.1 Einstein's Physics of Principles 255

cannot be jointly true if the right scheme for the description of motion
includes (5.1).
The transformation (5.1) involves essentially the quantity v, equal
to the distance in 2F that any point of 3F' traverses in unit time. Con-
sider the origin A' of the primed Cartesian coordinate system. At time
t = 0, it lies on the origin A of the unprimed system. So, if A' passes
by the point B in 9^ at time t = 1, then v equals the distance from A to
B. This statement makes no sense unless the time coordinate t associ-
ated with the unprimed frame 2F is defined at both A and B. Since B is
arbitrary, there must be a definite way of setting up t throughout 8F.
Since 3F is, by definition, an inertial frame of reference, t must be such
that a force-free particle traverses equal distances (in 3F) in equal t-
intervals. But this requirement leaves a wide latitude of choice. Suppose
that it is satisfied by a time coordinate function t. Then, evidently, it is
also satisfied by the coordinate function t* given by
t* = ax + by + cz + dt + k (5.3)
where a, fc, c, d, and k are arbitrary constants.
The problem of defining a time coordinate for an inertial frame of
reference is dealt with by Einstein in 1905r, §1. 9 He discusses it in
picturesque, seemingly down-to-earth terms, as a matter of synchro-
nizing distant clocks; but, of course, anyone trying to practice the
"operational" method of definition that he proposes would fail to
reach the perfect accuracy that he subsequently takes for granted (when
the time coordinate thus set up is used in the equations of physics). An
obvious method of distant clock synchronization is implicit in eqns.
(5.1). If t = t\ clocks placed along the x-axis of 3F can be synchronized
by comparing them with the clock at A' as it passes by them; clocks
placed along other axes through A can be set by a clock at the origin
of other suitably moving frames, whose time coordinate also agrees
with t by the appropriate analogues of eqns. (5.1). Since (5.1) precludes
the joint validity of Einstein's principles, this method of defining the
time coordinate function on the inertial frame 3F - known as synchro-
nization of distant clocks by clock transport - begs the question. The

mingle it with his work on the Relativity Principle. I therefore use 'photon' in this
chapter not as a term of art, but only as a useful abbreviation.
As far as I know, James Thomson (1884) was the first to mention this problem in
writing, but he glossed over it; cf. Chapter Two, note 16, and the text linked to it. It
was later discussed by Poincare (1898).
256 Relativity

method proposed by Einstein is, at first blush, no less obvious than the
former; indeed, the big surprise is that, as a matter of fact, they do not
agree with each other (unless clock transport takes place at infinitely
slow speed - cf. Eddington 1923, §§4, 11). The time coordinate t
defined at point A of the inertial frame 2F is diffused through all space
by means of rebounding photons. Let B be a point on 3F outside A. A
photon is sent from A to B at time t\. If the photon, after rebounding
at B, returns to A at time £3, the time t2 at which it reaches B is, by
definition,

t2=t1+-(t3-t1) (5.4)

Thus the photon takes the same time to go from A to B and from B to
A. Since B is arbitrary and ranges over all 3% the photons employed in
setting up a time coordinate function in this way satisfy the Light Prin-
ciple by definition - if indeed the procedure can be consistently per-
formed. A coordinate function t defined by this method will henceforth
be referred to as Einstein time. It must be understood that the refer-
ences to time implicit in the Relativity Principle - when it talks about
laws of change - and in the Light Principle - when it mentions light
velocity - are references to Einstein time.
The blatantly conventional character of the definition of Einstein
time took most physicists and philosophers by surprise and became the
source of endless debate. I shall deal with some of it in §5.3.2. But
before proceeding any further it is important to realize that, although
the definition of Einstein time rests on an agreement, this could not
work as expected if some things happened differently. As Einstein
says, one assumes that such a definition "is possible, without contra-
diction, for any number of points" (1905r, p. 894), that is, that no
inconsistencies will arise as points A and B range freely over space.
Specifically, (i) Einstein time defined from A at B should agree with
Einstein time defined from B at A; and (ii) if Einstein time is defined
from A at B and C, and then again from B at C, both time coordinates
should agree at C. Einstein does not add that (iii) Einstein times defined
from A by sending out photons at different moments must agree
with each another - but he presumably took it for granted. Evidently,
such agreements cannot be secured by convention. I shall refer to them
as 'the consistency conditions of Einstein time'. They can hold only
if all photons obey the Light Principle, not just the one by means of
 5.1 Einstein's Physics of Principles 257

which Einstein time is defined from a particular point at a particular

moment.
Suppose then that the coordinate functions t and t' are Einstein times
defined, respectively, for the inertial frames 8F and 3F', and that (x,y,z)
and (x\y\z') are Cartesian systems for each frame. Einstein was able
to prove, from his two principles, that the primed and unprimed coor-
dinates are then related by the following transformation, called a
Lorentz boost:10

(5.5)

vx
t' =

In his proof, Einstein assumes from the outset that the sought for trans-
formation must be linear "because of the properties of homogeneity
which we attribute to space and time" (1905r, p. 898). In fact, however,
linearity follows from homogeneity only if the latter is a feature of the
four-dimensional "space" compounded of space and time (i.e., the 4-
manifold charted by the coordinate systems (x,y,z,£) and (x\y\z\f) -
cf. §4.1.3), not just of space on one hand and of time on the other.
This is interesting because it shows that Minkowski's reading of SR
(§5.2) was already implicit in Einstein's first steps. Yet at this point he
could still have done without it, for the linearity of the transformation

10
A Lorentz transformation is any coordinate transformation that results from apply-
ing - in any order - one or more coordinate transformations chosen from the fol-
lowing two sets: (i) the Lorentz boosts, defined by eqns. (5.5), and (ii) the space
rotations such as that mentioned in (iii) of note 6. The Lorentz transformations form
a group, of which the Lorentz boosts are a subgroup. The Lorentz group, in turn, is
a subgroup of the Poincare group. A Poincare transformation is any compound
of Lorentz transformations and space and time translations, such as those described
in (ii) and (iv) of note 6. The designation does justice to Poincare's independent
discovery of the transformation (5.5) and the major group to which it belongs (see
note 12).
258 Relativity

(x,y,z9t) •-> (x\y\z\f) can also be inferred from Einstein's two princi-
ples (cf. Torretti 1983, pp. 75f.). Ignatowsky (1910, 1911) tried to
derive eqns. (5.5) from the Relativity Principle alone. However, to dis-
pense with the Light Principle he had to use the following

Principle of Reciprocity. If the inertial frame W moves with velocity v

relatively to the inertial frame 3% the inertial frame 8F moves with veloc-
ity -v relatively to the inertial frame 3F'.

Contrary to Ignatowsky's belief, the Principle of Reciprocity does not

follow from the Relativity Principle.11
It follows at once from eqns. (5.5) that two events that happen at the
same time t cannot happen at the same time f unless their x-
coordinate is the same. In other words, two events simultaneous in an
inertial frame of reference are also simultaneous in another frame
moving uniformly with respect to the former only if they take place on
a plane perpendicular to the direction of motion. Such relativity of
simultaneity flies in the face of ingrained commonsense representations
and is the source of a galling philosophical problem (see §5.3.4). Yet
without it the Light Principle and the Relativity Principle cannot be
jointly held. Consider a light wave emitted from A in every direction at
the one instant in which A coincides with A'. By the Light Principle, the
wave front takes up in 3F and 8F, at each subsequent instant, a sphere
centered at A and A', respectively. This would be absurd if each instan-
taneous location of the wave front in 2F were also an instantaneous loca-
tion of the wave front in 9*'. It is, however, perfectly possible if, due to
the frame dependence of simultaneity, the events that constitute the
wave front at a given time in 2F are not simultaneous in SF' and hence do
not constitute the wave front at any time in 2F'. The sphere in 3F centered
at A, which the wave front covers at some instant of Einstein time £, is

11
Ignatowsky (1911, p. 5) argues thus for the Principle of Reciprocity: Let AB and A'B'
be rods of unit length at rest, respectively, in the inertial frames <3> and W. As W
moves, A'B' slides along AB, so that B' passes A and B, respectively, at times tQ and
t\ of the unprimed system. Let t'o and t\ be the times of the primed system at which
A passes B' and A\ respectively. Then, according to Ignatowsky, the Relativity Prin-
ciple requires that At = ti - t0 = t\ - t'o - At'. Consequently, the speed 1/At of W in
^ equals the speed VAt' of 2£ in SF', and the velocities of each frame relative to the
other can only differ in sign. But Ignatowsky overestimates the strength of the Rela-
tivity Principle. Relativity entails that At is the same function of At' as Af is of At,
not that this function is the identity.
 5.1 Einstein's Physics of Principles 259

certainly not a sphere in W centered at A', but neither is it the shape of

the wave front at any instant of Einstein time t'.
Equations (5.5) also entail the relativity (frame-dependence) of three
physical quantities hitherto regarded as fundamental, viz., length, dura-
tion and inertial mass. This caused astonishment and much discussion.
I deal with these issues in §§5.3.1, 5.3.3, and 5.3.5.
Equations (5.5) lead to a rule for the transformation of velocities -
from one coordinate system to another - quite different from the simple
classical law of addition used in eqn. (5.2). An object moving in 3F
along the x-axis with velocity u = dx/dt, moves in 2F' along the x'-axis
with velocity u' - dx'ldt'. But uf is not equal to u - v. By substituting
from eqns. (5.5), we obtain after a short calculation:
u-v
-\uvjcl)
Note that, in consonance with the assumption that photons travel with
the same speed in all inertial frames, eqn. (5.6) implies that if u = c,
then u' - c as well.
The Relativity Principle, as stated above, is a straightforward gen-
eralization to all laws of physics of Newton's principle of relativity
embodied in Corollary V to his Laws of Motion. However, in combi-
nation with the Light Principle and with Einstein's introduction of a
physically significant, frame-dependent, universal time coordinate, the
Relativity Principle yields eqns. (5.5), which are incompatible with
Newton's Laws of Motion. In this new context, the Relativity Princi-
ple translates into the following
Principle of the Lorentz Invariance of the Laws of Physics. The laws
by which the states of physical systems change, expressed in terms of
Cartesian coordinates and Einstein time for an inertial frame of refer-
ence, are invariant under Lorentz transformations.
Now, the Maxwell equations of electrodynamics automatically meet
this requirement, as if by magic, if the electric and magnetic field
vectors are transformed in a certain way.12 But Newton's Laws of

12
Einstein (1905r, §6) showed this for the Maxwell equations as formulated by Hein-
rich Hertz. The transformation of the electric and magnetic field components by the
Lorentz boost (5.5) is given in eqns. (5.21). Continuing with notes 4 and 10, one
might wish to say that what Poincare (1905, 1906) and Einstein (1905r) discovered
in the wake of Lorentz's groping efforts (1895, 1899, 1904) was the symmetry group
260 Relativity

Motion - and his Law of Gravity - do not. So Einstein's principles

brought about at once a deep revision of classical mechanics (and soon
moved him to seek for a new account of gravitational phenomena).13
Still, if the relative speed v of the inertial frames being considered is
much smaller than the speed of light c, the Lorentz transformation (5.5)
does not differ significantly from the Galilei transformation (5.1),
under which Newton's laws are invariant. And the rule of transfor-
mation of velocities (5.6) practically agrees with the rule of addition u'
- u ± v, if both u and v are much less than c. So, whatever support
accrued to Newtonian physics from experiments involving speeds much
less than that of light stood ready to corroborate the new physics based
on Einstein's principles.14

5.2 Minkowski's Spacetime

In lectures delivered on 5 November 1907 (Minkowski 1915) and 21
September 1908 (Minkowski 1909), the mathematician Hermann
Minkowski proposed a cogent reading of Einstein's new physics that
clarified its baffling features. He showed how to substitute for the
£/?re£-dimensional space continuum and the owe-dimensional time
continuum with their separate metrics, presupposed by Newtonian
physics, a /bwr-dimensional continuum metrically structured in conso-
nance with the physical behavior of photons and free material parti-
cles. Einstein's principles follow at once from this new geometry.
Therefore, the experimental results that confirm these principles also
support Minkowski's declaration that, "from now on, space for itself

of Maxwell electrodynamics, which is now usually called the Poincare group. (Still,
Minkowski 1909, p. 106nl, traces this discovery back to Voigt 1887.) But, whereas
Einstein boldly proceeded to make this group into the symmetry group of nature,
Poincare squandered the glory of his discovery in a hopeless inquiry into the defor-
mation of electrons in motion.
13
Of course, the Maxwell equations cannot remain unscathed if mechanics is reformed,
for, in the classical understanding of them, they concern the relations between elec-
tric and magnetic forces (in the Newtonian sense of 'force'). So, even if in SR the
Maxwell equations retain their shape, they must change their meaning. I touch on
this matter in passing in §5.3.5; see also Torretti (1983, pp. 108ff.).
14
I should also mention that Einstein derived from his principles new, more accurate
formulas for the optical effects hitherto explained by a partial aether drag. Indeed,
after Einstein, twentieth-century physicists discarded the aether as lightheartedly as
their nineteenth-century predecessors had jetissoned caloric.
 5.2 Minkowski's Spacetime 261

and time for itself should completely reduce to shadows, and only a
sort of union of both ought to retain autonomy" (1909, p. 104). For
this union of space and time Minkowski used the term 'world', but
today we normally call it 'spacetime'.
Before proceeding any further with the explanation of spacetime, I
should emphasize that the idea of a four-dimensional continuum in
which we live, move, and have our being comes a good deal closer to
our experience than the traditional ideas of time separate from space
and hence from motion, and of space separate from time and hence
from all forms of change. Think of traffic on a motorway. One might,
with considerable effort, imagine the different geometrical figures that,
at successive instants, the vehicles cut in space; but the phenomenon
one perceives is a single coherent spacetime flow. Still, Minkowski's
idea of spacetime does not stem from a desire to grasp life's course less
abstractly, but from a consideration of what is actually involved in the
physicist's practice of describing the phenomena of motion by means
of four coordinate functions, especially if these are related to inertial
frames and to one another in the manner proposed by Einstein. The
familiar Cartesian coordinate system (x,y) used in plane geometry is a
mapping of the Euclidian plane onto the (structured) set of real number
pairs IR2. The coordinate systems (t,x,y,z) used in Newtonian and SR
kinematics - combining a universal time coordinate with Cartesian
space coordinates - are mappings onto the (structured) set of real
number quadruples R4. But what are they mappings of} What is the
domain of such kinematic coordinate systems? We can only think of it
as a (structured) set of points, each one of which is the possible loca-
tion of an instantaneous pointlike physical event. Minkowski showed
that by adopting Einstein's principles and using his coordinates one
automatically bestows a specific structure on this point set. A set of
points endowed with this specific structure is called Minkowski space-
time. Following the standard practice, I refer to spacetime points as
events.
The structure of Minkowski spacetime is a geometry in Klein's sense
(§4.1.2) and also in Riemann's (§4.1.3). Klein's approach is more eco-
nomical and in this case more fitting,15 but only Riemann's can enable
us to understand the transition from SR to GR and the conceptual rela-
tionship between these theories. I shall therefore base my presentation

15
Klein (1911) greeted SR as a most welcome confirmation of his insight.
262 Relativity

on both. Let me first show that spacetime is a 4-manifold. To see

this it is enough to recall the definition of w-manifold in §4.1.3 and to
note that any coordinate system that - like the primed and the
unprimed systems of §5.1 - combines Cartesian coordinates and
Einstein time tied to an inertial frame of reference is in effect a one-
one mapping of all spacetime onto R4. Therefore, a single coordinate
system of this type constitutes, all by itself, an atlas for spacetime
and determines - in the manner explained in Chapter Four, note 17 -
a maximal atlas, which in turn bestows on spacetime a definite mani-
fold structure.
In agreement with the literature on differential geometry, I shall
henceforth use lowercase Roman letters such as x, y, z to designate dif-
ferent spacetime charts, that is, one-one mappings of an open region
of spacetime into or onto R4. If x is such a chart, it labels each event
P with four real numbers or coordinates that I shall designate by x°(P),
x^P), x2(P), and x3(P). In this way, each coordinate function is denoted
by xk (with the index - not exponent! - k ranging from 0 to 3). The
index 0 corresponds to the time coordinate, and the other indices to
the spatial coordinates.
In the remainder of this section I shall only use spacetime charts
of a special kind, which I call Lorentz charts. Just as Cartesian
coordinate systems are especially useful in Euclidian geometry, so
Lorentz charts are particularly well suited for describing relations in
Minkowski spacetime. A Lorentz chart combines Cartesian coordinates
and Einstein time, defined as in §5.1 for an inertial frame to which the
chart is said to be adapted. As is usual in classical physics, the coor-
dinate values measure distances along the coordinate axes; but the mea-
surement units are chosen in such a way that the speed of light c is one
unit of length per unit of time.16 For this purpose we agree that all
Lorentz charts meet the following condition: The time unit is our
second (i.e., the duration of 9,192,631,770 periods of the radiation
corresponding to the transition between the two hyperfine levels of the
ground state of caesium-133), but the unit of length is the light-second
(i.e., the distance in vacuo traversed by a photon in one second), not
the all too human meter (defined since 1983 as 1/299,792,458 of the
said distance). Consider now the collection of all Lorentz charts x such

16
If space and time are fused, lengths and time intervals must, of course, be regarded
as aspects of a single metrical quantity. Therefore, speeds are pure numbers. But first
we must work our way to the vantage point from which this will be obvious.
 5.2 Minkowski's Spacetime 263

that (i) there is a unique event O at the origin of all coordinate axes
(in other words, there is an O such that, for all x, x°(O) =tf^O)= x2(O)
= x3(O) = 0); (ii) the time coordinate increases as time advances (for
all x, and any two events X and Y in the history of a given particle,
x°(X) > x°(Y) if and only if X occurs later than Y); and (iii) the Carte-
sian coordinates form a standard, right-oriented system. This collec-
tion is obviously an atlas for spacetime. I call it the standard Lorentz
atlas based at O and denote it by si^o). Note that if the primed and
the unprimed system of eqns. (5.5) meet the above condition on mea-
surement units, they both belong to s£#(O).
Let x and y belong to s&g(O)> x and y are one-one mappings of all
spacetime onto [R4, so they have inverses, x~l and y'1, that map IR4 onto
spacetime. The composite mapping y°x~l is therefore a one-one
mapping of IR4 onto [R4, that is, a one-one correspondence between
number quadruples. Consider the %-coordinates of an arbitrary event
P. They form the number quadruple (A: 0 (P) 5 ^ 1 (P) 5 X 2 (P) 5 ^ 3 (P)) = x(P). The
composite mapping y°x~1 assigns to x(P) the number quadruple y(P) =
<y°(P),y1(P),y2(P)O'3(P)> formed by the y-coordinates of P. Thus, yo*"1
is the coordinate transformation that substitutes the y-coordinates for
the x-coordinates. And, of course, x°y~1, that is, the transformation
that substitutes the x-coordinates for the y-coordinates, is the inverse
of y °x~1. If z is a third chart in s&${O), the composite mapping (z°y~1)0
(y0;*;"1) is identical with the coordinate transformation z°x~1. Thus, the
set of all transformations y°x~1, where x and y range over «94#(O)5 forms
a group, with the trivial transformation x°x~x as a neutral element (see
§4.1.2). In the light of note 10, it should be clear that this group is a
realization of what I called there the Lorentz group.17 Consider now
the composite mapping y~lox, where x and y belong to s&^o)- y~l°x is
a one-one mapping of spacetime onto itself, which assigns to each
event P the one and only event Q whose y-coordinates are identical
with the ^-coordinates of P.18 Such a mapping is called a point trans-
formation (to distinguish it from coordinate transformations, which
map number n-tuples to number w-tuples). The set of point transfor-
mations (y~lox), where x and y range over si^o), is of course another

17
Let slap be the union of all atlases sl%{P), where P ranges over all of spacetime. Then
the set of all transformations y°x~x, where x and y range over si®, is a realization of
the Poincare group, as defined in note 10. One may naturally call si® the standard
Poincare atlas of spacetime.
18
If Q = )Tlo%(P), y(Q) = x(P); therefore, y'(Q) = JC'(P) (i = 0, 1, 2, 3).
264 Relativity

realization of the Lorentz group19 acting on spacetime itself in the

manner contemplated by Klein. In accordance with Klein's Erlangen
program, the geometric structure of Minkowski spacetime consists of
- or rests on - the properties and relations of events that are Lorentz-
invariant.
One can easily verify that neither lengths nor durations remain
invariant under the Lorentz boost (5.5). To simplify calculations, let us
first rewrite eqns. (5.5) using the new nomenclature (x° for £, x1 for x,
etc.) and putting c = 1:

y
"

(5.7)

(where a superscript right after a letter is an index, but a superscript

after a right parenthesis is an exponent).
Consider two events, P and Q, at the origin of the Cartesian system
(x^x^x3). The time interval between them is given by \x°(P) - x°(Q)|
in terms of the x-chart and by |y°(P) - y°(Q)| in terms of the y-chart.
These two numbers are not equal, for

m £ > v < f )
•h-ivf
(unless v - 0).
Consider now a rod of length X, at rest on the y-frame. Assume that
it lies along the y^axis, with one end at y1 = 0 and the other end at y1
= X. So, in the x-frame the rod moves with speed v in the direction
along which it lies. We wish to ascertain the rod's length in the x-frame,
that is, the distance between two simultaneous positions of the rod's
ends on this frame. So let P and Q denote the events at which either
end of the rod takes those positions. The distance in question is equal

19
Let x and y range over s^(O). The mapping (y°x~l) •-* {y~l°x) is a group isomorphism.
 5.2 Minkowski's Spacetime 265

to the square root of (xa(Q) - ^(P)) 2 + (x2(Q) - x2(P))2 + (x3(Q) -

x3(P))2. Our assumptions concerning the position of the rod on the y-
frame imply that y2(P) = )>2(Q) and y3(P) = y3(Q). Hence, by virtue of
eqns. (5.7), the distance on the x-frame between the positions of the
rod ends at events P and Q equals |x*(Q) - x1{P)\. We have that

Vi-W1 Vl-M
Since x°(P) = x°(Q) (for P and Q are simultaneous on the x-frame), on
subtracting the first equation from the second we obtain:
\x\Q)-^(P)! = Wl-(v)2 < X (5.10)
(unless i/ = 0).
On the other hand, as the reader can easily verify, if P and Q are
any two events, the quantity o(P,Q), expressed below in terms of chart
x, is invariant under the Lorentz boost (5.7):20

= (x°(P)-x°(Q))2
(* 2 (P)-x 2 (Q)) 2 +(x 3 (P)-x 3 (Q)) 2 )

S1
(5.11)
Note that o(P,Q) is also invariant under spatial rotations; it is the dif-
ference between (x°(P) - x°(Q))2, which is the squared time interval
between events P and Q, and X3=i (x*(P) - x*(Q))2, which is the squared
spatial distance between their respective locations, and spatial rotations
preserve both these quantities. Since every Lorentz transformation is a
product of boosts and space rotations (see note 10), it follows that
o(P,Q) is Lorentz-invariant. I call a(P,Q) the spacetime interval between
events P and Q and regard it as the fundamental invariant of
Minkowski spacetime.
Since the difference between the frame-dependent quantities (x°(P)
- *°(Q))2 and I3=i (x'(P) - x;(Q))2 is Lorentz-invariant, so is the fol-
lowing classification of spacetime intervals:

Ignore the second and third spatial coordinates and just prove - by using (5.7) - that
266 Relativity

(i) o(P,Q) is timelike if the temporal separation |x°(P) - x°(Q)| pre-

vails over the spatial distance VX3=1(x'(P) - x'(Q))2, and o(P,Q) > 0.
(ii) a(P,Q) is spacelike if the spatial distance Vl3=i (xl(?) - x'(Q))2 pre-
vails over the temporal separation \x°(P) - x°(Q)\, and o(P,Q) < 0.
(iii) o(P,Q) is null if the spatial distance Vz3=1 (x'(P) -x'(Q)) 2 equals the
temporal separation |x°(P) - x°(Q)|, and a(P,Q) = 0.
Note that if a(P,Q) is null, P and Q could be two events in the history
of a photon (or any object traveling with the speed of light).21 Because
of this, some authors call zero-valued spacetime intervals 'lightlike'
instead of 'null'. On the other hand, if CJ(P,Q) is timelike, P and Q could
be two events in the history of an inertial particle traveling with a veloc-
ity less than that of light; and if o(P,Q) is spacelike, P and Q could be
two events in the history of an inertial particle traveling with a veloc-
ity greater than that of light.22
Pick an event P and consider all events X, such that o(P,X) = 0. These
are the events from which a photon received at P could have been sent,
or at which a photon sent from P could be received. The set of events
{X | G(P,X) = 0} is the null-cone (or light-cone) of P.23 It comprises three
mutually exclusive parts, viz., (a) P itself, (b) all events X from which
a photon could be sent for reception at P, and (c) all events X at which
a photon could be received from P. Part (b) is the past null-cone of P,
and part (c) is the future null-cone of P. Parts (b) and (c) are three-

21
Case (iii) obtains if and only if the ratio

is equal to 1, which, remember, is the velocity of light in the units we are using.
Faster-than-light particles are called tachyons (from the Greek Ta%i5<;,'fast'). Much
has been written about tachyons, although there is not a whit of evidence that they
exist. SR is not incompatible with them but imposes some constraints on their behav-
ior. Tachyons could never be slowed down to the speed of an ordinary massive par-
ticle, nor ordinary massive particles accelerated to tachyon speed. If real numbers are
employed, as usual, to express the inertial mass of ordinary particles, the inertial mass
of tachyons must be expressed by multiples of v—1. It is generally agreed that, if
tachyons existed, they could not be used to transmit information between ordinary
particles.
If we discard one spatial dimension and use our intuitive image of it to represent time,
the set {X | o(P,X) = 0} covers a two-sheeted cone with its vertex at P. By analogy,
the three-dimensional figure cut by {X | a(P,X) = 0} in four-dimensional spacetime is
also called a cone or, more strictly, a hypercone.
 5.2 Minkowski's Spacetime 267

dimensional submanifolds of four-dimensional spacetime (hypersur-

faces, in mathematical jargon). Let K(P) denote the null-cone of P. Let
Y and Z be events such that a(P,Y) > 0 and a(P,Z) < 0. Then, any
smooth curve joining Y to Z intersects K(P).24 Thus the null-cone K(P)
separates two mutually exclusive regions of spacetime, constituted,
respectively, by events Y such that CJ(P,Y) is timelike - which, we say,
are inside K(P) - and by events Z, such that o(P,Z) is spacelike - which
lie outside K(P).
The function a assigns real numbers to event-pairs much like the
ordinary distance function assigns numbers to point-pairs in space.
Indeed, if a(P,Q) is spacelike, the distance between the spatial location
of P and Q in an inertial frame in which both events are Einstein-
simultaneous is given precisely by Vo(P,Q). On the other hand, if o(P,Q)
is timelike, and P and Q are events in the history of the same inertial
particle, Va(P,Q) equals the time lapse between them in the rest frame
of that particle.25 But, in stark contrast with the ordinary distance func-
tion, G(P,Q) - with P ^ Q - can take negative and zero values, so it
cannot be used for comparing timelike with spacelike intervals and is
worthless as a measure of the separation between events on the same
null-cone. Still, we could use the null-cones for distinguishing between
spacelike, timelike, and null curves and then resort to a(P,Q) or Vo(P,Q)
for separately defining the length of timelike and of spacelike curves in
Minkowski spacetime, in the way that the ordinary distance function
is used for defining the length of curves in Euclidian 3-space. However,
the same purpose is achieved in a neater and - with a view to GR -
more illuminating manner by introducing the Minkowski metric T|.
This is defined on the analogy of a Riemannian metric g (see §4.1.3,
after eqn. (4.5)) but with one important difference: T| assigns a bilin-
ear function t|P to each event P, but T|P is not required to be positive
definite and nondegenerate; therefore, if v is a tangent vector at P, dif-
ferent from the zero-vector, it may happen that T|P(t/,t/) is greater than,

24
To prove this statement consider a curve y such that y(0) = Y and y(l) = Z. Let o>
denote the function that assigns to each event X the real number o(P,X). o> is smooth.
Therefore, the composite mapping 0>°y is a smooth real-valued function of one real
variable whose value at Y is o(P,Y) > 0 and whose value at 1 is o(P,Z) < 0. So there
must exist a real number a, between 0 and 1, such that a?°y(a) = 0. This implies that
a(P,y(tf)) = 0, so y(a) lies on K(P).
25
L e t P a n d Q b e e v e n t s i n t h e h i s t o r y o f a n inertial p a r t i c l e , a n d let x b e a L o r e n t z
chart adapted to the rest frame of that particle. Then, if P * Q, x°(P) * x°(Q), but
*»(P) = *»(Q) for n = 1, 2, 3.
268 Relativity

equal to, or less than 0. The definition of T| is made easy by the fol-
lowing basic fact: The Minkowski spacetime manifold admits global
charts (i.e., charts defined on the whole manifold), for example, the
Lorentz charts. Take any Lorentz chart x. x determines four curves
through each event P, viz., the parametric curves through P of the four
coordinate functions x°, x1, x2, x3.26 Thus, the mere existence of x
defines the assignment

a a a a (5.12)
dx° /a* 1 P '3x 2 P '3x 3
where d/dx% denotes the vector tangent at P to the parametric curve
of xl through P (0 < / < 3). The four vectors are linearly independent
and so constitute a basis of the tangent space at P, a so-called tetrad at
P. The mapping (5.12) is a tetrad field on the spacetime manifold. By
means of it, one can equate directions and orientations at different
points of the manifold. Because of this feature, the Minkowski space-
time manifold is said to be parallelizable. Since every vector that is
tangent to Minkowski spacetime is a linear combination of the four
vectors in a local tetrad, the Minkowski metric T| is fully determined
by defining, for each event P, the value of T|P at each vector pair formed
from the tetrad
9
dx°
The definition is as follows:

(5.13)
dxh
where h and k range over {0,1,2,3}, r\00 = 1, r\kk = - 1 if k > 0, and %*
= 0 if h ^ k. The metric K\ assigns a "length" of sorts to each vector v
at P, on the analogy of Riemannian metrics. If v = Y

(5.14)
h=0 k=0

26
As I explained in Chapter Four, note 21, given a point P in a manifold $Jl and a coor-
dinate function xk, defined at P and whose values range over an interval I, there is a
unique curve y: I —> 33? such that (i) P is the value of y at xk(P), and (ii) for every other
point Q in the range of y we have that Q = y(x*(Q)) and **(Q) = xh(?) if h * k. y is
the parametric curve of xk through P.
 5.2 Minkowski's Spacetime 269

I note for future reference that the Riemann tensor defined from the
Minkowski metric T| is everywhere 0. Therefore, on the analogy of
Euclidean space, the spacetime endowed with T| is said to be flat.
In contrast with proper Riemannian metrics, the Minkowski metric
T| is indefinite and degenerate, and it is therefore dubbed semi-
Riemannian. As noted above, for any given tangent vector v at P it may
occur that:
(i) T\p(v,v) > 0, in which case we say that v is timelike; or
(ii) T\v(v,v) < 0, in which case we say that v is spacelike; or
(iii) 1\P(V,V) = 0, in which case we say that v is null (or lightlike).
Through this tripartite classification of tangent vectors it is possible
to classify all spacetime curves into four kinds. (On curves and their
paths, see §4.1.3.) A curve y through P is timelike, spacelike, or null
at P if its tangent vector at P is timelike, spacelike, or null, respectively,
y is a timelike (spacelike, null) curve if it is timelike (spacelike, null)
everywhere, otherwise, y is said to be mixed. This classification is
extended in a natural way to two- and three-dimensional submanifolds
of spacetime; for example, a hypersurface is said to be spacelike if every
vector tangent to it is spacelike. The distinction between timelike,
spacelike, and null curves has a special physical significance. Ordinary
massive particles can never attain the speed of light (see §5.3.5), so all
events in their lives must lie on the path of a timelike curve (in the case
of all known particles) or on that of a spacelike curve (in the case of
tachyons, if such particles exist - see note 22). Null curves are reserved
for massless objects, traveling with the speed of light. Henceforth, I
shall use Minkowski's term 'worldline' for the curve - null or timelike
- that tracks the life history of a photon or an ordinary massive point-
particle. (Tachyons will be ignored.) With mild abuse of language I
shall occasionally speak of the 'worldline' of a larger object, such as a
clock or a planet, whose volume is irrelevant in the context.
If the curve y is defined on the interval (a,b), a measure of its
"length" is naturally obtained by substituting the metric T| in the inte-
grand of integral (4.3):

Note that integral (5.15) is positive if y is timelike, negative if yis space-

like, and zero if y is null. So this measure of "length" yields significant
quantitative information only if it is used for comparing a timelike
270 Relativity

curve with other timelike curves or a spacelike curve with other space-
like curves. Applied to null curves it yields the uninteresting informa-
tion that they all have "length" zero. And, of course, it cannot be
applied significantly to mixed curves. Now, once it is clear that a set
of spacetime curves cannot be compared as to "length" unless they are
all timelike, or all spacelike, it is perhaps preferable to measure them
by the following integral, which comes closer to proper Riemannian
length (4.6):

)(Y(«),Y(«))|d« (5.16)

Integral (5.16) is, of course, equal to 0 if y is null, and it conveys no

information if y is mixed. If y is timelike, the "length" measured by
integral (5.16) is called proper time. It is a remarkable fact of life that
- to the extent that gravitational fields are uniform - atomic clocks
keep proper time along their respective worldlines.27
Due to the nature of the Minkowski metric, the classical definition
of geodesic as a curve of extremal length cannot be universally applied
to spacetime curves. However, Levi-Civita (1917) showed how to
define geodesies without appealing to a concept of length. By his
definition, a geodesic is a curve whose tangent vectors are all parallel
to each other along it. A geodesic is therefore a curve of constant direc-
tion. The definition uses Levi-Civita's notion of parallel transport of
vectors along a curve, which I cannot explain here. But no explanation
is necessary in the case of Minkowski spacetime, which, as we saw, is
a parallelizable manifold. Two vectors v and w at different points of
such a manifold are parallel - in Levi-Civita's sense - along every line
joining those points if and only if they are parallel absolutely. This will
be the case if and only if they have proportional components relative
to the local values of a global tetrad field. To see this more clearly,
remember the tetrad field (5.12) determined by a Lorentz chart x. Let
v = Zj=0 ifdldx\ and w = EJ=OW/3/3X'|Q be vectors at spacetime points P
and Q, respectively, v and w are parallel if and only if vl = kwl for some

27
If gravity's actual lack of uniformity is brought back into the picture, the time mea-
sured by an atomic clock along its worldline still agrees splendidly with eqn. (5.16),
provided that the GR metric is substituted for the Minkowski metric (see §5.4).
Neither SR nor GR contains the slightest hint that could help one understand this
agreement; but were it not for it, they would surely be less interesting as theories of
the physical world.
 5.3 Philosophical Problems of Special Relativity 271

k > 0 (/ = 0, 1, 2, 3). Hence, if v and w are parallel, T|P(i/,^) =

Thus, a spacetime geodesic can be either spacelike, timelike, or null,
but not mixed.
Let me note further that every Lorentz chart x maps geodesic paths
onto straight lines in IR4, and that its inverse x~l maps straight lines in
R4 onto geodesic paths. This means, in particular, that the worldlines
of ordinary inertial particles and photons - which Lorentz charts evi-
dently map onto straight lines - are the paths of timelike and null
geodesies, respectively. Here we have spelled out the geometry of
Minkowski spacetime - somewhat artificially - in terms of the Lorentz
charts. From our present vantage point we realize that the Minkowski
structure of null-cones and timelike and null geodesic paths can be built
on a proper physical basis, supplied by the behavior of ordinary iner-
tial particles and photons (see Ehlers, Pirani, and Schild 1972).

5.3 Philosophical Problems of Special Relativity

In the central portions of this section I shall discuss some difficulties
raised by the seemingly arbitrary definition and the frame dependence
of Einstein time (§§5.3.2, 5.3.3, 5.3.4). Then I shall turn to the ques-
tion of conceptual change in physics, as illustrated by the contrast
between Newtonian mass and relativistic mass (§5.3.5). I begin,
however, with a problem that is more physical than philosophical
(§5.3.1), which will prepare us for dealing with the difficulties of Ein-
stein time.

5.3.1 The Length of a Moving Rod

Consider a rod, 10 meters long, moving with constant speed v along
the x-axis of our inertial frame 2F. To simplify calculations I put (v)2 =
0.75(c)2. By eqn. (5.10) the length of the rod in 9? is equal to 10VT/4
meters = 5 meters, so it fits comfortably in a rectangular barn that is
10 meters long, standing in 2F along the x-axis. On the other hand, in
the rod's own rest frame 2F'5 it is the barn that moves with speed v in
the opposite direction, so the barn's length is 5 meters, while the rod's
length is 10 meters. Thus, from this perspective, there is no way that
the rod can fit into the barn. At first blush this is baffling, for one tends
to think that a solid object either fits or does not fit into another, as a
matter of frame-independent fact. However, we would not say that the
rod fits into the barn, or that the barn surrounds the rod, unless there
272 Relativity

t= 0

t' = 0

Figure 13

are two simultaneous positions of the barn and the rod such that the
former lies within the latter. Now this is certainly the case if simul-
taneity is defined by Einstein's method in the barn's rest frame 2F: There
is a short interval of Einstein time adapted to 2F during which each suc-
cessive position of the moving rod lies wholly inside the barn. But the
situation is very different if simultaneity is defined by Einstein's method
in the rod's rest frame 2F': There is no instant of Einstein time adapted
to W at which both ends of the rod are contained within the barn;
when the forward end of the rod has already crossed the front and
back entrances of the barn and is moving away from it, the rod's rear
end still has not reached the barn and is advancing toward it. That this
follows necessarily from the principles of SR can be proved by straight-
forward calculation, by substituting the appropriate data into eqns.
(5.5) or (5.7). It can also be shown very easily if we represent the x-t
spacetime plane on the page, mark the loci of Einstein simultaneous
events on each frame, and draw the worldlines of the two ends of the
rod and of each entrance to the barn. This is done in Fig. 13, where
 5.3 Philosophical Problems of Special Relativity 273

the lines Ri and R2 represent, respectively, the worldlines of the forward

and the rear ends of the rod, while Bi and B2 represent, respectively,
the worldline of a point at the front and the back entrances of the barn.
The thick lines indicate the locus of events at time zero on the primed
and the unprimed frames. The relative speed of the rod and the barn
is represented by the angle between the B-lines and the R-lines, and
also by the angle between the line t = 0 and the line t' = 0. To keep the
line t' = 0 well distinguished from the R-lines, Fig. 13 represents a rel-
ative speed that is not as high as the one assumed in the text. Still, the
drawing shows clearly that Ri and R2 cross the line t = 0 at points lying
between the intersections of that line with Ba and B2, and they cross
the line t' = 0 at points placed outside the intersections of Bx and B2
with this line.

5.3.2 Simultaneity in a Single Frame28

Einstein's Light Principle is curiously intertwined with his definition of
a time coordinate function adapted to an inertial frame of reference.
On the one hand, without some such function the Light Principle is
meaningless. On the other hand, the Light Principle is true only under
Einstein time. The situation closely resembles Lange's handling of
Newton's First Law (§2.2). Lange used three free particles for defining
an inertial frame and an inertial time scale. For these particles the
First Law holds by definition. However, once the frame and its time
scale are fixed, all other free particles bear witness to the First Law's
validity. Likewise, Einstein picks a swarm of bouncing photons issuing
at a particular instant from a point P in an inertial frame. His defini-
tion of time implies that these photons move with the same speed in
every direction. But once Einstein time is fixed in an inertial frame, all
other photons bear witness to the validity of the Light Principle in that
frame.
To devise a scheme of description through which the universal facts
of nature shine splendidly is perhaps the most telling sign of scientific
genius. But many philosophers and scientists would rather draw a neat
line between hard, dependable facts and soft, whimsical conventions.
So some philosophers - notably Reichenbach (1924, §7; 1928, §19) -
have come up with the notion that the Light Principle, insofar as it

28
For a recent, lucid, and fair discussion of this subject, with references to the litera-
ture, see Redhead (1993).
274 Relativity

makes a factual claim, can only refer to what they call the "two-way
speed" of light - that is, the average speed with which a radar signal
travels in vacuo from its source to a reflecting body and back to its
source again -, whereas Einstein's application of the Light Principle to
the "one-way speed" of light, involving as it does the definition of Ein-
stein time, is purely conventional. Reichenbach treats Einstein time as
a particular case in an infinite family of admissible time coordinates,
each one of which assigns a different pair of values to the one-way
speed of light to and from a given point in space.
Of course, no amount of philosophical argumentation can do away
with the fact that the constant c that occurs in the equations of physics
designates an ordinary, one-way speed.29 Disturbed by the idea that this
fundamental constant of nature could be fixed by convention, other
authors have racked their brains in search for ways of measuring the
one-way speed of light that would not presuppose a universal time
coordinate. Unfortunately, any such procedure must at some point
resort openly or furtively to synchronized distant clocks,30 so their pro-
posals have about as great a chance of success as the familiar attempts
at squaring the circle. To expose the fallacy, proposals of this type must
be patiently scrutinized, preferably by the editors of the professional
journals to which they are regularly submitted. The following remark
might be helpful: As I noted above, one meter is, by definition,
1/299,792,458 of the distance traveled by a photon in one second.31
Therefore, no conceivable instrument can measure the one-way speed
of light as anything else than exactly 299,792,458 meters per second.
Under these circumstances, any attempt to measure the one-way speed
of light c without relying on synchronized distant clocks would, if suc-
cessful, do no more than establish experimentally that the distance trav-

29
Indeed, the very idea of a "two-way speed" evinces insensitivity to one of the great
achievements of modern mathematical physics, viz., the conception of motion as a
state of the moving body, fully actual at any instant (see in §1.3 the quotation from
Descartes (1644, II §39), and the remarks that follow).
30
A good example of this is Remer's method for measuring the speed of light, which
was discussed in §1.5.3. To calculate the one-way speed of light c between Jupiter's
satellite Io and the Earth, one needs to use the speed u with which the Earth travels
toward Io or away from it between successive eclipses of the latter. Any estimate of
u involves a comparison of time at the different places from which the said eclipses
are observed.
31
This convention is not a whim, but the outcome of a vast experience with other stan-
dards that proved to be harder to reproduce in a reliably stable way.
 5.3 Philosophical Problems of Special Relativity 275

eled by a photon in a certain direction is approximately equal - within

the limit of admissible experimental error - to the distance traveled by
it over the same path in the opposite direction. But this is ridiculous,
for, by dint of a universal linguistic convention, the distance between
two points in space is exactly the same, no matter in what direction
you travel it.
The definition of the meter ratifies Reichenbach's view of the one-
way speed of light as conventional but also undercuts his claim that its
two-way speed is factual. With a single clock affixed to a radar trans-
mitter at rest on an inertial frame one can measure the distance from
it to any radar-reflecting post in the frame, not the average speed with
which the signal travels to and fro. To be precise, one measures the
length of the two-way trip; that this is exactly twice the length of the
one-way trip is a consequence of the age-old convention noted above:
'Distance' is a symmetric real-valued function on point-pairs. More-
over, if the one-way speed of light is by definition the same everywhere
and in every direction, a photon takes exactly the same time in travel-
ing back and forth between two points at rest on an inertial frame. Evi-
dently, Einstein time is the only time coordinate function that agrees
with this.
Reichenbach wrote about time coordinates and the speed of light at
a time when the meter was identified with the distance between two
marks on a metal rod kept at Sevres. Although his concepts are out of
step with today's metrology, they still deserve attention. He concen-
trates on the definition of simultaneity between distant events. Let a
photon be issued from A to B, where it is reflected back toward A. Let
Ei be the emission of the photon and E3 its final reception at A. What
event at A shall we judge simultaneous with the bouncing of the photon
at B? Reichenbach says that any event E2 will do, provided that it comes
after Ei and before E3. This condition is satisfied if
t(E2) = t(E1) + e(t(E3)-t(E1)) (5.17)
where t assigns to events at A their time as given by a standard clock
at that point, and £ is any positive real number less than 1. By means
of eqn. (5.17), every moment of time t is - in Newton's phrase - "dif-
fused indivisibly throughout all spaces", as point B ranges over them.
For some unknown reason, when Reichenbach formally introduced
eqn. (5.17) he postulated that "£ is an arbitrary factor which, however,
must be equally large for all points B" (1924, §7, Def. 2; my notation).
By Reichenbach time I mean a universal time coordinate defined in this
276 Relativity

way. According to Reichenbach, Einstein time is just a special case of

Reichenbach time, in which 8 = 7 . Although eqn. (5.4) might seem to
indicate this, it is not true without further qualification. Einstein's
definition presupposes that A and B are points at rest on the same iner-
tial frame, a requirement that Reichenbach does not bother to mention
(so one is free to wonder how his definition of simultaneity might work
if A and B stood on the rim of a rotating disk). Moreover, Einstein
assumes that the universal time coordinate defined by his method is an
inertial time scale, so that free material particles traverse equal dis-
tances in equal times as measured by it; but if e takes the same value
in every direction issuing from A, the universal time coordinate defined
by eqn. (5.17) cannot be an inertial time scale unless 8 =j.32 In other
words, Einstein time as given by eqn. (5.4) is an instance of Reichen-
bach time as given by eqn. (5.17) only if Reichenbach time is so con-
strained that Einstein time is its sole instance.
The relevance of Reichenbach time to SR physics can somehow be
rescued if, following Reichenbach's own suggestion (1928, §26), we
forget his earlier demand that 8 be fixed and define it as a continuous
function of direction that (a) takes in each direction about A a value
equal to 1 minus the value it takes in the opposite direction, (b) ranges
between a maximal value 8max < 1 and a minimal value 8min = 1 - emax >
0, and (c) satisfies the condition 8 = \ in every direction perpendicular to
that in which 8 = 8max. Let us say that a universal time coordinate that
constitutes an inertial time scale, defined as in eqn. (5.17), with A and
B affixed to the same inertial frame and 8 subject to (a), (b), and (c), is
a modified Reichenbach time. Given the facts of nature that secure the
consistency conditions of Einstein time, a modified Reichenbach time
can also be defined consistently. This is no wonder, for if 8 varies with
direction in the manner I have just described, the modified Reichenbach
time associated by eqn. (5.17) to the inertial frame 3F in which A rests
agrees precisely with the Einstein time associated by eqn. (5.4) to an
inertial frame 3F' that moves relative to 3% in the direction in which 8 =
8max5 with a definite speed v depending on 8max.33 Thus, except in the case

32
For a proof of this statement, see Torretti 1983, pp. 224f.
33
Here is a proof-sketch. To make things clearer I describe any figure F in spacetime
by the same geometric terms - for example, 'straight', 'plane', 'hyperplane' - which
apply to the image of F by a Lorentz chart. Pick an event P. The condition 0 < e < 1
implies that any other event that is Reichenbach-simultaneous with P lies outside the
null-cone K(P). Let y be a spacetime chart combining a modified Reichenbach time
 5.3 Philosophical Problems of Special Relativity 277

that 8 = const. = y, modified Reichenbach time is, if I may say so, mal-
adapted Einstein time.34

5.3.3 Twins Who Differ in Age

Perhaps the most picturesque expression of conceptual novelty in SR
is the so-called paradox of twins. Jack and Jill are born on earth in
quick succession from the same mother. At an early age, Jill is placed
on a spaceship in which she tours the galaxy at enormous speed, say,
0.9c. When she returns, a handsome woman in her thirties, Jack is a
decrepit octogenarian. This is not what one would have expected, not,
at any rate, before Einstein. But then we have never met any woman
traveling at such speeds, so why should her aging process meet our
expectations? The worldlines of Jack and Jill are not the same, and
proper time as measured by eqn. (5.16) is a good deal shorter along
Jill's. So, if the twins' biological clocks run more or less like atomic
clocks, Jack must be a good deal older than his sister when she
returns.35

coordinate y° with Cartesian coordinates y1, y2, y3, all adapted to the same
inertial frame &*. The sets of Reichenbach-simultaneous events {X | y°(X) = const.)
are flat 3-manifolds, i.e., hyperplanes. Let us say that a hyperplane in spacetime is
spacelike if the spacetime interval o between any two events on it is spacelike. Then,
each hyperplane y° = const, is spacelike. Each family of parallel spacelike hyperplanes
corresponds precisely to a particular partition of spacetime into classes of Einstein
simultaneous events. Consequently, the modified Reichenbach time coordinate y° is
identical with the Einstein time coordinate x° of a particular Lorentz chart x. Of
course x is not adapted to 3£ but to some other inertial frame 8F' in which $* moves
with velocity v. If rmax denotes a unit vector in the direction in which e = emax,
v = tanh(arc tan |1 - 2emax|)rmax.
34
Someone who does not care for historical precedence might state this identity the
other way around: Each Einstein time coordinate x° is none other than the one
modified Reichenbach time coordinate that is well adapted to the inertial frame
for which it is defined. I say that this time coordinate is "well adapted" because it
does not require for its everyday use that one keep the bearings of a fixed direction
in space, a practically impossible thing to do given the spatial isotropy of inertial
frames.
35
That atomic clocks behave like our twins was shown by Hafele and Keating (1972).
Thirty years earlier, Rossi and Hall (1941) studied short-lived fast particles generated
in the upper atmosphere by cosmic rays, and showed that their decay rate, as mea-
sured by earthbound clocks, was - in good agreement with SR - less than it would
be if the particles were at rest on the earth.
278 Relativity

Obviously, this could not be true if biological clocks kept absolute

universal Newtonian time, but nobody claims that they do. Still, in the
literature the twins' story has taken an air of paradox due to misun-
derstandings. Thus, it was argued that if Jack ages sooner than
Jill when events are referred to his rest frame, Jill must age sooner
than Jack if events are referred to hers. After all, SR is a theory of
relativity, in which any inertial frame may be considered to be at
rest. This argument overlooks that Jack's life is not a mirror image
of Jill's. Suppose, for simplicity's sake, that he stays all the time in a
single inertial frame; then Jill must sit in at least two such frames as
she speeds away and then returns. Others contend that, due to this
asymmetry, the twin paradox lies outside SR's ken, for Jill must
undergo an acceleration that only GR can describe. But, as we saw
in §5.2, Minkowski geometry assigns lengths to timelike curves of
any shape, not just geodesies. Since only the latter represent the space-
time tracks of ordinary massive inertial particles, the often heard
remark that SR kinematics is unable to cope with accelerated motion
is groundless.
However, we can disregard Jill's acceleration if we suppose that she
travels successively in two inertial spaceships and jumps instanta-
neously from one to the other. This violent exercise may cause her to
look older than she is, but her chronological age, as measured by the
atomic clocks that surround her, will anyway be much less than Jack's
when they come together. Let x, y, and z be Lorentz charts adapted,
respectively, to Jack's inertial frame, to Jill's as she goes away, and to
Jill's as she returns. Let Jack stand all the time at the spatial origin of
chart x (so that, for any event E in his life, x^E) = x2(E) = x3(E) = 0).
Put c = 1 and let Jill travel on the plane x2 = x3 = 0 with speed 0.9. To
make the numbers smaller, I express time in years and distance in light-
years. Jack and Jill separate at P and reunite at Q. Let J denote Jill's
jump, midway between P and Q, from the outward bound to the
inward bound spaceship. For later reference, I designate by G and H
the events in Jack's life that are simultaneous with J by chart y and by
chart £, respectively (see Fig. 14.)
If Jack lives on earth 80 years while Jill is traveling, x°(Q) - x°(P) =
80, *°(J) = 40, and xx(J) = 36. Proper time along Jill's worldline
amounts to y°(J) - y°(P) plus z°(Q) - z°(J). These values can be obtained
by straightforward calculation after making the appropriate substitu-
tions in eqn. (5.8):
 5.3 Philosophical Problems of Special Relativity 279

z° = z°(J)

Figure 14

40-0.9(36) 7.6
= 17.4356
Vl-(0.9) 2 0-43589
(5.18)
2°(Q)-2°(J) = -

40 + 0.9(-36) 7.6
= 17.4356

So y°(J) - /(P) + z°(Q) - z°(J) = 34.8712, and Jill has lived less than
35 years while her brother lives 80.
Another seeming paradox might irk us if we allow Jill to keep track
of Jack's aging in her calendar. In Jill's own rest frame Jack's birthdays
occur less frequently than hers. Thus, in the 17.4356 years between
280 Relativity

P and J, Jack, by her reckoning, grows - from P to G - only 7.6 years

older.36 And again by her reckoning, his age increases by 7.6 years
- from H to Q - in the 17.4356 years between J and Q. So Jill can
celebrate no more than 15 of Jack's birthdays while she travels.
When - she might ask herself - does he complete, in her time, the 64.8
additional years he is burdened with at the end of her trip? If, as we
may well assume, she is cleverer than some philosophers of science,
it will surely dawn on her that her circumstances make it advisable
to use two methods of time-reckoning, adapted to her two successive
rest frames, namely, y° and z°. In y°-time, the 64 birthdays between G
and H are still to come when she jumps at J; but in £°-time they are,
at that very moment, already past. Jill could try a composite chronol-
ogy, viz., "y° before the jump and z° thereafter"; but such a non-
Einstein, noninertial, nonuniversal time coordinate cannot be defined
on all spacetime, and Jack's life from G to H certainly lies outside its
domain.37

5.3.4 Kinematical Determinism

Universal determinism has haunted European thought ever since Leu-
cippus, the inventor of atoms, proclaimed: "Nothing occurs at random;
but everything for a reason and by necessity" (DK 67B2). Some Greek
philosophers argued for determinism from logic: Since every declara-
tive sentence is either true or false, every fact - past, present, or future
- must be determined once and for all.38 But in modern times, deter-
minism has normally been based on dynamics (see the end of §2.5.3).
It is the privilege of our postmodern age to have produced an argu-
ment for universal determinism on purely kinematical grounds, apart
from any conception of dynamics that might be required to enforce it.
The argument, based on Einstein's SR scheme for the description of
motion, was independently put forward by Rietdijk (1966; cf. his

36
To verify this figure, substitute G for P and P for Q in eqn. (5.8). Remember that
x!(G) = x'i?) = 0. So x°(G) - *°(P) = 17.4356 x 0.43589 = 7.6.
37
My handling of Jill's time-keeping was prompted by Debs and Redhead (1996), but
I am not certain that it agrees with their approach.
38
Consider the sentence: 'On 4 July 2076, abortion will be illegal in the United States'.
If this is either true or false, there is a fact of the matter concerning the legality of
abortion in the United States in 2076 that is already determined, although unknown
to us and to the justices and legislators - still unborn perhaps - who will be respon-
sible for it.
 5.3 Philosophical Problems of Special Relativity 281

1976) and by Putnam (1967). It rests on the following unquestionable

facts of SR: Suppose that P and Q are two events in the worldline of
an inertial particle, such that Q lies in the future light cone of P, and
let x be a Lorentz chart adapted to the particle's inertial frame; then
there is
(i) an event E, such that x°(P) = x°(E), and
(ii) a Lorentz chart y, adapted to a different inertial frame, such that
/(E) = /(Q). 3 9
Now if P is present, and therefore fully determinate, in some inertial
frame, and E is simultaneous with P in that frame, E is just as present
as P in that frame and therefore no less determinate; but then, if Q is
simultaneous with E in another frame, it is no less present than E in
that frame and therefore quite as determinate. Note that in this argu-
ment, as I present it,40 'X is determinate' is a frame-independent pred-
icate, as it well may be, but 'X is present' and 'X is simultaneous with
Y' are frame-dependent, as required by SR. But is it right to assume
that two events must be equally determinate (absolutely) merely
because they are simultaneously present (in some frame or another)?
This is no doubt so in the natural philosophy of Kant, in which events
earn the same place in time through their thoroughgoing mutual deter-
mination (§3.4.4). But in SR it is just the opposite: Two events can be
simultaneous (in a frame) if and only if (absolutely) they do not
influence each other.
The story of Jack and Jill in §5.3.3 can throw some light on the issue
at hand. In the frame from which Jill jumps at J, her jump is simulta-
neous with event G in Jack's life; but in the frame to which she jumps,

39
This can be verified by putting numbers into eqns. (5.7). Let P and Q occur at the
spatial origin of chart x, adapted to frame SF. and let E occur at the spatial origin of
chart y, adapted to frame 2£' moving in SF with speed 0.9 along the .x^-axis, in the
direction of decreasing x1. Put x1(E) = 900 and ;t2(E) = x3(E) = 0. Under these con-
ditions, if P and E are ^-simultaneous and x and y are related by the Lorentz boost
(5.7), x°(P) = *°(E) = -1,000 and y°(E) = -190. But then / ( Q ) = y°(E), provided that
*°(Q) = -82.82 > -1,000 = *°(P). You can adjust this result to any pair of events P
and Q by rescaling the Lorentz charts (i.e., by changing the unit of time).
40
I give my own presentation, rather than follow Rietdijk or Putnam, because they use
more words than are necessary, including some that do not speak to their credit, as
when Putnam refers to "the coordinate system of x", where x is an event (1979, p.
200, line 15 from the bottom), or when Rietdijk mentions two observers "experi-
encing the same 'present'" while they rest at separate points on the same inertial
frame (1966, p. 342). See also Stein (1991).
282 Relativity

the jump is simultaneous with H. Jill's jump, completed in an instant,

is surely determinate there and then. Does this imply that H is deter-
minate when G is? Remember that Jack is 7.6 years old at G and 72.4
years old at H. It is ridiculous to expect that from the purely kinematic
information provided Jack can infer at G that he will - or will not -
live until H. Neither can Jill infer it at J; indeed, as she jumps she cannot
even know that Jack is alive at G.
Kinematic determinists cannot claim that the information required
for kinematic description in SR is also sufficient for determining the
course of events in full. So their claim must amount to this: SR kine-
matic description is not possible unless all events are fully determined
at all times.41 This, in my view, groundless claim stems perhaps from
a confusion between the ordinary and the relativistic sense of 'event'.
In the jargon of Relativity, 'event' is short for 'spacetime point'. These
abstract events should not be confused with the events of real life, nor
with their idealized version, the pointlike yet concrete events of physics.
Spacetime points are, of course, fully determined by the spacetime
structure (Minkowskian or otherwise); indeed, were they not fully indi-
viduated, they would not be available as arguments for spacetime
charts (Lorentzian or otherwise).42 The abstract events P, Q, J, G, and
H in Fig. 14 must be given from the outset or the kinematic relations
that we have assumed between them could not be specified. But this
says nothing about the concrete events at P and Q, J, G and H, viz.,
the twins' separation and reunion, Jill's jump and the contemporary
happenings in Jack's life. A concrete event is what it is, where, and
when it is. Through the laws of physics one can infer at least some of
its features from other concrete events on or inside its null-cone. But,
according to SR, nothing that happens outside an event's null-cone can
be a necessary or a sufficient condition of any concrete aspect of that
event. Thus, it is possible to argue - on the strength of some relativis-

41
Full determination would be the job of natural forces. In a similar vein, the Stoic
philosopher Chrysippus did not just let logical determinism stand by itself but invoked
it as a proof of dynamical determinism. "For Chrysippus argues thus: 'If there is
motion without a cause, not every proposition [. . .] will be either true or false, since
anything lacking efficient causes will be neither true nor false. But every proposition
is either true or false. Therefore, there is no motion without a cause. If this is so,
everything that happens happens through antecedent causes. If this is so, everything
happens by fate. Hence everything that happens happens by fate'" (Cicero, De fato,
20-21).
42
On individuation by structure see Newton in Hall and Hall (1962, p. 103) (quoted
in §2.2).
 5.3 Philosophical Problems of Special Relativity 283

tic theory of dynamics - from the determination of an event P to that

of another event Q on a worldline through P; but one cannot buttress
up such an argument by referring to a third event E that lies outside
the null-cones of both P and Q.

5.3.5 The Quantities We Call 'Mass'

Newton's Laws of Motion do not meet the requirement of Lorentz
invariance and are therefore incompatible with SR.43 So Einstein had
to develop a new mechanics. However, in stark contrast with Galileo
and Newton, he did not try his best to forget the received view and to
build from scratch a new physics of motion. Instead he reached for
equations of motion that take the Newtonian form at the low speeds
to which Newtonian mechanics had hitherto been successfully applied
and significantly diverge from it only at speeds closer to the speed of
light. This approach was reasonable but not inevitable. Evidently, the
new mechanics ought to yield the well-corroborated predictions of the
old, but their agreement need only be numerical and concern just
the experimental results. The new theory was therefore free to refer to
those results with new terms, embedded in new equations that do not
collapse to the earlier ones when one deletes every term that tends to
zero together with speed. Einstein exercised this kind of freedom a few
years later, when he produced a theory of gravitation constrained to
agree numerically with Newton's in the weak-field low-speed region
but which differs from it drastically even in typography. But in 1905
his new Lorentz invariant mechanics was formulated as a variation of
the old. Success has proved him right. Yet by using, say, the same old
word 'mass' or writing m (actually, |i) in his equations, he fostered the
illusion that he was merely proposing new laws for the familiar quan-
tity that Newton named thus, when he was in fact replacing it with
new quantities that were deeply at variance with it.
Einstein (1905r, §10) discusses the motion of a charged particle
in an electromagnetic field. Let x, y, z be Cartesian coordinates and t

43
See §5.1, at the end. Of course, the requirement of Lorentz invariance holds only for
physical laws referred to a Lorentz chart. Therefore, the requirement cannot prop-
erly apply to Newton's laws, for the time variable that occurs in them is not Einstein
time. However, the incompatibility of SR with Newton's Laws of Motion can be
restated more accurately as follows: In the light of Einstein's criticism, the time vari-
able in Newton's Laws is meaningless; if one charitably replaces it with Einstein time,
the Laws thus refurbished are not Lorentz invariant.
284 Relativity

Einstein time, adapted to an inertial frame SF. I denote by £ x , £y, and

Ez the electric field components relative to this coordinate system, and
I denote the magnetic field components by Bx, By and Bz. Let m be the
mass and q the electric charge of a particle momentarily at rest in 2F.
Then, says Einstein, "in the next small bit of time" ("im nachsten Zeit-
teilchen") the particle obeys the equations of motion:

d 2x d2y d2z _ ,_ 1 Q .
m— = qEx m-± = qEy m— = qEz (5.19)

Equations (5.19) combine Newton's Second Law wv= f (eqn. (2.1)),

with the electrodynamic "Lorentz force" law, f = q(E + (v x B)) (cf.
eqn. (7.3)). This move is admissible, despite all that has been said about
the incompatibility of Newtonian mechanics with SR, because "in
the next small bit of time" the particle is still moving very slowly
in frame 3% so the SR-compatible laws of mechanics, whatever they
might be, must agree well with Newton's Second Law in this case. Now
consider a time in which the particle moves in 2F with instantaneous
speed v along the x-axis in the direction of increasing x. There is an
inertial frame in which the particle's instantaneous speed is 0. Clearly
this frame in which our particle is momentarily at rest stands to 2F in
the same relation as the frame we called W in §5.1. Once again we
employ the primed letters x\ y\ z\ and t' to denote a system of Carte-
sian coordinates and Einstein time adapted to 3F, which share a
common origin and spatial orientation with the unprimed coordinates.
Let EV, £'/? ^V a n d ^V? B'y, B'z' denote, respectively, the electric and
the magnetic field components relative to the primed system. In the
immediately ensuing very short time interval, the particle obeys the
equations of motion

m dV= £, m 4V= F, m^L= £', (5 20)

As we know, eqns. (5.20) are not Lorentz invariant. What shape do

they take when referred to the unprimed coordinate system? This
depends on the way the electric field components transform under the
Lorentz boost (5.5). The Lorentz invariance of Hertz's version of the
Maxwell equations is secured if the electric and magnetic field com-
ponents relative to the primed and unprimed coordinate systems are
linked to one another as follows:
 5.3 Philosophical Problems of Special Relativity 285

£ x'' — V
— •*~J x
R' — R
^ x' — -^ x

Substituting from eqns. (5.5) and (5.21) in (5.20) and reshuffling terms,
we obtain:

m d2x
i\ 3
At1 Ex= qE'x>

,2 q\Ey—By
m &y V c
= qE'y>

c
q\Ez+-Bz
m dA2Az \ c

Einstein (1905r, p. 919) reminds us that qE'*, qE'y, and qE'* are "the
components of the ponderomotive force acting upon" the charged par-
ticle, referred to an inertial system that - at the moment considered -
moves with the same velocity as the particle. The said components can
be measured, for instance, with a spring dynamometer at rest in 8F'.
So, if 'mass' is inertia or resistance to acceleration, and it is therefore
equal to the ratio between the magnitude of the acting force and that
of the acceleration it produces in the direction in which it acts, we are
forced to distinguish between two kinds of mass, the longitudinal mass
Wfo or resistance to acceleration in the direction of motion, and the
transversal mass raT, or resistance to acceleration perpendicular to the
direction of motion. We gather at once from eqn. (5.22) that these
quantities are given by
286 Relativity

^ (5.23)

where v is the particle's speed relative to the inertial frame 3F from

which the accelerations are measured, and m denotes the particle's mass
in whatever inertial frame it happens to be at rest.
Einstein uses eqn. (5.22) to calculate the kinetic energy that the par-
ticle acquires as its speed relative to 3F increases from 0 to v. The work
W done by the force exerted on the particle in the direction of motion
is given by:

W = \qExdx -m (5.24)

Forces perpendicular to the direction of motion do no work on the par-

ticle, so the total kinetic energy acquired by the latter is W. The energy
obviously grows beyond all bounds as v approaches c. It is therefore
impossible for an ordinary massive particle to achieve the speed of light.
Since q in the above arguments can be any arbitrarily small charge,
Einstein concludes that eqns. (5.23) also hold for a particle with charge
equal to 0. Since every force can be balanced by an electromagnetic force,
they all must have the same effect on matter. So m^ and mx measure - in
the manner explained - our particle's resistance to acceleration by any
type of force. Einstein (1905r, p. 919) emphasizes that "with a different
definition of force and acceleration one would naturally obtain for the
masses other numerical values" than those in eqn. (5.23). In fact, a dif-
ferent definition eventually did prevail; it was put forward by Planck
(1906). While Einstein gave a unique, frame-independent representation
of the force accelerating a particle in any inertial frame by equating it
formally with the force defined according to Newton's Second Law in
the inertial frame in which the particle is momentarily at rest, Planck pro-
vided a physically warranted representation of the force peculiar to each
frame. This rests on Einstein's discovery that "the mass of a body is a
measure of its energy content" (1905s, p. 641 ).44 If m denotes the mass

On the derivation of mass-energy equivalence in Einstein (1905s), see Stachel and

Torretti (1982). Other derivations are found in Einstein (1906e, 1935).
 5.3 Philosophical Problems of Special Relativity 287

of a particle in its rest frame, its rest energy equals me1. The total energy
E(v) of a particle moving with speed v is the sum of its rest energy and
its kinetic energy. So, by eqn. (5.24),
me1
(5-25)

Prompted by this result we define the relativistie mass of our particle by

m
m(v) =^ = , , ,N (5.26)

The particle's relativistie momentum p, in an inertial frame in which it

moves with velocity v (with |v| = i/), is equated with m(v)v, and the rel-
ative foree k acting on the particle in that frame is defined, in formal
agreement with Newton's Second Law, by

k= | = i L m ( l , ) v (5.271

where t is Einstein time adapted to the frame.45

Evidently, relativistie mass, a function of speed, cannot be the same
physical quantity as Newtonian mass, the absolute measure of matter.
In fact, m(v) is not normally the ratio of force to acceleration and thus
not even a measure of inertia in the received sense.46 The quantity that

45
These definitions have some very nice consequences. For instance, (i) given a system
of freely colliding but otherwise unrelated particles, such that v,- is the velocity and
mj(Vi) the relativistie mass of the ith particle at any given moment in the chosen iner-
tial frame of reference, the conservation of energy (L-rnhv^c1 = const.) and the con-
servation of momentum (Z/W^f/Jv,- = const.) hold if and only if relativistie mass is
defined by eqn. (5.26). Also, (ii) the power or rate at which the relative force k does
work on a particle moving with velocity v is

v k = —m(v)c2
at
Moreover, (iii) the relative force exerted by the electric and magnetic fields E and B
on a particle with charge q that moves with velocity v is k = q(E + v x B). In other
words, if 'force' is defined by (5.27), the Lorentz force law (cf. eqn. (7.3)) holds in
each inertial frame as a corollary of Einstein's Relativity Principle and the Maxwell
equations.
46
By result (ii) in note 45,
288 Relativity

we have called m - known in the literature as 'rest mass' or 'proper

mass' - is of course constant, frame-independent, and a measure of
inertia (in the particle's momentary inertial rest frame), and it is thus a
closer analogue of Newtonian mass; but it is still a far cry from it. New-
tonian mass is additive in a literal physical sense, and proper mass is not.
If several bodies are put together to make a bigger body, the Newton-
ian mass of the latter equals the sum of the masses of the former. Like-
wise, if a body explodes and breaks asunder, the Newtonian masses of
the splinters add up to the original mass of the body. In stark contrast
with this, the proper mass of the bigger body includes, besides the
proper masses of the ingredient bodies, the mass equivalent of the
energy invested in bringing them together. And the sum of the proper
masses of the splinters equals the proper mass of the original body
minus the mass equivalent of the energy released in the explosion.
These deep differences in the meaning of a central term of physics,
which is common to SR and Newtonian mechanics, fueled the claim
made by Kuhn (1962) and others that revolutionary science changes the
referents of scientific discourse and therefore, since it no longer speaks
about the same things, is "incommensurable" with the system it seeks
to displace. Countering this extravagant claim, Putnam promoted the
doctrine of reference without sense, which he later abandoned, but
which is still endorsed by some philosophers. According to this doc-
trine, a reference is assigned to a general term of science - or ordinary
language - in an original act of name-giving, patterned after Genesis
2:19-20, and is subsequently transmitted from generation to generation
by word of mouth, independently of any connotations accruing to the
term. We can readily imagine Adam pointing with his forefinger at a tall,
dappled, long-legged creature with a towering neck as he mutters
"giraffe, giraffe". But what could Newton - or the medieval inventor of
the term - have pointed at to provide 'quantitas materiae' with a refer-
ence? It is hard to believe that by mere gestures and without elucidating
its sense he could get anyone to understand what he was talking about.
And, had he succeeded, it is not at all likely that the reference thus
bestowed on 'quantity of matter' or 'mass' would be transferred by

i d , . . ,dv / i N _?
k = — m(v)v = m{v)— + \(\-k)c
at at
So m(v) is the ratio of force to acceleration if and only if v(vk) = 0, that is, if and
only if the particle is at rest or the relative force is perpendicular to the direction of
motion.
 5.4 Gravitation as Geometry 289

Einstein or Planck to 'longitudinal mass', or 'transversal mass', or

'relativistic mass', or 'proper mass', or to all these terms at once when
they defined them in the manner explained above.
I have discussed with some care the introduction of new concepts
of mass in SR so that the reader can form a judgment on this question.
Apparently no finger-pointing was done to secure the continuity of ref-
erence despite the momentous shift in meaning. But the definitions were
so contrived by Einstein and his followers that all mass measurements
recorded in a Newtonian setting could be received and utilized in SR.47
Analogous precautions have surrounded the main conceptual changes
in twentieth-century physics.

5.4 Gravitation as Geometry

It is clear that Newton's law of gravity does not meet the requirement
of Lorentz invariance. The attractive force exerted now by a massive
particle on another depends on their present distance, and as that dis-
tance varies the force changes instantaneously.4* Einstein soon per-
suaded himself that a Lorentz invariant theory of gravitation was not
likely to succeed and adopted a wholly new approach that he fondly
regarded as an extension or "generalization" of his original Principle
of Relativity. Einstein's search lasted many years, from 1907 to the
hectic month of November 1915, when he finally struck gold.
Throughout that time he remained faithful to one guiding insight that
he had in the fall of 1907 and which he later described as his "happi-
est" - or "luckiest" - idea.49 The gist of it can be expressed thus: A
man falling freely, say, from a tall building, must feel completely
weightless, and so will all the other heavy objects - purse, watch, key-
holder - feel that fall together with him. A faint-hearted scientist would
have viewed this merely as a manifestation of the fact that gravity

Indeed, Newtonian measurements made in the so-called Newtonian limit can be

employed to test SR only after an SR-meaning has been imposed on all the basic con-
cepts, starting of course with time (cf. note 43). So Kuhn is right in saying that
Newton's laws cannot be regarded as a "special case of the laws of relativistic
mechanics" unless they "are reinterpreted in a way that would have been impossible
until after Einstein's work" (1970, p. 101).
See eqn. (2.3). The independent variable t in this equation is absolute Newtonian time
and cannot be subjected to a Lorentz boost.
See Pais (1982, p. 178). Einstein says that the idea came to him while he was working
on his (1907j), that is, between 25 September and 4 December 1907.
290 Relativity

imparts the same acceleration on all bodies, regardless of the stuff that
they are made of.50 But Einstein, true to the spirit of Newton's Rules
of Philosophy, boldly jumped to the conclusion that a frame of refer-
ence falling freely in a uniform gravitational field is fully equivalent
to an inertial frame. This is Einstein's Equivalence Principle.51 It implies
that an astronaut shut up in a capsule in interstellar space cannot
tell, by means of physical experiments, whether the capsule moves iner-
tially or falls freely, unless the gravitational field it traverses varies
significantly in space and time. The Equivalence Principle suggests that
inertia is just a limiting case of gravity, in agreement with Mach's idea
(§4.4.3) that inertial phenomena - for example, the deformation of the
liquid surface in Newton's rotating bucket - reflect the presence of
distant matter.
Einstein promptly derived two testable consequences of the Equiv-
alence Principle, viz., (i) that gravity bends light rays,52 and (ii) that the

50
See Chapter Two, note 34. This is sometimes described as the equality of inertial and
gravitational mass, as if there were two distinct Platonic ideas of mass that happen
to be always realized in equal amounts in each piece of matter.
51
Professional literature distinguishes between the strong Equivalence Principle that I
have just stated and the so-called weak Equivalence Principle, which equates inertial
with (passive) gravitational mass. Note that the latter does not assert the equivalence
of certain types of frames and of the physical experiments referred to them, but merely
the equality of mass in its twofold Newtonian role of resistance to acceleration and
susceptibility to gravity (cf. eqns. (2.2*) and (2.3)). Einstein usually formulated his
Equivalence Principle from the point of view of observers at rest - not freely falling
— in a uniform gravitational field (cf. 1907j, p. 454). He repeatedly gave the follow-
ing example: A man inside an elevator completely shut up to the outside world cannot
tell by means of physical experiments whether the elevator is at rest in a uniform
gravitational field that pulls everything toward the elevator's floor with a force —g per
unit mass, or whether the elevator is moving in gravitation-free space with accelera-
tion g in the opposite direction. However, the formulation that I chose above not only
exactly expresses the import of the daydream that he described as his most fortunate
thought, but prepares us for GR in a way in which the alternative formulation does
not.
52
Einstein (1907j, p. 483). In Einstein (1911h, §2), he showed by means of one of his
clever thought-experiments that radiation must gravitate if, in agreement with SR, a
body's mass increases when it absorbs radiant energy. If radiation were not affected
by gravity and could be lifted at no cost, say, from the ground level A to a point B
at a higher gravitational potential, we would be able to create energy by radiating it
from A to B and returning it from B to A stored in a falling body. After a full cycle
the initially radiated energy E would increase to £(1 + ylc'2), where y is the magni-
tude of the gravitational acceleration, / is the height of B above A, and c is the speed
of light.
 5.4 Gravitation as Geometry 291

frequency of radiation varies with the gravitational potential.53 Again,

someone more timid than Einstein might have conceived this gravita-
tional frequency shift as a distortion that, if not corrected, disqualifies
the readings of atomic clocks. But in Einstein's thinking it became tan-
gible evidence that gravity and spacetime geometry are intimately
bound together. Atomic clocks measure proper time along their respec-
tive worldlines. If they seem to go faster or slower in the presence of
different gravitational potentials, it is simply because the length of
intervals along timelike spacetime curves depends on local gravity.
Gravity does not therefore impair the timekeeping function of natural
clocks; it shapes the several strands of time measured by differently
placed clocks. In Einstein's eyes this unexpected turn of thought com-
bined well with Mach's approach to inertia. In SR, the motion of free
particles is determined by the spacetime geometry, which prescribes
which worldlines are timelike or null geodesies. So, in this theory - as
Einstein would later complain to Born (Einstein and Born 1969, p. 258)
- spacetime, a ghostly presence, acts on matter and radiation, and yet
is not acted upon. This asymmetry is overcome if the spacetime geom-
etry, in turn, embodies the gravitational field structure determined by
the distribution of matter.
The real gravitational field is, of course, uniform only within an
agreed approximation, inside a more or less small neighborhood of each
spacetime point. If we lift these restrictions, we must face the fact that
existing fields are far from uniform, with the field here and that in the
antipodes obviously pointing in opposite directions. Einstein was
undoubtedly aware of this when he began to work on gravity. However,
we do not know exactly when he first had the idea of piecing together
the one highly diversified gravitationalfieldof the universe from all those
gravitationally almost uniform small neighborhoods, on the analogy of
a curved surface, pieced together from the almost flat neighborhoods of
its points. The first published testimony of this idea is the long paper he
wrote with Grossmann in 1913. But Einstein declared later that the said
analogy occurred to him shortly after arriving in Zurich in August 1912,
before Grossmann introduced him to the mathematics with which he

53
Einstein (1907j, pp. 458f.; 191 lh, §3). The gravitational frequency shift follows at
once from the argument in note 52 combined with the quantum equation E = hv
(where h is Planck's constant and E and v are, respectively, the energy and the fre-
quency of a photon). Einstein (1905i) was the first to postulate this equation with
full generality, but he never appealed to it in his papers on Relativity (cf. note 8).
292 Relativity

worked it out (CP 6, 535, n. 4). And it is hard to imagine that the Equiv-
alence Principle, the link between gravity and proper time, and the famil-
iar fact of the nonuniformity of gravity could live long together in his
mind without combining to yield some such insight.54
The Einstein-Grossmann paper (1913) presents a theory of gravity
that agrees with GR in intent and in many basic features but sports a
different set of field equations. The central ideas shared by both theo-
ries can be summarily stated thus:
(A) Spacetime is a 4-manifold endowed with a semi-Riemannian metric
g, which agrees well with the Minkowski metric T| in a small neigh-
borhood of any spacetime point P (in the same way that a proper
Riemannian metric agrees with the Euclidian metric - see Chapter
Four, text linked to note 25).55 By virtue of this property, spacetime
curves can be classified - just as in SR - as timelike, spacelike, or
null according to the nature of their tangent vectors. Natural clocks
measure proper time along their respective worldlines.
(B) The metric field g guides free fall, just as the Minkowski metric T|
guides inertial motion in SR: A chargeless nonrotating freely
falling particle - an ordinary massive particle that has no angular
momentum and is not subject to any external influence except that
of gravity - describes a timelike geodesic. Photons describe null
geodesies.
(C) The metric tensor field g is therefore none other than the gravita-
tional field, so it must depend on the distribution of matter and
nongravitational energy in spacetime.
(D) The dependence of g on matter-energy is governed by the gravita-
tional field equations, a set of second-order differential equations
in the metric components that are designed to agree, in the limit-
ing case in which the gravitational field is weak and the bodies

54
The delay until 1912 in reaching this insight was probably due to Einstein's initial
resistance to Minkowski's approach to SR (cf. Einstein CP 5, 121nl2). He was finally
persuaded of its usefulness by Sommerfeld's papers on four-dimensional vector
algebra and vector analysis (1910a, 1910b; cf. Einstein's letter to Sommerfeld of July
1910 in CP 5, 246). Later he acknowledged that without Minkowski's contribution
GR "would probably never have got out of its nappies" (1917a, §17, p. 39).
55
This implies, among other things, that on a small neighborhood of any spacetime
point P one can define a chart relative to which the components of the metric at P
satisfy the relations g/;-(P) = %. Thus, dgij/dxk\P = 0; therefore, the metric g agrees to
first order with T| at P. Of course, the second derivatives of the g{j need not vanish at
P, so the Riemann tensor is not identically 0; the spacetime characterized by the metric
g is generally not flat (or parallelizable).
 5.4 Gravitation as Geometry 293

concerned move slowly, with the Poisson equation of Newtonian

theory (viz., V2O = 4rcp, where O is the gravitational potential
introduced in eqns. (4.13) and p is the density).
Points (A)-(D) require further elucidation. Note first that the metric
g is not known a priori. Hence it is generally impossible to use coordi-
nates that are metrically significant, like the Lorentz coordinates of SR.
From now on, coordinates are just numerical labels with no quantita-
tive meaning.56 Physical quantities must therefore be represented, not by
real-valued functions of the coordinates, as in earlier physics, but by
tensor fields or other geometric objects that can be specified in terms
of arbitrary coordinate systems, but are themselves coordinate-
independent. Equations between such objects are therefore generally
covariant, that is, they hold in every coordinate system. As explained in
§4.1.3, a Riemannian (or semi-Riemannian) metric is a tensor field, and
so is the Riemann tensor, which measures the metric's departure from
flatness. Generally speaking, a tensor field A of rank n on spacetime
assigns to each spacetime point P a tensor AP of rank n at P, that is, an
n-linear function on the tangent space at P (if A and hence AP are covari-
ant), or on its dual, the cotangent space at P (if A and hence AP are con-
travariant).57 Because of linearity, a tensor is fully specified by its action
on a basis of the vector space on which it acts. The local tetrad

56
Of course, this condition does not hold in those special cases in which some fea-
tures of the metric are known or postulated. Thus, in the spherically symmetric
Schwarzschild solution of the Einstein field equations applied in the study of plane-
tary motion, one can define a radial space coordinate that measures the distance from
the planet to the central body. In Friedmann solutions in which there is a singularity
in the past of every worldline of matter (see §5.5), one can define a universal time
coordinate that measures the proper time elapsed since the singularity along each
worldline of matter.
57
Multilinear functions on vector spaces are explained in Supplement 1.5. The above
description can be made more precise as follows: A (p,q)-tensorfieldA on space time
assigns to each spacetime point P a (p,q)-tensor AP at P. This is a {p + g)-linear func-
tion acting on (p + g)-tuples of vectors, q of which are drawn from the tangent space
at P and p of which are drawn from the cotangent space at P. If p = 0, A and AP are
said to be covariant of rank q. If q = 0, they are said to be contravariant of rank p.
If both p and q differ from 0, A and AP are said to be mixed, p times contravariant
and q times covariant, of rank (p + q). Of course, the definition of a mixed tensor
must specify the position of the p tangent vectors and the q cotangent vectors in the
294 Relativity

defined at P by an arbitrary chart x is a basis of the tangent space at P

(see above, after eqn. (5.12)). Likewise, the corresponding quadruple of
coordinate differentials (dx^dx^dx^dx3) is a basis of the cotangent
space at P. Suppose then that A is a tensor field of rank 2. The real-
valued function that assigns to each point P in the domain of x the
number AP(3/3x',3/3x;) - if A is covariant - or the number AP(dxz,dx;) -
if A is contravariant - is the (z,/)th component of A relative to the chart
x. When there can be no confusion as to the chart in question, this func-
tion is denoted by Atj if A is covariant and by Alj if A is contravariant.
The set of functions A{j (respectively, Alj) fully specify the tensor field A
on the domain of x.58 But it is not enough to use tensor fields or other
coordinate-free geometric objects for representing physical quantities.
The physicist also needs a coordinate-free representation of the way
these quantities vary continuously over spacetime. Partial derivatives
with respect to the coordinates will not do if the coordinates are physi-
cally meaningless. So Grossmann drew Einstein's attention to a paper
by Ricci and Levi-Civita (1900) that presented an absolute - that is,
coordinate-independent - differential calculus for tensor fields (now
known as the tensor or Ricci calculus). Given a tensor field A of rank n
on a Riemannian manifold, the Ricci calculus constructs from it a new
tensor field of rank n + 1, the covariant derivative of A, which suitably
reflects A's change along any curve in the manifold.
The geodesic law of motion (B) was introduced by Einstein as a pos-
tulate. However, Einstein and Grossmann (1913, p 10) proved it as a
theorem for matter distributed in a particularly, simple way (as pres-

(p + q)-tuples that are its arguments. One speaks then, for example, of a tensor of
rank 6, covariant in the first and third, and contravariant in the second, fourth, fifth,
and sixth indices. The origin of the terms 'covariant' and 'contravariant' is too con-
torted to deserve explanation. The term 'index' comes from the habit of represent-
ing tensors by their components relative to a chart and labeling those components
with indices ordered like the vectors in each argument of the tensor.
Let y be another chart defined on the same spacetime region as x. Then, the vector
fields 3/dy (resp. dy') are linear combinations of the d/dx' (resp. the dx'), so the com-
ponents of A relative to y can be readily computed from its components relative to
A through formulas involving the partial derivatives dx'/dy' (resp. dyl/dxj). This led to
the popular description of tensor fields as "arrays of functions that transform accord-
ing to such-and-such rules" (the transformation formulas being given). Einstein
warned Besso against this unperspicuous description on 31 October 1916: "Definition
of tensors: not 'things that transform thus and thus'. But: things that can be described,
with respect to an (arbitrary) coordinate system by an array of quantities (A^v)
obeying a definite transformation law" (Einstein and Besso 1972, p. 85).
 5.4 Gravitation as Geometry 295

sureless dust). In the context of GR it was readily shown that photons

describe null geodesies (von Laue 1920; Whittaker 1928). Finally, in the
late 1930s, Einstein, in collaboration with several young mathemati-
cians, inferred the geodesic law of motion for massive particles from the
field equations of GR by a difficult and controversial argument.
Point (C) raises the following question: How shall the distribution of
matter and nongravitational energy be represented in a manner that is
appropriate for conveying the metric tensor field's dependence on it? In
the context of SR, Minkowski had devised a tensorial representation for
the energy of the electromagnetic field, whose components relative to
a Lorentz chart were familiar classical quantities: the charge density,
the components of the Poynting vector (representing energy flux), and
the components of Maxwell's stress tensor /initially conceived for
representing tensions in the aether). Inspired by this example, von Laue
(1911) constructed, again in the context of SR, the mechanical
energy-momentum tensor, a symmetric tensor field representing the
dynamical state of a material continuum whose components relative to
a Lorentz chart are equal to or built from classical quantities: the mass-
energy density, the components of momentum density, and the compo-
nents of the classical stress tensor.59 Von Laue argued that in each
domain of physics there is a tensor field whose components have a phys-
ical meaning corresponding to the above. As these diverse domains
interact in a region of spacetime, their respective tensor fields should be
added to define a single symmetric tensor field of rank 2, representing
the presence of matter-energy in that region. Einstein seized on this idea.
In his geometric theories of gravity, which culminated with GR, the dis-
tribution of matter and nongravitational energy is represented by just
such a general, generally unspecified, energy-momentum tensor, which
I shall denote by T. Late in life Einstein declared that "all attempts to
represent matter by an energy-momentum tensor are unsatisfactory"
(Einstein and Infeld 1949, p. 209), for all such tensor fields "must be
regarded as purely temporary and more or less phenomenological
devices for representing the structure of matter, and their entry into the
equations makes it impossible to determine how far the results obtained
are independent of the particular assumption made concerning the con-

59
It is perhaps worth noting that the classical stress tensor - employed since the nine-
teenth century by structural engineers for representing tensions, that is, strains and
stresses, inside a beam, column, etc., of a building - motivated the adoption of the
name 'tensor' for the geometric objects under discussion here.
296 Relativity

stitution of matter" (Einstein, Infeld, and Hoffmann 1938, p. 65). Still,

as we shall see in §5.5, simple, highly idealized models of T play an
essential role in relativistic cosmology.
Finally we come to point (D), the field equations of the new theory
of gravity. One would expect that, once the gravitational field and the
distribution of gravitational sources are represented, respectively, by
the tensors g and T, the field equations would equate geometric objects
constructed from these tensors. This would ensure that the equations
are generally covariant and thus physically meaningful even if the coor-
dinate systems are not. However, Einstein and Grossmann, while insist-
ing on the value of a generally covariant formulation of the laws of
physics (1913, pp. 7, 10, 18), make an exception with the equations
of gravity, which, in the form proposed by them, hold only for a certain
ample but anyway restricted family of coordinate systems. Einstein
was distressed by it, but he soon found consolation in an argument
by which he purported to prove that a generally covariant system of
gravitational field equations would be incapable of determining the
course of events under its sway. The argument - later dubbed "the hole
argument" - runs as follows: Suppose that T is identically zero on a
compact, open, simply connected spacetime region $£, a veritable hole
in the fullness of matter. Let y be a timelike curve that crosses 3^, joining
two points P and Q on ffls border. Let f be a smooth one-one mapping
of spacetime onto itself that agrees outside %C with the identity mapping
but differs from it on <K (f(X) = X if X e 3€; f(X) * X if X e W). Then,
if the field equations are generally covariant and y is the worldline of
a freely falling test particle, so is /"°y. Thus a generally covariant system
of gravitational field equations cannot determine the trajectory that,
given the matter distribution T, a freely falling particle would follow
from P to Q, but leaves more than one way open to it. To see the force
of this argument one must bear in mind that a smooth one-one
mapping f oi an ^-manifold onto itself "drags along" all the geomet-
ric objects defined on the manifold; in other words, if A is such an
object, f defines an object f^A of the same type that behaves at f(X)
just like A at X.60 If the field equations are generally covariant and the
metric g is a solution for the matter distribution T, then, under the
conditions prescribed above for T and /*, the metric /*g is also a solu-
tion. And, of course, if y is a geodesic of the metric g, f°y is a geodesic
of the metric /*g. So we have two equally admissible worldlines for our

60
To be specific, let A be a covariant tensor field of rank 2. Then, if V and W are two
vector fields and P is any point in the manifold, ^A^(p)(^V^p),^W/r(P)) = AP(VP,WP).
 5.4 Gravitation as Geometry 297

freely falling test particles starting from the same event P (indeed
infinitely many, since fis arbitrary on $£). So far, so good. But the argu-
ment forgets the fact, so clearly set forth by Newton, that points in a
structured manifold have no individuality apart from their structural
relations.61 Let X be a point in the hole *K that is equidistant by the
metric g from five points A, B, C, D, and E outside 3€. Then f(X) is
equidistant, by the metric /*g, from A, B, C, D, and E. In what sense
can f(X) stand in the manifold (3€,/*g) for a different physical event
than the one X stands for in the manifold (9€,g)? An answer eludes me.
Nor can a physicist conceive of any difference between a test particle
aptly represented by the geodesic y in ($t u {P,Q},g) and one aptly rep-
resented by the geodesic f°y in (3€ u {P,Q},/*g}.62 After Einstein dis-
covered tensorial equations for gravity in the fall of 1915, the hole
argument was no longer necessary. He never publicly retracted it, but
in letters to Ehrenfest (26 December 1915) and Besso (3 January 1916)
he explained that physical reality consists of spacetime coincidences,
which are preserved by all point transformations; thus, "to demand
that the laws do not determine anything beyond the aggregate of space-
time coincidences is the most natural thing in the world".63
Einstein submitted to the Prussian Academy a system of generally
covariant equations for the gravitational field on 11 November 1915
(1915g). They equate the Ricci tensor - a symmetric tensor of rank 2
built from the metric tensor and its first and second derivatives - with
the energy-momentum tensor T multiplied by a constant:
Rjk = -KT ; * (5.28)
These equations responded to a very questionable physical assumption
and Einstein very soon discarded them, but not without first having
derived from them - in a paper submitted on 18 November (1915h) -
61
See the long quotation from Hall and Hall (1962, p. 103), in §2.2. I learned this in
my view decisive objection to Einstein's hole argument from John Stachel, before
either of us became aware of Newton's text. Of course, in 1914 Einstein did not have
access to it.
62
Of course, each geodesic can represent more than just one thing. Mathematical
models fit many applications, to such-and-such a degree of accuracy, at this or that
level of abstraction. But whatever is well represented by y in ("K u {P,Q},g) is equally
well represented by / ° y in (M u {PjQh/^g).
63
Einstein and Besso (1972, p. 39). I quote the relevant passage of the letter to Ehren-
fest in English translation in Torretti (1983, p. 166). At the time of writing neither
letter had yet been printed in Einstein CP. Einstein's hole argument attracted the atten-
tion of philosophers after Earman and Norton (1987) reformulated it as a refutation
of absolutist theories of spacetime.
298 Relativity

a correct prediction of the observed motion of Mercury, which New-

tonian celestial mechanics did not fully account for.64 On 25 Novem-
ber, Einstein (1915i) proposed to the Academy another system of
generally covariant equations. These are the Einstein field equations
(EFE), the core of GR. They agree with eqns. (5.28) in free space -
where T is identically zero - so they also support Einstein's solution of
the anomaly of Mercury. In the form in which they were first published,
the EFE equate the Ricci tensor with a tensor field constructed by Ein-
stein from the energy-momentum tensor T:

(5.29)

Equations (5.29) are algebraically equivalent to the following system,

in which T multiplied by the gravitational constant -K stands alone on
the right-hand side and is equated with the so-called Einstein tensor -
built from the Ricci tensor - on the left:

The Einstein tensor has a mathematical property that T must have, if

energy is conserved. This is a strong motive for equating these two
tensors fields.65 Due to the symmetry of the tensors involved (Rjk = Rkp
etc.), the EFE are only ten distinct equations, with ten unknowns (the
metric components gjk). However, due to the said mathematical prop-
erty, there are no more than six independent equations. Ten indepen-
dent equations with ten unknowns would unduly constrain the choice
64
The perihelion of Mercury - the point where the planet comes closest to the Sun -
advances each year on the celestial sphere some 56" (seconds of arc). Almost 90%
of this advance merely reflects the slow precession of the Earth's axis. Over 9% can
be attributed, in Newtonian celestial mechanics, to the perturbing action of the other
planets. But there remains a perihelion advance of some 43" per century for which
no satisfactory Newtonian explanation was ever found. (For more precise figures see
§7.2.) Einstein expressed his interest in this anomaly in connection with his research
on gravity in the letter that he wrote to Habicht on Christmas eve, 1907 (CP 5, p.
82). His disappointment with the Einstein-Grossmann field equations stemmed in
part from the fact that they gave the wrong number for the unexplained part of
Mercury's perihelion advance (letter to Sommerfeld of 28 November 1915; Einstein
and Sommerfeld 1968, p. 32). Einstein's work on the perihelion of Mercury is dis-
cussed again in §7.2.
65
For a stricter explanation of this point, see Torretti (1983, p. 323n34, and the text
(on p. 175) to which this note is appended).
 5.5 Relativistic Cosmology 299

of a coordinate system. But this is not the right place for discussing the
mathematics of the EFE. In the following section I shall touch on some
of their wholly unexpected physical implications.

5.5 Relativistic Cosmology

Less than two months after the publication of system (5.29), Schwarz-
schild (1916a, b) produced an exact solution. It provided a model of the
motion of a small planet or satellite in the gravitational field of a large
central body. Schwarzschild assumed a field static in time and spheri-
cally symmetric in space that converges to the flat Minkowski metric at
spatial infinity and is free of matter except at most on a small neigh-
borhood of the axis of symmetry. The Schwarzschild solution ratifies
Einstein's (1915h) calculation of Mercury's perihelion advance through
the approximate solution of system (5.28). To obtain this particular
application one assumes that the Sun is alone in the universe, on the
symmetry axis of the Schwarzschild field. The constant of integration
that occurs in the solution is put equal to two solar masses66 and a test
particle - that is, a nonrotating object of negligible mass - is imagined
circulating around the Sun, under the sole influence of the Schwarz-
schild field. Then, if the test particle is initially placed at Mercury's dis-
tance from the Sun and given Mercury's velocity, it will go around the
Sun in 80 days, tracing an ellipse in space with the Sun at one focus, and
a perihelion advancing slightly more than 43" per century.
The Minkowski metric at spatial infinity implicit in the Schwarz-
schild solution made good sense in applications to the Sun, moving
through empty space at great distance from comparable bodies. The
GR spacetime curvature on a sphere drawn about the Sun at one-half
the distance from a Centauri is surely insignificant compared to that
on Mercury's orbit. On the other hand, flatness at true spatial infinity
is incompatible with Einstein's Machian view of inertia, for it prescribes
a definite worldline - in agreement with SR - to any free particle
infinitely far from other matter. Now, in accordance to the Machian
view, the worldline of a free particle can only be prescribed by the
matter outside it. It is all right for very distant matter to guide the free
particle along a (nearly) Minkowski geodesic, but infinitely distant
matter can have no such effect. A free particle that has no matter-

66
Bartels (1994, §2.3), "The Interpretation of the Schwarzschild Mass", provides a
lucid, philosophically motivated analysis of this move.
300 Relativity

dependent gravitational field to guide it should not know what to do.

Einstein expressed his Machian standpoint as follows:
In a consistent theory of relativity there can be no inertia relative to
"space", but only an inertia of masses relative to each other. Hence if I
take a mass sufficiently far away from all other masses in the world its
inertia must fall down to zero.
(Einstein 1917b, p. 145)
He therefore explored together with Jakob Grommer solutions of eqn.
(5.29) that are (i) static, (ii) spherically symmetric in space, and (iii)
such that the metric components gtj relative to a suitable chart do not
converge at spatial infinity to the values r^ but degenerate in a way
that entails the vanishing of inertial masses. They concluded that these
requirements could not be reconciled with the fairly slow motion that
astronomers attributed to stars.67
Confronted with this difficulty Einstein realized, in a stroke of
genius, that in the context of GR the problem he was trying to solve
could be simply done away with. A Riemannian metric of positive cur-
vature can be defined on a boundless yet finite space in which no lump
of matter is infinitely far from the others and the question of the behav-
ior of free particles and the metric at infinity does not even arise.
Assuming that matter in the large is evenly distributed and can there-
fore be passably represented by a homogeneous and isotropic fluid,
Einstein searched for a solution of the EFE such that spacetime is uni-
formly filled with matter and can be partitioned into finite spacelike
slices of the same constant curvature. When he saw that eqns. (5.29)
admit no such solution, he added a new term to the left-hand side, to
make them read as follows:

(5.31)

The factor X is a constant, known as the cosmological constant, which

in the light of observation must be so small as to be negligible.68 Still

67
In fact they were pursuing a will-o'-the-wisp, for - as G. D. Birkhoff would prove
shortly (1923) - a solution of the EFE that is spherically symmetric in space must
converge to flatness at spatial infinity (besides being necessarily static).
68
Current data imply that |A,| ^ 3 x lO^m" 2 (Ciufolini and Wheeler 1995, p. 209). In
what other branch of physics would an adjustable parameter with such a tiny upper
bound be pronounced different from 0?
 5.5 Relativistic Cosmology 301

eqns. (5.31), with X > 0, do admit the spatially finite static solution
that Einstein was looking for.
Einstein's static world model marks the beginning of modern cos-
mology. The notion that the world fills only a finite space, even though
it has no limits, resolved in an unexpected, intellectually most satis-
factory way one of the antinomies that - according to Kant (§3.5) -
banned cosmology from the realm of rational inquiry. On the other
hand, the modified EFE did not serve Einstein's Machian motivation
very well. First, a nonzero cosmological constant constitutes precisely
the kind of sourceless universal field acting on matter and not acted
upon that the Machian approach was supposed to avoid. Second, de
Sitter (1917a, 1917b, 1917c) soon found a nonflat solution of eqns.
(5.31) that is totally exempt of matter. As originally presented, the solu-
tion is static, but a handful of test particles scattered in de Sitter space
will fly away from each other. De Sitter, Weyl, and others noted the
agreement of this surprising prediction with the no less surprising data
on the recession of galaxies that were coming in from California.
The next momentous step in the creation of modern cosmology
was hardly noticed at the time. The Russian meteorologist Alexander
Friedmann (1922, 1924) discovered a family of nonstatic, spatially
homogeneous and isotropic solutions of the EFE, of which the static
solutions of Einstein (1917b) and de Sitter (1917c) were particular -
and, as it turned out, unstable - cases. Friedmann's first paper appeared
in the Zeitschrift fur Physik followed by a note in which Einstein
(1922) asserted, without any further explanation, that Friedmann's
work rested on a mathematical error. That the paper was printed
despite this authoritative disclaimer speaks well of German science in
the 1920s. Einstein (1923) soon conceded that the error was his, but
Friedmann earned little or no credit for his world models until long
after Lemaitre (1927) had independently rediscovered them.
The Friedmann world models are solutions of the modified EFE -
eqns. (5.31) - admitting any value for the cosmological constant A,,
including 0. Like all particular solutions of a system of partial differ-
ential equations, they involve certain special assumptions. These
concern the geometric structure of spacetime and the state and motion
of matter. As to the former, Friedmann assumed (a) that spacetime
admits a global time coordinate x°; (b) that the parametric curves of
x° are everywhere at right angles to the spacelike hypersurfaces x° =
const.; and (c) that on each hypersurface x° = const, the spacetime
metric amounts to a proper, time-dependent, Riemannian metric of
302 Relativity

constant curvature.69 With regard to matter, Friedmann assumed (d)

that it can be represented as a pressureless fluid whose worldlines agree
with the paths of the parametric curves of x°. Assumption (c) can be
seen as a geometric statement of the spatial homogeneity and isotropy
of matter in the large. On the other hand, according to Friedmann,
assumption (b) is to be adopted only because it simplifies calculations.
Now, assumption (b) is a strong requirement that severely restricts the
available solutions. However, Robertson showed that, if the worldlines
of matter are geodesies - a natural assumption in GR, given assump-
tion (d) -, then assumptions (b) and (c) follow from the philosophical
demand that no observer shall be able to distinguish between any two
directions about him by some intrinsic property of spacetime, or to
"detect any difference between his observations and those of any con-
temporary observer" (1929, p. 823 - this is known in the literature as
"the Cosmological Principle").
Except for the limiting case in which spacetime is static, the
Friedmann world models share a remarkable feature: The worldlines
of matter either diverge from one another, or converge toward one
another, or first diverge and then converge. Suppose that we live in a
Friedmann universe in which the matter worldlines are currently
diverging, either forever or until such time as they will begin to con-
verge. Suppose that we mentally go backwards in time along one such
worldline y. Then, as we delve into the past, second by second, year
by year, all the other worldlines will converge towards y, so that matter
in every spatial neighborhood of y will be denser and denser, its average
density increasing beyond all bounds as the time we have gone back-
wards approaches a certain value (the currently fashionable figure is
12 to 15 billion years). Of course, the density of matter cannot be actu-
ally infinite at a given spacetime point, since neither the energy-
momentum tensor nor the metric would be defined at such a point. On
the other hand, in a universe of this type, if the time coordinate x° is
defined so as to measure the temporal distance between events along
each worldline of matter, the range of x° obviously cannot be the entire
real number field R, but only a finite connected open interval in R.
Thus, if you choose a time unit, assign time 0 to the present, and use
the negative reals to label the past, the entire past of each worldline of

69
Let K(t) denote the constant curvature of the hypersurface x° = t. Friedmann (1922)
considers only the case K(t) > 0; the case K(t) < 0 is the subject of Friedmann (1924).
He did not consider the case K(t) = 0 countenanced by Einstein and de Sitter (1932).
 5.5 Relativistic Cosmology 303

matter will be mapped onto some interval (-T, 0). You may then, with
Friedmann, call T "the time since the creation of the world" (die Zeit
seit der Erschaffung der Welt; Friedman 1922, p. 384n.). But you
would be wrong to think that you have found the date of creation.
- T is the greatest lower bound of the range of your time coordinate.
But it is not the date of any event, for events can only occur on matter
worldlines, and therefore they must carry a date greater than -T.
Indeed, in a Friedmann universe every process whose beginning can be
dated is preceded by some length of time.
Relativistic cosmology achieved somewhat greater generality at the
hands of Lemaitre (1927), who proceeded from the weaker assump-
tion that matter is a perfect fluid (with constant, isotropic pres-
sure), not a presureless one. Einstein and de Sitter (1932) proposed an
expanding universe with flat (Euclidean) spacelike slices orthogonal
to the worldlines of matter. Robertson (1935) and Walker (1935,
1937) derived the most general expression for the metric of a spatially
homogeneous and isotropic spacetime, which is independent of GR
(not necessarily a solution of the EFE) but agrees with SR on each
tangent space (is "locally Minkowskian"). Spacetimes with this metric
are termed FRW worlds (for Friedmann, Robertson, and Walker), or
FLRW worlds (if one is mindful of Lemaitre), and constitute the so-
called standard model of current cosmology. The success of a world
picture so inimical to traditional myth and science is mainly due to the
accumulation of favorable empirical evidence. By the end of the 1920s,
Hubble and his colleagues at the Mt. Wilson observatory had estab-
lished that the Galaxy, with its estimated hundred million stars, is only
a speck in the universe, and that most of the observable nebulae are in
fact other such star groupings, at enormous distances from each other.
The systematic study of the light received from these sister galaxies
showed that, except in the case of a few nearby ones, they are reced-
ing from us at speeds that - according to Hubble (1929) - are pro-
portional to their distance. Lemaitre (1927) proposed his version of the
Friedmann solutions as a theoretical explanation of Hubble's empiri-
cal results and was the first professional scientist to speak of the dense,
hot beginning of the universe (the poet Edgar Allan Poe had done so
in 1848).
For 20 years, the GR theory of the expanding universe ran against
one grave obstacle. According to Hubble's calculations, the time
elapsed since the start of cosmic expansion was about 2 billion years,
a good deal less than the age of some rocks, as calculated by geolo-
304 Relativity

gists. To overcome this difficulty, Milne proposed that there are two
types of natural clocks, embodied in different processes and running at
different rates, that yield two seemingly inconsistent series of chrono-
logical data. The steady-state cosmology put forward in two different
versions, by Bondi and Gold (1948) and by Hoyle (1948), simply
denied the finite age and evolution of the universe, and compensated
for its manifest expansion by postulating the constant creation of
matter everywhere at a rate too low to be detected by us.70 The
cosmological and the geological time scales were reconciled when
Baade (1956) showed that intergalactic distances had been underesti-
mated due to an error concerning the luminosity of Cepheid variable
stars. Baade's recalibration lengthened the age of distant light signals
by a factor of 2, and subsequent findings have further increased this
factor.
Still, the great majority of physicists did not take GR cosmology very
seriously until quantum physics - or rather, a factual discovery that
they understood in the light of it - persuaded them that the world we
live in is well portrayed, in the large, by one or the other of the expand-
ing FLRW models. I think that this turn of events is highly instructive,
for GR, which provides the framework for understanding the expan-
sion of the universe, is in fact incompatible with quantum physics. But
let me just tell the story very briefly here and leave the philosophical
morals for Chapter Seven. In the 1940s Gamow and others speculated
on the origin of the elements.71 The relative abundance and uniform
distribution of helium could be readily understood if it was generated
by thermonuclear reactions when the universe was very dense and hot.
If that was the case, matter and radiation must have once been coupled
together in continual mutual exchange, and a relic of radiation as it
stood at the time of decoupling should still be observable. This radia-
tion will now be cooled down to almost OK due to expansion, but it

70
Bondi and Gold (1948, §4.1) put the average rate of creation at 10~43 grams per cubic
centimeter per second. The steady-state theory was not in itself a response to the time-
scale problem. It issued from a purely methodological consideration. "The universe
is postulated to be homogeneous and stationary in its large-scale appearance as well
as in its physical laws" not because the authors claim that this so-called Perfect Cos-
mological Principle must be true, but because "if it does not hold, one's choice of the
variability of the physical laws becomes so wide that cosmology is no longer a science;
one can then no longer use laboratory physics without relying on some arbitrary prin-
ciple for their extrapolation" (1948, §1.2).
71
Alpher, Bethe, and Gamow (1948), and Gamow (1948). For a lively and highly
instructive report on these developments, see Kragh (1996, Ch. 3).
 5.5 Relativistic Cosmology 305

will retain its isotropy and universal presence, as well as its thermal
character (i.e., the distribution of energy flux at each frequency will
agree with Planck's law). Alpher and Hermann (1948) predicted a
current temperature of 5K. Little was made of this until a decade
and a half later, when the idea was revived just in time to furnish a
cosmological explanation of the pervasive unremitting isotropic
apparently thermal radiation at c. 2.7K discovered by Penzias and
Wilson (1965; cf. Dicke et al. 1965) while they investigated back-
ground microwave noise in radio communications. Very accurate and
detailed measurements have since confirmed that this radiation is
indeed thermal and almost perfectly isotropical.72 This is generally
regarded as overwhelming evidence for the hot dense beginning of the
universe and its subsequent cooling by expansion (of course, there are
dissenters).
In the standard model all worldlines of matter proceed as if from
a common point at which, if it existed, the energy density would be
infinite and the spacetime metric would be undefined. In the early days
of modern cosmology this "initial singularity" was glossed over as an
idealization that would be automatically avoided in a more realistic,
not quite perfectly isotropic and homogeneous world model. For
example, de Sitter wrote that "the conception of a universe shrinking
to a mathematical point at one particular moment of time [. . .] must
[. . .] be replaced by that of a near approach of all galaxies during a
short interval of time" (1933, p. 631). But since after 1965 there has
been a tendency to take the singularity literally, due to the almost
perfect isotropy of the cosmic background radiation and to the singu-
larity theorems proved by Penrose, Hawking, and Geroch.73 Some
believe that this raises a problem. If matter has existed only for a finite

72
A small but systematic anisotropy is attributable to the motion of the Solar System
across this ocean of radiation. When that is subtracted there remain anisotropies of
up to 1 part in 10,000, short of which the articulation of matter into galaxies and
stars would perhaps have been impossible.
73
In a series of papers culminating with Hawking and Penrose (1970). For a textbook
exposition, see Wald (1984, Ch. 9). For a philosophically minded discussion, see
Earman (1995, Ch. 2). The singularity theorems establish the existence of incomplete
timelike geodesies in every GR spacetime that meets certain plausible physical con-
ditions, when such geodesies converge to one another as in FLRW worlds. An incom-
plete geodesic is an inextensible geodesic defined on an interval that has a greatest
lower bound, or a least upper bound, or both. A geodesic y defined on an interval I
is said to be inextensible unless there is another geodesic y' defined on an interval I'
such that I is a proper part of I' and y' takes the same values as y on I.
306 Relativity

time in an FLRW world, some parts of it have never had an opportu-

nity to interact with others. For example, the most distant galaxies that
we barely manage to observe at opposite sides of the sky cannot yet
be observed from each other. How could they get to be so similar? In
particular, when could the cosmic background radiation in those non-
interacting regions of the world reach the state of thermal equilibrium
that we ascribe to it? Guth (1982) thought that he could give an unex-
pected solution to this question - and a few others - by boldly assum-
ing the EFE to work in a cosmic setting subject to a "Grand Unified"
quantum field theory of nuclear forces. Under his assumptions the uni-
verse expands exponentially during a very short "inflationary" period
(less than 10~30 second), after which the standard FLRW expansion
takes over. Before the latter begins, even the more distant parts of the
currently observable universe have been sufficiently close to mix
together by normal thermal processes. However, Guth's original model
leads to gross inhomogeneities, that render it untenable (Blau and Guth
1987, p. 550). To overcome this difficulty other inflationary schemes
have been proposed, the most remarkable of which is due to Linde
(1983, 1986). Here "the local structure of the universe is determined
by inflation," governed by the EFE; however, "its global structure is
determined by quantum effects" (Linde 1987, p. 607).
It proves that the large-scale quantum fluctuations of the scalar field $
generated in the chaotic inflation scenario lead to an infinite process of
creation and self-reproduction of inflationary parts of the universe. In
this scenario the evolution of the [. . .] universe has no end and may have
no beginning. As a result, the universe becomes divided into many dif-
ferent domains (mini-universes) of exponentially large size, inside which
all possible (metastable) vacuum states are realized.
(Ibid.)
Such mini-universes can possess very different basic physical proper-
ties (e.g., different dimension number). One of them evolves through
inflation into the four-dimensional roughly FLRW world in which we
live, which, in stark contrast with its siblings, sports the very unlikely
combination of features required for human life.74

74
Guth (1997) explains inflationary cosmology to nonspecialists, with ample references
to the original literature. For a less sanguine appraisal, see Earman and Mosterin
(1999).
 CHAPTER SIX

Quantum Mechanics

The word 'quantum' was used in Latin as a relative and interrogative

adjective, adverb, or pronoun, to mean 'how much' or 'how many'.
Before the twentieth century it was occasionally used in English, as in
German,1 to mean a definite amount of something. It was in this sense
that Planck (1900, in PAV I, 706) spoke of "the elementary quantum
of electricity e", meaning "the electric charge of a positive univalent
ion or an electron". In that paper Planck derived the law of thermal
(or "black body") radiation that now bears his name. He used as a
model a collection of many linearly oscillating monochromatic res-
onators enclosed in a cavity with reflecting walls,2 and he assumed that
the energy of the resonators oscillating with a particular frequency v
was an integral multiple of the quantity
Ev=kv (6.1)
where h is a constant of nature with the dimension of action (= energy
x time), subsequently known as Planck's constant. It is unlikely that
Planck meant this assumption as a general physical hypothesis and not
just as a prop for his argument, a peculiarity of his fictitious model. At
any rate, he did not then call h an 'elementary quantum of action' nor
Ev one of energy. Five years later, Einstein - building on Planck's work
- conjectured that all electromagnetic radiation consists of "energy
quanta (Energiequanten) localized at points in space, which move undi-

1
See, for example, Kant (1787, p. 224), quoted at the beginning of §3.4.2.
2
He could profitably adopt this far-fetched and seemingly arbitrary model because in
such a cavity - as Kirchhoff (1860) proved from the conservation of energy princi-
ple - the distribution of energy over frequency in thermal equilibrium does not depend
on the nature of the radiating bodies.

307
308 Quantum Mechanics

vided, and can only be absorbed or generated as wholes" satisfying the

condition (6.1) (1905i, p. 133). Thereafter, the word 'quantum' was
increasingly used as a noun to designate the minimum amount in which
some physical quantity is found in nature and by multiples of which it
increases or decreases, as well as an epithet for hypotheses, theories,
and the like, that imply the reality of such quanta and, in particular,
the "quantization" of energy in accordance with eqn. (6.1). Following
standard usage, I speak of 'quantum physics' in this general sense and
reserve the name Quantum Mechanics (QM) for the theory formed in
the late 1920s by the conflation of Heisenberg's "matrix mechanics"
of 1925 and Schrodinger's "wave mechanics" of 1926. QM involves a
wholly new way of understanding the purpose and the basic concepts
of physics, which has become the subject of endless philosophical
debate. This state of affairs is all the more irksome in view of the
theory's unblemished record of experimental success. The equations of
QM are not Lorentz invariant and therefore can hold well only in sit-
uations in which relative speeds are small, but most authors agree that
its philosophical difficulties are inherited by the Lorentz invariant
quantum theories developed in its wake. The philosophical problems
of QM are no doubt by far the chief source of current interest in
the philosophy of physics, especially since the - mostly imaginary -
philosophical problems of Relativity have been tamed.
In the space available here I can only introduce the reader to the
main disputed questions and opposing theses. I begin with a summary
of events leading to the birth of QM (§6.1). In §6.2 I refer to matrix
and wave mechanics, the equivalence of both theories, Born's
probabilistic interpretation of Schrodinger's \|/-function, and Heisen-
berg's indeterminacy relations; before explaining the latter I sketch the
QM formalism of vectors and operators in Hilbert space (§6.2.5).
In §6.3 I discuss the two chief philosophical problems of QM, viz., the
EPR paradox (thus called after Einstein, Podolski, and Rosen 1935)
and the problem of measurement. In §6.4 I consider a few of the
philosophical proposals concerning the right meaning of the theory. I
close the chapter with a few remarks on relativistic quantum physics
(§6.5).3

1
§§6.1 and 6.2 owe much to Jammer (1966), Darrigol (1992), and the editor's intro-
duction to Van der Waerden (1967). Readers who are unsatisfied with my sketches
should turn to these works. For greater detail and abundant references, see Mehra
and Rechenberg (1982-88).
 6.1 Background 309

6.1 Background

6.1.1 The Old Quantum Theory

Quantum physics before QM is loosely referred to as the Old Quantum
Theory. It was the theoreticians' response to the surprising news about
the fine structure of matter that began to flow from physical laborato-
ries at the end of the nineteenth century. In 1895 Rontgen discovered
X-rays, which, after long discussions, were finally classified as high-
frequency electromagnetic radiation when von Laue and his associates
succeeded in diffracting them in 1912. In 1896 Becquerel discovered
radioactivity. In 1898 Rutherford analyzed radioactive output into oc-
rays (to be identified many years later as twice ionized Helium atoms)
and |3-rays (soon assimilated to cathode rays), to which in 1900 Villard
added y-rays (ultra-high-frequency radiation). At about the same time,
Rutherford established that a given amount of a radioactive element
loses one-half of its radioactivity in a period of time that depends solely
on its nature and is not affected by changes in the surrounding cir-
cumstances. In 1899 J. J. Thomson proved that cathode rays - first
observed by Faraday in the 1830s and instrumental to Rontgen's
discovery - are beams of extremely light particles, which he dubbed
'electrons' and were thereafter regarded as the ultimate free carriers of
negative electric charge.4 The use of these powerful effects as labora-
tory tools produced a flood of information. Side by side with them, the
spectroscopists supplied increasingly accurate measurements of the
spectral lines characteristic of the several chemical elements, which dis-
played enticingly simple yet utterly incomprehensible regularities.
Philosophers who reflect on quantum physics should not lose sight
of these events. In stark contrast with the birth of mathematical physics
in the seventeenth century, which consisted for the most part in recon-
ceiving familiar facts under new standards of rigor, the "quantum rev-
olution" was forced on a fairly conservative scientific establishment by
a surfeit of unexpected phenomena. By stressing this I do not mean to

4
For the date of J. J. Thomson's discovery of the electron, see Pais (1986, pp. 78ff.).
Other authors give other dates, not because of any disagreement about laboratory
records, but because they have different notions of what it takes to discover an ele-
mentary particle (for an instructive discussion, see Arabatzis 1996). The term 'elec-
tron' had been introduced in 1891 by Stoney, who since 1874 had argued for the
existence of an indivisible unit of electric charge.
310 Quantum Mechanics

endorse thoughtless empiricism. To be sure, experiments were designed

and results gleaned from them in terms of some preconceived theoret-
ical scheme. Even Becquerel's photographic plates, which were black-
ened by radioactivity when accidentally exposed to uranium salts while
they lay idle in a drawer, could only convey their startling revelation to
someone versed in the carefully articulated late nineteenth-century view
of nature. Democritus or Aristotle would probably have been content
with a more casual explanation of a much lesser consequence. But late
nineteenth-century physics, the classical system that some had come to
think would thenceforth progress only by making better measurements
of well-defined, well-understood quantities, proved unable to cope with
the new phenomena, and the efforts of the early quantum theorists were
driven by their perception of this somehow paradoxical state of affairs.
This explains their partial yet persistent reliance on classical mechanics
and electrodynamics, whose concepts and principles were, after all,
built into the experimental and computational procedures by which the
new information was obtained. It also accounts for their readiness to
combine seemingly inconsistent notions, to try out obscure algebraic
relations in lengthy and laborious calculations, in short, to follow every
lead that might take physics out of the morass in which it was caught.
The Old Quantum Theory can be traced back to the work by Planck
(1900) and Einstein (1905i) mentioned above, but it began in earnest
with Bohr's three-part series, "On the Constitution of Atoms and
Molecules" (1913). By that time it was generally agreed that matter
consisted of atoms of a number of (the order of 102) different kinds.
Despite the etymology of their name (< 'a-' privative + 'tomos' = 'slice'),
atoms were supposed to have parts. These were required in view of the
complex structure of atomic spectra and because atoms, which under
ordinary conditions are electrically neutral, acquire a positive or nega-
tive charge through ionization. After Thomson's discovery of the elec-
tron, the opinion prevailed that every atom contains such particles, that
ionization is due to the atom's catching or shedding one or more of
them, and that the spectral lines characteristic of each element reflect
the periodic motion that is proper to the electrons in its atoms. Two
alternatives were contemplated: Either the extremely light, negatively
charged electrons circulate around a massive, positively charged
nucleus, like planets around the sun, or they are buried, like raisins in
a cake, inside a positively charged sphere. The latter arrangement had
greater hope of being proved stable according to classical electrody-
namics, but it foundered on some stunning experimental results. Geiger
 6.1 Background 311

and Marsden (1909), working under Rutherford in Manchester, bom-

barded thin sheets of metal foil with a-rays, or oc-particles, as they were
already called by then. Most of them went through, but gold foil 6 x
10~5 centimeters thick deflected about 1 particle in 8,000 by an angle
>90°. Years later Rutherford described his reaction as follows: "It was
quite the most incredible event that has ever happened to me in my life.
It was almost as incredible as if you fired a 15-inch shell at a piece of
tissue paper and it came back and hit you" (quoted by Pais 1986, p.
189). He concluded that almost all the gold mass was concentrated in
pointlike positively charged nuclei separated by comparatively enor-
mous distances; this explained both the overall transparency of the foil
to the a-particles and the energetic rebound of the few that hit the mark.
Rutherford therefore embraced the planetary model of the atom.
However, according to classical electrodynamics, an electron moving
circularly, and thus acceleratedly, around the atomic nucleus would con-
tinually emit radiation, thereby losing energy, so the time should soon
arrive when the electron, no longer able to resist the electric attraction
of the nucleus, would plunge into it. Even if this process, as some con-
jectured, was infinitely slow, radiation from a body of continuously
decreasing energy could hardly be responsible for the neatly defined
spectral lines characteristic of each element; nor could one reasonably
hope to explain the stability of such lines from classical premises if, as
all agreed, every atom in an incandescent gas is being struck by others
at the rate of 100 million collisions per second.
This led, in Bohr's words, to "a general acknowledgment of the inad-
equacy of the classical electrodynamics in describing the behavior of
systems of atomic size" (1913, p. 1). It was "necessary to introduce in
the laws in question a quantity foreign to the classical electrodynam-
ics, i.e. Planck's constant, or as often it is called the elementary
quantum of action" (Ibid.). In Bohr's new scheme of things an atom
of a given element can exist in different "stationary states" in which
each electron circulates around the nucleus at a characteristic energy
level. The electrons absorb or emit radiation only during the brief tran-
sition from one stationary state to another. Bohr assumed that:

(i) classical mechanics can be of help in discussing the dynamics of

the stationary states, but it does not apply to the transition from
one such state to another; and
(ii) the transition of an electron from a higher energy level En to a
lower energy level Em is accompanied by the emission of mono-
312 Quantum Mechanics

chromatic radiation whose frequency v(n,m) is related to the

energy difference En - Em by
En-Em=hv(n,m) (6.2)
Equation (6.2) agreed beautifully with the Ritz Combination Principle.
According to this rule, which epitomized the experience of spectro-
scopists, every frequency of the observable spectrum can be expressed
as the difference between two terms, and any difference between two
such terms corresponds to a possibly observable difference. Thus, given
spectral lines of frequency v(n,k) = Tn- Tk and v(fe,m) = Tk- Tm, one
may expect a line of frequency
v(n,m) = v(n,k) + v(k,m) (6.3)
Bohr's scheme secured the atom's stability. Its straightforward appli-
cation to the single-electron hydrogen atom enabled him to calculate
the frequencies of the well-known Balmer series of spectral lines. At
first Bohr ignored the fact, already established by Michelson in 1891,
that most of the lines in the hydrogen spectrum can be resolved into
multiplets, that is, sets of two or more lines with slightly different fre-
quencies. However, Sommerfeld (1915a, 1915b, 1916) was able to
account for them. To do so, he assumed that electrons move, like
planets, on elliptic orbits; he took into consideration the relativistic
variation of inertia with speed, and, most importantly, he postulated
the famous quantum condition reminiscent of Pythagoric arithmology.
By virtue of it, the stationary states of a (multiply) periodic mechani-
cal system with r degrees of freedom satisfy the equations
§pkdqk=nkh (l<k<r) (6.4)
where the qk and pk are the generalized position and momentum coor-
dinates employed for describing the system (§2.5.3), each integral is
taken over a period of the respective qk, and the nk are nonnegative
integers, sometimes referred to as "quantum numbers".5
Still, Bohr's atom theory had great difficulty in dealing with the

5
The quantum condition implies that, among the continuously many mathematically
admissible solutions of the classical equations of motion for a system of bound elec-
trons orbiting under Coulomb's Law around an atomic nucleus, only those that satisfy
eqns. (6.4) are physically possible. Wilson (1915) introduced these equations a few
months before Sommerfeld and unbeknownst to him, but he did not calculate spec-
tral lines from them.
 6.1 B ackgr ound 313

observed effects of magnetic and electric fields on atomic spectra, and

it never yielded accurate values for the spectrum of helium - element
number 2 - not to mention anything more complex. But it did give
a promising overall qualitative understanding of atomic structure,
valence, and the periodic system. So Bohr and his followers persisted
in trying to extend the scope of their quantitative predictions. The
perturbation methods of classical celestial mechanics were skillfully
applied by Sommerfeld and Born to the calculation of electron orbits
in many-electron atoms, while Bohr conducted with admirable tact the
guesswork prompted by his Correspondence Principle. As explained
by him, this is "a law of the quantum theory" (1924, p. 22n), which
asserts "a far-reaching correspondence between the various types of
possible transitions between the stationary states on the one hand and
the various harmonic components of the motion on the other hand"
(1922, pp. 23f.; cf. Bohr 1934, p. 37). Its sole motivation (or justifi-
cation) was to secure that, in contexts in which Planck's constant h is
insignificant, the quantum theory would agree with classical electro-
dynamics, "according to which the nature of the radiation emitted by
an atom is directly related to the harmonic components occurring in
the motion of the system" (Ibid.). But Bohr and his collaborators
wielded the Correspondence Principle, not without success, like a flex-
ible magic wand, also in situations where the finite value of h makes
a difference, as a guide to quantitative predictions and in selecting the
stationary states of an atom that are physically possible from among
the continuously many that are mechanically conceivable.6

6.1.2 Einstein on the Absorption and Emission of Radiation

In 1916 Einstein, who was long busy with his theory of gravity, again
entered the lists of quantum physics. On 11 August he wrote to Besso:
"A splendid idea on the absorption and emission of radiation has
dawned on me". Its fruits were two papers (1916), 1916n) in which
Einstein derives Planck's Law by a statistical argument that does not
depend on the premises from classical mechanics and electrodynamics
that Planck had used. The argument concerns a gas consisting of equal
molecules in statistical equilibrium with thermal radiation. Mindful of
Kirchhoff's theorem on thermal radiation (see note 2), Einstein does

6
For a fairly clear illustration of the latter use of the Correspondence Principle, see
Bohr (1918) in Van der Waerden (1967, pp. 110f.).
314 Quantum Mechanics

not assume anything about the constitution of such molecules, except

that each will go from a state Zm with energy Em to a state Zn with
energy En by emitting or absorbing radiation of a definite frequency
vnm. The argument yields not only a formula for the density of thermal
radiation equivalent to Planck's, but also a proof that the difference
Em - En\ between the energy levels must be proportional to the fre-
quency vwm, the proportionality constant filling precisely the place of h
in Planck's formula. What mainly matters for our present story is
Einstein's innovative approach to energy emission, in case Em > En. He
recalls that an oscillating Planck resonator would radiate energy
whether or not it is excited by an external field. Correspondingly - he
assumes - each of his molecules may go from a state Zm to a state Zn
emitting radiation energy Em — En with frequency vnm without excita-
tion from external causes. "One can hardly think of it otherwise
than on the analogy of radioactivity" (1916j, p. 321); this refers to
Rutherford's findings on radioactive decay mentioned at the beginning
of §6.1.1. Let M be a mass of some radioactive element L. After time
tL - the so-called half-life of L - one-half of the atoms in M will yield
their radioactive output and transmute into another element; tL is char-
acteristic of L and does not depend at all on external circumstances.
Since all atoms of L are equal, one is bound to think of each as a chance
setup with a 0.5 probability of decaying within time tL. Based on this
analogy Einstein equates the probability dW that the transition from
Zm to Zn occurs within time dt with A^dJ, where A^ is a constant char-
acteristic of the index combination (m,w). Turning now to energy
absorption, he naturally assumes that it depends on the surrounding
thermal radiation, which will do work on our molecule in proportion
to the radiation density of the relevant frequency. The work done on
a Planck resonator by the surrounding classical electromagnetic field
can be either positive or negative, depending on their respective phases.
Correspondingly, Einstein introduces "the following quantum-theoret-
ical hypotheses" (1916n, p. 51): Under the action of radiation density
pnm of frequency vnm a molecule can either pass from state Zn to state
Zm by absorbing energy Em - En or pass from Zm to state Zn by releas-
ing energy Em - Em the probability that transition Zn —> Zm occurs in
time dt being B%pmndt and the probability that transition Zm —» Zn
occurs in time dt being B^pmndt (where B™ and Bnm are constants char-
acteristic of the respective index combinations). Now, in thermal equi-
librium, the number of transitions from Zm to Zn (due to negative
absorption or spontaneous emission) must balance the number of tran-
 6.1 B ackgr ound 315

sitions from Zn to Zm (due to absorption). So, denoting by N7 the

number of molecules in state Z;, we have that
(AZ +Bnmpmn)Nm = B:PmnNn (6.5)
Let the probability W; that a molecule finds itself in the state Z; be
given, as in classical statistical physics, by
W;-=frexp(-E,y*T) (6.6)
where p ; is "a constant, characteristic of the molecule's quantum state
Zj and independent of the gas temperature".7 From eqns. (6.5) and
(6.6), Planck's Law and the relation Em - En = hvnm follow - as
Einstein announced to Besso - by "bafflingly simple" algebra.
In the second paper Einstein also investigates the motion of his
molecules under the influence of radiation. The brilliant discussion
leads to the following conclusions:
If a radiation bundle strikes a molecule causing it through an elemen-
tary process to receive or release the quantity of energy hv in the form
of radiation, a momentum hv/c is always transferred to the molecule, in
the direction of propagation of the bundle if energy is received, and in
the opposite direction if energy is released. If the molecule is acted upon
by several directed bundles of radiation, there is always only one of them
taking part in a particular elementary process of irradiation; this bundle
alone determines then the direction of the momentum transferred to the
molecule.
If the molecule undergoes without external excitation a loss in energy
of magnitude hv by emitting this energy in the form of radiation, this
process is also a directed one. There is no outgoing radiation in spheri-
cal waves. In the elementary process of emission the molecule experi-
ences a recoil of magnitude hv/c in a direction which, in the present state
of the theory, is only determined by "chance".
(Einstein 1916n, p. 61)
Einstein expressly regrets that the time and direction of emission are
being left to "chance" ("Zufall" - a word he pointedly surrounds with
scare quotes). Nevertheless, in the light of the above conclusions, the
establishment of a "proper quantum theory of radiation" - in accor-
dance with his earlier conjecture (1905i; see above) - "appears almost

Einstein (1916j, p. 320), where he stresses that eqn. (6.4) can also be obtained "from
thermodynamical considerations" (thus without assuming the validity of classical
mechanics).
316 Quantum Mechanics

inevitable". On the other hand, he admits that he is not a bit closer to

harmonizing this approach with the wave theory of radiation, based
on undeniable facts of radiation interference.

6.1.3 Virtual Oscillators

It was this latter, apparently insurmountable difficulty that moved the
leading scientists to resist Einstein's proposal on light quanta. Planck
himself, when recommending Einstein's incorporation to the Prussian
Academy, noted it as a forgivable blemish in an otherwise extraordi-
nary creative career.8 And Bohr - who continued to spurn a corpuscu-
lar view of radiation even after Compton arranged X-rays to ricochet
from electrons like tiny bullets9 - is said to have replied, as late as 1924,
to critical remarks by Einstein that if Einstein telegraphed him that he
had found an irrevocable proof of the existence of light quanta, "the
telegram could only reach me by radio on account of the waves which
are there".10 So one had to devise some way of linking the discontin-
uous emission and absorption of discrete amounts of radiation energy
by atoms with the supposedly continuous propagation of radiation
through space. In the paper that motivated the said exchange with
Einstein (Bohr, Kramers, and Slater 1924), Bohr appeared ready to
sacrifice the conservation of energy and momentum,11 which would
henceforth hold only in the statistical average but not for every par-

8
"That he may sometimes have missed the target in his speculations as, for example,
in his hypothesis of light quanta, cannot really be held against him. For without taking
a risk no innovation can really result in the most exact natural science" (from a
request to the Prussian Ministry of Education, dated 12 June 1913, handwritten by
Planck and signed by him, W. Nernst, H. Rubens, and E. Warburg; quoted in Seelig
1957, p. 145).
9
Compton showed that X-rays scattered at less than a right angle by a block of paraf-
fin have a smaller frequency than the incident radiation. This result is readily
explained on the hypothesis of light quanta: if a light quantum strikes an electron,
it will transmit to it a quantity of energy A£, so its frequency will decrease by
Av = AE/h.
10
Quoted in Jammer (1966, p. 187), from an interview with Werner Heisenberg on 15
February 1963.
11
Bohr had been toying with this idea for some time. Here is a passage from an early
draft of Bohr (1918) that was not included in the final version: "It would seem that
any theory capable of an explanation of the photoelectric effect as well as the inter-
ference phenomena must involve a departure from the ordinary theorem of conser-
vation of energy as regards the interaction between radiation and matter" (quoted
by Darrigol 1992, p. 214).
 6.1 Background 317

ticular energy-momentum transfer. This seemingly desperate expedient

was combined with the novel idea of "virtual" oscillators creating
"virtual" fields, contributed by Slater.12
We will assume that a given atom in a certain stationary state will com-
municate continually with other atoms through a time-spatial mecha-
nism which is virtually equivalent with the field of radiation which on
the classical theory would originate from the virtual harmonic oscilla-
tors corresponding with the various possible transitions to other sta-
tionary states. Further, we will assume that the occurrence of transition
processes for the given atom itself, as well as for the other atoms with
which it is in mutual communication, is connected with this mechanism
by probability laws which are analogous to those which in Einstein's
theory [described in §6.1.2 - R. T.] hold for the induced transitions
between stationary states when illuminated by radiation. On the one
hand, the transitions which in this theory are designated as spontaneous
are, on our view, considered as induced by the virtual field of radiation
which is connected with the virtual harmonic oscillators conjugated with
the motion of the atom itself. On the other hand, the induced transitions
of Einstein's theory occur in consequence of the virtual radiation in the
surrounding space due to other atoms.

We shall assume an independence of the individual transition processes,

which stands in striking contrast to the classical claim of conservation
of energy and momentum. Thus we assume that an induced transition
in an atom is not directly caused by a transition in a distant atom for
which the energy difference between the initial and the final stationary
state is the same.
(Bohr, Kramers, and Slater 1924, in Van der Waerden 1967, pp.
164-65, 166)
The authors grant that "at present there is unfortunately no experi-
mental evidence at hand which allows to test these ideas," but they
insist that the independence that they ascribe to the transition processes
"would seem the only consistent way of describing the interaction
between radiation and atoms by a theory involving probability con-
siderations" (Ibid., in Van der Waerden 1967, pp. 166-67). This BKS
theory - as it is known in the literature - was forcefully criticized by

12
On Slater's original idea, a new conception of light with "both the waves and the
particles", see Darrigol (1992, pp. 218f.). When Slater went to Copenhagen, Kramers
and Bohr apparently persuaded him to get rid of light quanta by sacrificing
energy-momentum conservation.
318 Quantum Mechanics

Einstein and Pauli and finally wrecked by tests of energy conservation

in individual Compton scattering events performed by Geiger and
Bothe and by Compton himself. Bohr graciously accepted defeat. The
BKS scheme, he wrote to Geiger on 21 April 1925, "was more an
expression of an endeavor to attain the greatest possible applicability
of the classical concepts than a completed theory" (CW, 5, 353). There-
after, Bohr rejected all "the space-time pictures previously used in the
quantum theory: electronic orbits in stationary states, trajectories in
collision processes, radiation fields, and corpuscular light-quanta"
(Darrigol 1992, p. 252).
Nevertheless, Slater's virtual oscillators still played a role in the
theory of dispersion put forward by Kramers in two communications
to Nature (1924a, 1924b) and a longer article coauthored by Heisen-
berg (1925). According to Kramers, such oscillators were introduced
not as an "additional hypothetical mechanism" but "only as a termi-
nology suitable to characterise certain main features of the connexion
between the description of optical phenomena and the theoretical inter-
pretation of spectra" (Kramers 1924b, p. 311; my italics) - in other
words, they were just a manner of speaking. Kramers and Heisenberg's
work on dispersion was only partially successful. The "translation" of
classical into quantum terms and relations pursuant to Bohr's Corre-
spondence Principle attained here new levels of virtuosity. Finite differ-
ence quotients systematically replace the derivatives in the classical
formulas. A similar "transition from differential to difference equa-
tions" was advocated by Born in "Uber Quantenmechanik" (1924), the
first paper to use this name in print.13 Heisenberg (1925) carried such
translation methods one long step forward, with amazing success.

6.1.4 On Spin, Statistics, and the Exclusion Principle

I shall finish this sketch of the immediate background of QM with a
short note on three novelties that made their appearance more or less

13
Born described this work as "a first step towards a quantum theory of coupling"
(1924, p. 379). He endorsed Kramers's handling of dispersion caused by the inter-
action between atomic electrons and radiation and sought to extend it to the inter-
action between the several electrons in a single atom (heavier than hydrogen).
Born's method of substituting finite differences for classical differentials is concisely
explained by Tian Yu Cao (1997, p. 134) in a manner that clearly brings out its affin-
ity with Heisenberg's (1925).
 6.1 Background 319

at the same time as the new theory, namely, Pauli's Exclusion

Principle, the "spin" of the electron, and the new approach to physi-
cal statistics initiated by Bose. None of them played a role in the foun-
dation of QM, but they enabled it to cope with experimental data that
would otherwise have eluded it, so they were received within it as
welcome additions. Later they found in the context of relativistic
quantum theory a rational justification that bound them together.
Pauli lighted on the Exclusion Principle in his endeavor to under-
stand the so-called anomalous Zeeman effect. Spectral lines are split
when the source of light is placed in a magnetic field (Zeeman 1897).
A threefold split could be readily accounted by classical electrody-
namics (Lorentz 1897) and also by Bohr's atom theory. But experi-
ments with heavier elements and better instruments disclosed further
splits, which were dubbed "anomalous" because they did not fit the
available explanations. When Pauli tackled the problem in the 1920s
the developing Old Quantum Theory characterized the stationary state
of a bound electron by three quantum numbers, which were thought
to reflect, respectively, the size of the electron's orbit, its shape, and its
inclination with respect to the external magnetic field. Pauli (1925a, p.
385) concluded that the doublet structure of alkali spectra is due to "a
peculiar kind of duplicity (Zweideutigkeit) in the quantum-theoretic
properties of the optical electron, which cannot be described in classi-
cal terms". Shortly thereafter, in a second, brilliant paper, Pauli added
a fourth, two-valued quantum number to the former three and postu-
lated his Exclusion Principle:
There can never be two or more equivalent electrons in an atom for
which the values of all [four] quantum numbers coincide in strong fields.
If there is an electron in an atom for which these quantum numbers take
certain values (in the external field), this state is "occupied".
(Pauli 1925b, p. 776)
With this powerful and very original assumption Pauli was able to
explain both the "anomalous" Zeeman multiplets and the number of
electrons allowed in each group or "shell" surrounding the atomic
nucleus.
In his two papers Pauli carefully avoided any suggestion that might
help to visualize the physical meaning of the quantum numbers by asso-
ciating them with the area, excentricity, tilt, etc., of an electron orbit.
He was probably convinced that - as he put it in the survey of quantum
theory that he wrote in the same year, 1925 - "one must renounce the
320 Quantum Mechanics

practice of attributing to electrons in the stationary states trajectories

that are uniquely defined in the sense of ordinary kinematics" (1926a,
p. 167). This may have motivated his initial rejection of the solution
of the Zeeman anomalies put forward by Uhlenbeck and Goudschmidt
(1925), which bestowed a definite - classical - physical meaning on
Pauli's fourth quantum number. From a classical standpoint, the atom's
behavior in a magnetic field depends on its angular momentum. By eqn.
(6.4), the angular momentum due to the electron's orbital motion must
be an integral multiple of h/2n (henceforth abbreviated ft). Uhlenbeck
and Goudschmidt assumed that the electron also has an intrinsic
angular momentum or spin, which can be conceived classically as due
to the electron's rotation about a stable axis and takes the value +till
or -fill (depending on the sense of rotation). By taking spin into
account they obtained the number of distinct atom states required to
match the observed Zeeman multiplets. The spin hypothesis accounts
precisely for the peculiar duplicity or twofoldness in the quantum-
theoretic properties of the electron that Pauli had noted (1925a, p. 385;
quoted above). However, the conception of the electron as a finite
rotating sphere is fraught with difficulties, for not only is it hard to
understand what keeps its charge together, but its equatorial velocity
would have to exceed the speed of light to yield spin = ±fill. After the
image of the stationary atom as a classical mechanical system was scut-
tled by Heisenberg (§6.2.1) it was no longer necessary or even possi-
ble to understand the spin of the electron as a manifestation of rotation.
Just as Pauli anticipated, spin is now conceived as an irreducible
quantum property of matter, but the name 'spin' has stuck.
Finally, I turn to the matter of quantum statistics. Bose (1924)
showed that Planck's Black-Body Radiation Law could be established
by a purely statistical argument, without appealing to the classical elec-
trodynamic assumptions invoked by Planck (1900), if one regarded
photons of a given frequency as utterly undistinguishable for statisti-
cal purposes. This has the effect of reducing the number of distinct
states available to a photon gas. A simple, abstract example will show
why this is so. Consider a system consisting of two objects, each of
which can be in one of three possible states. If each object has an indi-
viduality of its own, we may denote them by proper names, say a and
b. The system can be in one of nine different states, viz., [tf]-[b]-[ ], [b]-
[<*]-[ ], i Ha]-[b], [ ]-[b]-[a], [a]-[ ]-[b], [b]-[ ]-[a], [ab}-[ ]-[ ], [ ]-[ab]-
[ ], and [ ]-[ ]-[<zfc], where [a]-[b]-[ ] is the systemic state in which a is
in the first object-state and b is in the second, and so on. (This approach
 6.2 The Constitution of Quantum Mechanics 321

is characteristic of the classical, Maxwell-Boltzmann statistics exam-

ined in §4.3.) But if the objects are indistinguishable and are denoted,
say, by o and o, there are no more than six different states in which
the system can find itself, viz., [o]-[o]-[ ], [ ]-[o]-[o], [o]-[ ]-[o], [oo]-
[ ]-[ L [ ]-[°°]-[ L and [ ]-[ ]-[oo]. Bose's approach was generalized and
fruitfully applied to several pending problems by Einstein (1924,
1925a, 1925b) and is commonly known as Bose-Einstein statistics.
Note, however, that the last three states mentioned are not possible if
the objects obey Pauli's Exclusion Principle. In this case, the available
systemic states reduce to three, viz., [o]-[o]-[ ], [ ]-[o]-[o], and [o]-[ ]-
[o]. This statistical approach was introduced by Dirac (1926a) and
Fermi (1926) and is known as Fermi-Dirac statistics. The physical par-
ticles are currently classified into fermions (such as the electron and the
proton), which obey the Exclusion Principle, and bosons (such as the
photon), which do not. Fermi-Dirac statistics apply to the former, and
Bose-Einstein statistics to the latter. In QM one postulates that the spin
of a boson is an integral multiple and the spin of a fermion a half-
integral multiple of h. In relativistic quantum theory this connection
between spin, statistics, and the Exclusion Principle is proved as a
theorem (Pauli 1940).

6.2 The Constitution of Quantum Mechanics

6.2.1 Matrix Mechanics

According to classical electrodynamics, the frequency of radiation
emitted or absorbed by an atom is directly related to the frequency
with which its electrons vibrate or circulate. In Bohr's theory, this con-
nection is suppressed. The measurement of spectral lines can now yield
no information at all about the periodic motion of electrons. Bohr and
his followers continued to speculate about electron orbits to calculate
energy levels. By 1925, however, it was clear that this activity was yield-
ing quickly diminishing returns. Born and Jordan sternly warned that
the "quantities that enter the true laws of nature" must all be "observ-
able and ascertainable in principle" (1925a, p. 493), while Kramers
and Heisenberg boasted that the formulas of their dispersion theory
"contain only the frequencies and amplitudes which are characteristic
for the transitions [between stationary states of the atom], while all
those symbols which refer to the mathematical theory of periodic
systems will have disappeared" (1925, in Van der Waerden p. 234). In
322 Quantum Mechanics

line with this trend, Heisenberg resolved "to give up altogether the
hope of observing the hitherto unobservable quantities (such as the
electron's position and period)" and "to try to develop a quantum-
theoretical mechanics, analogous to classical mechanics, in which only
relations between observable quantities occur" (1925, p. 880).14
He secured the analogy between the new and the old mechanics by
what he called a "reinterpretation of kinematic and mechanical rela-
tions". The classical equations of motion were preserved, but the letters
that occur in them were made to stand for new-fangled mathematical
objects, each of them representing not a single real number but an infi-
nite array of complex numbers. A reinterpretation of the mathemati-
cal operations at play was therefore inevitable: One continued to talk
of multiplication or differentiation and to use the familiar symbols,
but they now meant something quite different. To say that kinematic
and dynamic relations were merely being "reinterpreted" is perhaps an
understatement. At any rate, the equations of motion did not apply
now to something one could call "motion" in a standard sense. The
name "mechanics" could be retained, insofar as the new theory was
intended, like Newton's, as a basic framework for describing and
explaining all forms of physical change; but physical change was no
longer to be explained by the displacement of point-masses in three-
dimensional space. A failure to accept this, or perhaps even to realize
it, may account for some of the difficulty in understanding QM.
One could not say that Heisenberg had a rational justification for
proceeding as he did, but here are some hints as to the intellectual
motives that guided him. Any periodic function corresponding to a
time-dependent physical quantity can be represented by a Fourier
series. In the Old Quantum Theory one considered time-dependent

14
Heisenberg's stance on this question should be taken with a pinch of salt. On the one
hand, his "observables" include the very amplitudes that Born and Jordan (1925a)
left aside as unobservable; on the other, although Heisenberg discarded electron posi-
tions as unobservable, in mature QM the three Cartesian coordinates of a particle
are considered observables. Many years later, Heisenberg (1969, p. 95) approvingly
recalled some remarks on observability that he heard from Einstein in 1926: Obser-
vation presupposes a known, unambiguous relation between the phenomenon to be
observed and our sense perceptions. We can be certain of this relation only if we
know the laws that determine it. If the laws of nature are in doubt - as they obvi-
ously were for Heisenberg in 1925 - the concept of observation has no clear meaning.
The theory must then by itself determine what is observable. (Indeed, it was Bohr's
atom theory that made electron orbits unobservable; cf. Bohr 1934, pp. 12, 36.)
 6.2 The Constitution of Quantum Mechanics 323

quantities pertaining to each stationary state of an atom. Thus, a single

electron's displacement xn(t) in the nth state could be expressed as:

*«(*)= ^xnkexp(2nivnkt) (6.7)15

Now, "in quantum theory it has not been possible to associate the elec-
tron with a point of space, considered as a function of time, by means
of observable quantities; however, even in quantum theory it is possi-
ble to ascribe to an electron the emission of radiation" (Heisenberg
1925, p. 881). Radiation is emitted (or absorbed) in the transition
between two stationary states and is related by eqn. (6.2) to the dif-
ference between the respective energy levels; thus,
v(n,n-k) = ^(En-En_k) (6.8)
Bohr and his school took the frequencies v(n,n - k) as the quantum-
theoretic analogue of the classical frequencies vn. Heisenberg intro-
duces quantum-theoretic analogues x(n,n - k) for the classical
displacement components xnk. So each Fourier term xnkcxp(2nivnkt) in
eqn. (6.7) is to be replaced in the quantum mechanics by x(n,n - k)
exp(2niv(n,n - k)t). Due to the symmetric role of the indices n and (n
- k), these new terms cannot be meaningfully gathered into an infinite
series such as (6.7). So, Heisenberg says, one ought to regard the whole
array of complex amplitudes x(n,n - k) exp(2niv(n,n - k)t) - with
« = 1, 2 , . . ., and with k ranging over all the integers - as the quantum-
theoretic analogue of the classical quantity x(t). This is the backbone
of his new "kinematics". He then asks: "What will represent the quan-
tity [x(t)]1}" (Ibid., p. 882). The answer to this question will yield the
quantum-theoretic analogue of [x(t)]r for every integer r and hence of
any function f[x(t)] that can be expanded in a power series. Now, clas-
sically, one got the square of the left-hand side of eqn. (6.7) by multi-
plying term by term the Fourier series on the right-hand side by itself.
This can be written as

[xn(t)f=f,cnkexp(2nivnkt) (6.9)
where

cnkexp(2nivnkt) = ^xnjxn{H)exp(2nivn(j+ k-j)t) (6.10)

;=-oo

15
Remember that exp(z'a) = cos a + / sin a.
324 Quantum Mechanics

According to Heisenberg, in quantum theory, instead of eqn. (6.10),

one must write:
c(n,n- k)exp(2niv(n,n - k)t)
~ (6.11)16
= 2^ x(n, n - j)x(n -j9n- k)exp(2niv(n, n - k)t)
y=-oo

Indeed, according to him, the Ritz Combination Principle leads to "this

form of composition almost compulsorily" (Ibid., p. 883). Generally
speaking, the product xy of two quantities x and y, represented in
quantum theory by arrays with typical elements x(n,k) and y(m,j),
respectively, should be represented by an array z whose typical
element is
) (6.12)17

Equation (6.12) evidently implies that xy ^ yx, except in special cases.

In contrast with ordinary multiplication, the product of two quantum
quantities is not commutative. Heisenberg initially had some difficulty
in accepting this. Equation (6.12) states the rule of matrix multiplica-
tion, which is nowadays familiar to countless professionals who use
linear algebra in their work (Supplement 1.6). But in 1925 Heisenberg
had no idea what a matrix was. It was Born who, after pondering
Heisenberg's manuscript for a whole day and night, remembered a
course that he had taken in his youth with the mathematician Rosanes
and recognized Heisenberg's complex number arrays as matrices.
Heisenberg tried his translation rules on a simple mechanical
problem: the anharmonic oscillator. From the matrix analogue of the
classical equation
x + a2ox + Xx2=O (6.13)
he obtained good values for the energy levels. Papers by Born and
Jordan (1925b) and by Born, Heisenberg, and Jordan (1926 - the
"three-men paper") developed Heisenberg's scheme into a system of
matrix mechanics. The authors gave plausible rules for symbolic matrix
differentiation - a different one in each paper - and adopted the fol-
lowing precise measure of the noncommutatitivity of the matrices q

16
By eqn. (6.8), v(n,n - j) + v(« - j,n - k) = v(n,n - k). So
exp(2niv(n,n - j)t)exp(2niv(n - j,n - k)t) = exp(2niv(n,n - k)t)
17
That eqn. (6.11) is a special case of eqn. (6.12) can be seen by substituting / for
n - j and m (or n - k.
 6.2 The Constitution of Quantum Mechanics 325

and p, which replace the classical position and momentum coordinates

qk and pk:
q p - p q = /7rt (6.14)
(1 denotes the unit matrix, with diagonal elements equal to 1 and all
other elements equal to 0). Equation (6.14) was discovered in rapt med-
itation by Born, who dubbed it the "sharpened quantum condition",
because it replaces in the new theory the Sommerfeld quantum condi-
tion (6.4). On the analogy of the Hamilton equations (2.41), the
quantum-mechanical equations of motion take the following canoni-
cal form (where the dot signifies matrix differentiation with respect to
time and H is the matrix representing the classical Hamiltonian func-
tion (2.38)):

The new mechanics won general approval when Pauli (1926b) cal-
culated from it the hydrogen spectrum in agreement with experience
both in conditions already covered by the Old Quantum Theory and
in the presence of crossed electric and magnetic fields, in which the Old
Quantum Theory failed.

6.2.2 Wave Mechanics

In the same year Schrodinger published the four installments of a long
paper, "Quantisation as a problem of proper values" (1926a, 1926b,
1926d, 1926e),18 in which he tackled the unsolved problems of atomic
physics in a way that seemed completely different from and almost
opposite to that of matrix mechanics. Schrodinger drew his inspiration
from the bold ideas put forward by Louis de Broglie in two commu-
nications to the Paris Academy (1923a, 1923b) and in his doctoral
thesis (1925). Rather than seek, like Bohr, to reconcile by means of an

18
'Proper values' of linear operators are defined in Supplement 1.6. The following
example illustrates the use of the term in the present context. Let D stand for the dif-
ferential operator dl&x and let f be a unknown function of x. Consider the equation

D2f-Xf = 0 (*)
It can be readily verified that for each positive integer n the function fn = exp(nx)
satisfies eqn. (*), with X = n2. The functions fu f2, f^ .. . are proper functions of the
operator D2, and I 2 , 22, 3 2 ,. .. are the corresponding proper values.
326 Quantum Mechanics

opportunistic hypothesis the new evidence for light quanta with the
well-established fact of radiation interference, de Broglie extended the
difficulty to ordinary massive matter, thus pressing the need for a
radical solution. He recalled that Hamilton had been led to his equa-
tions of motion (eqns. (2.41)) by a formal analogy between geometri-
cal optics and analytical mechanics, and proposed that the new
mechanics, required by atomic phenomena, should stand to the classi-
cal mechanics of corpuscular trajectories in a relation similar to that
of the wave theory of light to the geometrical optics of light rays.
Special Relativity (SR) established that 'matter' and 'energy' are syn-
onymous expressions that denote the same physical reality. Since
quantum physics associates to every isolated portion of energy E a
certain frequency v = E/h, it is plausible to think that every material
particle of rest mass m0 is associated with a periodic phenomenon of
frequency v0 = m0c2/h. De Broglie proved from SR that an observer
relative to whom a particle moves uniformly with velocity v must
assign to the said wave a "phase velocity" (velocity of propagation) V
= cNl-(v/c)2 > c. On the other hand, if two or more such waves are
superposed, the said observer will see the maximum beat formed by
them move with a "group velocity" that is precisely the same as the
classical velocity of the particle. De Broglie therefore suggested that so-
called material particles are simply the traveling beats formed by very
narrow packets of superposed waves. Since energy and momentum
stand to one another in SR as the timelike and the spacelike parts of
a single spatio-temporal reality, the same relation that holds between
the energy E and the frequency v - which is the number of wave cycles
per unit of time - must hold between the momentum p and the recip-
rocal value of the wavelength X - which is the number of cycles per
unit of length -, viz., p = hlX. De Broglie's examiners applauded his
brilliant mathematics but met his physical suggestions with scepticism.
Einstein (1925a, p. 9) was more receptive and contributed to their dif-
fusion. Two years later, Davisson and Germer published their discov-
ery of electron diffraction, that is, the interference of electron beams.19

19
Born says that the discovery by Stern and associates, in 1932, that molecular beams
of hydrogen and helium also show diffraction phenomena when reflected by crystals
was particularly impressive. "De Broglie's equation [p = h/X] was confirmed for these
particles with an accuracy of about 1 per cent. Here, surely, we are dealing with mate-
rial particles, which must be regarded as the elementary constituents not only of gases
but also of liquids and solids" (1962, p. 96).
 6.2 The Constitution of Quantum Mechanics 327

Schrodinger seized on de Broglie's idea before its spectacular exper-

imental confirmation. It furnished him a welcome means of reconcil-
ing the continuity of nature with the presence of integers in quantum
physics, instead of linking these to "jumps" or some other unpalatable
discontinuity. It is a well-known fact that a vibrating string with both
ends fixed displays a definite number of motionless points or nodes.
Likewise, if bound electrons are conceived as standing waves around
the atomic nucleus, the differential equation governing the wave
motion could yield the integers characteristic of their state (cf. eqns.
(6.4)).20 So Schrodinger set out to find the equation of the matter
waves. He met great difficulties in setting up a Lorentz invariant equa-
tion, so, despite, the fact that de Broglie's arguments were firmly rooted
in SR, Schrodinger decided to work provisionally on a wave equation
valid in the nonrelativistic approximation. To find it he resorted to
familiar classical methods. It is not necessary to explain them here. I
give, however, a superficial description of Schrodinger's procedure to
show the strong roots of QM in analytical mechanics. This may help
dispel some common misrepresentations concerning revolutions in
physics.
Schrodinger replaces the action S in the classical Hamilton-Jacobi
equation (2.43) with K log\|/, where i|/ is an unknown function and K
is a constant with the dimensions of action. If we ignore the relativis-
tic change of mass with velocity, the equation thus obtained can then
be reformulated as Q(\|/) = 0, where Q(\|/) is a quadratic form of \|/ and
its first derivatives. Schrodinger assumes that the \j/-functions of inter-
est to physics are twice continuously differentiable finite functions
defined on the entire configuration space of the mechanical system
under study.21 In the tradition of classical variational principles (see
remark (iii) after eqns. (2.33)), Schrodinger postulated the following
requirement: The integral of Q(\|/) over the entire configuration space
is stationary (takes either a minimum or a maximum value). The \|/-

20
"It is hardly necessary to emphasize how much more congenial it would be to imagine
that at a quantum transition the energy changes over from one form of vibration to
another, than to think of a jumping electron. The changing of the vibration form can
take place continuously in space and time, and it can readily last as long as the emis-
sion process lasts empirically" (Schrodinger 1926a, in WM, pp. 10-11).
21
Schrodinger (1926a) also required that the y-functions be real-valued (WM, p. 2);
this requirement was omitted in 1926b (WM, p. 28) and explicitly left aside in 1926e
(WM, p. 104), where the Schrodinger equation is said to allow complex-valued
solutions.
328 Quantum Mechanics

functions that satisfy this requirement are the solutions of a certain

differential equation, the time-independent form of the famous
Schrodinger equation that holds center stage in QM.22 Its solutions rep-
resent waves in configuration space, with |\(/(q)|2 the wave intensity at
point q. The mysterious integers of quantum theory make their appear-
ance in a natural way in the equation's proper values. So Schrodinger
could justifiably regard his variational approach as a substitute for -
and a welcome improvement on - the arbitrarily postulated "quantum
condition" (6.4).
Schrodinger (1926a) applied the procedure sketched above to the
motion of a single electron in the Coulomb field of a much heavier
nucleus with the opposite charge. He showed that \|/-functions satisfy-
ing the variational postulate exist for every conceivable positive energy
value but only for a discrete set of negative energy values.23 If we put

22
When one speaks nowadays of the Schrodinger equation, without further qualifica-
tion, one usually refers to the time-dependent equation first introduced in Schrodinger
(1926b, eqn. 18). This can be schematically written:

(ESI)
at
where \\f(t) denotes the state of the system at time t and H stands for the Schrodinger
Hamiltonian operator, obtained from the classical Hamiltonian H(qhpk) of the
mechanical problem under study by substituting -ihd/dqk for pk. The time-indepen-
dent equation of Schrodinger (1926a) can be summarily written:
Hy = wy (ES2)
where \j/ denotes a stationary state and w is a real number interpreted as a proper
value of energy. Nowadays one views eqn. (ES2) as a special case of eqn. (ESI) - thus
Mainzer (1995, p. 430). However, Schrodinger originally derived eqn. (ESI) from
eqn. (ES2) as is explained in note 23.
They are the solutions of the time-independent Schrodinger equation

^(EV)\|/ 0 (ES2*)

with the potential V = -e2/r. (As Falkenburg 1995, p. 154 n. 20, pointedly remarks:
"Quantum Mechanics is on the whole somewhat of a hybrid theory, which describes
a quantum object in a classical potential.") Here is how Schrodinger (1928, §6)
derived from eqn. (ES2*) the corresponding form of his time-dependent equation (cf.
note 22). The time dependence of the wave function is expressed by i|/ ~ exp(/co£),
where co = 27W = Elfi. Therefore, d\\f/dt = (E/h)i\\f and E\\f = (h/i)d\\f/dt. Substituting for
Ex\f in (ES2*) and putting fi for K (as indicated in the main text), one obtains
 6.2 The Constitution of Quantum Mechanics 329

the constant K = h, the said negative values agree numerically with the
Bohr energy levels corresponding to the Balmer terms of the hydrogen
spectrum. The continuous spectrum of positive energy levels suits the
hyperbolic orbits of unbound electrons that pass by the nucleus, from
infinity to infinity, like comets go past the sun. In the next installments
of his paper (1926b, 1926d) Schrodinger solved other simple examples
and developed a perturbation theory, which he applied to the effect of
an electric field on the Balmer lines (Stark effect). All results agreed
admirably with experimental data, in the limit in which SR effects are
negligible; they also tallied with the predictions of matrix mechanics,
even where these differed from those of the Old Quantum Theory. This
was very surprising, for, as Schrodinger noted, due to "the extraordi-
nary difference between the starting-points and between the methods"
of Heisenberg's program and his own, one might expect them "to sup-
plement one another" (1926b, in WM, p. 30) but not to coincide.

6.2.3 The Equivalence of Matrix and Wave Mechanics

The mysterious coincidence was explained by Schrodinger himself in
a paper, "On the Relation Between the Heisenberg-Born-Jordan
Quantum Mechanics and Mine" (1926c), which he published after the
second part of his longer work and before the third. He argued in it
that the theories are built around isomorphic mathematical structures,
so that "from the formal mathematical standpoint one might well
speak of the identity of the two theories" (WM, p. 46). The matter was
later explained with much greater clarity and precision by the mathe-
matician John von Neumann (1932), whose exposition I summarize.24
Faced with a quantum-mechanical problem, matrix mechanics first

24
According to Muller (1997) the equivalence wasfirstproved by von Neumann (1932),
who established it not for wave mechanics and matrix mechanics as they existed in
1926 - which were not equivalent - but for theories developed from them in the next
year or two, in part under the influence of Schrodinger's purported proof. Strictly
speaking, Muller is right, but perhaps one ought not to speak too strictly about his-
torical creatures such as physical theories. While a theory is still growing in the minds
of its authors one cannot simply equate it with what has appeared in print. It is
nonetheless remarkable that the community of physicists, including the matrix
mechanicists in Gottingen, accepted Schrodinger's proof of equivalence, and oblig-
ingly strengthened both theories until the conclusion followed. Note also that von
Neumann (1927) already presents, quite rigorously, matrix and wave mechanics as a
single theory.
330 Quantum Mechanics

solves the canonical equations (6.15) for the Hamiltonian matrix H.

The q and p are, as we know, infinite matrices of complex numbers,
which satisfy the commutation rule (6.14). H is then transformed to a
diagonal matrix, whose diagonal elements are the admissible energy
levels of the quantum system under study. To achieve this one must find
a matrix S such that the matrix W = S^HS is diagonal, with real-valued
diagonal elements W/;. Thus, one seeks an invertible matrix S such that
HS = SW, or, in other words, such that, for each index pair (/,/),

JTHaS^S^ (6.16)
k=l

Therefore, each column Slp S 2 ; ,... of S is a solution of the following

problem: To find the sequences x = (xi,x2,. . .) such that Hx is a mul-
tiple of x, say wx, with w a real number. Every such sequence x is said
to be a proper sequence of H, the factor w being the corresponding
proper value. Equations (6.16) constitute a complete solution of this
problem in the following sense: Every proper value of H occurs in the
sequence Wlu W22,. . .; and if w is a proper value of H, the corre-
sponding proper sequences are the linear combinations of the columns
Slp S2p . . . such that W/; = w.
Wave mechanics explicitly seeks the proper values of a differential
operator. The Schrodinger equation can be summarily written as
H\|/ = MA|/ (6.17)
where w is a real number and H is Schrodinger's Hamiltonian opera-
tor (see note 22). Each solution \|/ is a proper function of H and w is
the corresponding proper value (which may or may not be different for
different proper functions). Again, the w's are the energy levels of the
system.
The energy levels obtained from eqns. (6.16) for a particular
quantum-mechanical problem coincide with the energy levels given for
the same problem by eqn. (6.17). This alone cannot justify the con-
fusing use of the same term 'proper value' in two apparently so diverse
mathematical contexts. The fact is, however, that from a mathemati-
cian's standpoint the two contexts are intimately related. The sequences
of complex numbers x = (xi,x 2 ,. . .) that we met above can be regarded
as vectors in a vector space © characterized as follows:25

25
General information about vector spaces is provided in Supplement I. The Dirac
symbol (x\y) for the inner product of vectors x and y is explained in 1.5.
 6.2 The Constitution of Quantum Mechanics 331

(@1) Vector addition is term-by-term addition: x + y = (xx + y^x2 +

yi,...).
(©2) Scalar multiplication is term-by-term multiplication by a
complex number: ax = (ax^ax2^. . .).
(©3) There is an inner product defined by (x|y) = Zk = iXk*yh where x^
denotes the complex conjugate of xk.
(©4) For every sequence x in ©, the series X^=1|x^|2 converges to a
finite value.
The Hamiltonian matrix H is plainly the matrix of a linear mapping
© —> ©. On the other hand, the arbitrary complex-valued functions \|/
among which the solutions of eqn. (6.17) are to be found, can be
regarded as vectors in a vector space $a characterized as follows:
Each vector in $Q is a twice continuously differentiable finite
complex-valued function on the configuration space Q of a spe-
cific mechanical system.
Vector addition is pointwise addition: For each q in Q, (\|/! +

Scalar multiplication is multiplication by a complex number: For

each a in C and q in Q, (a\\f)(q) = a\\f{q).
There is an inner product defined by (\\f\§) = JQ\|/*(q)(p(q)dq, where
\|/* denotes the complex conjugate of \|/.26
Every function \|/ in $Q is square integrable: The integral
L q is well defined.
The vector spaces © and $ Q are isomorphic: There are one-one map-
pings of each onto the other that preserve vector addition, scalar mul-
tiplication, and the inner product. This was proved first in 1906 by
Hilbert for © and a certain part of ^Q, and shortly thereafter by Riesz
and Fischer for the whole of ^^. The abstract mathematical structure
that is common to these vector spaces is called Hilbert space. Since
both matrix and wave mechanics have isomorphic Hilbert spaces at
their core, it is no wonder that both theories make precisely the same
quantitative predictions.

6.2.4 Interpretation
One often hears that matrix and wave mechanics, although formally
equivalent, differed profoundly in physical contents. Two theories can

26
In other words, for every q in Q, if \|/(q) = a + bi, i|/*(q) = a - hi.
332 Quantum Mechanics

certainly share the same mathematical structure and yet have each an
empirical meaning of its own. However, if the structures underlying
two theories are isomorphic - and so, formally indistinguishable - any
physical interpretation bestowed on one of them is, via the isomor-
phism, applicable to the other. Indeed, in the case in point, the surmise
that both theories were closely related was prompted by the fact that
they assigned the same values to certain physical quantities. This could
happen only if some identical numbers obtained by calculation from
either theory were being interpreted in the same way, for example, as
the frequencies and intensities of some definite spectral lines. Of course,
such shared meanings involve only the points of contact of matrix and
wave mechanics with laboratory data, and the philosophers who see a
big semantic gap between both theories must have something deeper
in mind. As we know, Heisenberg scorned the pretension of reaching
with his theory anything beyond what is actually observable, while
Schrodinger was deliberately trying to conceive and describe continu-
ous (wavelike) processes responsible for the quantum integers and the
appearance of "jumps". To this extent, the philosophers are clearly
right.
However, Schrodinger's early attempts at a physical interpretation
of his \|/-function ended in failure.27 He had expected that the solutions
of his equation would describe the undulatory processes envisioned by
de Broglie, which, superposed in narrow stable packets, would be the
carriers of particle-like phenomena. But he had to admit that, except
in the simplest cases, all wave packets spread out and are therefore
incapable of simulating particles.
Moreover, the \|/-function is defined on the configuration space of
the system under study, which possesses as many dimensions as the
latter has degrees of freedom. This in itself is neither new nor baffling:
In §2.5.3 we saw how Lagrange represented the evolution of a mechan-
ical system by the trajectory of a point in the system's configuration
space. However, if the system consisted of n particles one could always
- at least in principle - calculate their n trajectories in Newtonian
3-space from the trajectory of its representative in 3^-space. In wave
mechanics things are quite different. If the system consists of two or
more interacting subsystems, they are usually so entangled that it is
impossible to assign a \|/-function to each. Even if the \|/-function of the

27
Another, more promising interpretation advanced in his Dublin seminars of the 1950s
never attained a final, sufficiently definite form. See Schrodinger (1995).
 6.2 The Constitution of Quantum Mechanics 333

system is initially constructed from \|/-functions describing the sub-

systems right before the interaction, its subsequent evolution under
Schrodinger's equation precludes its analysis into distinct subsystemic
\|/-functions. Thus, the wavelike process constituted by \|/ in configura-
tion space cannot normally be translated into an equivalent, equally
continuous process in ordinary space.
Nor could Schrodinger successfully carry through the "heuristic
hypothesis on the electrodynamical significance" of \|/ (1926e, in CW,
p. 108) that he subsequently proposed. This consisted in the general-
ization to any number of particles of an interpretation he found to
work for a single electron, viz., that the squared amplitude |\|/|2 of the
complex wave function \|/ represents the electrical density as a function
of the space coordinates and the time.
In July 1926, Schrodinger met Bohr and Heisenberg in Copenhagen.
Bohr argued forcefully to persuade him to give up every hope of
understanding quantum physics in classical terms. At one point
Schrodinger shouted in despair: "If all this damned quantum jumping
were really here to stay then I should be sorry I ever got involved with
quantum theory" (quoted from Heisenberg's recollections in Pais 1991,
p. 299).
The standard interpretation of the \|/-function was proposed by
Born (1926a, 1926b), in full agreement with the spirit of Heisenberg's
approach. According to it, the squared amplitude |\|/p measures
the probability that certain observable quantities might take certain
values at the location represented by the \|/-function's argument. The
individual process, the "quantum jump", is a chance event whose
antecedent probability is determined by Schrodinger's equation. "The
jump therefore spans a considerable abyss; what happens during the
jump [. . .] perhaps cannot be described at all in a language which
suggests pictures to our visualizing faculty" (Born 1927a, p. 172).
In this way, the conception of atoms and their component particles
as chance setups, which was foreshadowed in Rutherford's law of
radioactive decay and was openly applied by Einstein to the absorp-
tion and emission of radiative energy (§6.1.2), now became universal.
In Born's revolutionary view, the paths of particles "are determined
only insofar as they are constrained by the principle of energy and
momentum conservation; apart from this, the value distribution
of the \|/-function determines only the probability that a particle will
follow a particular path. [. . .] The motion of particles obeys the laws
of probability, but probability itself spreads in accordance with the
334 Quantum Mechanics

principle of determinism [embodied in Schrodinger's differential equa-

tion - R. T.]."28
Born did not impose the said meaning on \\f by an arbitrary decision
- as one is supposed to do when interpreting an uninterpreted formal
system - but read it off a. particular application of wave mechanics. He
noted that the new quantum mechanics, in both the matrix and the
wave versions, had only been applied to stationary states, and that
many assumed that additional conceptual developments would be nec-
essary to deal with transitions. He himself, "impressed with the closed
character of the logical structure of quantum mechanics, conjectured
that the theory is complete and must encompass the problem of tran-
sitions" (1926a, p. 863). He was now ready to prove it from the appli-
cation of wave mechanics to collisions.29
Born considers an electron coming from infinity that strikes an atom
and rebounds toward infinity. Clearly, both before and after the colli-
sion, when the electron is far enough and the coupling is small, a def-
inite state must be specifiable for the atom and a definite rectilinear
motion for the electron. Now,
according to Schrodinger, the atom in its n-t\\ quantum state is a vibra-
tion of a state function of fixed frequency [. . .] spread over all of space.
An electron moving in straight line is [. . .] a vibratory phenomenon cor-
responding to a plane wave. When two such waves interact, a compli-
cated vibration arises. However, one sees at once that one can determine
it through its asymptotic behavior at infinity.
(Born 1926a, p. 864)
One must therefore solve the Schrodinger equation for the atom-plus-
electron system subject to this boundary condition: In the particular
direction of electron space from which the electron is assumed to come
the solution goes over asymptotically into a plane wave propagating
from exactly this direction. In the solution thus selected "we are inter-
ested mainly in the behavior of the 'scattered' wave at infinity, for it
describes the behavior of the system after the collision" (Ibid.).
As usual in physics, the mathematics is introduced not as a mean-

28
Born (1926b, p. 804). I translate Born's term "das Kausalgesetz" as "the principle of
determinism". See Chapter Three, note 33, and the main text leading to it.
29
"Of the different forms of the theory only Schrodinger's has proved suitable for this,
and precisely for this reason I would regard it as the deepest formulation of the
quantum laws" (Born 1926a, p. 864).
 6.2 The Constitution of Quantum Mechanics 335

ingless calculus, but as a way of conceiving a specific physical situa-

tion. This in turn is considered abstractly, in the simplified and ideal-
ized form required for it to be conceived in that way. Let \|/?(q), \|/2(q),
. . . denote the proper functions of the unperturbed atom. A free elec-
tron of mass me moving in from infinity with energy E in the direction
of the unit vector with components a, p, and y is assigned the contin-
uous spectrum of proper functions sin(27i/^)(ax + Py + yz + 8), where
the wavelength X satisfies the de Broglie relation E = p2/2me = h2/2meX2
and the phase 8 takes all possible values. If the electron comes from
the direction +z, the atom-plus-electron system, before interaction,
should be assigned the proper function:

< ( q , z ) = ¥«(q)sinyz (6.18)

Born figured out by simple perturbation calculations that, for a given
interaction potential, there is a unique solution of the Schrodinger
equation which at +z —> °° goes over asymptotically into the said func-
tion \|/°(q) sin(2n/X)z. Further calculation shows that "after collision"
the scattered wave goes over asymptotically into this superposition of
solutions of the unperturbed process:

^ ax+fiy+yz>0 '^nEm

(6.19)
where dco is the element of solid angle in the direction (oc,P,y) and
V«£m((X,p,y) is what is now called the differential cross section for that
direction. Born then draws the momentous conclusion: "If one is to
understand this result in corpuscular terms, only one interpretation is
possible: \|/W£W(a,(3,y) determines the probability that the electron coming
from the ^-direction be thrown in the direction specified by the angles
a, p and y, with the phase change 8" (1926a, pp. 865f.; my notation).
Two questions are in order here: (i) Why must one reach for a cor-
puscular understanding? and (ii) why, if one does reach for it, is it nec-
essary to accept Born's interpretation? Question (ii) is fairly easy: If the
electron is a corpuscle, it can only move in a single direction at a time;
the function \|/n£W(oc,P,y) can take nonzero values in continuously many
different directions and so reflects a property that is tied not to the elec-
tron's actual path to infinity, but rather to the whole set of its possible
paths; whence Born's conclusion that \|/W£m(a,p,y) measures the proba-
bility that the electron takes the particular direction designated by the
336 Quantum Mechanics

respective argument (a,(3,y).30 Question (i) is harder. The only answer

that comes to my mind is that we are dealing with a physical situation
that was understood in corpuscular terms to begin with. A plane wave
incoming from infinity in a definite direction was associated with the
electron precisely because the latter was assumed to be a corpuscle all
along. The output of our calculations must of course be interpreted in
the same corpuscular terms as the input, and therefore the value of the
functions v|/W£W(a,P,y) for each set of angles a, (3, and y must have the
said probabilistic meaning.
One may wish to know how Born could be so sure that the electron
is a corpuscle. Late in life, he gave this explanation: He worked in Got-
tingen in the same building as James Franck, so he "was witnessing the
fertility of the particle concept every day in Franck's brilliant experi-
ments on atomic and molecular collisions and was convinced that par-
ticles could not simply be abolished" (1968, p. 55).

6.2.5 Quantum Mechanics in Hilbert Space

Once it was understood that Heisenberg's approach was equivalent to
Schrodinger's, they became fused into the theory that we call QM,
which promptly attained its mature, streamlined mathematical form
through the efforts of Jordan (1926, 1927), London (1926a, 1926b),
Dirac (1926a, 1926b, 1930), and von Neumann (1927,1932). The fol-
lowing sketch will be useful in the discussion of philosophical prob-
lems in the next two sections.31
Like all theories of mathematical physics, QM comes to life through
its application to physical systems, that is, simple or complex, small or
large chunks of the world we live in, which are picked out and con-

30
In a note added in proof, Born observes that the said probability must be propor-
tional to |\|/M£W(a,p,y)|2.
31
The sketch is rather too abstract, but I expect that, combined with Supplement I at
the end of the book, it will be sufficient for our purposes. Good philosophical books
on QM, such as Redhead (1987) and Hughes (1989), provide more informative -
although scarcely less abstract - summaries of the mathematics. For concreteness one
must turn to the standard textbooks, for example, Messiah (1961) or Cohen-
Tannoudji et al. (1977); however, philosophy students will probably feel overwhelmed
by their length. Sudbery (1986, Ch. 2-5), provides a shorter, clean and clear exposi-
tion; Chapter 4 works out some of the theory's simplest applications. The new text-
book by Hannabuss (1997) has a manageable size and looks very attractive to me,
but at the time of writing I had not yet formed an opinion about it.
 6.2 The Constitution of Quantum Mechanics 337

ceived by the physicist, under suitable idealization, in terms of the

theory.32 We saw in §5.5 that GR has been fruitfully applied to systems
that are drastically simplified models of the whole world. And there
are authors who write *F, with a capital letter, for the \|/-function of
the universe. They might even write down a mock Schrodinger equa-
tion for *F, but they come nowhere near a point where they could
attempt to solve it. Others think, with good reason, that the very nature
of QM precludes such attempts at globalization. For one thing,
although QM applies to single systems - one hydrogen atom, one pair
of photons - it makes statistical predictions, so it must be tested on
ensembles of such systems, that is, collections of noninteracting
instances of them. Now, it does not seem likely that one will ever test
a theory on an ensemble of worlds.33 Moreover, the following consid-
eration should make it clear that every scientifically significant appli-
cation of QM inevitably refers to a more or less conventionally
delimited fragment of the world. In QM one distinguishes between the
deterministic evolution of chances and the random occurrence of
chance events (see above, the quotation linked to note 28). But a chance
distribution, deterministically evolving in isolation, does not translate
into actual, particular, definite outcomes. These, it would seem, can
only be effected as the deterministic system impinges on its boundaries
with the rest of the world.34

32
J. S. Bell (1990, p. 19) places "system" at the head of a "list of bad words", followed
by "apparatus" and "environment". Such words, he says, "imply an artificial divi-
sion of the world, and an intention to neglect, or take only schematic account of, the
interaction across the split". Such artificial divisions, however, are inevitable in exper-
imental physics and the key to its success, as opposed, say, to Presocratic or Aris-
totelian cosmology. Bell's curious stance contrasts with that of the mathematicians
Birkhoff and von Neumann: "The concept of a physically observable 'physical system'
is present in all branches of physics, and we shall assume it" (1936, p. 823).
33
Peirce expressed it forcefully (CP 2.684): "The relative probability of this or that
arrangement of Nature is something which we should have a right to talk about if
universes were as plenty as blackberries, if we could put a quantity of them in a bag,
shake them well up, draw out a sample, and examine them to see what proportion
of them had one arrangement and what proportion another. But even in that case, a
higher universe would contain us, in regard to whose arrangements the conception
of probability could have no applicability."
34
Think of the little balls dancing in a modern lottery urn. There is a digit on each ball.
A prize is awarded to the number formed by the digits on, say, the first six balls to
jump out of the urn. Evidently, if the urn encompassed the entire world, there could
be no winners.
338 Quantum Mechanics

The duality of determinism and randomness underlies one of QM's

most remarkable features. QM provides a separate, mathematically dis-
tinct representation for the state of a physical system and for the
observable physical quantities by measuring which we get to know the
system. By this expedient the theory can predict future states with cer-
tainty while assigning definite probabilities to future results of mea-
surement. The probability distributions for the possible measured
values of the different physical quantities are encoded in the mathe-
matical representation of the state and can be readily retrieved from it.
In Hamilton's formulation of classical mechanics (see remark D at
the end of §2.5.3), a system with n degrees of freedom is associated
with a copy of R2n, the system's 2^-dimensional phase space. The state
of the system at any given time is represented by a point P in this space.
The coordinates of P are the (generalized) position and momentum
coordinates of the system, and every other physical quantity of inter-
est to classical mechanics must be defined as a function of these coor-
dinates. The past and future evolution of the state is represented by the
unique solution through P of the Hamilton equations for the system.
This solution encapsulates the past and future values of every physical
quantity of interest.
A quantum-mechanical system is associated with a - generally infi-
nitely dimensional - Hilbert space 3^.35 The state of the system is rep-
resented by a nonzero vector in 3€. If |\|/) represents the state, so does
#|\|/), where a is any complex number. From now on, unless otherwise
indicated, it is understood that any vector |\|/) chosen as state repre-
sentative is normalized, that is, such that |\|/|2 = 1. It is assumed that
every vector in 3€ represents a possible state of the system. This implies
that, if |\|/i) and |\|/2) represent two possible states, and a and b are any
complex numbers, the linear combination tf|\|/i) + b\\\f2) represents a
third state, the "superposition" of the former two (for which one can
readily find a normalized representative). The physical quantities of
interest - usually called 'observables' - are represented by self-adjoint
operators in 3€. The seemingly plausible assumption that every such
operator represents a measurable quantity has turned out to be unten-
able (Wigner 1952; cf. Stein and Shimony 1971).
To see how these representations of states and observables work

35
See Supplement 1.7 and §6.2.3. If this will help you, you may think of %C as the space
5Q of complex-valued square-integrable functions on a suitable configuration space
Q.
 6.2 The Constitution of Quantum Mechanics 339

together, consider for the moment a Hilbert space W with finitely many
dimensions. Suppose that a certain observable Q is represented by the
self-adjoint operator Q on <3t. The possible values of Q are the proper
values of Q. If 'K has n dimensions and there are n such values
qu . . ., qm there are corresponding proper vectors, |\|/i),. . ., |\|/w), that
form an orthonormal basis of 3^. Any vector |\|/) representing a state
of our system can be expressed as a linear combination of these vectors,
viz., |\|/) = If- = i|\|/jX\|/;|i|/)- If no two proper values correspond to the
same proper vector, we say that Q is nondegenerate. Then, K\|/,]\|/)|2 is
the probability that, if Q is measured on a system that happens to be
in state |\|/), the measured value of Q is found to be g,.36 If Q is degen-
erate and the proper value g& corresponds, say, to r linearly indepen-
dent normalized proper vectors |\|/*i),. . ., |\|/&r), the probability that the
value of Q is qk if Q is measured on a system in state |\|/) is equal to
Zf=1|(\|/J\|/)|2. If %C is infinite-dimensional, there are other complications.
In the first place, the concept of a basis must be widened (see Supple-
ment I, after eqn. (S5)). Even so, not every self-adjoint operator on Hi
will have a family of proper vectors constituting an orthonormal basis.
It is usually assumed that only those that do are apt for representing
observables (Messiah 1961, vol. I, p. 188; Cohen-Tannoudji et al.
1977, vol. I, p. 137). Finally, the basic notions of proper vector and
linear combination of vectors must be adjusted - or replaced - to cope
with the fact that some observables have a continuous spectrum of pos-
sible values (Supplement 1.7).
The above gives an inkling of QM's probabilistic account of actual

36
The condition "if Q is measured" will surely trouble some readers. They expect a
physical theory to predict, at least with probability, the value that a physical quan-
tity also has when it is not measured. QM generally cannot do this, essentially due
to the fact that many pairs of self-adjoint operators p, q, representing physical quan-
tities, satisfy the relations (6.14). Cf. Dirac 1958, pp. 46-47:
The expression that an observable 'has a particular value' for a particular state
is permissible in quantum mechanics in the special case when a measurement
of the observable is certain to lead to the particular value, so that the state is
an eigenstate of the observable.

In the general case we cannot speak of an observable having a value for a par-
ticular state, but we can speak of its having an average value for the state. We
can go further and speak of the probability of its having any specified value
for the state, meaning the probability of this specified value being obtained
when one makes a measurement of the observable.
340 Quantum Mechanics

measurement results. I turn now to the system's evolution in time. In

the so-called Heisenberg picture, the state |\|/) of the system is fixed and
self-ad joint operators representing the quantum-mechanical analogues
of classical position and momentum coordinates evolve deterministi-
cally according to eqns. (6.15). This is equivalent - via a suitable math-
ematical transformation - to the more manageable Schrodinger picture
in which the state |\|/) of the system evolves deterministically in $£
according to Schrodinger's time-dependent equation (note 22) and the
observables are represented by fixed self-adjoint operators. Writing H
for the Hamiltonian operator of the system and |\|/(£)) for the
state of the system at time £, Schrodinger's equation can be written
schematically:

^|V(*)> = -£H|V(*)> (6.20)

Therefore, if |\|/(0)) is the state of the system at some chosen time
t = 0, its state |\|/(£)) at any arbitrary time t satisfies the relation:

(6.21)

where exp(-(i/ti)Ht) is a unitary operator on 3£, commonly denoted by

Ut. Since exp(0) is the identity and

- that is, Uo = I and UtlUt2 = Uh+t2 - it is clear that the operators Ut

form a continuous group parametrized by t. The evolution of states in
the Hilbert space 3€ under eqn. (6.20) is appropriately described by the
action of this group of unitary operators on 3€, and it is therefore often
referred to as 'unitary evolution'.
As even this bald sketch shows, the quantum-mechanical scheme of
description and prediction is one of the most elegant creations of the
human spirit. But it has implications that numerous philosophers and
a few physicists find objectionable. I shall deal with their views in
§§6.3-6.4. But let me show here, by means of a simple example, why
the sort of situation that they have difficulty in assimilating is inevitable
in the quantum-mechanical scheme. Consider a suitably prepared phys-
ical system S that, according to QM, will be in state |\|/) at time t.
Arrangements are made to measure at t on S the observable (repre-
sented by the operator) Q. Assume that |\|/) is not a proper vector of Q
and that Q is nondegenerate, with proper values a, uniquely associated
 6.2 The Constitution of Quantum Mechanics 341

with the elements of an orthonormal basis of proper vectors |(p,) (/ = 1,

2 , . . .). Under such circumstances, QM predicts that the measurement
will record the particular value Uk with probability |(cp^|\(/)|2. This is typ-
ically a real number greater than 0 and smaller than 1. However, as
soon as the measurement is completed, its result turns out to be either
ak with certainty or some other value a, & a^ so that ak is excluded.
There is no contradiction here, for the prediction is compatible with
both results. The prediction would be questionable only if a set of Q-
measurements performed on a very large ensemble of systems in state
|\|/) yielded a proportion of ^-values that is significantly different from
|((pfe|\|/)|2. Still, a difficulty apparently arises if the matter is put as
follows. Suppose that the measured value in effect is ak. Then, unless
the system S is destroyed during measurement (a common occurrence
in microphysics), one ought to say in accordance with QM that, right
after the measurement, S is in state |(p*)5 the single proper state of Q
corresponding to the proper value ak. Thus S "jumps" at t from state
|\|/> = Z7=i|<PiX<Pij\|/) to state |q>*). Such a "jump" cannot be derived from
eqn. (6.20). So - it appears - the state of a quantum-mechanical system
evolves, in different circumstances, in one or the other of two differ-
ent ways, governed by different laws. One of these laws, the
Schrodinger equation, is well known and similar to other familiar laws.
The other law is unknown and apparently unfathomable. I dare say
that this way of putting things involves either unwillingness to accept
the reality of chance or inability to intellectually cope with it. The
"jump" in question is the transition from an initial situation in which
a chance event is being expected to a final situation in which the
outcome of chance is already given. QM describes the former by means
of a state vector from which different probability distributions can be
extracted, one for each physical quantity that one might be willing to
measure. Thus we have the following statement:

(A) 'S will attain state |\|/) at time t\

The final situation, however, does not require such an abstruse descrip-
tion. The definite outcome of a definite measurement can be straight-
forwardly described like any ordinary matter of fact:
(B) 'Value ak was obtained when Q was measured on S at f.
If S was not destroyed by the measurement and Q is a nondegenerate
operator with a discrete spectrum, (B) entails in QM a description of
S at t by means of a state vector, viz.,
342 Quantum Mechanics

(C) The state of S at £, right after measurement, is |(p&)\

Note that - within QM - (C) is strictly weaker than (B). (C) entails
that 'the value of Q measured on system S at t is a^ with probability
K^l^)! 2 = 1'J but this is not equivalent to (B) (although philosophers
tend to forget it). Thus, if we only wished to express what we know
about S right after the measurement at £, (C) would be pointless. Its
real utility is in predicting the outcome of subsequent measurements
on S. (This is done by calculating the evolution of |cp*) under
Schrodinger's equation and expressing the resulting vector as linear
combinations of proper vectors of the relevant operators.) In other
words, (C) - like (A) - is appropriate for speaking about expected
chance events. Now if (A) and (C) are read, in straight succession, omit-
ting (B), it sounds like the story of a state evolution that is not subject
to Schrodinger's equation. At first blush it might seem that physics
should be committed to explain this story. However, (C) is just the pre-
dictive content of (B), under our special assumptions regarding S and
Q. To say why and how (A) leads to (C) one must explain the "jump"
from (A) to (B). To give such an explanation, however, would amount
to denying that (B) describes the outcome of chance.

In the remainder of this section I introduce some notions that we shall

need in §§6.2.6, 6.3, and 6.4, namely, (a) the handling of compound
systems, (b) projectors, (c) the expected value and the variance of an
observable, (d) mixtures, and (e) the statistical operator. The explana-
tions are dry, and I suggest that readers turn to them as and when they
need them.
(a) If the states of systems A and B are represented by vectors in
spaces "KA and 2£B, respectively, the states of the compound system
A + B are represented in the tensor product space 3€A ® $£B. See Sup-
plement 1.8. Remember that ^A <8> %tB contains every conceivable linear
combination of the tensor products |a) ® |(3), for every vector |a) in
VtA and |p> in 3KB.
(b) Projectors are self-adjoint operators with peculiar properties
that make them suitable for representing states. Let P be a linear oper-
ator on the Hilbert space $£, with adjoint P+. P is a projector if and
only if P+P = P. This implies that P = P+ (P is self-adjoint) and that PP
 6.2 The Constitution of Quantum Mechanics 343

= P (P is idempotent).37 It can be shown that a projector P maps -

"projects" - the whole of "M onto a subspace of it. This range of P is
not shared by any other projector and is therefore uniquely associated
with P. In particular, if |\|/) is a normalized vector, there is a unique
projector P ^ that sends every vector in $£ to the subspace spanned by
|\|/). The projection of an arbitrary vector |(p) along the direction of |\|/)
is equal to the result of multiplying |\|/) by the inner product of |\|/) and
|(p), viz.,

|V> (6.22)
We may therefore write
(6-23)
Equation (6.23) defines P|v> uniquely for each normalized |\|/) in M and
may therefore be used for representing the same state as |\|/). Note
further that if the vectors \|/,- (for every index / in some set $) form a
basis, the operator defined by adding their respective projectors sends
each vector in 'K to itself. In other words,
ZjV.XV.hl (6-24)
where I denotes the identity operator on 3€.
Consider now an observable Q represented by an operator Q whose
proper value qk corresponds to r linearly independent normalized
proper vectors |\|/*i), . . ., |\|/*r). Let P(#*|\|/) be the probability that the
value of Q is qk if Q is measured on a system in state |\|/). Then

where P*. stands for the projector onto the subspace spanned by the
vectors |\|/*i),. . ., |\|/*r) corresponding to the proper value qk.
(c) The predictions of a probabilistic theory must be tested on very
large collections of like systems (ideally, on infinitely many of them).
One compares the predicted mean or expectation value of a quantity

37
If P = PfP , then P+ = (Ptp)t = ptptt = P t P = P> so P is self-adjoint. Therefore P+P =
PP= P. So P is idempotent.
344 Quantum Mechanics

with the actual average of a long series of measurements. One also

compares the predicted and the actual spread of the particular results
about the mean. Suppose that we measure an observable Q, repre-
sented by the operator Q, on a large collection of identical systems -
for example, silver atoms, or circularly polarized photons - prepared
in a state represented by the normalized vector |\|/). The expectation
value, denoted by (Q)|¥>, is then given by:

<Q)M = <i|/|Ch|/> = <QV| V ) = <V|Q| V ) (6.26)

(The second and third expressions are equal because Q is self-adjoint;
the last expression serves as a conventional reminder of it.) Equation
(6.26) can be easily proved if Q has a discrete, nondegenerate spectrum
of proper values {q{\ i e 3} corresponding to the orthonormal basis
{\|/,: / G $}. Then the probability PQ(<j'i|\|/) of obtaining the value q{ in a
measurement of Q on a system in state \|/ is, as we know, equal to
>|2. Thus,

(6.27)

where the last step in the derivation is based on eqn. (6.24).

A good measure of the dispersion of measured values about the mean
is the variance (called "uncertainty" in some older books). The variance
AQy of the quantity Q as measured on systems in state |\|/) is the square
root of the expectation value of (Q - (Q)|v>)2. Hence, by eqn. (6.27),

(AQ¥)2 =((Q-(Q) |¥) ) 2 > |¥> =<V|(Q-<v|/|Q|v|/»2M

= <vt/|Q2|v|/> - 2(V|/|Q|V|/)(V|/|Q|VK) + (V|Q|¥)2(VK|V|/) (6.28)

(d) It is not always possible to prepare a large collection of objects

in a state that is representable by a particular vector. A method of
preparation might yield a mixture of like objects in different states, rep-
resented by the vectors |\|/i),. . ., |\|/M), each with a known - or conjec-
tured - relative frequency. If the collection is very large, the relative
frequency of a given state approaches the probability that a randomly
chosen object in the collection is in effect in that state. It would seem,
 6.2 The Constitution of Quantum Mechanics 345

at first sight, that the probability that an object picked at random from
such a mixture is in the "pure" state represented, say, by \\\fk) (1 < k <
n) ought to be carefully distinguished from the probability |((p;j^)|2 that
an object that happens to be in the "pure" state represented by |\|/*)
will sport, upon measurement of an observable Q with discrete non-
degenerate spectrum {|cpt): / e $}, the proper value corresponding to
|(p7). Yet these two kinds of probabilities must surely obey the same
mathematical rules, for they are freely combined in QM calculations.
The probability pk that, say, the feth proper value of an observable will
be measured on a mixture is computed by adding up the probabilities
pku • • • >pkr that the said value will turn up in each of the r pure states
in the mixture, weighted by the probabilities wu . . ., wr that an object
in the mixture is in one of these states: pk = If^Wipki.
(e) Pure states and mixtures can be represented in a uniform way
by means of the statistical operator (or density operator), which I shall
now define. Consider first a system that at time t is in the pure state
represented by the normalized vector \\\f(t)). The statistical operator p(t)
representing this state is then simply the projector P|V(*)>:
(6.29)
Pick an orthonormal basis {\|/,: / e$}, such that \\\f(t)) = Z/e<* c,-(
where c{{t) stands for (xi/^f^)). Then the matrix of p{t) in the basis
{|\|/f): / e 3} has the typical element:
= cl(t)c*(t) (6.30)
We have that
X Xk,-W| 2 =l (6.31)
since |i|/(£)) is normalized. The expectation value (Q)|V(t)> of an observ-
able Q represented by the operator Q can now be expressed in terms
of the statistical operator p(t):

jet
346 Quantum Mechanics

By combining eqn. (6.32) with (6.25) we obtain an expression in terms

of p(t) for the probability VQ(qk\\\f(t)) that the observable Q takes the
value qk if measured on a system in state |\|/):
PQMVW) = (vWPtlvW) - Tr(p(t)Pk) (6.33)
The evolution of the statistical operator p(t) can be derived from the
Schrodinger equation (6.20):

{634)

I now turn to an ensemble £ in a mixture of states |\|/i),. . ., \\\fn)

and denote by pr the probability that a randomly chosen object be in
state |\|/r) (0 < pr < 1; 1 < r < n; lnl=1 pi =1). Let P Q (^|£) be the proba-
bility that the observable Q takes the value qk if measured on an object
of this ensemble. To calculate PQ(qk\E) we multiply the probability pr
that an object is in state |\|/r) by the probability Pa(^|\|/r) that Q takes
the value qk if measured on an object in that state, and add up the
products thus obtained. Writing pr for the statistical operator repre-
senting the state |\|/r) and p for the weighted sum X?=i piPi of such sta-
tistical operators, we have that

(6-35)

Evidently p, as a linear combination of projectors, is a self-adjoint oper-

ator (although generally not a projector). This is, by definition, the sta-
tistical operator that represents our mixture:
n

P = 2>,P, (6.36)
Equations (6.31)-(6.34) continue to hold if p(t) is the time-dependent
statistical operator of a mixture (see Cohen-Tannoudji et al. 1977,
vol. I, pp. 301f.).
The uniform representation of mixtures and pure states by statisti-
cal operators paved the way for a different conception of quantum-
 6.2 The Constitution of Quantum Mechanics 347

mechanical states. One regards "the state of a system, whether pure or

not, as defined by its previous history, i.e., by the method of its prepa-
ration" (Fano 1957, p. 76). Thus, the general concept of a quantum-
mechanical state is that of a mixture, while the so-called pure state is a
limiting case. "A pure state is characterized by the existence of an exper-
iment that gives a result predictable with certainty when performed on
a system in that state and in that state only".38 Such an experiment, pro-
viding a maximum of information about the state prepared, is called
'complete'. "A pure state can then be identified by specifying the com-
plete experiment that characterizes it uniquely. Mathematically one can
construct a variety of [self-adjoint] operators which have a given pure
state as a [proper state]" (p. 75). But there are systems "for which no
complete experiment gives a unique result predictable with certainty".39
Still, the state of such systems is "fully identified by any data adequate
to predict the (statistical) results of all conceivable observations of the
system. [. . .] Indeed, 'state' means whatever information is required
about a specific system, in addition to physical laws, in order to predict
its behavior in future experiments" (p. 76). Nonpure states are called
'mixtures' because they can be represented by the incoherent superpo-
sition of pure states.40 But one must bear in mind that such a represen-
tation is nonunique. "There is in general no reason, for example, why
unpolarized light should be described as a mixture of two linear polar-
izations rather than of two circular ones". From this standpoint, one
considers "only the broader set of all fluctuations among the experi-
mental results obtained with an ensemble of systems prepared accord-
ing to identical specifications", without analyzing those fluctuations
into subsets "when this analysis is not unique and does not correspond
to observable characteristics of the situation". In other words, one deals
"with a single statistical ensemble of quantum mechanical systems pre-

"For example, linear polarization of a light beam in a given plane is characterized

by 100% transmission of each photon through a suitably oriented Nicol prism; no
other state of polarization is fully transmitted by the same prism" (Fano 1957, p.
75).
"For example, no polarization analyzer admits or rejects with certainty photons of
partially polarized light" (Fano 1957, p. 76).
"Incoherent superposition means, by definition, that to calculate the probability of
finding a certain experimental result with a system in the mixed state one must first
calculate the probability for each of the pure states and then take an average, attribut-
ing to each of the pure states an assigned 'weight'" (Fano 1957, p. 76). See the main
text, under (d).
348 Quantum Mechanics

pared by identical procedures, not with a statistical ensemble of

quantum mechanical ensembles" (Ibid.).

6.2.6 Heisenberg's Indeterminacy Relations

Heisenberg's paper, "The Intuitive Content of Quantum-Theoretical
Kinematics and Mechanics" (1927), is regarded by some as the begin-
ning of a new era in physics. He showed there that, if QM is right, it
is impossible to establish with perfect or near-perfect accuracy the posi-
tion and momentum of one or more particles at a given time. Heisen-
berg's language, both in this paper and in his Chicago lectures (1930),
suggests that this impossibility reflects a limit to the precision attain-
able by experimental measurement, due to the existence of the finite
quantum of action h. To the artless reader it might seem as if all that
is being proved is that human observers cannot get to know accurately
the position and momentum of a physical object because every suc-
cessful attempt to measure the former will blur the latter, and vice
versa, although the positions and momenta are still there, inscribed in
real things, in their full classical sense. But the relations of indetermi-
nacy (or "uncertainty") between certain pairs of observables in QM
are a corollary of the theory's mathematics. Thus the simultaneous
assignment of dispersion-free values to any such pair is incompatible
with the theory's conceptual framework and therefore meaningless in
its context.41 Heisenberg's thought experiments were definitely not
meant as proofs of the indeterminacy relations. As indicated by the title
of his paper of 1927, they were proposed in order to make QM intu-
itively plausible despite its - for the early twentieth-century physicist -
implausible implications.
Consider any two quantum-mechanical observables Q and P, rep-
resented by the self-adjoint operators Q and P. Write [Q,P] for QP -
PQ. It can be easily proved that, for any ensemble of systems in an
arbitrarily chosen state |\|/),

41
Heisenberg (1927, pp. 179f.) puts this rather differently, but the upshot, I think, is
the same: The definition of physical quantities rests on the experiments by which we
measure them; "every experiment which we can use for defining these words ['elec-
tron position' and 'velocity'] necessarily contains the imprecision indicated by the
equation [AQ¥-APV~ h]". Equation (6.14) could not be valid if this imprecision did
not obtain. "If there were any experiments which allowed a 'sharper' simultaneous
determination of P and Q . . . , Quantum Mechanics would be impossible."
 6.3 Philosophical Problems 349

AQy-APv>i|</[Q,P]),J (6.37)
Thus, if Q and P satisfy eqn. (6.14) - as they must if they happen to
be the quantum-mechanical analogues of a pair of conjugate classical
position and momentum coordinates - the product AQ^-APy > ~-h.
Therefore the variance of one of these observables is infinite if that of
the other is zero.
To prove the inequality (6.37) recall eqn. (6.27) and the definition of
variance, put A = Q - (Q)|¥>, B = P - (P)|v>, note that A and B are self-
adjoint operators, and bear in mind that (Q)|v> and (P)|v> are complex
numbers that commute with every observable (so that [A,B] = [Q,P]).
Let |(p) denote the vector A|\|/) + z'AB|\|/), with X an arbitrary real number:

The last expression is a second-degree polynomial in X of the standard

form c + bX + aX1. Since ((p|(p) > 0 (Supplement 1.4), the polynomial has
either no zeroes or equal zeroes and the discriminant b1 - Aac < 0.
Thus, </[Q,P])jv> < 4(AQVj/)2(APV|,)2. The inequality (6.37) follows.

6.3 Philosophical Problems

6.3.1 The EPR Problem

It was Einstein who taught Bohr and his followers to conceive the tran-
sitions between atomic stationary states as chance events (see §6.1.2).
But he could never bring himself to see this view as final. For him, one
resorted to probability - in quantum as in classical physics - due to
human ignorance, not because the physical world is ultimately a gigan-
tic chance setup.42 So he favored the notion that QM, despite its great
successes, is an "incomplete" theory, in other words, that there must be
some "elements of reality" that the theory ignores. In 1935 Einstein pub-
lished, together with his younger collaborators Podolsky and Rosen, the
42
Whence Einstein's insistence, in the very paper in which he gave chance a leading role
in the quantum theory of radiation, that this was a "weakness of the theory" and a
feature of its "present state" (1916n, p. 62).
350 Quantum Mechanics

famous EPR paper in which they purport to show this. Their allegation
was promptly impugned by Bohr (1935), with an argument that effec-
tively persuaded the great majority of physicists (cf. §6.4.1). The present
fame of EPR is not due, however, to its immediate effect on physical
research - which was virtually nil - or to its logical excellence - which
is questionable - but to the proof, found 30 years later by J.S. Bell (1964)
that a theory that met the EPR demands would not just teach us more
things than QM, but would make predictions that flatly contradict QM.
It is not clear how such a theory could cope with the enormous mass of
empirical evidence that supports QM. Leaving this question aside,
several research teams designed and performed clever experiments to
test specific predictions in which QM and an EPR-style theory must run
foul of each other. The results were ambiguous at first, but in the end,
as one might well have expected, the superiority of QM was established
by the group led by Alain Aspect (references in note 45).
Einstein, Podolsky, and Rosen built their case on two seemingly
obvious premises:
1. If a physical theory gives a complete description of reality, every
element of the physical reality must have a counterpart in the theory.
2. "If, without in any way disturbing a system, we can predict with
certainty (i.e with probability equal to unity) the value of a physi-
cal quantity, then there exists an element of physical reality corre-
sponding to this physical quantity" (1935, p. 777).
They proposed a thought experiment that, according to QM, enables
one to predict with certainty, without in any way disturbing a quantum
system, the value of certain physical quantities pertaining to that
system. Since the \|/-function representing such a system in QM does
not furnish us with a definite counterpart for all these quantities, Ein-
stein, Podolsky, and Rosen concluded that the \|/-function "does not
provide a complete description of the physical reality" (p. 780). The
EPR thought experiment involves noncommuting operators with
continuous spectra. Bohm replaced it with a conceptually equivalent
thought experiment involving noncommuting operators with two-
valued nondegenerate spectra (viz., mutually perpendicular spin com-
ponents). This was the subject of Bell's proof of 1964 and also the
ideal model for the experiments actually performed. It is also easier to
explain, so I shall refer to it.
This is how Bohm (1951, p. 614) describes his experimental setup:
A molecule consisting of two atoms with spin +till (see §6.1.4) is in a
state in which the total spin is 0. The molecule is disintegrated by a
 6.3 Philosophical Problems 351

process that does not change the total spin. The atoms move apart and
cease to interact appreciably. Spin is conserved, so the spin of the two
atoms continues to add up to 0. The spin component of either atom
in a given direction takes one of the two possible values, +fill or -fill.
Thus, if we measure the spin component of the first atom in a partic-
ular direction and find it to be +ft/2, we can predict with certainty and
without in any way interfering with the second atom that its spin com-
ponent in the said direction is -fill. Therefore, the spin component of
the second atom in the chosen direction is an element of reality accord-
ing to premise 2. Since the direction in question was arbitrary, this is
true of every direction. The trouble is that, for a particle with spin ±ft/2,
the spin components in mutually perpendicular directions are repre-
sented in QM by noncommuting linear operators, and the state of the
particle at a given time can be represented by a proper vector of one
or the other of two such operators, but not of both. Therefore, given
a system of Cartesian coordinates x, y, and z9 if we know with cer-
tainty that the second atom has spin -fill in the direction parallel to
the z-axis, QM will not allow us to assign to it spin -fill or spin +fill
in the direction parallel to the x-axis. Yet these are the only two pos-
sible values of the spin component in this direction, and, according to
the EPR argument, the spin component of the second atom in every
direction is an element of reality. Therefore, by premise 1, QM does
not give a complete description of reality.
This can be explained more formally as follows. If one temporarily
forgets all physical properties except spin, the state of a spin ±fill par-
ticle a can be represented in a two-dimensional Hilbert space $£a. The
spin components parallel to the Cartesian coordinate axes x, y, and z
are represented by three linear operators on 9^a that I shall denote by
S*, Sy, and Sz, respectively.43 Each one of these operators has two
proper vectors, corresponding, respectively, to the proper values +h/l
and -fill. To simplify formulas, I choose units such that fill = 1 and
denote the proper vectors of each operator by |+) and |-), followed by
a subscript that indicates the direction. In this notation, Sz|+)z = |+)z
and Sz|-)z = -\-)z and similarly for x and y. The two proper vectors of
each operator span the space Wa. The proper vectors of Sx and Sy are
therefore linear combinations of |+)z and |-) z . In particular,

l+>, =^fl+>,+h>J l->, = ^ ( l + > H - U (6.39)

43
These operators do not commute. In effect, [SX,SJ = ihSz, [Sy, SJ = ihSx and [Sz, SJ
= ihSy.
352 Quantum Mechanics

Now let a and p be two spin ±fill particles forming together - as in

Bohm's setup - a 0-spin system that we denote by a + (3. Insofar as
spin alone is concerned, the state of a + (3 is represented by a vector
|\|/) in the product space 3^a ® 3^p. This space is spanned by the four
vectors |+ +)z = \+)z ® |+>« |- -)z = \-)z ® K , |+ - ) , = \+)z ® \-)a and
|- +)z = |-) z ® \+)Z9 so \|/ is a linear combination of these vectors.44 a +
P has an even chance of being found on measurement in state |+ -)z or
|- +)z and no chance at all of being found in state |+ +)z or | — ) z . We
assume that system a + (3 is in the so-called singlet state, in which

IV> = ^ l + - > , - ^ | - + > , (6-40)

Suppose that a and p are now far apart and that we measure P's spin
component parallel to the 2-axis. There is the same probability 1/2 that
the measurement will show that p is in spin state \+)z - so that a is in
state \-)z - or that P is in spin state \-)z - so that a is in state |+)z. In
either case it is evidently impossible, by eqn. (6.39), that a is in state
\+)x or in state \-)x. However, according to the EPR argument, one of
these two states is no less an element of a's reality than one of the two
states \+)z and \-)z.
We are thus driven to the conclusion that, if the EPR argument is
right, the description of reality provided by QM is not just incomplete
- as Einstein and his associates claimed - but simply wrong. As I men-
tioned earlier, this is also the upshot of Bell's discovery. Bell (1964)
proved by a fairly straightforward mathematical argument that statis-
tical predictions based on the EPR conception of elements of reality
are subject to an inequality that the statistical predictions of QM do
not satisfy. Laboratory tests of Bell's inequality showed that real ele-
mentary particles in Bohm-type experiments do not satisfy them
either.45 More recently, Greenberger, Home, and Zeilinger (1989)
proved that by considering a system formed by three or more corre-
lated particles of spin ±h/2 - instead of just two, as in Bohm's thought
experiment and in Bell's original proof - a contradiction between the

44
This sentence and the next remain true, of course, when the subscript z is consistently
replaced with x (or with y).
45
Aspect, Grangier, and Roger (1982) and Aspect, Dalibard, and Roger (1982). Their
results are not immune to nitpicking, but - combined with all the other available evi-
dence for QM - they would in the normal practice of science have laid the issue to
rest if the dispute were purely about matters of fact and did not impinge on deeply
seated metaphysical prejudices.
 6.3 Philosophical Problems 353

EPR assumptions and the predictions of QM can be derived not just

in the case of statistical, but also of perfect, correlations (by which the
authors mean arrangements in which the result of a measurement on
one particle can be predicted with certainty given the outcomes of mea-
surements on the other particles of the system). A satisfactory expla-
nation of these matters would take us too long, and it would also be
somehow redundant, for there are several good expositions in the
literature.46
From a philosophical standpoint it is important to realize that the
EPR reasoning - as formulated by its authors - is fallacious. Accord-
ing to premise 2, a quantitative attribute of a physical object is - or,
in EPR jargon, "corresponds to" - an element of the reality of that
object if it is possible to predict its value with certainty without in any
way disturbing the object. In the Bohm-type situation described above,
this criterion of reality was applied to the spin components of atom a
in two mutually perpendicular directions. But, of course, according
to QM, precisely when it is possible to predict with certainty the z-
component of a - because one has measured the ^-component of (3 -,
it is not possible to predict with certainty the x-component of a, and
vice versa. Thus, of two mutually perpendicular spin components of
our atom a, no more than one at a time can satisfy the EPR criterion
of reality under the principles of QM. Surely, while the ^-component
of a is being predicted with certainty without disturbing a, one can
also predict with certainty the x-component of a similar atom a' -
which is entangled with an atom p' like a is with (3 - again without in
any way disturbing a'. But this fact about a' does not say anything
about the present reality of the x-component of a; to think otherwise
presupposes a different criterion of reality that the EPR authors do not
make explicit.47

46
My own favorite is still Wigner (1970) (but read '03i > TC on p. 1,008, line 3 from
below). See also Shimony (1990). A very readable analysis of the Aspect experiments
will be found in Ruhla (1992), Ch. 8. Greenberger et al. (1990) give a perspicuous,
self-contained exposition of the Greenberger-Horne-Zeilinger theorem and its back-
ground; see also Clifton, Redhead, and Butterfield (1991a, 1991b).
47
It lurks perhaps in the following sentence: "Since at the time of measurement the two
systems no longer interact, no real change can take place in the second system in con-
sequence of anything that may be done to the first system" (Einstein, Podolsky, and
Rosen 1935, p. 779). Yet this manner of change is common in social systems. For
example, Judge Jones becomes Smith's father-in-law, and therefore incompetent - in
some legal systems - to judge a case in which Smith is the plaintiff, as soon as, unbe-
354 Quantum Mechanics

There is one critical aspect of the EPR problem that I have side-
stepped until now. I have unquestioningly accepted that in Bohm's
experiment the measurement of a spin component on particle (3 does
not "in any way" disturb particle a. This was the general attitude while
the discussion was concerned only with thought experiments: One
simply assumed that the condition prescribed in premise 2 was fulfilled.
But since Bell's inequality was tested by real experiments, we often hear
that its effective violation reveals the presence of a faster-than-light
action-at-a-distance of some sort. Now the said experiments are
designed so that no light signal can link the measurements on (3 with
the correlative measurements on a (by which the QM predictions based
on the former are confirmed). However, this precaution was not taken
to test for faster-than-light signals, but to ensure that no influence went
covertly from p to a; for if it did the data would be worthless. Indeed,
if the experiments - as designed - did show the existence of such influ-
ence, this alone would lead to the conclusion that Einstein and his col-
laborators had in mind, albeit by a path that was different from theirs.
Since QM itself does not contemplate any form of physical action
through which the measurement performed on particle (3 might bring
about a change in particle a, if such an action does exist, the quantum-
mechanical description of reality is certainly incomplete. But what kind
of action would this be? Shall we postulate a new kind of field through
which the act of measuring on (3 the ^-component of spin fixes the z-
component of a? And devise some costly artifact that is capable of
detecting the corresponding "particle"? And try to persuade the Con-
gress of a wealthy and intellectually avid nation to finance its con-
struction? In Bohm-type experiments the measurement of a particular
spin component on particle (3 is deliberately arranged so that it does
not cause any physical change in a. The act of measurement does
however bring about a resetting of the terms in which - pursuant to
QM - the spin of a must be described. Given that the total spin of a

knownst to Jones, Smith marries his daughter. However, before the advent of QM,
it was unfamiliar to physicists. Right after the quoted sentence, Einstein et al. remark
that it "is, of course, merely a statement of what is meant by the absence of interac-
tion between the two systems". Of course, if one understands 'interaction' in this
broad sense, instead of taking the word in its usual physical meaning of 'energy-and-
momentum exchange', the sentence in question does indeed become trivially true, but
it also becomes impossible to ensure by physical means - say, by separation, or shield-
ing - that our two particles "no longer interact".
 63 Philosophical Problems 355

+ P is 0, as soon as, say, the ^-component of p's spin is definitely known,

so is the ^-component of a. That the latter, if measured, obligingly com-
plies with the quantum-mechanical description is one more reason for
adopting QM.48
The violation of Bell-type inequalities in Bohm-type experiments
according to QM has led philosophers to say that QM is a nonlocal
theory. This is their way of expressing the fact - noted above for par-
ticle a - that the quantum-mechanical state description of an object
sometimes must be changed even though one is not prompted to do so
by what actually goes on in its immediate surroundings. I tend to think
that this novel terminology is unnecessary, if not downright mislead-
ing. The correlation between the spin components of a and P discussed
above was quite fitly described by Bohm himself as noncausal (1951,
p. 430). And surely we ought not to expect that QM, having so
famously made a mockery of causation, should submit to its traditional
demands in this particular case.49

6.3.2 The Measurement Problem

QM began with Heisenberg's decision to pay attention exclusively to
observable - that is, measurable - quantities and to refrain from trying
to reconstruct in thought the unobservable "reality" that supposedly
lies beneath them. It is therefore vexing - although perhaps only
natural - that the problems regarding the microstructure of matter and
radiation - quantum jumps, wave-particle duality, and so on - that
beset the Old Quantum Theory and which Heisenberg's approach dis-

48
Falkenburg (1995, pp. 295, 296) fittingly associates "the non-local EPR-correlations
of particles detached from a coupled quantum system" with the breakdown of the
classical concept of a particle and, more generally, of "all the traditional representa-
tions of natural philosophy regarding the constitution of empirical reality in the
small".
49
Niels Bohr, who understood that "the very existence of the quantum of action" entails
"the necessity of a final renunciation of the classical ideal of causality" (1935, p.
697), had this to say about the EPR experiment: "It obviously can make no differ-
ence as regards observable effects obtainable by a definite experimental arrangement,
whether our plans of constructing or handling the instruments are fixed beforehand
or whether we prefer to postpone the completion of our planning until a later moment
when the particle is already on its way from one instrument to another" (1949, p.
230). This trenchant remark can be profitably contrasted with recent quibbling over
"quantum nonlocality" (and its purported conflict with Relativity).
356 Quantum Mechanics

pelled should be replaced in QM with a problem of measurement. QM

is the source of several such problems - indeed the EPR problem might
count as one of them - but 'the measurement problem' designates in
the literature the problem complex that I shall now discuss.50 It arises
at the interface between the deterministic evolution of the state of a
quantum system and the measurement on it of a particular physical
quantity with definite, probabilistically distributed values. As Wigner
pithily remarks, "this is hardly surprising because the deterministic
nature of the equations of motion prevents them from accounting for
a probabilistic result" (1983, p. 285). The measurement problem is
especially severe if one views the state vector and its unitary evolution
as a representation of underlying reality. But it also demands consid-
eration and solution if the observations results are the sole "reality" of
interest for QM, and the state vector is only a store of information
from which to calculate the probabilities of the several possible out-
comes of different types of measurement. As we shall see, when QM
is applied to the composite physical object formed by a system under
study and the apparatus employed for studying it, the theory has - or
so it seems - bewildering implications.
To make the following discussion easier to read - and to write - I
shall no longer distinguish between physical states and observables, on
the one hand, and the Hilbert space vectors and self-adjoint operators
that represent them, on the other. Unless otherwise noted, I shall write
as if every observable had a discrete, nondegenerate spectrum. This sim-
plification - which is quite common in philosophical literature - is per-
missible because the fact that in real life some important observables
have a degenerate spectrum, or a partly or completely continuous one,
does not make the measurement problem any easier to deal with.51

For a mathematically more advanced and thereby more satisfactory and much more
complete up-to-date discussion of the measurement problem, see Busch, Lahti, and
Mittelstaedt (1996). This is also the chief subject of Jeffrey Bub's prize-winning book
(1997). For other problems of measurement in QM see Wigner (1983, pp. 275-76,
297-313).
Indeed, an observable Q with a continuous spectrum cannot be measured exactly, for
every measurement process yields rational numbers differing among themselves by
multiples of some rational £ > 0, fixed by the instrument's power of resolution. So
what one measures in fact is an observable f(Q) with discrete spectrum, where f is a
step function. For example, the typical clinical thermometer does not give us the
values of the continuous quantity temperature, but of a discontinuous function of it,
which jumps in steps of 0.1 °C. I owe this important remark to Daneri, Loinger, and
Prosperi (1962, p. 659).
 6.3 Philosophical Problems 357

Let S be a quantum-mechanical system, and let Ws denote its state

space. If Q is an observable, with proper vectors {|\|/?): i e $} forming
an orthonormal basis of 3€5, the state of S is always equal to a linear
combination 2,ie$ c,-|\|/?), such that, for each / e $, |c,|2 is the probabil-
ity that, if Q is measured on S, it sports the proper value q{ corre-
sponding to the proper vector |\|/?). Only in special circumstances will
it happen that Ck = 1 for some k e $ and c, = 0 for every other index
/, in which case the state of S is a proper vector of Q, viz., |\|/°). In the
general case, in which c{ & 0 for two or more indexes, I shall say that
the state described by E/e<* £*M°) ls a nontrivial superposition of the
states |\|/P). To measure Q on S one must make S interact - perhaps
very briefly - with an apparatus A. Let ra(Q,S,A) denote the interac-
tion by which Q is measured on S with A. It makes good sense to think
of A as a quantum-mechanical system with state space fflA, which
together with S forms a larger system S + A whose state space is Ws ®
3€A. To furnish measurements of Q the apparatus A should admit a zero
state - represented by a normalized vector |oc0) - in which it does not
display any value of Q, and a set of states {|oc/): i e $} - orthogonal
among themselves52 and with |a0) - that are such that, for each k e $,
ra(Q,S,A) brings A from state |oc0) into state |oc*) if and only if the value
of Q measured by A on S is qk.sz We view the states |a0) and {|a,): / e
3} as the proper vectors of an operator R on HtA\ since A does its job
of measuring Q only by adopting one of these states, I refer to them
as the characteristic states - or characteristic vectors - of A.
Standard discussions of the measurement problem presuppose that,
if S enters the measurement interaction in a proper state of the observ-
able Q, then it retains this state right after the measurement.54 This

52
In order that A can do its measurement job, ((tylai) must be 0 whenever ; * i; other-
wise, there would be a nonzero probability |(a,-|a/)|2 that when A is in state |a,-) it is
also in a different state |oc;-).
53
This corresponds to the ideal case in which every possible value of Q is reflected
uniquely by a state of A. Usually, of course, a distinct state |a*) of A - for example,
a distinct position of the arrow in A's dial - will correspond to a subset {|\|/?): / e $k
c $} of the proper states of Q. This is, indeed, inevitable if we drop the simplifying
assumption that Q has only a discrete spectrum.
54
See London and Bauer (1939, §11), van Fraassen (1972, p. 327), Wigner (1983, p.
281), Sudbery (1986, p. 186), Redhead (1987, p. 52), and Bub (1997, p. 52). Wigner
stresses the "highly idealized" nature of the description provided by eqns. (6.41) and
(6.43), and mentions some problems that this raises (1983, p. 284). On the other end
of the spectrum, Shimony (1963, p. 5), simply defines a proper state of an observ-
able F as "a state which is unchanged when a sufficiently careful measurement of F
358 Quantum Mechanics

excludes the common case in which the measurement procedure

destroys the object on which a measurement is performed (e.g., a
photon is converted into chemical energy upon arrival in the photo-
graphic plate that records its final position). The standard analysis does
not apply to such measurements, or, in general, to measurements that
significantly disturb the object, but only to repeatable measurements,
which, when performed twice on the same system within a very short
time, yield with certainty, on the second performance, the same result
as on the first.55
Therefore, we assume that, if S and A go into m(Q,S,A) in states
|\|/°) and |a0), respectively, the state of S + A evolves through m(Q,S,A)
from |\|/°) ® |oc0) to |\|/°) ® \ak). So, if m(Q,S,A) is a quantum-
mechanical interaction governed by Schrodinger's equation and lasting
a short time £, the apparatus A must be so contrived that
(6.41)
Now consider the general case, in which S goes into the interaction
m(Q,S,A) in a state |\|/) that is a nontrivial superposition of the |\|/P). In
other words, |\|/) = E/eI c,-|\|/?), with c{ ^ 0 for more than one index /.
The initial state of the compound system S + A is then:
| V ) ® |cco> = Xc,-| V ?) <8> loco) = 5>.-(k?> ® | a o » (6.42)
Since the unitary operator UE is linear, the state of S + A when the inter-
action ceases will be

(6.43)

is performed". If we adopt this physical definition of proper state, the usual algebraic
definition - 'a proper state of observable F represented by linear operator F is a state
represented by a vector that F maps to a multiple of itself - becomes a prescription
for the due representation of proper states. (Compare Shimony's definition with the
mathematical theorem stated in note 55.)
Note the following important mathematical result: "No continuous observable
admits a repeatable measurement." This is Corollary 8.1.1 in Busch, Lahti, and
Mittelstaedt (1996), who comment: "This result causes difficulties in our under-
standing of the operational definition of continuous observables, among them posi-
tion, momentum and energy - observables which are most important for the concept
of a particle in quantum physics" (p. 84).
 6.3 Philosophical Problems 359

To appreciate the implications of eqn. (6.43), let us put E ^ ,(||?)

at)) = ]*¥) and figure out the expectation value of the operator Q' = Q
R on I s (8) I A when S + A is in state |*F). This is given by

(6.44)

The last term on the right-hand side - viz., the double summation over
/, / G 3 (j ^ /) - conveys the interference between the several proper
states of Q'. Since the set {|\|/?) ® |a,): / G $>} is orthonormal and con-
sists of proper vectors of Q', every term in this summation vanishes.
Thus, according to eqn. (6.44), the probability that system S + A will
be found right after m(Q,S,A) in state |i|/°) ® \ak) is |c*|2.|*F°) ® \ak) is
a proper state of Q' in which S is in state |\|/£) and A is in state |a^).
The latter is the state that the apparatus reaches when it measures the
value qk of observable Q. So our result agrees with the QM prediction
that if Q is measured on S when the latter is in state |\|/) = Z ^ c,-|\|/P),
the probability of obtaining the value qk is precisely \ck\2.
Despite the pleasant consistency of these results, the transition
described in eqns. (6.42) and (6.43) has often been judged objection-
able. Equation (6.43) equates the final state of the macroscopic system
S + A with a nontrivial superposition of states; yet in real life - it is
said - macroscopic objects are, at any given time, in one definite state
or another, never in a superposition of them. To my mind, this objec-
tion reveals unreadiness to live with the quantum-mechanical concept
of physical states. If physical states are properly represented by nonzero
vectors in a Hilbert space of more than one dimension, every physical
system is under every circumstance in nontrivial superpositions of dif-
ferent sets of states (some of which sets may well consist of proper
states of some interesting observable). This follows at once from the
elementary mathematical fact that every nonzero vector in such a space
can be expressed in infinitely many different ways as a sum of two or
more noncollinear vectors. Therefore, it is not through its entangle-
ment with the - usually microscopic - observed system S that the
macroscopic apparatus A becomes "infected with superposition". If
superposition is a disease, it is one from which A suffers congenitally,
if it is in effect a quantum system.
It is true, on the other hand, that according to QM, when one mea-
sures a discrete nondegenerate observable on an individual physical
system - no matter what its size - one will always find the system in
360 Quantum Mechanics

a definite proper state of that observable and never in a nontrivial

superposition of such states. But the tension, if there is one, between
this truth and the information contained in eqns. (6.42) and (6.43)
merely reflects a certain laxity in the expression 'to find a system in a
proper state of an observable'. What one actually finds through the
measurement carried out on the system is a definite proper value of the
measured observable. Based on this value, one assigns to the system,
at the time of measurement, the corresponding proper state, which can
then be used for calculating the probability of finding this or that
proper value of the same or a different observable if it is measured on
the system soon thereafter or subsequently. In the context of QM we
must inevitably distinguish between the continuous evolution of states
and the punctual detection of values. We care for the former only
insofar as it provides the means of anticipating the probabilistic dis-
tribution of the latter; but ultimately what matters to us are the values.
And, of course, no differential law can bridge the gap between a deter-
ministic evolution and its chancy manifestations: this is precisely what
real chance consists in (see pp. 340-42).
John von Neumann, who apparently was the first to notice a
problem here, regally cut the Gordian knot. He postulated that the
interactions between quantum-mechanical systems and measurement
apparatuses are not governed by Schrodinger's equation, but instead
they constitute a different sort of quantum-mechanical change. One
must therefore distinguish between "two fundamentally different types
of interventions which can occur in a system S or in an ensemble {Si,
. . ., Sw}. First, the arbitrary changes by measurements [...]. Second,
the automatic changes which occur with passage of time" (1955, p.
351). The latter is the unitary evolution that we have considered to this
point, while the former results from "discontinuous, nondeterministic56
and instantaneously acting experiments or measurements" (1955, p.
349). By virtue of it, when a system in the state |\|/) interacts with an

56
Instead of "nondeterministic" the English translation has "non-causal". Given the
connotations of 'kausal' in German philosophical and scientific literature c. 1930, I
believe that my rendering is more accurate (see note 28 and Chapter Three, note 33;
cf. von Neumann's own discussion of "causality" in his (1955, pp. 302f.)). Two lines
later, the adverbs "continuously and causally" are applied in one breath to unitary
evolution. In the light of what I said in §3.4.3, it should be clear that 'causally' here
means 'deterministically', for causal change, in the ordinary acceptance of this expres-
sion, cannot obey a differential equation. Indeed, it is the other type of change, by
"discontinuously and instantaneously acting experiments", that better fits the com-
monsense idea of causation.
 6.3 Philosophical Problems 361

instrument designed to measure the observable Q, the vector |v|/) at once

becomes - "the wave function collapses into" - one of the proper
vectors of Q. Collapse into the proper vector \\\ft) occurs with proba-
bility |(\|/;j\|/)|2 if the spectrum of Q is discrete and nondegenerate. If the
spectrum of Q is discrete but degenerate, and there are r normalized
linearly independent vectors |\|/*i),. . ., |\|/*r) corresponding to the
proper value qh |\|/) collapses with probability SLi|(Vfe/JV)|2 ^ nto some
normalized vector in the subspace spanned by these r vectors.57 Things
get more complicated if Q has a partially or totally continuous spec-
trum. But besides such purely technical - and manageable - difficul-
ties, von Neumann's solution faces serious obstacles of another sort.
To begin with, the very suggestion that QM countenances physical
interactions that are not governed by Schrodinger's equation is dis-
concerting. If QM is any good, it should apply - within a satisfactory
margin of error - to every physical situation in which h differs signif-
icantly from 0 and the speed of light can be regarded as practically infi-
nite. If the system S accidentally interacts with the apparatus A, say, in
a laboratory junkyard, then, according to QM, S + A should undergo
a unitary evolution. It would seem therefore that von Neumann's alter-
native form of evolution, involving the "collapse" of S's latest state to
a proper state of the measured observable, can only occur if the com-
pound system somehow knows that the process it is going through is
meant to serve as a measurement. Evidently this can only happen if the
compound includes a knowing mind. Of course, measurements are con-
trived by human beings and are completed only when one of them takes
note of the results. Only they can find the value of a certain observ-
able to be this or that when it is measured on a physical system. So
perhaps one ought to include the human observer O, besides the system
S and the apparatus A, among the interacting parts of a proper mea-
surement process.
Regardless of what metaphysical materialists may think, the truth is
that there is no evidence that interactions involving O are governed by
Schrodinger's equation, nor have we the slightest inkling of what a dif-
ferential equation for the system S + A + O would be like. On the other
hand, we do know that a measurement interaction involving an
observer O should normally lead him or her to a state of awareness

57
According to the much quoted - and disputed - Liiders' Rule (Liiders 1951), in the
degenerate case \\\f) collapses with the stated probability precisely into its projection
on the subspace spanned by the proper vectors corresponding to the proper value in
question.
362 Quantum Mechanics

that reflects the finding of a particular proper value of the observable

that the interaction was supposed to measure on S. In the light of
these facts, it is not altogether unreasonable to distinguish between the
unitary evolution of S + A and the "collapse of the wave function"
brought about by the interaction between S + A and O. The unana-
lyzed concept of "collapse" nicely fits both our knowledge and our
ignorance. Indeed, if the information content of the state vector |\|/)
consists of probability distributions - one for each family of commut-
ing observables - it is not extravagant to say that, when the observer
perceives a definite proper value of a particular observable Q, |\|/) col-
lapses forthwith to a matching proper vector of Q - just as a gambler's
hopes collapse when he sees the roulette ball stop, say, at double-zero.
Still, these considerations show only that von Neumann's solution
is not so wanton as some philosophers have made it out to be, not that
it solves the measurement problem. Indeed, the transition described in
eqns. (6.42) and (6.43) has an unwanted consequence that von
Neumann's solution does nothing to explain away. It turns out that,
following the entanglement of apparatus A with system S, the predicted
state of S + A precludes the assignment of definite probabilities to the
proper values of some observables. Let T be an observable of A that
does not commute with the observable R that we considered above.
Let I denote the identity operator on 3£s. By replacing Q' with I (E) T in
eqn. (6.44) we obtain the expectation value of this operator on 3^5 (8)
%tA when S + A is in state |*F):

(6.44 *)
Due to the condition we imposed on T, the interference terms - the
summands led by the factors c% (i ^ /') - do not all vanish, so there is
no definite probability that S + A, in state *F, sports a particular proper
value of I ® T. Such a predicament is not easy to reconcile with our
experience of macroscopic objects.58

58
In my view, Schrodinger's renowned cat paradox constitutes a serious problem only
if it is made to turn around the difficulty just stated. He imagined the following "dia-
bolical device": A cat is penned up in a steel box together with a Geiger counter con-
taining "a tiny bit of radioactive substance, so small, that perhaps in the course of
 6.3 Philosophical Problems 363

It is often said that Bohr managed to sidestep this difficulty by assert-

ing that QM applies solely to microscopic objects and that the behav-
ior of macroscopic systems must be described in classical terms. Bohr's
actual views were subtler (see §6.4.1), but let me just say here why the
said assertion cannot solve the measurement problem. For one thing,
the border between "small" and "big" objects is fuzzy. What physical
laws apply in this no-man's land? Moreover, physicists have discovered
and incorporated in their research equipment many macroscopic effects
that can only be understood in quantum-physical terms (e.g., the laser).
Indeed many laboratory procedures are currently used that depend
explicitly on quantum interactions (see Braginsky and Khalili 1992).
One must therefore search for another way out of this conundrum.
Von Neumann himself was well aware of this need and pointed out the
direction to follow:
We must show that [the unitary evolution of the compound system] gives
the same result for S as the direct application of [wave function collapse]
on S. If this is successful, then we have achieved a unified way of looking
at the physical world on a quantum mechanical basis.
(Von Neumann 1955, p. 352; my notation)
Right after these lines he refers us to the last section of the book (VI.3),
but I do not see that the discussion there really proceeds in the said
direction. It rests entirely on the conception of measurement as an inter-
action of the observed system with an observing system that includes
a human consciousness, and von Neumann obviously knows no way
of incorporating the latter in a unitary evolution. However, we do not
expect physics to account for the occurrence of states of awareness -
such as the perception by someone of a pointer over a dial mark or of
a row of digits in a computer printout - but only for the behavior of
the physical objects involved in measurement. To perform a measure-
ment on a quantum system we must put it in interaction with another
system that undergoes, through the interaction, macroscopic modifica-

an hour one of the atoms decays, but also, with equal probability, perhaps none; if
it happens, the counter tube discharges and through a relay releases a hammer which
shatters a small flask of hydrocyanic acid". Schrodinger comments: "If one has left
this entire system to itself for an hour, one would say that the cat still lives // mean-
while no atom has decayed. The first atomic decay would have poisoned it. The \j/-
function of the entire system would express this by having in it the living and the
dead cat (pardon the expression) mixed or smeared out in equal parts" (1935; quoted
from Wheeler and Zurek 1983, p. 157).
364 Quantum Mechanics

tions depending on the state of the first system. Thus, the aim of QM
is "essentially to make predictions on the trace that our microscopic
body will leave in the macroscopic world, when the trace left at a
certain time is known" (Daneri, Loinger, and Prosperi 1962, p. 298).
Clearly, then, "in order to build up a satisfactory quantum theory of
measurement, it is necessary to have a theory, at least schematic, of the
large bodies, i.e. a theory which gives the connection between the
macro-properties of these bodies and their microscopic structure
described by quantum mechanics" (Ibid.). This theory ought to prove,
from the laws of quantum mechanics and the structure of macroscopic
bodies, that states which are not compatible with the actual macro-
scopic observations are, in effect, impossible. In other words, it ought
to imply that the interference terms in eqn. (6.44*) must promptly
become zero, at least for all practical purposes.
Glinther Ludwig did valuable work along these lines since the
1950s.59 The paper by Daneri et al. (1962), from which I have just
quoted, was an important step forward. Recent efforts by Gell-Mann
and Hartle, Omnes, and others turn around decoherence, "a dynami-
cal effect taking place in the bulk of matter, [which] happens so quickly
that one cannot catch it while it is acting and one can only accede most
of the time to a situation where it has already occurred" (Omnes 1994,
p. 269). The name 'decoherence' is linked to 'coherences' - which is
what some authors call the off-diagonal elements of the matrix (6.30)
(cf. Cohen-Tannoudji et al. 1977, I, 303) - and signifies the vanishing
of the coherences of some suitably chosen statistical operators as a
consequence of the said effect.60 There is still no general theory of
decoherence, but detailed studies of several models have produced
promising results. I cannot dwell on them here. The following indica-
tions may stimulate some readers to pursue this fascinating matter on
their own.
In this approach, the solution of the measurement problem is acces-

Ludwig (1953, 1954). A revised and enlarged second edition of Ludwig (1954) was
published in English in 1984/85. Together with Ludwig (1985/87) it offers a grand
synthesis of Ludwig's lifework on the foundations of QM, in four imposing volumes,
which regrettably, I must confess, I have not been able to penetrate. A less "griindlich"
but much more accessible presentation of his ideas on the subject is given in vols. 3
and 4 of his Introduction to the Foundations of Theoretical Physics (1984, 1979; in
German).
See Gell-Mann and Hartle (1993, p. 3347n.), on another related but significantly
different sense of 'decoherence'.
 6.3 Philosophical Problems 365

sory to that of another, more general problem, which can be stated

thus: Since the objects that surround us are supposed to be aggregates
of mutually interacting quantum systems, and yet they obey the laws
of classical mechanics to an excellent approximation, QM should
account for this "quasiclassical" behavior of everyday objects. In par-
ticular, it should explain why things of our size such as a chair or a
ship are always found in states that, quantum-mechanically, can only
be conceived as proper vectors of pairwise commuting observables, and
why no interference of state vectors is ever noted in such objects (except
in some specially devised superconducting systems and, of course, in
the familiar case of radiation). By solving this problem one would
obtain, as a supplementary bonus, a quantum-mechanical explanation
of the peculiar behavior of apparatuses. Gell-Mann and Hartle deal
with the problem in their paper "Classical Equations for Quantum
Systems" (1993); and Omnes deals with it in Chapter 6, "Recovering
Classical Physics", of his book (1994). Omnes begins by noting that
only a few properties are normally sufficient for describing a macro-
scopic object as such and its motion. "They are easily identified when
one actually sees a real object: they are the coordinates of the center
of mass, some orientation angles, some distances between outstanding
reference points (as, for instance, the distance between the ends of a
spring), the orientation of a wheel in a clockwork, the electric charge
on a capacity, and so on" (1994, p. 205). He calls them "the collec-
tive observables". He denotes by Hc the part of the Hamiltonian oper-
ator that depends only on the collective observables. If there is no
thermal dissipation, the total Hamiltonian H of the object - regarded
as a closed system S - is the sum of the "collective Hamiltonian" Hc
and a "microscopic Hamiltonian" He that depends only on the non-
collective or "microscopic" observables. Hc represents the mechanical
energy and He the internal energy (in the sense of thermodynamics).
However, if dissipation occurs, H = Hc + He + Him where the coupling
Hamiltonian Hint depends on both kinds of observables and expresses
the exchange between the two kinds of energy. Obviously, the coupling
"must be taken into account in order to describe the existence of fric-
tion effects in classical physics. Far less trivial is the fact that it has also
a dramatic effect upon a superposition of two macroscopically differ-
ent states, which is called decoherence. It destroys quantum interfer-
ences at a macroscopic level..." (Omnes 1994, p. 269). Decoherence
is formally derived for several special cases by constructing a complete
statistical operator p for the system under consideration and then
366 Quantum Mechanics

extracting from it a "reduced" or "collective" statistical operator pc

(by performing a partial trace of p). Under unitary evolution the off-
diagonal terms of the matrix of pc vanish very rapidly. Omnes calcu-
lates the decoherence time for a pendulum with a mass of 1 gram, a
period of 1 second, and damping time of 1 hour. If the pendulum is let
go at zero temperature in a superposition of two states with initial posi-
tions differing only by 1 micron, 1 nanosecond later the off-diagonal
matrix terms of the reduced statistical operator will be smaller than
exp(-103). Indeed the decoherence time is less the shorter the damping
time and the greater the mass, the initial temperature, the squared fre-
quency, and the squared separation between the initial positions of
the superposed states. "No effect so spontaneous, so efficient, and of
so frequent occurrence is known in the whole field of physics", says
Omnes (1994, p. 291).
The mathematical procedure that I have adumbrated only shows
that the systems to which it is applicable cannot display any interfer-
ence of proper vectors of a collective observable. Still, someone could
object that
there always exist more subtle observables that would be able to show
it. By measuring such an observable, as always was assumed to be pos-
sible in principle, one will exhibit the existence of a surviving quantum
superposition of macroscopically different states. The answer provided
by decoherence is therefore valid "for all practical purposes", because
one can only measure in practice a collective observable of a macroscopic
object, but it is not a valid answer as far as basic principles are
concerned.
(Omnes 1994, p. 305)
This objection, which was actually made by Bell (1975),61 is firmly
rejected by Omnes. He estimates that a typical measurement appara-
tus A has 1027 continuous degrees of freedom (including the noncol-
lective ones). He reckons roughly that a second apparatus B capable
of measuring an arbitrary, noncollective observable on A would require

61
Commenting on "a very elegant and rigorous paper" by K. Hepp (1972), Bell con-
ceded that certain formal results show "that any fixed observable Q will eventually
give a very poor (zero, in [the case considered]) measure of the persisting coherence".
However, "nothing forbids the use of different observables as time goes on". Thus,
"while for any given observable one can find a time for which the unwanted inter-
ference is as small as you like, for any given time one can find an observable for
which it is as big as you do not like" (1975, in Bell 1987, pp. 48-49).
 6.4 Meta-Physical Ventures 367

no less than 101018 degrees of freedom. B would therefore be too large

to fit within our horizon.62 According to Omnes this implies that such
an observable cannot be measured in principle. Philosophers will prefer
to use the words 'in principle' more strictly, but surely impossibility for
all practical purposes should be good enough for them. As Cini aptly
noted in a paper that anticipated many of the above ideas, "this situ-
ation implies a close analogy with the second law of thermodynamics,
which is valid to a very high degree of approximation in spite of its
being incompatible with the time reversibility of the equations of
motion" (1983, p. 30). We may rest assured that we shall never see a
cup of white tea spontaneously split into separate layers of milk and
tea, although this is possible in principle. Analogously, even though
every macroscopic object should teem with interferences between the
proper states of some observable or other, this state of affairs cannot
come to light if the conditions for decoherence are satisfied.

6.4 Meta-Physical Ventures

The conceptual difficulties of QM have prompted a wide variety of
purported solutions. Those suggested in §6.3 are perhaps the most eco-
nomical, namely, to suppress the EPR problem by accepting that the
idea of causation - despite its invaluable contribution to the intelligent
handling of our everyday life - is quite insufficient for the description
of natural phenomena, and to tame the measurement problem by
strictly physical means, developed as far as possible from within QM.
In this section I shall comment on a sample of other conceptions,
chosen mainly for their influence and their diversity. I collect them
under the epithet 'meta-physical' (meta - 'beyond') for they view the
meaning and scope of QM from standpoints outside empirical science.
Such is clearly the case of Bohr's epistemology (§6.4.1) and of quantum
logic (§6.4.3), if indeed this is more than a misnomer; but it is also the
case of Bohm's theory of hidden variables (§6.4.2), which is, no doubt,
a proper physical theory, but was proposed as a substitute for QM not

62
Omnes does not explain whether he refers to our so-called particle horizon or to our
event horizon. If the center of gravity of B stood within the Solar System, its outer
rim would, in the former case, be further away than the most distant galaxies from
which a light signal can - in principle - reach us now; in the latter case, the outer
rim of B would be beyond the most distant galaxies from which a light signal will
ever reach the earth.
368 Quantum Mechanics

because of any experimental results unaccountable by the latter, but on

unabashedly metaphysical grounds. As to Everett's "theory of the uni-
versal wave function" (§6.4.4), we shall see that - at any rate in Bryce
DeWitt's "many-universes" version - it makes QM into a metaphysi-
cal theory of the less reputable sort.

6.4.1 Complementarity
The most influential philosophy of quantum mechanics is the "Copen-
hagen interpretation" put forward by Niels Bohr in Como at the Volta
centennial conference of 1927, and further elucidated and elaborated
by him in numerous lectures and papers.63 The very difficulty that
philosophical readers usually find in ascertaining Bohr's exact meaning
probably made it easier for other outstanding physicists - Heisenberg,
Born, von Weizsacker - to equate their views with his. Bohr extended
his philosophy from microphysics to biology, psychology, and cultural
anthropology. In its most general version it turns around the idea that
no single coherent system of human concepts can cope with the com-
plexity of things, so that in each field of intellectual endeavor we must
resort to pairs of concepts that afford mutually inconsistent but com-
plementary perspectives (e.g., "thoughts" and "feelings", "instinct"
and "reason" - Bohr 1958, p. 27). On the face of it, this looks like the
perfect recipe for woolly thinking, and it is fondly cherished by some
as such. But in its specific application to quantum physics Bohr's con-
ception of complementarity achieved a quite definite and, for most
practicing physicists, quite convincing formulation. Thus, in his book
on Bohr, the distinguished particle physicist Abraham Pais says that
"Bohr's exegesis of the quantum theory is the best we have to date"
(1991, p. 435). According to Pais, Bohr became in effect, through his
idea of complementarity, the successor to Kant in philosophy (Ibid., p.
23). Although somewhat exaggerated, this statement contains a valu-
able suggestion, viz., that we use Bohr's tacit link with Kant to throw
light on complementarity.
Kant maintained that natural phenomena can be perceived and
described as such only if the sense impressions that disclose them are
combined and ordered under concepts, the most general of which are
contributed by human reason. Kant distinguished two kinds of ratio-

63
Many of them were collected in Bohr (1934, 1958, 1963), recently reprinted as The
Philosophical Writings of Niels Bohr (Woodbridge, CN: Ox Bow Press, 1987; 3 vols.).
 6.4 Meta-Physical Ventures 369

nal principles controlling the constitution of human experience, viz.,

the "mathematical" principles of the "axioms of intuition" and the
"anticipations of perception" (§3.3), and the "dynamical" principles
of the "analogies of experience" (§3.4). The former preside over the
determination of distances, areas, and volumes in space and of inter-
vals in time, and prescribe the continuity of all intensive magnitudes.
The latter regulate the world wide web of causal relations, subject to
major conservation principles and resting on the universal interaction
of all things. This conception of experience beautifully fits Newtonian
science, as it was shaped in the eighteenth and early nineteenth cen-
turies (§2.5.2 and 2.5.3). But Kant and his followers were apparently
convinced that our everyday experience is also organized - less pre-
cisely, perhaps, but no less consistently - after the same patterns. The
"mathematical" and "dynamical" principles of reason are supposed to
work together, and Kant does not for a moment imagine that there
could arise a conflict or incompatibility among them. Moreover,
although he emphatically asserted the role of sensation in providing
the "matter" that is to fill the structure projected by reason - so that,
for example, the gravitational constant or the boiling point of alcohol
certainly could not be ascertained otherwise than by experiment -, he
does not seem to have seriously contemplated that the "manifold of
sense" might turn out, at some point, not to fit that structure. Yet this
is precisely what, according to Bohr, has happened with the discovery
of the quantum of action.
The progress of experience has compelled physicists to acknowledge
that physical action never occurs in quantities less than Planck's con-
stant h. This implies that the interference of laboratory equipment with
the physical objects under study cannot be indefinitely reduced but
shall always have a lower bound. According to Bohr, this is the source
of Heisenberg's indeterminacy relations, by virtue of which there is a
lower bound - viz., h (more precisely, h/4n) - to the precision with
which positions and momenta can be simultaneously assigned to par-
ticles and bodies. As an illustrative example, Bohr cites the "acute dif-
ficulties" that stood in the way of a "consistent description" of the
Compton effect (see note 9 and §6.1.3):

Any arrangement suited to study the exchange of energy [£] and momen-
tum [P] between the electron and the photon must involve a latitude in
the space-time description of the interaction sufficient for the definition
of wave-number [a] and frequency [v] which enter into the relation
370 Quantum Mechanics

[E = /?v, P = ho]. Conversely, any attempt of locating the collision

between the photon and the electron more accurately would, on account
of the unavoidable interaction with the fixed scales and clocks defining
the space-time reference frame, exclude all closer account as regards the
balance of momentum and energy.
(Bohr 1949, p. 210; my italics)

The situation thus revealed has momentous consequences:

On the one hand, the definition of the state of a physical system, as ordi-
narily understood, claims the elimination of all external disturbances.
But in that case, according to the quantum postulate, any observation
will be impossible, and, above all, the concepts of space and time lose
their immediate sense. On the other hand, if in order to make observa-
tion possible we permit certain interactions with suitable agencies of
measurement, not belonging to the system, an unambiguous definition
of the state of the system is naturally no longer possible, and there can
be no question of causality in the ordinary sense of the word.
(Bohr 1928, in Bohr 1934, p. 54; my italics)

Thus, it turns out that "our usual causal space-time description" - that
is, the joint application of Kant's mathematical and dynamical princi-
ples - has hitherto been "appropriate" only because h is negligible "as
compared to the actions involved in ordinary sense perceptions" (Ibid.,
p. 55). Still, Bohr does not think that we must now scuttle the Kantian
categories and create a new framework for the description of phe-
nomena. Although we must be prepared, as our knowledge grows, "to
expect alterations in the points of view best suited for the ordering of
our experience" (1934, p. 1), and although "the modern development
of physics" is forcing on us a "thoroughgoing revision of our concep-
tual means of comparing observations" (1949, p. 239), it would be,
according to Bohr,

a misconception to believe that the difficulties of the atomic theory may

be evaded by eventually replacing the concepts of classical physics with
new conceptual forms. [. . .] The recognition of the limitation of our
forms of perception by no means implies that we can dispense with our
customary ideas or their direct verbal expressions when reducing
our sense impressions to order. No more is it likely that the fundamen-
tal concepts of the classical theories will ever become superfluous for the
description of physical experience.
(Bohr 1934, p. 16)
 6.4 Meta-Physical Ventures 371

Not only does the very recognition of the quantum of action and the
determination of its magnitude "depend on an analysis of measure-
ments based on classical concepts, but it continues to be the applica-
tion of these concepts alone that makes it possible to relate the
symbolism of the quantum theory to the data of experience" (Ibid.).
On this point, Bohr is emphatic: "However far the phenomena tran-
scend the scope of classical physical explanation, the account of all evi-
dence must be expressed in classical terms". The reason for this is
simple:

By the word "experiment" we refer to a situation where we can tell

others what we have done and what we have learned and [...], there-
fore, the account of the experimental arrangement and of the results of
the observations must be expressed in unambiguous language with suit-
able application of the terminology of classical physics.
(Bohr 1949, p. 209)

For "only with the help of classical ideas is it possible to ascribe an

unambiguous meaning to the results of observation" (1934, p. 17).
Therefore, "the space-time coordination and the claim of causality,
the union of which characterizes the classical theories" must now be
regarded "as complementary but exclusive features of the description"
(1928, in 1934, p. 54). Because it is impossible "in the field of quantum
theory" to accurately control "the reaction of the object on the mea-
suring instruments, i.e., the transfer of momentum in case of position
measurements, and the displacement in case of momentum measure-
ments," one has to renounce in each experimental arrangement to "one
or the other of two aspects of the description of physical phenomena,
the combination of which characterizes the method of classical physics,
and which therefore in this sense may be considered as complementary
of each other" (1935, p. 699).

In fact, it is only the mutual exclusion of any two experimental pro-

cedures, permitting the unambiguous definition of complementary
physical quantities, which provides room for new physical laws, the
coexistence of which might at first sight appear irreconcilable with the
basic principles of science. It is just this entirely new situation as regards
the description of physical phenomena, that the notion of complemen-
tarity aims at characterizing.
(Bohr 1935, p. 700)
372 Quantum Mechanics

According to Bohr, "the quantum-mechanical formalism" precisely

offers "an adequate tool" for this kind of description, inasmuch as it
is "a purely symbolic scheme permitting only predictions, on lines of
the correspondence principle, as to results obtainable under conditions
specified by means of classical concepts" (1949, pp. 210f.). But "there
can be no question of any unambiguous interpretation of the symbols
of quantum mechanics other than that embodied in the well-known
rules which allow to predict the results to be obtained by a given
experimental arrangement described in a totally classical way"
(1935, p. 701). "The appropriate physical interpretation of the sym-
bolic quantum-mechanical formalism amounts only to predictions, of
determinate or statistical character, pertaining to individual phenom-
ena64 appearing under conditions defined by classical physical con-
cepts" (1949, p. 238).65
Bohr's approach entails, of course, that the unitary evolution of state
vectors in Hilbert space does not represent a process in the real world,
but is merely a computational link between two sets of data described
in classical terms, viz., those concerning the preparation of an experi-
ment and those concerning its results. This view has offended the meta-
physical conscience of most philosophers and some physicists. In
particular, the utilization of QM in astrophysics and cosmology seems
questionable or at any rate funny if the talk about vectors and linear
operators in Hilbert space is meaningful only as it relates to experi-
mental arrangements describable by classical concepts that do not fit
the conditions in the early universe and in the interior of stars. This
has led some to "interpret" unitary evolution as more than just an algo-
rithmic bridge between data input and data output (§6.4.4), and others
to seek for a proper physico-mathematical description of the real
processes that supposedly underlie the statistical correlations correctly
predicted by QM (§6.4.2). Their qualms are not altogether unjustified
for, although every physical theory obtains its meaning and truth

64
Note that Bohr advocates "the application of the word phenomenon exclusively to
refer to the observations obtained under specified circumstances, including an account
of the whole experimental arrangement" (1949, pp. 237f.).
65
In other words, "quantum mechanics speaks neither of particles the positions and
velocities of which exist but cannot be accurately observed, nor of particles with
indefinite positions and velocities. Rather, it speaks of experimental arrangements in
the description of which the expressions 'position of a particle' and 'velocity of a par-
ticle' can never be employed simultaneously" (Frank 1935; quoted in Jammer 1974,
p. 200).
 6.4 Meta-Physical Ventures 373

through those few points where it comes in contact with human expe-
rience, the theories of Newton and Einstein provided a conceptual
framework encompassing both this experience and the natural world
beyond it, while QM, as seen from Copenhagen, depends on the admit-
tedly insufficient classical framework for describing macroscopic events
in the laboratory and has nothing to say about anything else.
How do the philosophical problems of §6.3 fare under Bohr's
approach? He had little trouble dispelling the EPR paradox (Bohr
1935), but that is no great feat if, as Bohr himself implied, the EPR
argument is fallacious. At first blush it might seem that Bohr also dis-
solves the measurement problem. As David Bohm pointedly notes,
"according to Bohr's interpretation nothing is measured in the
quantum domain [. . .] because all 'unambiguous' concepts that could
be used to describe, define, and think about the meaning of the results
of such a measurement belong to the classical domain only" (1980, p.
75). On a closer look, however, I find that the measurement problem
remains very much alive. The interference terms in eqn. (6.44*) are
predicted by the quantum-mechanical formalism, so they ought to
show up somehow in laboratory results. Bohr owes us an explanation
of why this does not happen.

6.4.2 Hidden Variables

In the "final remarks" of the paper in which he established the prob-
abilistic interpretation of Schrodinger's \|/-function (§6.2.4), Born says
that anyone who will not rest content with indeterminism is naturally
free "to assume that there are additional parameters, not yet introduced
in the theory, which determine the individual event" (1926b, p. 825).
In a lecture delivered at Oxford 20 days after submitting that paper,
Born compared the use of probability in QM and in the classical theory
of gases (§4.3.3). The latter conceives a gas as a collection of mole-
cules whose positions and momenta are determined at every instant by
their initial values and their evolution according to the differential
equations of classical mechanics (§2.5.3). The theory, however,
works with probabilistic assumptions leading to statistical predictions,
because it is impossible to know the exact positions and momenta of
the molecules at any given time. Thus "the classical theory introduces
the microscopic coordinates which determine the individual process,
only to eliminate them because of ignorance by averaging over their
values; whereas the new theory gets the same results without intro-
374 Quantum Mechanics

ducing them at all. Of course, it is not forbidden to believe in the exis-

tence of these coordinates; but they will only be of physical significance
when methods have been devised for their experimental observation"
(1927b, in Born 1969, p. 10). Some physicists, however, have exercised
the freedom allowed by Born, despite the total absence of such exper-
imental methods. The "hidden variables" - or "hidden parameters" -
approach to QM consists in postulating the existence of hitherto unob-
served and presumably unobservable physical quantities whose evolu-
tion under suitably designed laws exactly determines the outcome of
the individual quantum processes.66 In 70 years, theories countenanc-
ing such hidden variables have not led to the discovery of a single new
physical effect. Indeed, their partisans are content if the consequences
of their hypotheses agree with the predictions of QM. I shall refer
to hidden variables theories that satisfy this requirement as 'HV-
extensions of QM'.
In his book of 1932 (IV, §§1-2), von Neumann argued that no HV-
extension of QM is possible. He discussed the following situation,
which is characteristic of QM: The measurement of a particular quan-
tity on an ensemble of systems is found to yield different values,
although every system is in the same quantum-mechanical state. There
are two explanations for this: Either (a) the systems are in different
states, which QM is incapable of distinguishing, so it represents them
all by the same \|/-function; or (b) the systems are really in the same
state, and the dispersion of measured values is due not to our igno-
rance and to the grossness of the quantum-mechanical representation,
but to "nature itself, which has disregarded the 'principle of sufficient
cause'" (von Neumann 1955, p. 302). Alternative {a) would imply that
our ensemble comprises as many subensembles as there are different
results of the said measurement, and that every system in any one
subensemble is in a particular dispersion-free state, characterized by a
distinct value of a list of "hidden variables" unknown to QM. Von
Neumann proved from assumptions he considered plausible that,
under QM, no dispersion-free states are possible. The main assump-
tion is that, if the quantities A, B,. . ., K are represented by operators
A, B,. . ., K, and a, & , . . . , k are any real numbers, then (i) the quan-
tity a A + bB + . . . + kK is represented by the operator a A +foB+ . . .

66
See the massive survey by Belinfante (1973). It is perhaps worth noting that, with the
exception of Louis de Broglie (1956), Nobel prize winners have stood aloof of such
efforts.
 6.4 Meta-Physical Ventures 375

+ &K, and (ii) its expectation value (aA + bB + . . . + &K)|V> in an ensem-

ble prepared in a given state |\|/) is equal to a(A)\^ + b(B)\^ +. . . +
k(K)\y). Now, this assumption holds, without question, for quantum-
mechanical states, but there is no reason why it should also apply to
the dispersion-free states foreign to QM that von Neumann is arguing
about, and to the "hidden variables" that would specify them. There-
fore, von Neumann's formal proof does not justify his informal con-
clusion that "the present system of quantum mechanics would have to
be objectively false, in order that another description of the elementary
processes than the statistical one be possible" (von Neumann 1955,
p. 325; see Bell's discussion in his 1987, pp. 4-5).
Although von Neumann's argument is unable to prove that QM
precludes the existence of well-defined, deterministically evolving
"hidden variables", other mathematical theorems discovered by
Gleason (1957), Bell (1964), and Kochen and Specker (1967) impose
severe conditions on HV-extensions of QM.67 However, the theory put
forward by David Bohm (1952) meets those conditions and makes the
same predictions as QM. To get a feeling of how it works, let us take
a look at Bohm's treatment of a system consisting of a single particle.
Schrodinger's equation for one particle of mass m moving in a classi-
cal potential V is eqn. (ESI*) of note 23, which Bohm writes thus:

ifi^- = - — V > + V(x)\|/ (6.45)

at 2m
\|/ is a complex-valued function on three-dimensional Euclidean space.
It can be expressed in terms of two real-valued functions, R and S, on
the same space:
y = Rexp(/S/ft) (6.46)
By substituting eqn. (6.46) in (6.45) and separating real and imaginary
parts we obtain the equations:

67
To agree with QM, a hidden variables theory must be "nonlocal" and "contextual",
in the senses indicated in notes 68 and 69. The theorems of Gleason, and of Kochen
and Specker are often mentioned and discussed in philosophical literature, and they
deserve more than a bare mention here. However, to state them intelligibly I would
have to fill a few pages with mathematical definitions. I refer the curious reader to
Redhead (1987) (who devotes pp. 119-52 to "the Kochen-Specker paradox") and
Hughes (1989a) (who reprints a very readable proof of Gleason's theorem on pp.
321-46). As to Bell's theorem, it is none other than the famous inequality that was
mentioned in §6.3.1; for references, see note 46.
376 Quantum Mechanics

dR _ 1
dt ~ 1m
dS
v
dt [ 2 m ' 2m R
Writing P for R2, eqns. (6.47) become

—1 = 0 (6.48)
m )

d* 2ra 4ra^ P 2 P2
Now, if A = 0, eqn. (6.49) reduces to eqn. (2.50), the classical Hamil-
ton-Jacobi equation for our system. Then, as Bohm recalls, given an
ensemble of particles whose trajectories are solutions of the classical
equations of motion, every one of which is normal to a given surface
S = const., every surface of constant S is normal to every trajectory and
VS(x)/ra equals the velocity vector v(x) of any particle passing the point
x. Equation (6.48) can therefore be rewritten thus:

^ + V(Pv) = 0 (6.48a)
at
Pv may then be regarded as the mean current of particles in the en-
semble and P(x) as the probability density, so eqn. (6.48a) expresses the
conservation of probability. Bohm's decisive move was to extend this
interpretation to the case fi > 0. He did it by assuming that, besides the
classical potential V, there is a quantum potential U acting on the par-
ticles that is given by the last term on the left-hand side of eqn. (6.49):

"The Eq. [6.49] can still be regarded as the Hamilton-Jacobi equation

for our ensemble of particles, VS(x)/m can still be regarded as the par-
ticle velocity, and Eq. [6.48] can still be regarded as describing con-
servation of probability in our ensemble. Thus, it would seem that we
have here the nucleus of an alternative interpretation for Schroedinger's
equation" (Bohm 1952, p. 170). In this interpretation, particles are
endowed with "precisely definable and continuously varying values of
position and momentum"; they move "under the action of a force
which is not entirely derivable from the classical potential, V(x), but
 6.4 Meta-Physical Ventures 377

which also obtains a contribution from the 'quantum-mechanical'

potential, U(x)"; since the latter is a function of R(x), which is the
modulus of the wave function \|/(x) - see eqn. (6.46) - "we have effec-
tively been led to regard the wave function of an individual [particle]
as a mathematical representation of an objectively real field" that acts
on the particle in a way that is similar to, although not identical with,
the way the electromagnetic field acts on a charge (p. 170).
Bohm lists three assumptions under which his theory yields the same
predictions as QM, viz., (1) the \|/-field satisfies Schrodinger's equation;
(2) the particle's momentum equals VS(x); and (3) we do not predict
or control the precise location of the particle but have, in practice, a
statistical ensemble with probability density P(x) = |\|/(x)p (1952, p.
171). He emphasizes, however, that these assumptions do not belong
to the conceptual structure of his theory and may therefore be relaxed.
He takes pride in the fact that "there are an infinite number of ways
of modifying the mathematical form of the theory that are consistent
with our interpretation and not with the usual interpretation" (p. 179).
In particular, by modifying the differential equation for \|/ - say by
making it inhomogeneous or by including nonlinear terms that are
large only for processes involving small distances - one could perhaps
account for "phenomena associated with distances of the order of
10~13cm or less, which are not now adequately understood in terms of
the existing theory" (p. 179). Such modifications would generate dis-
crepancies between the predictions of QM and of Bohm's theory,
which, if testable, would facilitate a choice between them. Still, the
enormous flexibility of the latter can hardly count as an asset in the
actual practice of physics.
Bohm (1952) extends his approach to the many-body case and
presents in Part II a theory of measurement. According to it, "the at
present 'hidden' precisely definable particle positions and momenta
determine the results of each individual measurement process, but in a
way whose precise details are so complicated and uncontrollable, and
so little known, that one must for all practical purposes restrict oneself
to a statistical description of the connection between the values of these
variables and the directly observable results of measurements. Thus,
we are unable at present to obtain direct experimental evidence for the
existence of precisely definable particle positions and momenta" (p.
183; my italics). This reminds one of the Cartesian adversaries of
Newton, who, to avoid gravity's "magical" action-at-a-distance, tried
to explain planetary motions by means of fantastic hypotheses. On the
378 Quantum Mechanics

other hand, Bohm's new force is in a way more "magical" than

Newton's. The quantum potential U is not linked to any known source
and, in the w-body case, it is defined, of course, on 3w-dimensional
space. Although logically consistent, these notions "are difficult to
understand from a physical point of view" and should be regarded "as
schematic or preliminary representations of certain features of some
more plausible physical ideas to be obtained later" (Bohm 1980, pp.
80f.). Moreover, "the 'quantum-mechanical' forces may be said to
transmit uncontrollable disturbances instantaneously from one parti-
cle to another through the medium of the \|/-field", as Bohm notes in
connection with the EPR experiment (1952, p. 186).68 As a conse-
quence of the mutual entanglement of all interacting particles through
the \|/-field, the results of measurement depend on context: Measure-
ment of the same observable A on two systems prepared in the same
way may lead to different results if A is measured on one of them
together with an observable B and on the other together with an
observable C, and B and C are noncommuting.69 Therefore, "the mea-
surement of an 'observable' is not really a measurement of any physi-
cal property belonging to the observed system alone. Instead, the value
of an 'observable' measures only an incompletely predictable and con-
trollable potentiality belonging just as much to the measuring appara-
tus as to the observed system itself" (1952, p. 183).70

6.4.3 Quantum Logic

Since Lobachevsky and Bolyai achieved posthumous fame by breaking
the Euclidian monopoly of truth in geometry, some philosophers have
dreamed of going one step further and shattering the uniqueness of
logic. The first nonstandard logic was formulated - according to
Jammer (1974, p. 342) - in 1910 by N.A. Vasil'ev, a professor of phi-
losophy at Kazan University. Vasil'ev's "imaginary logic" - as he called
it to draw attention to the analogy with Lobachevsky's "imaginary
geometry" - was based on the denial of the so-called law of bivalence,
by virtue of which every statement must have one and one only of the

68
Bohm's theory is "nonlocal"; see note 67.
69
Bohm's theory is "contextual"; see note 67.
70
Bohm's last thoughts were published, after his death, in Bohm and Hiley (1993). I
have not been able to find in this book any important idea that was not already
present in Bohm's brilliant paper of 1952.
 6.4 Meta-Physical Ventures 379

two admissible truth values true and false, Vasil'ev added a third value,
indifferent, and, denying the law of contradiction, concluded that a
sentence of the form 'S is P and S is not P' is neither true nor false, but
indifferent. A more plausible three-valued logic was proposed by Jan
Lukasiewicz (1920), who saw in it the proper way of formally dealing
with contingent statements about the future, such as 'A sea-battle will
be fought tomorrow' (cf. Aristotle, De Interpretation, Chapter 9). The
following are his truth tables for negation (—•) and material implication
(->), with the three truth values by 1, 0 and y-.71

1 1 0 p 1 l 1 21 21 21 0 0 0
q 1 21 0 1 "J 0 1 "2" 0
1 1
p —^ *? 1 "2~ 0 1 2 2 1 1 1

In 1931, Zygmunt Zawirski recommended using a three-valued logic

to cope with the baffling features of QM, such as the apparent impos-
sibility of simultaneously assigning a precise position and a precise
momentum to a particle. This suggestion was later taken up by Paulette
Fevrier (1937a, 1937b, 1951) and by Hans Reichenbach (1944), who
developed it in different ways. Their systems found a cool reception
and I shall not comment on them.72
Quantum logic, as it is usually understood today, stems from a joint
paper by Garrett Birkhoff and John von Neumann on "The Logic of
Quantum Mechanics" (1936). They saw that the subspaces of a Hilbert
space W form a lattice that I shall call L(W).73 This is similar to but sig-

71
Let me recall that in standard logic, with two truth-values, T ('true') and F ('false'),
the truth table of material implication is

T F
F T

T F T T
The importance of material implication in standard logic lies in its close connection
with entailment: Given (p —> q), p entails q; that p entails q entails (p —> q).
Jammer (1974, pp. 361-79) discusses the systems of multivalued quantum logic of
Fevrier, Reichenbach, and von Weizsacker and the chief objections leveled against
them. See also Putnam (1957), Feyerabend (1958), and Haack (1974, Ch. 8).
Supplement II contains all the information about lattices that is required to under-
stand the remainder of this section.
380 Quantum Mechanics

nificantly different from the Boolean lattice of subsets of the phase

space of a classical physical system, which in turn is isomorphic to the
lattice of propositions asserting or denying that the state of the system
lies in this or that part of the phase space. Birkhoff and von Neumann
noted that the propositions concerning the state of a quantum system
with Hilbert space 'K formed a lattice isomorphic to L(^C). Now, the
Boolean structure of the lattice of propositions concerning the state of
a classical system reflects the meaning of the classical logical operators
'and', 'or', and 'not'. So, by analogy, the non-Boolean structure of the
newly built lattice of propositions about quantum states in 5^ was said
to reflect - or to constitute - a nonclassical logic appropriate for think-
ing about micro-objects. The novelty of the logic could then explain
away some of the conceptual perplexities associated with QM.
To see what this means, let us consider the lattices involved, begin-
ning with the Boolean lattice of propositions in classical logic. (For
symbolism and terminology, see Supplement II.) By proposition I mean
a class of logically equivalent sentences. I say that the proposition p is
true if any - and therefore every - sentence in p is true, and that p is
false if any sentence in p is false. The denial of /?, symbolized by —ip,
is the proposition that is true if and only if p is false, and is false if and
only if p is true (law of bivalence). The lattice of propositions is defined
by: (i) For any propositions p and q,p < q'\i and only p entails q, that
is, if no assignment of truth values is possible in which p is true and q
is false; (ii) the maximal element 1 is the class of tautologies, that is,
the sentences that are true in every possible assignment of truth values;
(iii) the maximal element 0 is the class of contradictions, that is, the
sentences that are false in every possible assignment of truth values;
and (iv) the orthocomplement p' = —<p. It follows at once from condi-
tion (i) that the meet of propositions p and g, denoted in lattice theory
by '/? A q\ is the proposition that is true if and only if both p and q
art true and otherwise is false; and that the join of p and q, denoted
by 'p v q\ is the proposition that is false if and only if both p and q
are false and otherwise is true. (Due to this, the same symbols A and
v employed in lattice theory to signify, respectively, meet and join, are
used for conjunction and disjunction in standard logic.) Let (p <-> q)
stand for the proposition that is true if and only if p and q are either
both true or both false. It is easily proved by means of truth tables that
b o t h (p v (q A r)) <-> ((/? v q) A (p v r)) a n d (p A (q v r)) <-> ((p A q)
v (p A r)) are tautologies, so the distributive laws Dv/A and DA/v hold in
the classical lattice of propositions, which thus is a Boolean lattice.
 6.4 Meta-Physical Ventures 381

The Boolean lattice B(L) of subsets of a set X is defined in Supple-

ment II.4. B(L) is defined on the power set 2P(S) and is partially ordered
by the relation of inclusion c. Note in particular that the meet of two
subsets P and Q is their intersection P n Q and the join of P and Q
is their union P u Q. We are interested in the case in which £ is the
phase space of a classical system S, which, if S has « degrees of freedom,
is U2n. I say that a proposition is a state description of S if it contains
a sentence of the form 'the state of S is in P', for some P c R2". Let
B(S) be the classical Boolean lattice of such propositions. Let p be the
equivalence class of 'the state of S is in P' (likewise for q, etc.). Since
'the state of S is in P' is not logically equivalent to 'the state of S is in
Q' unless P = Q, the mapping fip^Pisa one-one mapping of B(S)
2n
onto B(R ). This mapping is an isomorphism of lattices, as the fol-
lowing considerations will show. P c Q if and only if every conceiv-
able state of S that lies in P also lies in Q; this is necessary and sufficient
for p to entail q; thus, p < q in B(S) if and only if /"(/?) < f(q) in B(R2n).
Moreover, 'the state of S is in U2m is true under every circumstance and
therefore belongs to the maximal element of B(S), and 'the state of S
is in 0 ' in not true under any circumstance and therefore belongs to
the minimal element of B(S). Finally, 'the state of S is not in P' is log-
ically equivalent to 'the state of S is in P", where Pf denotes the com-
plement of P in B{Uln), so f(p') = /H>) = F = (f(p)Y.74
I now turn to the lattice L(9€) formed by the subspaces of the Hilbert
space lH. Its maximal element is 3^, its minimal element is the zero
vector, the meet A A B of two subspaces A and B is their intersection
A n B, and their join A v £ is the subspace A 0 B spanned by them
(|\|/) is in A © B if |\|/) is a linear combination of vectors in A and vectors
in B). If A is a subspace of $£, there is a subspace A' formed by all the
vectors that are orthogonal to A (that is, orthogonal to each vector in

To simplify my exposition I have adopted above a rather unrealistic form of state

description for classical systems. In real life, neither the data obtained by observation
nor the predictions achieved by solving the classical equations of motion for a system
of more than two bodies enable us to locate the state of a system at a point of phase
space. In the actual practice of physics, especially in classical statistical mechanics,
two measurable subsets of phase space are equated when their symmetric difference
- i.e., the set of points belonging to one or the other but not to both - has measure
zero. Anyway, the collection of (classes of purportedly equal) subsets obtained in this
way is again a Boolean lattice, which is isomorphic to the classical lattice of state
descriptions referred to them. Thus, for our purposes, a less simplistic presentation
would not make any difference.
382 Quantum Mechanics

A). A' satisfies conditions OC1-OC3 in Supplement II.3 and is there-

fore the orthocomplement of A. So L(H€) is an orthocomplemented
lattice, but it is not Boolean, as the following example shows. Let |cp)
and |\|/) be orthogonal vectors in a Hilbert space Hi and put |0) = a\q>)
+fc|\|/),where the scalars a and b are both different from 0. Let A<p, Av,
and A0 be the subspaces of Hi spanned, respectively, by |cp), |\|/), and |0).
Then, clearly, A0 A (A9 V AV) = Ae * 0, but (Ae A AJ v (Ae A A¥) = 0
v 0 = 0. Thus, L(H€) does not obey the distributive law DA/v. The struc-
ture of L(Hi) can be reproduced in an obvious way in the set of pro-
jectors of HI, that is, the linear operators that map the whole of Hi onto
each of its subspaces (see §6.2.5). Let PA denote the projector onto sub-
space A. Put PA v PB = PAvB, P A A P B = PAAB, (PA)'=PA>. These condi-
tions define a lattice isomorphic to L(Hi) with minimal element Po and
maximal element P^.
The very idea of a quantum logic depends on the construction of a
lattice that
(LI) contains all the propositions suitable for conveying information
- data, predictions - about a quantum system S,
(L2) is partially ordered by entailment - or at least by something that
can pass for entailment if one makes allowance for a change in
logic - and
(L3) is isomorphic with the lattice L(His) of subspaces of the system's
Hilbert space Hts.
By virtue of (L3) the lattice in question is non-Boolean and therefore
essentially different from the classical lattice of propositions. Without
(L2) there could hardly be any reason for regarding the quantum lattice
of propositions as a logical system. Birkhoff and von Neumann's pro-
posal for a lattice meeting these three requirements is somewhat less
perspicuous than one might wish. G. W. Mackey (1963), who brought
great clarity and rigor into this subject, proceeded in the opposite direc-
tion: He constructed a partially ordered structure that satisfies (LI) and
(L2) and then postulated that it is isomorphic with the lattice of sub-
spaces of a Hilbert space. From this standpoint, the lattice of proposi-
tions concerning a quantum system appears as a structured expression
of experience that is more basic and hence, presumably, more lasting
than the Hilbert space of standard QM. Unfortunately, it is not pos-
sible to deal here with Mackey's work. Anyway, his followers are not
in the business of revolutionizing logic; they explore an alternative
mathematical structure for the formulation and eventual advancement
 6.4 Meta-Physical Ventures 383

of QM, which they call "quantum logic" mainly in deference to Birk-

hoff and von Neumann (cf. Jauch 1968, p. 77; Beltrametti and Casinelli
1981, p. xxiii).
A simpler scheme was adopted by Putnam (1969, 1974), with
the explicit aim of revising logic proper - comprising, I presume,
the general theory of deductive inference - in the light of physical
experience.
Suppose we are willing to adopt the heroic course of changing our logic.
What then? It turns out that there is a very natural way of doing this.
Namely, just read the logic off from the Hilbert space %. Two proposi-
tions are to be considered equivalent just in case they are mapped onto
the same subspace of 2CS, and a proposition p is to be considered as
'implying' a proposition q just in case Ap is a subspace of Aq.
(Putnam 1979, p. 179; my notation)

Putnam's intent can, I think, be spelled out as follows: Let A, B,. . .

denote the subspaces of the Hilbert space 9€5, and let a, b,. . . be the
propositions whose informative content is that the state vector of
system S lies in A, B,. . ., respectively. Such propositions form a lattice
L(S) if meet, join, and orthocomplement are defined thus:
a A b is the proposition that the state of S is in A n B;
a v b is the proposition that the state of S is in A 0 B;
a' is the proposition that the state of S is orthogonal to A.
The mapping a •-» A is an isomorphism of L(S) onto L($£5), so L(S)
fulfills requirement (L3). It also satisfies (L2): Partial order in L(S) con-
notes a form of entailment, for a < b is tantamount to 'the state of S
is in A only if it is in B and the state of S is orthogonal to B only if it
is orthogonal to A\ Does it also meet (LI)? Do the elements of L(S)
represent, as Putnam says, "all possible physical propositions about S"
(1979, p. 179)? In standard QM, information about a system S has to
do chiefly with the probability that this or that physical quantity rep-
resented by a self-adjoint operator in 3€5 sports a certain value - or a
value in a certain range - if it is measured on S when S is prepared in
such-and-such a state. Surely, such information is not conveyed by
propositions that do no more than report with certainty to what sub-
spaces of 3^5 the state of S belongs or is orthogonal. But in Putnam's
view, "probability enters in quantum mechanics just as it entered in
classical physics, via considering large populations [and] whatever
problems may remain in the analysis of probability, they have nothing
384 Quantum Mechanics

special to do with quantum mechanics" (1979, p. 186; see the expla-

nation preceding this conclusion). So the primary - and proper -
quantum-mechanical information about S concerns only matters that
can be asserted with certainty.
An example will show how this works. Let P and Q be two non-
degenerate observables on %£s that do not commute with each other.
Let each one have only two linearly independent proper vectors. For-
getting all other observables, we view 3ts as two-dimensional. Put P|\|/,)
= pi\yt) and Q|cpz> = q^) (i = 1,2). I denote by \|/ the subspace spanned
by the vector |\|/) and by \\f(t) the proposition that S is in \\f at time t.
Clearly, $£5 = \\f1 0 \|/2 = (pi 0 cp2; therefore, we may assert with cer-
tainty that (\\f\(t) v \\f2(t)) A (<fi(t) v cp2(*)) at all times. If, at time tm9 Q
is measured on S with result qu this certifies that ((\|/i(£w) v \|/2(£m)) A
(pi(£m). But, of course, this does not entail that (i|/i(£m) A q>i(tm)) v (\|/2(£w)
A cpi(£w)), for the lattice of quantum propositions is not Boolean. Thus,
from the fact disclosed by measurement, that the state of S at tm was
in subspace q>i and, of course, trivially, also in \\f1 0 y 2 , it does not
follow that the said state was then either in \|/a n q^ or in \|/2 n q>i. This
would indeed be absurd, for \\ft n q>i = {0}.
Putnam insisted that the logical operators symbolized by v, A, and
—i (or ') have essentially the same meaning in the new and the old logic.
This is surprising, for the classical operators are truth-functional - the
truth value of any proposition formed by using one of these operators
follows, by definition, from the truth value of the proposition or pro-
positions to which the said operator is applied - but the quantum oper-
ators are not.75 But Putnam surely had in mind that the meaning of the
logical operators can be gathered - after Gentzen (1934) and Jaskowski
(1934) - from the rules that govern their use in inference. For, if entail-
ment is represented in each logic by partial order in the respective
lattice, then clearly the following laws of entailment hold in both: (i)
p entails p v q, and q entails p v q; (ii) if p entails r and q entails r,
then p v q entails r; (iii) p and q jointly entail p A q; (iv) p A q entails
both p and q; and (v) (p')' entails and is entailed by p. In classical logic,
(i)-(v) translate into sound rules of inference - for example, 'given /?,
infer p v q" - sufficient to establish every logical truth. However, if the
same rules of inference were applicable in quantum logic, the distrib-

75
No truth-functional characterization of meet, join, and orthocomplement can be
given in a lattice isomorphic to L(W) if the Hilbert space ^t has more than two dimen-
sions (Kochen and Specker 1967; for a related result, see Jauch and Piron 1963).
 6.4 Meta-Physical Ventures 385

utive laws Dv/A and DA/v could without more ado be established by
means of them; therefore, some important change must have affected
the very meaning of entailment. Moreover, despite (v), there must be
some significant difference in the meaning of negation represented by
orthocomplementation in both B(S) and L(S), for in a Boolean lattice
orthocomplementation is fixed by the partial order, whereas in a non-
Boolean lattice it must be introduced on its own (Bell and Hallett 1982;
cf. Supplement II.4).
For Putnam the main attraction of quantum logic was the hope it
offered of preserving the purportedly realist commonsense view of
physical properties. For common sense, each thing is characterized by
a series of properties - shape, color, weight, and so on - that change
with the circumstances but are supposed to be there, fully determined
at all times while the thing exists. Philosophy and physics have under-
mined this view by conceiving most - if not all - properties of a thing
C as the manifestation of relations that it has with other things, or with
surrounding fields, and which must vanish with the latter, even if C
remains. Thus, in Newtonian physics, a particle can be meaningfully
said to possess weight only in the presence of gravitational sources.
Mach and Einstein formed a relational conception of inertia, a quan-
tity that Newton still conceived as intrinsic to each thing. Even so, since
some relations are never lacking, classical physics could apparently
retain the core of commonsense "realism", by conceiving things
through the properties that correspond to such universal relations and
treating the others as derivative or even as subjective. But according to
QM it is not possible to assign at the same time definite values to
certain pairs of physical quantities, including the two - admittedly rela-
tional but universal - properties of position and momentum, which in
classical mechanics jointly characterize the state of each thing. What
shall we make of a particle that has a definite velocity but is not located
at any particular place? Or one that occupies a definite position but
neither rests there nor moves with a definite velocity? Should we believe
that physical properties are had only insofar as they are observed}
Putnam (1969) thought that quantum logic would rescue us from this
undesirable conclusion.
To facilitate his exposition, he invited the reader to "pretend that all
physical magnitudes have finitely many values, instead of continuously
many" (1979, p. 178). Suppose, then, that for our system S the admis-
sible values of position and momentum are, respectively, pi,. . ., pm and
<7i,..., qn. Let Pi (Qi) denote the subspace of Ws spanned by the proper
386 Quantum Mechanics

vectors of position (momentum) corresponding to pt (qt). Write pi(t) and

qi(t) for 'S has position pt at time f and 'S has momentum q{ at time t\
respectively. Then, if it is verified that S has position pk at time t, we
may confidently assert that pk(t) A (qi(t) v . . . v qn(t)) but certainly not
that (pk(t) A q\(t)) v . . . v (/?&(£) A #„(£)), for the distributive laws are not
valid in quantum logic. Putnam (1974) informally introduces the exis-
tential quantifier (by using which, indeed, one could drop the pretense
that physical quantities have only finitely many values). With it, (q\(t)
v . . . v qn(i)) is shortened to 3xqx(t), which may be read as "there is a
definite value of momentum that S has at t." Interestingly, in quantum
logic, 3xqx(t), 3xpx(t), and hence 3xqx(t) A 3xpx(t) are logically true, for
?! v . . . v Pm = Qj v . . . v Qn - 3^s. However, from 3xqx(t) A 3xpx(t)
one may not infer 3x3z(qx(t) A pz(t)). Moreover, the last expression is
logically false, since, for every pair of indices / and k, Q; A Pk = 0. Thus,
thanks to quantum logic we can rest assured that, at any time £, there
exists a definite momentum possessed by S at t and there exists a
definite position possessed by S at £, and yet assent to the quantum-
mechanical theorem that S does not possess a definite position and a
definite momentum at the same time t. I wonder how many common-
sense realists will feel that their craving for determinate properties is
relieved by this result. Like other Putnam efforts to save this or that
face of realism, his quantum logic is a tantalizing jeu d'esprit.76

76
Hilary Putnam acknowledges his debt to David Finkelstein (1962/63, 1969; see also
Finkelstein 1973) for some of the above ideas on quantum logic. For extensive,
enlightening criticism of both authors, see Stachel (1986). In the sixties and seven-
ties, Peter Mittelstaedt sought to vindicate quantum logic in the context of the "dia-
logical logic" of Lorenzen and Lorenz (cf. their book of 1978). In this approach to
the foundations of logic, an implication is proved logically true through a schematic
dialogue in which someone proposes it, an opponent proves the antecedent yet ques-
tions the consequent, and the proponent succeeds in showing that the consequent has
already been granted in the course of the opponent's argument. (For example,
someone proposes p —> (q —> p); the opponent asserts p, whereby the proponent is
constrained to assert (q —> p); the opponent asserts q and questions p, only to be told
that p has been already asserted by him.) In Mittelstaedt's quantum-logical dialogues
the proponent is often unable to turn the opponent's own words against him, because
they are no longer "available" when the proponent requires them. For example, if
the opponent has invoked at some stage of the dialogue a definite position value, the
proponent cannot quote him later if a definite value of momentum has in the mean-
time been asserted of the same system. Because of this, implications that have been
dialogically proved under more familiar conditions - such as the two-way implica-
tions in the distributive laws - cannot be established in a quantum-logical setting. See
Mittelstaedt (1978, 1979).
 6.4 Meta-Physical Ventures 387

6.4.4 "Many Worlds"

My last example of meta-quantum-physics is Everett's solution of the
measurement problem. It was the subject of his doctoral thesis (1957),
written under the supervision of J. A. Wheeler. He developed it at
greater length in a manuscript that was not published until 1973. I
shall follow both. According to Everett, his approach provides "a refor-
mulation of quantum theory in a form believed suitable for applica-
tion to general relativity" (1957, p. 454). He believes that it will enable
one to "consider the state function of the whole universe", from which
"all of physics is presumed to follow" (1973, p. 9). (Pity he does not
deign to offer us any quantum-mechanical predictions regarding "the
whole universe", a system on which indeed no quantum-mechanical
observable can be measured.)
Everett recalls the two fundamentally different ways in which
quantum states can change according to von Neumann (see §6.3.2),
viz., "Process 1: The discontinuous change brought about by the obser-
vation of a quantity with eigenstates |(|)i), l^),. . ., in which the state
|\|/) will be changed to the state |(j)7) with probability |(\|/|<|>;)|2"; and
"Process 2: The continuous, deterministic change of state of the iso-
lated system with time according to a wave equation d\\\f)/dt = U|\|/),
where U is a linear operator" (1957, p. 454; cf. 1973, p. 3; I supply
Dirac notation). After vigorously criticizing the very idea of Process 1,
Everett undertakes to describe observation processes "completely by
the state function of the composite system which includes the observer
and his object-system, and which at all times obeys the wave equation"
(1973, p. 8). From his standpoint, there is no fundamental distinction
between "measurement apparatuses" and other physical systems. "A
measurement is simply a special case of interaction between physical
systems - an interaction which has the property of correlating a quan-
tity in one subsystem with a quantity in another" (1973, p. 53). Since
almost every interaction between systems produces some correlation,
one could indeed take the view "that the two interacting systems are
continually 'measuring' one another" (Ibid.). However, this view does
not correspond closely to our intuitive idea of measurement as a
process yielding measured values. So Everett specifies several conditions
that characterize measurements. Without going into the technical
details, let me just say that his description of the interaction between
an observed system S and a measuring apparatus A is quite similar to
that given above, in the text surrounding eqns. (6.41)-(6.44). From this
388 Quantum Mechanics

he concludes "that for any possible measurement, for which the initial
system state is not an eigenstate, the resulting state of the composite
system leads to no definite system state nor any definite apparatus state.
The system will not be put into one or another of its eigenstates with
the apparatus indicating the corresponding value, and nothing resem-
bling Process 1 can take place" (1973, p. 60). Thus, "it seems as though
nothing can ever be settled by such a measurement" (p. 61). Moreover,
this conclusion has nothing to do with the size of the apparatus. And
yet, "macroscopic objects always appear to us to have definite posi-
tions" (Ibid.).
However, before dismissing his objective reading of QM "because
the actual states of systems as given by [it] seem to contradict our obser-
vations" one ought to investigate - says Everett - "what the theory
itself says about the appearance of phenomena to observers" (p. 63).
So he summons us to "the task of making deductions about the appear-
ance of phenomena on a subjective level, to observers which are con-
sidered as purely physical systems and are treated within the theory"
(Ibid.; my italics). Reading these words one thinks uncharitably: "Ain't
this a bid to square the circle}" But we must be patient. Everett himself
concedes that "in order to accomplish this it is necessary to identify
some objective properties of such an observer (states) with subjective
knowledge (i.e., perceptions)."77 He is content, however, to let his
observers have "memories" (in the clearly nonsubjective way in which
computer hard disks have them). "When the state \|/° describes an
observer whose memory contains representations of the events A, B,
. . ., C we shall denote this fact by appending the memory sequence in
brackets as a subscript, writing: \|/{A,B, . ,q" (PP- 64f.). I shall combine
this notation with the Dirac brakets that we have been using. Let an
observer O in initial state |\|/°..]) measure on system S the observable
Q with proper vectors {|(p*)K€ <*. If S is initially in state \q>k) {k e $), then,
through the measurement interaction, the compound system S + O
undergoes the unitary evolution described in eqn. (6.41). In our new
notation:

77
"Thus, in order to say that an observer O has observed the event a, it is necessary
that the state of O has become changed from its former state to a new state which
is dependent upon <z" (Everett 1973, pp. 63f.). Necessary indeed; but is it sufficient?
Certainly not. Why should such a change of state generate a perception of a} In fact
"the theory itself" has not a single word to say about the subjective appearances of
phenomena. To make his deductions Everett must act the ventriloquist.
 6.4 Meta-Physical Ventures 389

where oc& denotes the memory mark associated with state |(p*) (say, a
recording of the proper value corresponding to |(pjj). In the general
case, in which S is not initially in a proper state of the observable being
measured, the compound system undergoes the evolution described in
eqn. (6.43). The state reached by S + O is suitably described in the new
notation by:
(6-52)

Everett comments:
There is no longer any independent system state or observer state,
although the two have become correlated in a one-one manner.
However, in each element [\yt) ® |\|/°..,«,])] of the superposition [6.52],
the object-system state is a particular eigenstate of the observation [sic],
and furthermore the observer-system state describes the observer as
definitely perceiving that particular system state. This correlation is
what allows one to maintain the interpretation that a measurement has
been performed.
(Everett 1957, p. 459; cf. 1973, p. 68)
The words italicized by Everett deserve careful study, for his solu-
tion of the measurement problem turns on them. We must bear in mind
that the right-hand side of eqn. (6.52) expresses the state vector |*ps+o)
as a linear combination of vectors belonging to a particular basis of
%€s+o-> the Hilbert space of the (object + observer)-system. The elements
of that basis are all the tensor products formed by a proper vector of
the measured observable Q, on the one hand, and a characteristic
vector of the observer O, on the other (cf. §6.3.2). Interestingly, in the
said linear combination every basis vector of the form |(p*) <8> |\|/°), with
h & k, is multiplied by the scalar 0. This condition is part of the stan-
dard description of QM measurement (cf. London and Bauer 1939,
§11). |*FS+O) can of course be expressed as a superposition in terms of
each and every basis of %€s+o-> infinitely many of which are formed from
some other, arbitrarily chosen basis of Ws and the $Co-basis of charac-
teristic states of O. But few if any of these alternative expressions will
share with eqn. (6.52) this remarkable feature: Each nonzero term com-
bines a basis vector of 3^5 with a single basis vector of Hlo, which occurs
in the superposition only in that term. By virtue of it, the superposi-
tion on the right-hand side of eqn. (6.52) is certainly privileged among
390 Quantum Mechanics

the countless superpositions equal to l1?^0). Yet it might not be the

only one to be so privileged. For all we know, there may be other bases
of 3€5 and 3^o that combine together in a similar way to yield linear
combinations equal to l1?^0). This is of no consequence in the stan-
dard formulation of QM, in which a group of persons, the physicists,
express l ^ 0 ) in terms of the basis {q | >/> ® |\|/f)},-,/g* because the |(pf)'s are
the proper vectors of the observable that they propose to measure and
the |\|/f )'s represent the characteristic states marked on the dial of their
equipment. But in Everett's purportedly cosmic, anti-anthropocentric
formulation of QM, every expression of l^5*0) as a superposition with
the said feature stands on a par with the one in eqn. (6.52), and if
the latter is not unique, his "deductions about the appearance of
phenomena on a subjective level" become highly questionable.
I now turn to Everett's key statement: The observer-system state
\ty?[... ,<x,]) describes the observer as definitely perceiving the particular
object-system state |(f>,). Never mind the idiosyncratic use of 'perceiv-
ing'. It stands here for the fact that the physical object O bears the mark,
denoted by an of its interaction with another object S when S was in
state \(pj). That bearing such a mark should involve some awareness of
it surely does not follow from the theory. More important is this: The
theory does not predict that O will bear that mark after interacting with
S when the latter is in a state different from \<pt). In the notation the brack-
eted suffix a, in |\|/°... jOti]) denotes the most recent mark that O bears
right after interacting with S while the latter is in state |cp,). This is not a
mark that O should sport as the latest when S + O reaches the state
*F5+O) * Icp,) ® |\|/£ .. ><X|]). (If it did so, O could not pass for a good
observer.) The suffix at occurs in eqn. (6.52) simply because, following
Everett, I have affixed it once and for all to the symbol for the ket |\|/p),
which is a factor in one of the tensor products in terms of which eqn.
(6.52) spells out l1^^0). But in this use the suffix a, says nothing about
the marks present in O. It is just not true that eqn. (6.52) displays "a
superposition of states . . . for each of which the apparatus has recorded
a definite value" a, (1957, p. 457); all we can say is that eqn. (6.52) dis-
plays a superposition of states for each of which the apparatus would
record the definite value a, //the compound system were in that state -
which, of course, cannot be the case when it is in the state described in
eqn. (6.52), unless we assume that all the scalars q = 0, except one.
I apologize for explaining at such length these fairly obvious
minutiae. However, I feel that they are neglected by Everett and by the
numerous writers who either endorse his views or treat them with
 6.4 Meta-Physical Ventures 391

deference. Everett refers to expressions like the right-hand side of eqn.

(6.52) as "the superposition" that is "the final result" of a measure-
ment interaction. This is not inaccurate, but it glosses over the fact that
there are in every such case many equally valid expressions of the said
result as a superposition of other vectors. He also speaks of the vectors
iW?l.. ,at])}ie$ that occur as factors of the summands on the right-hand
side of eqn. (6.52) as if they represented states in which the observer
O actually finds him- or herself after the interaction with S.78 After dis-
cussing the case of repeated measurements - after which "every element
of the resulting final superposition will describe the observer with a
memory configuration [. . .] in which the earlier memory coincides with
the later" (1957, p. 459) - Everett writes:

We thus arrive at the following picture: Throughout all of a sequence of

observation processes there is only one physical system representing the
observer, yet there is no single unique state of the observer (which follows
from the representations of interacting systems). Nevertheless, there is a
representation in terms of a superposition, each element of which con-
tains a definite observer state and a corresponding system state. Thus
with each succeeding observation (or interaction), the observer state
"branches" into a number of different states. Each branch represents a
different outcome of the measurement and the corresponding eigenstate
for the object-system state. All branches exist simultaneously in the
superposition after any given sequence of observations.
(Everett 1957, p. 459)
In the footnote appended to this paragraph Everett remarks:
From the viewpoint of the theory all elements of a superposition (all
"branches") are "actual", none any more "real" than the rest. It is
unnecessary to suppose that all but one are somehow destroyed, since
all the separate elements of a superposition individually obey the wave
equation with complete indifference to the presence or absence ("actu-
ality" or not) of any other elements. This total lack of effect of one
branch on another also implies that no observer will ever be aware of
any "splitting" process.
(Everett 1957, pp. 459f., note %)
These quotations give the gist of Everett's solution to the measure-
ment problem. To establish quantitative results he skillfully develops a

Besides the two passages I quote next, see the footnote in Everett (1973, p. 68), which
was mercifully omitted in the published dissertation.
392 Quantum Mechanics

measure applicable to the elements of a superposition of orthogonal

vectors and applies it to ensembles of observation sequences. He con-
cludes that the statistical assertions of the standard formulation of QM
"will appear to be valid to almost all79 observers described by separate
elements of the superposition [. . .], in the limit as the number of obser-
vations goes to infinity" (1973, p. 74).
Everett's reformulation of QM was taken up by Bryce DeWitt, who
contributed the metaphor of the splitting universe, under which it is
mostly known. He wrote:
Our universe must be viewed as constantly splitting into a stupendous
number of branches, all resulting from the measurement like interactions
between its myriads of components. Because there exists neither a mech-
anism within the framework of the formalism nor, by definition, an entity
outside the universe that can designate which branch of the grand super-
position is the 'real' world, all branches must be regarded as equally real.
To see what this multiworld concept implies one need merely note
that because every cause, however microscopic, may ultimately propa-
gate its effects throughout the universe, it follows that every quantum
transition taking place on every star, in every galaxy, in every remote
corner of the universe is splitting our local world on earth into myriads
of copies of itself.
(DeWitt 1971, p. 222)
DeWitt contrasts "the mixture of metaphysics with physics" (1970, in
DeWitt and Graham 1973, p. 35) that he attributes to the Copenhagen
interpretation with the "remarkable metatheorem" proved by Everett,
viz., that "the mathematical formalism of the quantum theory is
capable of yielding its own interpretation" (1971, p. 212). I confess
that I cannot understand how the formalism of vectors and linear oper-
ators in complex Hilbert space can yield the following key elements of
the interpretation proposed by Everett:
(i) An observer - in Everett's sense, that is, a purely physical automa-
ton - perceives - in the ordinary sense, that is, becomes conscious
of- each "memory mark" associated to the tensor products of cor-
related object-system and observer-system states, a linear com-
bination of which is the state of the compound object-observer
system right after the measurement.

79
"Except for a set of memory sequences of measure zero" (Everett 1957, p. 461).
 6.5 On Relativistic Quantum Theories 393

(ii) If there is more than one tensor product with a nonzero scalar
factor in the said linear combination, the observer develops a
separate stream of consciousness for each, which is absolutely
impervious to the others.

Everett's bizarre reading of vector sums may have been prompted

by the classical Fourier analysis of waves in 3-space. Telephone
companies literally superpose the electromagnetic renderings of many
simultaneous long-distance messages in a single wave train that is
echoed by a satellite and then automatically analyzed at the destina-
tion exchange into its several components, each one of which is trans-
mitted over a separate private telephone line. No doubt we may speak
in this case of genuine splitting of what was initially added up. Still,
the signal could also be split into other, meaningless components if the
analysis were not guided by human interests and aims.

6.5 A Note on Relativistic Quantum Theories

The relativistic corrections ably introduced by Sommerfeld in Bohr's
hydrogen atom model (§6.1.1) contributed not only to the early success
of the Old Quantum Theory but also, decisively, to the general accep-
tance of SR by physicists. Since QM does not comply with SR require-
ments, its creators knew from the outset that QM was a provisional
first approximation that would not be tenable in the presence of high
energies. Indeed Schrodinger sought for a Lorentz invariant wave equa-
tion before settling for the Lorentz noninvariant one that bears his
name (§6.2.2). And it was clear to all that only a relativistic quantum
theory could suitably deal with radiation.
Efforts toward such a theory began as early as 1926, while nonrel-
ativistic QM had not yet settled down. A useful starting point was sup-
plied by Jordan in the final section of the three-men paper (Born et al.
1926, Section 4.3). Jordan complained later that nobody read that
section, nobody took notice of it, and nobody wanted to believe it.80
However, it probably exerted some influence on Dirac's paper, "The
Quantum Theory of the Emission and Absorption of Radiation"
(1927), which in turn was utilized by Jordan in his trailblazing work

80
1963 interview with T. S. Kuhn, quoted in Schweber (1994, p. 11).
394 Quantum Mechanics

on matter fields.81 Dirac's Lorentz invariant equation for the electron

was a major breakthrough (Dirac 1928). The fine structure of the
hydrogen spectrum - the subject of Sommerfeld's relativistic correc-
tions - flowed from it at once. More surprisingly, so did electron spin
(cf. §6.1.4). Yet the Dirac equation involved such difficulties that
Heisenberg - in a letter to Pauli of 31 July 1928 - referred to it as "the
saddest chapter in modern physics" (quoted in Miller 1994, p. 30). The
following comment, made by Heisenberg in 1963, explains his initial
gloom:
Until that time I had the impression that in quantum theory we had come
back into the harbor, into the port. Dirac's paper threw us out into the
open sea again. Everything got loose again and we got into new diffi-
culties. Of course at the same time, I saw that we had to go that way.
There was no escape from it because relativity was true.
(Interview with T. S. Kuhn, quoted in Schweber 1994,
p. 5; my italics)
The theory centered in Dirac's equation is known as Quantum Elec-
trodynamics (QED). Its most conspicuous difficulty led to its greatest
triumph. The equation has solutions that represent particles with
negative energy. Dirac conjectured that every negative energy state is
normally occupied by an electron, so that transitions to such states are
forbidden by the Exclusion Principle and therefore are never recorded
in spectrographs. However, transitions from negative to positive energy
states are allowed. If, perchance, such a transition occurs, creating a
hole, so to speak, in the sea of negative energy electrons, this would
show up as "a new kind of particle, unknown to experimental physics,
having the same mass and opposite charge to an electron. We may call
such a particle an anti-electron" (Dirac 1931, p. 61). The theory
implied that such positively charged, low-mass particles would be dif-
ficult but not impossible to observe with the then available tools. Black-
ett and Occhialini, who worked like Dirac in Cambridge, had in fact
found some evidence of their existence and told Dirac about it, but
they did not dare to print it without further confirmation, so the first
photograph of a positron track to appear in print was taken at Caltech
by Anderson (1932). Blackett's diffidence was true to the spirit of
his working place, as expressed by Rutherford: "They [the theorists]

Papers by Jordan and Klein, Jordan and Wigner, and Jordan and Pauli; for references,
see Miller (1994, p. 115).
 6.5 On Relativistic Quantum Theories 395

play games with their symbols but we in the Cavendish turn out the
real facts of nature" (Schweber 1994, p. 67). Yet it was Dirac who,
playing with symbols, brought to light ordinary matter's elusive
Doppelgdnger}1
The other major difficulty besetting QED is, if I may say so, intrin-
sic to the Dirac equation and cannot be solved by experimental
research. The differential equations of physics usually cannot be solved
exactly, but one can approximate a solution as closely as one wishes
by evaluating the successive terms of an infinite series. However, in the
case of the Dirac equation, the second term of the relevant series is
already apt to be a divergent integral. A term of this sort is not only
impossible to calculate but, it would seem, also physically meaningless.
Before the Second World War this difficulty caused most theorists to
be pessimistic about QED. In 1930 Bohr wrote to Dirac that he
expected the theory to fail "for energies of order 137me2", and in 1935
Heisenberg wrote to Pauli that, "with respect to QED we are still at
the stage in which we were in 1922 with regard to quantum mechan-
ics: We know that everything is wrong" (letters quoted in Schweber
1994, p. 84). However, right after the war, Tomonaga, Schwinger,
Feynman, and Dyson developed a method for handling the divergences
that tranquilized most physicists and generated predictions of unrivaled
precision.83 With the flair for euphemism that is so typical of our time,
the method was dubbed 'renormalization'. The gist of it is well
explained - for philosophical use - by Teller (1995, Chapter 7; 1988).
Thus refurbished, QED was celebrated as the most accurate theory
in physics. It set the pattern after which two other successful relativis-
tic quantum theories were built in the 1960s: the unified theory of elec-
tromagnetic and nuclear weak interactions and the "chromodynamics"
of nuclear strong interactions. Celebration, however, was not unani-
mous. The following remarks were made by Dirac himself, in one of
his last papers, as he spoke of the "infinite factors" that appear in QED
"when we try to solve the equations":

82
The story has been variously told. See, for instance, de Maria and Russo, "The dis-
covery of the positron" (1985) and Roque, "The manufacture of the positron"
(1997). For a fine philosophical analysis, see Falkenburg 1995, pp. 119-27.
83
I take the following information from Weinberg (1995/96, I, 490). Let m^ and ^
denote the muon's mass and magnetic moment, respectively, while e stands for the
charge of the electron. Then, the predicted value of u^, to fourth order, is 1.00116546
ellm^ and its current experimental value is 1.001165923 ellm^.
396 Quantum Mechanics

These infinite factors are swept into renormalization procedures. The

result is a theory which is not based on strict mathematics, but is rather
a set of working rules. Many people are happy with this situation
because it has a limited amount of success. But this is not good enough.
Physics must be based on strict mathematics. One can conclude that the
fundamental ideas of the existing theory are wrong. A new mathemati-
cal basis is needed.
(Dirac 1984; quoted in Schweber 1994, p. 71)
According to Steven Weinberg, "we have learned in recent years to
think of our successful quantum field theories, including quantum elec-
trodynamics, as 'effective field theories', low-energy approximations to
a deeper theory that may not even be a field theory, but something dif-
ferent"; therefore, "we think differently now about some of the prob-
lems of quantum field theories [. . .] that used to bother us when we
thought of these theories as truly fundamental" (1995/96, I, xxi).
I cannot go further into this subject. To discuss it meaningfully, even
at a superficial level, requires a much greater proficiency in mathe-
matical analysis than I expect from the readers of this book and can
myself offer. Partly because of it, but partly, no doubt, because they are
sickened by untidy math, most philosophers of physics tend to neglect
QED.84 Yet there are at least three reasons why they ought to pay heed
to it, and to the kindred quantum field theories of chromodynamics
and the electroweak field. First, quantum field theories have been the
working theories at the frontline of physics for over 30 years. Second,
these theories appear to do away with the familiar conception of
physical systems as aggregates of substantive individual particles. This
conception was already undermined by Bose-Einstein and Fermi-Dirac
statistics (§6.1.4), and greatly impaired by QM, according to which
the so-called particles cannot be assigned a definite trajectory in ordi-
nary space. But quantum field theories go a long step further and - or
so it would seem - conceive "particles" as excitation modes of the field.
This, I presume, motivated Howard Stein's saying that "the quantum
theory of fields is the contemporary locus of metaphysical research"

84
There have been exceptions, of course; see Redhead (1988) and Teller (1995). Auyang
(1995) is a philosophical book by a physicist who is exceptionally well versed in good
philosophy. Added in proof: After writing the above, I received a copy of Tian Yu
Cao's Conceptual Developments of 20th Century Field Theories (1997), a remark-
able historico-critical study, two-thirds of which are devoted to the quantum field
theories.
 6.5 On Relativistic Quantum Theories 397

(1970, p. 285). Finally, the very fact that physicists consciously and
fruitfully resort to unperspicuous theories can teach us something
about the aim and reach of their science. Here is how physicists work,
dirty-handed, in their everyday practice, a far cry from what is taught
at the Sunday school of the "scientific worldview".
 CHAPTER SEVEN

Perspectives and Reflections

7.1 Physics and Common Sense

Modern mathematical physics began in open defiance of common
sense. Galileo declared - through his spokesman Salviati - that he could
not "sufficiently admire the outstanding acumen" of the heliocentrist
astronomers, who, "through sheer force of intellect," had "done such
violence to their own senses as to prefer what reason told them over
that which sensible experience plainly showed them to the contrary"
(EN VII, 355; Drake translation). Furthermore, he judged color and
sound, heat and cold to be mere affections of the human senses, like
the tickling one feels when a feather is introduced into one's nose,
which, of course, lies not on the feather but on the nerves stimulated
by it (1623, §48). The most conspicuous features by which we perceive
and classify in everyday life the objects that surround us were thus pro-
nounced mind-dependent or "subjective" and banished from the stock
of notions that the new physics would employ to describe and under-
stand the real, "objective" nature of things.1 Physical discourse retained
many terms from ordinary language - arithmetic terms familiar in

1
The quotation marks surrounding objective and subjective are meant to make the
reader wary of these terms, not to indicate that Galileo used them in this way. In his
time, those words meant the opposite of what they mean today. "The Scholastic
Philosophy made the distinction between what belongs to things subjectively (Lat.
subjective), or as they are 'in themselves', and what belongs to them objectively (Lat.
objective), as they are presented to consciousness" (OED, s.v. 'objective', 2). However,
in the first quarter of the eighteenth century the modern meaning was already estab-
lished. The OED cites Watts 1725: "Objective certainty, is when the proposition is
certainly true in itself; and subjective, when we are certain of the truth of it. The one
is in things, the other is in our minds" (ii. ii. § 8).

398
 7.1 Physics and Common Sense 399

housekeeping and trade; geometric terms developed in carpentry, archi-

tecture, and land-surveying; chronometric terms used in lithurgy, sea-
faring, and, increasingly since the Renaissance, also in daily business;
terms applicable to machines found in harbors and building sites -,
which physicists sought only to define better and to apply with greater
precision. However, with hindsight we are bound to view the whole-
sale dismissal of core ingredients of ordinary language right at the start
of modern physics as being only a first step, a preparation for and antic-
ipation of what would come later. As we saw in Chapters Five and Six,
since 1900 the notions of "classical" geometry, chronometry, and
dynamics have been found wanting and have been replaced in physics
by concepts drawn from novel mathematical theories that are entirely
foreign to commonsense "intuitions" and the ordinary conduct of
human life.
More significantly perhaps, modern physics did away with the view
of nature as forming - much like our social environment - a network
of references and meanings, means and ends, values and counterval-
ues. This perception of nature, shared until 1600 by all nations, is now
regarded as "primitive" and superstitious by people educated in
our civilization. One of the first to say so clearly and forcefully was
Spinoza, in the Appendix to Book I of his Ethics, in which he sought
to discredit the common supposition that "all natural things, just like
men themselves, work to some end" and "indeed . . . that God Himself
directs everything to some sure end", and to show "why so many assent
to this prejudice, and why all are naturally inclined to embrace it".

After men have persuaded themselves that everything which happens,

happens for their sake, they must judge that which is most useful to them
to be what matters most in each thing, and they must esteem that to be
most excellent by which they are most beneficially affected. In this way
they must form those notions by which they explain nature, namely
good, evil, order, confusion, heat, cold, beauty and ugliness.

For example, if the motion which the nerves receive from objects repre-
sented through the eyes conduces to good health, the objects by which
it is caused are called beautiful-, while those exciting a contrary motion
are called ugly. Those things, too, which move the senses through the
nostrils are called odorous or fetid; those which do so through the taste
are called sweet or bitter, savory or insipid; those which act through the
touch, hard or soft, heavy or light; those, lastly, which stimulate hearing
400 Perspectives and Reflections

are said to make noise, sound, or harmony, the latter having deranged
men to such a point that they have come to believe that even God is
delighted with it.
(Spinoza, Opera I, 81-82)

Spinoza's radical stance was not endorsed by seventeenth-century

physicists, who, however, professed that physical inquiry must ignore
natural goals and values. These were held to be accessible to God alone,
and out of bounds for human science, although the pervasive beauty
and occasional ugliness of nature surely was no less evident then than
it is now. Indeed, as the quick progress of physics showed, it was wise
to define its subject matter by abstracting from good and evil, order
and confusion, the beauty and the ugliness of things. A difficulty arises,
of course, if one is committed to give a physical explanation of the
presence of mind and values in nature. If what we call 'physical' has
been deliberately defined so as to exclude them, it stands to reason that
only magic can put them back into it. Fortunately, no such explana-
tion is needed for the fruitful pursuit of physics. (This, I dare say, is
the redeeming insight in Descartes's otherwise perverse mind-body
dualism.)
The human mind certainly takes pride of place in the physical lab-
oratory, through the purposeful design and performance of experi-
ments and the value-laden appraisal of results. But physics managed to
ignore this bewildering presence at the core of its practice for 300 years,
and the philosophy of physics generally turned a blind eye on it. Only
in the last few decades have some cosmologists argued that the actual
existence of physicists must make a difference to physics - and they
have been fiercely resisted by philosophers. The argument involves the
so-called Anthropic Principle, in several, more and less sober versions
(for encyclopedic treatment see Barrow and Tipler 1986). They all turn
around the following fact: According to the currently accepted theo-
ries of physics, living organisms can exist in the universe in some places
and for some time only if the values of certain fundamental constants
of nature lie within certain fairly narrow intervals. Their measured
values do, of course, fulfill this requirement. The Anthropic Principle
is meant to explain this arguably unlikely coincidence. The "strong"
version demands the intervention of an intelligent demiurge to fix those
constants at the beginning of time and thus prepare the terrain for life
on earth. The "weak" version plays the Darwinian spoilsport: the coin-
 7.1 Physics and Common Sense 401

cidence requires no explanation at all, for obviously, if it did not

happen, we would not be here to wonder at it.
In their great treatise, The Large Scale Structure of Space-Time
(1973), Hawking and Ellis invoke the prerequisites of laboratory life
to settle a specific physical question. The Einstein field equations (eqns.
(5.29) and (5.31)) admit solutions containing closed timelike world-
lines (Godel 1949):

However the existence of such curves would seem to lead to the possi-
bility of logical paradoxes: for, one could imagine that with a suitable
rocketship one could travel round such a curve and, arriving back before
one's departure, one could prevent oneself from setting out in the first
place. Of course there is a contradiction only if one assumes a simple
notion of free will; but this is not something which can be dropped lightly
since the whole of our philosophy of science is based on the assumption
that one is free to perform any experiment. It might be possible to form
a theory in which there were closed timelike curves and in which the
concept of free will was modified [. . .] but one would be much more
ready to believe that space-time satisfies what we shall call the chronol-
ogy condition: namely, that there are no closed timelike curves.
(Hawking and Ellis 1973, p. 189)

Hawking and Ellis's remark is apt and trenchant, yet few have been
stirred by it, probably because not many working physicists pay atten-
tion to closed timelike worldlines except when they read science fiction
on holidays. On the other hand, in the context of QM a debate has
been raging for two-thirds of a century over the need and the right to
include the decisions and perceptions of experimentalists in nature's
bookkeeping. There is a difficulty in combining the deterministic evo-
lution of chances that is so brilliantly described by QM with the occur-
rence of definite chance events. Predicted probabilities must indeed
eventually give way to actual outcomes, but physicists and philoso-
phers have a hard time conceiving the transition. The incongruence of
expectation and fulfillment is one of the commonest facts of life, yet
not one that modern physics was well prepared to handle. As we saw
in §6.3.2 and §6.4.4, some philosophies of quantum mechanics try to
deal with it in terms of the subjective-objective dichotomy (and ulti-
mately as a mode of mind-body dualism). However, the classical oppo-
sition of subjective prediction versus objective realization is now turned
upside down: Expectations are epitomized in the evolving \|/-function,
402 Perspectives and Reflections

which stands for objective becoming, while fulfillment, the chance real-
ization of one or the other definite outcome, comes in the guise of sub-
jective awareness.
The QM debate, however, lay still far in the future when Descartes
and his followers conceived the subject matter of physics as a mindless
and aimless realm. Initially they had hoped to pierce through the veil
of "obscure and confused" sense appearances by "sheer force of intel-
lect". However, it was soon realized that the "clear and distinct ideas"
of men would not carry physics very far. Except for at most a few
general truths, physics must learn from experience. The problem is that
in human experience, as Berkeley forcefully noted, the purportedly
objective attributes of bodies are inextricably interwoven with the
allegedly subjective features of perception (see §3.1). The Cartesian
quest for certainty then took a different direction. The positivist tradi-
tion stemming from Berkeley sought to build the foundations of science
not on abstract ideas of space, time, and matter, but on the rock-solid
ground of elementary sense impressions, unpolluted by intellect and
free, of course, from the "idols" of common sense. Every scientific truth
ought to be derivable by deductive or inductive reasoning from ele-
mentary statements about irreducible sense data. This was easier said
than done. As we saw in §4.4.3(iii), Mach's valuable critical work did
precious little to further his sensationist ideology. Carnap's Aufbau
(1928) and its sequels (Goodman 1951; Moulines 1973) dauntlessly
undertook to carry out the foundationist program of positivism. When
completed it would have provided a procedure for inferring say, eqns.
(5.29) and (6.20) from premises in sense-datum language. Needless to
say, all three books, for all their rebarbative display of misplaced pre-
cision, do not come anywhere near this Utopia.
In the early 1930s, under pressure from Otto Neurath, Carnap and
his friends in the Vienna Circle converted from "logical positivism" to
"logical empiricism". The results of observation and experiment that
form the evidence on which scientific generalizations rest must be
shared by the community of researchers. Therefore, they should be
stated in the public - "intersubjective" - language in which we talk to
each other about things and events in the world, not in terms of private
sense impressions. Neurath insisted that the "physicalistic" language
that he advocated is only a fragment of ordinary language, for, in the
interests of communication, it ought to be understandable by those
born blind and deaf. But this requirement is surely exaggerated and
laboratory protocols rarely comply with it. Laboratory assistants will
 7.1 Physics and Common Sense 403

probably avoid color words when reporting to a color-blind professor,

but such situations are exceptional. Although its name suggests inno-
vative artifice, physicalism in effect signified the philosophical rein-
statement of commonsense language in scientific discourse (from
which, indeed, it had never been excluded in real life). Neurath saw
that this spelled the end of foundationism. His seminal paper, "Proto-
col Sentences", includes the famous ship metaphor:
There is no way of taking conclusively established pure protocol sen-
tences as the starting point of the sciences. No tabula rasa exists. We are
like sailors who must rebuild their ship on the open sea, never able to
dismantle it in dry-dock and to reconstruct it there out of the best mate-
rials. [. . .] Vague linguistic conglomerations always remain in one way
or another as components of the ship.
(Neurath 1932/33; in Ayer 1959, p. 201)
But most Vienna Circle philosophers still hoped to satisfy their
craving for certainty. Carnap demanded that scientific discourse be
reduced to thing-language - that is, "that language which we use in
everyday life in speaking about the perceptible things surrounding us"
(1936/37, p. 466) - and, more specifically, to the "observable predi-
cates of the thing-language" (p. 467). The key notion of an observable
predicate was explicated thus:
A predicate 'P' of a language L is called observable for an organism (e.g.
a person) N, if, for suitable arguments, e.g. 'b', N is able under suitable
circumstances to come to a decision with the help of a few observations
about a full sentence, say 'P(b)', i.e. to a confirmation of either 'P(b)' or
c
~P(b)' of such a high degree that he will either accept or reject 'P(b)'.
(Carnap 1936/37, pp. 454ff.)

Presumably such predicates ought to be value-free, although there are

suitable circumstances in which anyone acquainted with this or that
physical instrument is able to decide after a few observations that it is
not working well (so its data output should be thrown out). Carnap
granted that this notion of observable predicate was imprecise, for dif-
ferent persons are more or less able to decide a given sentence quickly.
He would, however, conventionally draw a sharp distinction between
observable and nonobservable predicates (p. 455). Many important
terms of physics are nonobservable; they were dubbed 'theoretical'. For
the new foundationism to work, the meaning and use of theoretical
terms should fully rest on that of observable terms. Carnap's attempts
404 Perspectives and Reflections

to "reduce" the former to the latter involved ever less stringent crite-
ria and resulted in ever more tenuous links (see in particular Carnap
1956). Pliable and open-minded though they were, the logical empiri-
cists remained adamant on one point: Cognitive meaning could not
accrue to observational terms through their joint employment with the-
oretical terms. This rigidity may have been due in part to their general
disregard of the history of science and the details of scientific practice,
but it was quite essential to the foundationist orthodoxy. As a conse-
quence of it, when Feyerabend, Hanson, and Kuhn made clear c. 1960
that the meaning and use of observational terms in physics depend on
the theories in which they are embedded, logical empiricism fell down
like a house of cards.
The quick transition from Carnapian foundationism to Kuhnian his-
toricism generated a pseudoproblem that made philosophy of science
the laughing stock of practicing physicists. Carnap conceived a physi-
cal theory as a seamless linguistic structure anchored to experience
through its observational predicates. But if these - as the saying goes
- are "theory-laden", they secure no outside links. The theories of
physics rise then as close, self-contained edifices that cannot commu-
nicate or be compared with one another on shared terms. On this view,
a physicist cannot teach Hamiltonian mechanics from 9 to 10 and
quantum mechanics from 11 to 12 unless he is afflicted with a double
personality. Such a ridiculous conclusion could, of course, have been
avoided by taking Neurath's metaphor more seriously. In Neurath's
ship, the neat steel turrets of theory are built on and bridged by the
wooden planks of common sense, which may be worn and musty yet
are indispensable to keep afloat the enterprise of knowledge. Physicists
who advocate different theories do not "practice their trades in differ-
ent worlds" (Kuhn 1962, p. 149), for there is but one world for them
to wake up to, namely, the world they are in together with the persons
they love and the goods they yearn for, some aspects and fragments
of which are represented by physical theories in their abstractive and
simplifying fashion while its concrete reality is talked about - often
unperspicuously yet mostly to the point - in everyday language. If our
ordinary understanding were a systematic affair - the "theory of
common sense" that some philosophers have tried in vain to make
explicit - it would stand aloof, impermeable to intellectual innovation
and incapable of furthering it. However, thanks to the vices of ambi-
guity, vagueness, and ductility - which the Vienna Circle sought to
correct through the use of formal language - it affords the nourishing
 7.2 Laws and Patterns 405

ground on which physical theories and other such intellectual systems

can grow and across which they can communicate.

7.2 Laws and Patterns

We have all heard that the aim of physics is the discovery and formu-
lation of the laws of nature. The phrase 'law of nature' is first attested
in Plato's dialogue Gorgias (483e), where it is pointedly used as an oxy-
moron by Callicles, an otherwise unknown young Athenian right-
winger. Fifth-century sophists regularly opposed 'nature' (tyixjiq) and
'law' (vojLioc;), the latter being the product of human consensus and the
defining mark of civilized society.2 Callicles complains that laws are
made by the majority who are weak, and therefore decree that the
ambitions of the strong are unjust, and remain content with equality;
whereas "nature itself proclaims that it is just that the better man have
more than the meaner one and the abler more than the less able". After
asking rhetorically on what principle of justice Xerxes campaigned
against Greece and Xerxes's father against the Scythians, Callicles
offers this answer: "They acted according to the nature of what is just
(mxa (ptxnv xf|V xov SIKOCIOI)) and indeed, by Zeus, according to the
very law of nature (Korea v6(iov ye xov xfjq qyuaeax;)." Despite this
inauspicious beginning, the Greek expression vojioq (ptiaecx; ('law of
nature') - and its translations into Latin and the modern European lan-
guages - made a splendid career. It fitted admirably with the Stoic idea
of a rational world order of which legitimate human laws are a man-
ifestation, "the right reason of nature . . . being a divine law by which
whatever belongs to each thing or concerns it was assigned to it"
(Chrysippus, FM 337; cf. FM 323: "The laws of the several poleis are
but appendages to the right reason of nature"). Independently of this,
it was also used to convey the medical concept of normality, of the
workings of the animal body in good health.3 The expression vojioq
qyoaecoc; ('law of nature') is often found in the Greek Church fathers,

2
Cf. Plato (Gorg. 482e): "Nature and law are contrary to one another in many
respects"; (Protag. 337d): "Law, the tyrant of mankind, enforces many things con-
trary to nature". Antiphon, fr. 44 (DK), is the locus classicus on "necessary nature"
versus "adventitious law".
3
This use, repeatedly attested in Galenus (second century A.D.), is found already in
Plato (Timaeus 83e). Note that the "laws of nature" in this sense can be actually vio-
lated, just like the laws of man.
406 Perspectives and Reflections

who readily associated it with their idea of God as creator and sover-
eign of the world;4 but they also fondly employed it to denote the famil-
iar necessity of things to which God Himself was not bound, although
He voluntarily submitted to it in the person of Jesus Christ.5
The founders of modern physics took Plato's metaphor quite liter-
ally, and set out to find the articles of nature's legal code. In the writ-
ings of these Christian authors the word 'law' does not signify the
universal scope of the prescribed regularities, but rather the legislative
authority of their divine source; thus, 'Kepler's Laws' became the stan-
dard name for what at most might qualify as local traffic regulations.
However, all the founders shared the quaint belief that God, although
infinitely powerful, is a paradigm of good husbandry and therefore has
used the thriftiest means to achieve the richest variety and abundance
of effects. They understood this to imply that all the lowly local laws
of nature must follow from a few all-embracing ones. Descartes pro-
claimed three universal laws of nature, two of which he derived from
the very fact that they were willed by God (1644, II, arts. 37, 39,
quoted in §1.3). In a similar vein, Newton put forward his three laws
of motion, and showed that Kepler's Laws (duly corrected) could be
derived from a specification of the second. Thus, the particular facts
of nature came to be conceived as instances of universal statements or
laws organized into a deductive system or theory. From the writings
of Galileo, Descartes, Huygens, and Newton we gather that there was
one exemplary theory in everyone's mind, viz., the Elements of Euclid,
in which a wealth of geometrical laws, exactly obeyed - or so it seemed

Cf. St. Basil, Epistle 302, a letter of condolence. The saint reminds a recent widow of
"the legislation (VOUOOEGIOC) of our God, in force from the beginning," by which
"whoever comes to birth must necessarily depart from life at the proper time" - and
advises her not to be "vexed with the common laws of nature (\n\ dyavaKXCOjiEV mi
xoTq KOIVOU; T?j<; (puaeox; voum<;)".
St. John Chrysostomus, commenting on Genesis 1 : 1 - "In the beginning God made
heaven and earth" - bids us to "observe the order and sequence: first the roof, and
then the ground. As I said, He was not slave to the law of nature, or the order of art,
but [exercised] the authority of power" (Migne PG, 59, 507). The Council of Ephesus
(431 A.D.) emphatically asserted - against Nestorianism - that Christ was "under the
law of nature", and was therefore really and fully born of a woman and experienced
"the taste of death" (Schwartz ACO, vol. 1.1.6, pp. 38 [line 43], 19 [line 35], 103
[line 39]). See also St. Athanasius, Exp. in Psalmos (Migne PG, 27, 460 [lines 26-32]);
St. Gregory Nazianzenus, In sanctum pascha, orat. 45 (Migne PG, 36, 660 [lines
31-33]): "as a man," Christ "toiled and hungered and thirsted and agonized, by law
of nature".
 7.2 Laws and Patterns 407

- by all bodies, was inferred - or so it was thought - from a small

number of fairly simple principles.
The scheme of ascent from facts to laws, from laws to theories was
still canonic in the philosophy of physics long after its theological moti-
vation had been discarded. For John Herschel (1830) a law of nature
is either "a general proposition, announcing, in abstract terms, a whole
group of particular facts relating to the behavior of natural agents in
proposed circumstances," or "a proposition announcing that a whole
class of individuals agreeing in one character agree also in another" (p.
100). Like his French contemporary, Comte, Herschel enjoins us to
concentrate our attention on such laws, "dismissing then, as beyond
our reach, the enquiry into causes" (p. 91). To find the laws of nature
we must rely entirely on experience, on "the observation of facts and
the collection of instances" (p. 118). "As particular inductions and
laws of the first degree of generality are obtained from the considera-
tion of individual facts, so Theories result from a consideration of these
laws" (p. 190). Ascent to theory is now justified on purely method-
ological grounds:

The analysis of phenomena, philosophically speaking, is principally

useful, as it enables us to recognize, and mark for special investigation,
those which appear to us simple; to set methodically about determining
their laws, and thus to facilitate the work of raising up general axioms,
or forms of words, which shall include the whole of them; which shall,
as it were, transplant them out of the external into the intellectual world,
render them creatures of pure thought, and enable us to reason them out
a priori.
(Herschel 1830, p. 97)

Herschel's description of the direct objects of a physical theory as "the

creatures of reason rather than of sense" (p. 190) betrays his profes-
sional acquaintance with them. Still, it is clear that - as was custom-
ary in his century and during most of the next - Herschel viewed
physical theories as collections of statements intended to be true of the
actual course of the world. This is the distinctive feature of what
Putnam (1962) dubbed "the received view of theories". On this view,
physics ultimately aims to fuse all such collections into a single
consistent one, a "theory of everything". For, as John Stuart Mill
explained, "the question, What are the laws of nature? may be stated
thus: What are the fewest and simplest assumptions, which being
granted, the whole existing order of nature would result? [or] thus:
408 Perspectives and Reflections

What are the fewest general propositions from which all the unifor-
mities which exist in the universe might be deductively inferred?"
{System of Logic ffl.iv.l; 1874, p. 230).
I do not plan to discuss the "received view" in detail.6 I only wish
to stress one aspect of its twentieth-century history that, I think, has
not been sufficiently emphasized. As noted above, on the "received
view" physical theories should follow the example of Euclid's
Elements. Indeed, until the end of the nineteenth century, almost every-
one took for granted that the final theory of everything would include
Euclidian geometry as a subtheory, that is, that the latter's axioms
would either follow deductively from the "fewest general propositions"
put forward by the former, or would be counted among them.
However, the evolution of geometry during that same century (§4.1)
led to a completely different reading of mathematical axiom systems
in general and of Euclid's in particular.
The new understanding of axioms and their consequences is implicit
in Hilbert's masterly Foundations of Geometry (1899) and caused his
clash with Frege.7 Hilbert invites us to conceive "three different systems
of things". He calls them "points", "straights", and "planes", but only
as lip service to tradition. He does not explain what these things are,
but introduces five groups of axioms that impose conditions on their
mutual relations. (Examples: Any two points determine a unique
straight on which they are said to lie; any three points, not all of which
lie on the same straight, determine a unique plane on which they are
said to lie.) Frege thought that these unexplained terms referred to
things that Hilbert assumed to be well-known and that the axioms
were intended as true statements about such things. In a letter of 29
December 1899 Hilbert disabused him: "I do not wish to presuppose
anything as known; I see in my declaration in § 1 the definition of the
concepts 'points', 'straights', 'planes', provided that one adds all the
axioms in axiom groups I-V as expressing the defining characters"
(Frege KS, p. 411). Frege had complained that Hilbert's concepts were

6
I refer the reader to the excellent critical surveys by Frederick Suppe (1977, pp.
3-118, 619-32; 1989, pp. 38-77).
7
The Frege-Hilbert controversy has been repeatedly discussed by philosophers who,
however, are loath to acknowledge that on this occasion their sharp-witted hero Frege
displayed unusual obtuseness. A commendable exception is Shapiro (1997, pp.
157-70). Cf. Torretti (1978, pp. 249-52).
 7.2 Laws and Patterns 409

equivocal, the predicate 'between' being applied to genuine geometri-

cal points in §1 and to real number pairs in §9. Hilbert replied:
Of course every theory is only a scaffolding or schema of concepts
together with their necessary mutual relations, and the basic elements
can be conceived in any way you wish. If I take for my points any system
of things, for example, the system love, law, chimney-sweep,. . . and
I just assume all my axioms as relations between these things, my
theorems - for example, Pythagoras's - also hold of these things. In other
words: every theory can always be applied to infinitely many systems of
basic elements. One needs only to apply an invertible one-one transfor-
mation and to stipulate that the axioms for the transformed things are
respectively the same. [. . .] This feature of theories can never be a short-
coming8 and is in any case inevitable.
(Hilbert to Frege, 29.12.1899; in Frege KS, pp. 412-13)
Besides the undefined terms 'point', 'straight', and 'plane', Hilbert
employs five undefined relational predicates, viz., 'between' (a ternary
relation among points), two kinds of incidence (binary relations
between a point and a straight, and between a point and a plane), and
two kinds of congruence (between segments and between angles,
respectively). Axiom groups I-V jointly characterize a structure built
from the said three systems by means of these five relations. By using
the idioms of a later age this characterization can be formulated in the
traditional shape of a definition, as follows: A Euclidian 3-space is an
ordered octuple (L,A,Yl,B,IuI2,CuC2) satifying the Hilbert axioms J-V,
where X, A, and II are three nonempty, nonintersecting sets, B is a
triadic relation among elements of X, Ii is a dyadic relation between
an element of X and an element of A, I2 is a dyadic relation between
an element of X and an element of n, Q is a dyadic relation
between two unordered pairs ("segments") of elements of X, and C2 is
a dyadic relation between certain equivalence classes ("angles") formed
by ordered triples of elements of X.9 There are indeed many conceiv-

"But a rather powerful advantage" (Hilbert's footnote ad loc).

Hilbert defines an angle as an (unordered) pair of rays issuing from the same point.
The clumsier definition that I have in mind when I describe angles as equivalence
classes of ordered point triples may be stated as follows: Let Bxyz stand for 'point y
is between point x and point z\ Assume that B satisfies Hilbert's axioms for between-
ness (group II, which adequately characterizes the intuitive relations of three collinear
points, one of which lies between the other two). I say that two ordered triples of
410 Perspectives and Reflections

able octuples that satisfy this definition, and yet, surprisingly, the
Hilbert axioms are so contrived that all these octuples embody one and
the same structure, in the following precise sense: If S and S' are any
two such octuples, there are one-one mappings fu f2, and f3 that map
the first, second, and third terms of S onto, respectively, the first,
second, and third terms of S\ sending congruent segments to congru-
ent segments and congruent angles to congruent angles.10
On this novel approach, as Paul Bernays neatly expressed it, "an
axiom system is regarded not as a system of statements about a subject
matter but as a system of conditions for what might be called a rela-
tional structure" (1967, p. 97; my italics), that is, as the specification
of a concept of a certain type. Thus, if philosophers had promptly
assimilated Hilbert, the "received view" of theories would not have
made it to the twentieth century. But its modern advocates, notably the
logical empiricists, although thoroughly acquainted with Hilbert's
book and the developments in mathematics that motivated it, did not
understand it in the manner explained above and so were able to
defend the "received view" in good conscience. For them, an axiomatic
mathematical system is not the specification of a concept, which may
or may not have one or more real instances, but an uninterpreted cal-
culus, that is, a system of meaningless ink or chalk marks forming
strings - "sentences" - and strings of strings - "proofs" - according
to strict rules of syntax. Never mind that the stuff printed in mathe-
matical books and journals does not satisfy this description. Anyone
who doubted that all of mathematics was ideally reducible to uninter-
preted calculi was invited to read the three volumes of Principia
Mathematica by Whitehead and Russell.11 From this standpoint, the

points (x,y,z) and (u,v,w) are A-equivalent if and only if x = u and either Bxyv and
Bxzw, or Bxyv and Bxwz, or Bxvy and Bxzw, or Bxvy and Bxwz, or Bxyw and Bxzv,
or Bxyw and Bxvz, or Bxwy and Bxzv, or Bxwy and Bxvz. An angle is a class of A-
equivalent point triples. (The reader should verify that 2^yxz is identical with ^vuw
if and only if vertex x = vertex u and one of the eight stated alternatives is fulfilled,
and that A-equivalence is indeed an equivalence, i.e., a reflexive, symmetric, and tran-
sitive dyadic relation.)
0
It is not hard to see that the mappings in question will not preserve the congruences
unless they also preserve the relations of incidence and betweenness (for betweenness
this condition follows at once from my definition of 'angle' in note 9).
1
Of course, Whitehead and Russell did not understand their formal system as an unin-
terpreted calculus. Quite the contrary. Following Frege, they construed the numerals
as proper names (of certain classes). The good thing about this was that one could
 7.2 Laws and Patterns 411

"received view" can be straightforwardly restated thus: A physical

theory - and, indeed, any properly formulated scientific theory - is an
axiomatic mathematical system suitably interpreted so that its "sen-
tences" are meaningful sentences and its "proofs" are valid logical
proofs concerning a definite domain of reality. Now, the rules of syntax
of the standard calculi - stemming from Frege (1879) and Peano
(1895-1908) through Whitehead and Russell (1910-13) -were devised
to ensure that the "proofs" would work in just this way. Thus, to make
any such calculus into a physical theory in the stated sense, it is
sufficient to bestow a physical meaning on the so-called nonlogical
symbols of the calculus.12 According to logical empiricism, this can only
be done by associating such symbols with terms and predicates of the
physicalistic language (§7.1). However, even the informal versions of
accepted physical theories teem with notions that cannot be rendered
accurately in physicalistic terms;13 and this situation cannot be reme-
died by translating those theories, say, to the formal language of
Principia Mathematica. (Imagine, say, that eqn. (6.44) has been rewrit-
ten as a sentence in this formal language; try to imagine a physicalis-
tic rendering of this sentence.) So the logical empiricists settled in the
end for a diminished version of the "received view": They equated
physical theories with partially interpreted calculi (Carnap 1956).
Briefly, it comes down to this. While the logical symbols of an unin-
terpreted calculus receive their intended logical force from the rules of
syntax, the nonlogical symbols are divided into two classes: Those in
the first class, dubbed "observational", are associated with terms and
predicates of the physicalistic language, but those in the second class,

now be quite certain that 'two' does not denote Julius Caesar (cf. Frege 1884, p. 68).
The bad thing was that in Whitehead and Russell's construal, 'two' - at least in unreg-
imented ordinary English - is - like ' J o n n ' - the proper name of countless entities, a
different one for each level in the logical hierarchy of types introduced by Russell to
avoid the contradiction that he had discovered in Frege's system. (In a duly regimented
language one might distinguish the different twos by means of indices, usually numer-
als; one ought to refrain, however, under pain of circularity, from giving them the
said meaning in this use.)
12
Evidently, in an uninterpreted calculus the distinction between logical and nonlogi-
cal symbols is wholly out of place. On the other hand, Frege (1879) made the seman-
tic intent of his "conceptual script" clear to all by using quaint geometrical figures
as logical symbols and the letters of various alphabets as nonlogical symbols.
13
Consider acceleration (as in classical mechanics). How would you explicate it in phys-
icalistic language? By equating it with average increases of average speed over a short
sequence of short time intervals?
412 Perspectives and Reflections

dubbed "theoretical", remain uninterpreted. So, strictly speaking, the

"theoretical" symbols mean nothing, although they do acquire a
shadow of meaning through their logical connections with the "obser-
vational" symbols. Typically, the "theoretical" symbols of such a par-
tially interpreted calculus would include all the undefined primitives
that occur in the axioms, the "observational" symbols - which turn up
in remote logical consequences of the axioms - being defined in the
calculus in terms of those primitives, through usually long chains of
"theoretical" intermediaries. This version of the "received view" is a
far cry from Herschel's. A logical empiricist can surely reach for a
"theory of everything" in which every observable regularity of nature
can be deductively inferred. However, the axioms of such a theory
could not pass for true statements about the universal order of things.
They would just be master recipes for the computation of empirical
predictions (and retrodictions) from empirical data. To call them "laws
of nature" would be a colossal abuse of language.
In the 1960s and 1970s a different view of physical theories came
to the fore in America. As is usual in philosophy, there is more than
one version of it. However, for simplicity's sake, I shall bring them all
together under the single portmanteau term structuralism.14 This term
does justice to what I think is the gist of this view, which may be sum-
marily stated thus: At the core of a physical theory there is always a
coherent piece of mathematics that is intended to throw light on the
processes and states of affairs in the theory's chosen field of study. Any
coherent piece of mathematics can be articulated - as Hilbert did for
Euclidian geometry - as the conception of a relational system or struc-
ture. Such a conception throws light on a physical domain when this
is grasped as an instance or family of instances of the structure.
Structuralism originated with Patrick Suppes (1960, 1962, 1967,
1969) and gained strength through the early writings of Joseph Sneed
(1971) and Bas van Fraassen (1972).15 Suppes took his cue from

14
Frederick Suppe (1967, 1977, 1989) calls it 'the semantic conception of theories'. I
think that this appellation can only make sense at a time when there are people who
are willing to countenance a conception of theories that is not semantic. In saner
times, the term will not pick out a specific conception of theories.
15
A structuralist approach to physics can be discerned at a much earlier date in the
physical writings of a few great mathematicians who taught in Hilbert's Gottingen.
Thus, Minkowski (1909) went straight for the structure at the core of Special Rela-
tivity, WeyPs Space-Time-Matter (1918) can be readily given a structuralist reading,
 7.2 Laws and Patterns 413

Bourbaki, whose reconstruction of mathematics as the science of

abstract structures achieved the peak of its influence and began to spill
out into undergraduate textbooks in the 1950s.16 Bourbaki endeavored
to bring all the seemingly disparate branches of contemporary mathe-
matics under a single unifying perspective by treating each as the study
of a particular "espece de structure". Bourbaki gave a precise set-
theoretic definition of "species of structure". It would be out of place
to repeat it here.17 I think the reader can get a clearer and firmer idea
of Bourbaki's meaning from my presentation of several such species of
structure in the Supplements (group and fields in 1.2, vector space in
1.3, inner product space in 1.4, Hilbert space in 1.7, poset in II.1, lattice
in II.2, and topological space in III.l). Another example is Euclidian
3-space as defined above, viz., as an ordered octuple whose first three
components are abstract sets and whose remainingfivecomponents are
drawn, in fulfillment of Hilbert's axioms, from some of their Cartesian
products and their power sets. As explicated by Suppes and his fol-
lowers, a physical theory consists of a species of structure plus
the "empirical claim" that certain physical systems - the theory's
"applications" - are instances of that species.
Some qualifications must be added to this streamlined statement to
appreciate its actual import and its clarifying power. First, this is what
a physical theory is as explicated by the structuralists, not as it exists

and von Neumann (1932) established the equivalence between matrix mechanics and
wave mechanics by proving that the underlying structures are isomorphic (see §6.2.3).
Nicholas Bourbaki is the pseudonym of a group of French mathematicians - includ-
ing Henri Cartan, Claude Chevalley, Jean Dieudonne, Laurent Schwartz and Andre
Weil, among others - who undertook a systematic exposition of the whole of math-
ematics, which has been appearing since 1939 as Elements de mathematiques in an
open-ended series of fascicles. I must say that I first saw a Bourbakian approach
applied to all the major theories of physics in Bunge's Foundations of Physics (1967).
I am therefore inclined to regard this book as a classic of structuralism, although the
author would probably disavow me.
I tried my hand at it in Torretti (1990, pp. 84-86). For the original, see Bourbaki
(1970, ch. IV, § 1). The main idea is, roughly speaking, that any instance of a given
species of structure can be described as an ordered «-tuple (Au .. ., Ak, Ak+U . . . ,
Att), where k and n are positive integers (k < n), Au . . . , Ak are arbitrary nonempty
sets, and Ak+U • • •, An are distinguished elements meeting definite conditions and
picked out from Au .. ., Ak or from other sets obtained from the former by way of
the set-theoretic operations of forming Cartesian products and ascending to power
sets. (These two operations are defined in Supplement I.I.)
414 Perspectives and Reflections

in real life, in the discussions and writings of the physicists. Not every
physical theory in circulation has been given a structuralist explication.
Moreover, the treatment is only available for theories that have attained
sufficient clarity and precision. Physical theories do not fall like ripe
fruits from a "place beyond heaven" (Plato, Phaedrus 247c), but are
gropingly fashioned by physicists on earth. Their concepts are not all
found ready-made on the smorgasbord of extant mathematics, but have
often been created by the physicists themselves, or, at any rate, adapted
to their needs. This is done with a view to the applications that they
have in mind, which, in turn, are delineated with increasing distinct-
ness in nature's flux by their being grasped with those concepts. The
structuralist explication pairs a structure specified by axioms with a
family of physical complexes conceived as particular instances of it.
Live physics is murkier. Still, structuralism offers a picture of physical
theories in an idealized state of maturity that throws much light on
their actual development, their mutual relations, and their semantic
links with their referents in the real world.
There is one more qualification that I should add before proceeding
any further. Suppes and his followers present the mathematical core of
a physical theory as a Bourbakian species of structure. This is a fairly
straightforward method of presentation that was readily available to
them and facilitated a uniform approach to the mathematical concepts
employed in physics. But they do not hold to Bourbakism as a dogma.
Any viable way of understanding mathematics as the study of abstract
patterns can yield good structuralist explications of the mathematical
ingredient in physical theories (cf. Balzer, Moulines, and Sneed 1983;
Mormann 1996). Structuralism does not depend on the set-theoretic
reformulation of the mathematical theories of physics. Its mainspring
is the vision of physics as an endeavor to grasp the patterns of nature
as concrete embodiments of the abstract ones conceived by such
theories.18
Having made this clear, I shall assume, for definiteness, that struc-

Two further qualifications express only my own judgment: (i) The structuralist analy-
sis of physical theories by Joseph Sneed (1971) was constrained by the desire to solve
his "problem of theoretical terms"; I do not think that this is a real problem (see Tor-
retti 1990, pp. 109-30). (ii) Sneed's followers claim that in sciences that are very dif-
ferent from physics, such as biology, sociology, and even the history of science, every
proper scientific theory can be explicated as a Bourbakian species of structure coupled
with an empirical claim; I do not share this view.
 7.2 Laws and Patterns 415

turalist explications are made in the usual set-theoretic way. Imagine

for a moment that a demon of uncommon intelligence undertook to
do physics in this style. The set-theoretic hierarchy is so rich that he
or she could well build from it a species of structure of which nature
in all its complexity is an instance. Human physicists, however, are
bound to be more modest. Indeed, modern mathematical physics was
born of the realization that earlier attempts to grasp the entire fabric
of the world in one fell swoop had failed altogether. Since the seven-
teenth century, physicists have been focusing on particular patterns that
they isolate and grasp by means of suitably contrived mathematical
concepts. Surely, they have thrived not only by their modesty, but also
and chiefly by their uncanny eye for patterns. Thanks to it they ably
discerned those features in actual phenomena that make together a dis-
tinctive pattern from the rest, which only mess it up, and they recog-
nized structural affinities and identities where formerly one had seen
only abysmal differences (e.g., between the fall of apples and the cir-
culation of the moon). In this they were well served by their decision
to geometrize physics, to conceive the patterns in nature as instances
of mathematical structures.19 However, to do it they had to replace in
their minds the concrete physical processes that they intended to study
- which were unmanageably complex and entangled with the rest of
the world - with simplified models that truly instantiated such abstract
mathematical structures as they could handle.20 The gap between rep-
resentation and reality was then filled by tactfully introducing correc-

19
Thus, as the reader will recall, Maxwell was able to bring, as if with a magic wand,
all the phenomena of optics into the fold of electrodynamics, because his mathemat-
ical theory of the latter predicted waves propagating in vacuo with the speed of
light.
20
The word 'model' is used in current philosophical literature in two somehow con-
trary senses. In the sense used above - and in many other passages of this book - a
model is a representation of an individual or generic object by a real or ideal object
of a different sort (e.g., a plastic model of an airport terminal; the familiar model of
a pendulum, consisting of a weightless inextensible string with a massive dimension-
less particle at one end, affixed by the other end to a frictionless nail). In the sense
that is common in logic, models are models of structures, and a model of a structure
of a given species is any set endowed with structural features satisfying the require-
ments of that species. Both senses can be nicely combined in one breath in a modified
version of the above text, to wit: "Physicists had to replace concrete physical processes
with simplified models (first sense) which are in effect models (second sense) of such
abstract structures as they could handle." In this book I usually say 'realization' or
'instance' for 'model' in the second sense.
416 Perspectives and Reflections

tions and agreeing on admissible margins of error. This devious

approach to truth, which Aristotle would have spurned, enabled
modern physics to rise far above Aristotle's level of ignorance. Thanks
to it we can cope with two inescapable human limitations, namely, our
incapacity to make error-free measurements and to work with mathe-
matics that is too complex. But it also entails that the "empirical claim"
of a physical theory mentioned above, viz., that the theory's applica-
tions are actual examples of its characteristic mathematical structure,
must be taken with a sizable pinch of salt, for the structure is instan-
tiated by the models, not by the physical realities that the models rep-
resent only to within some specified or unspecified but anyway finite
degree of approximation. Thus, no astronomer in his right mind would
claim that the solar system is exactly pictured by the Newtonian model
that he uses to calculate the future motions of the planets. (The model
certainly does not include representatives of every existing asteroid and
comet.) What he claims instead - with abundant empirical support -
is that the planetary positions and velocities recorded in fact and those
computed from the model will not differ, within some agreed future
time, by more than some agreed margin. Until the end of the nineteenth
century one could maintain in good faith that the real solar system can
be approximated with increasing accuracy by a series of ever more
complex models instantiating the Newtonian theory of gravity.
However, we cannot say this any longer. It is generally agreed that for
predictions surpassing a certain level of accuracy the astronomer must
resort to models that are instances of General Relativity. Indeed, in the
case of Mercury, due to its smallness and its short distance from the
Sun, Newtonian models fail already at a level of accuracy that astro-
nomical observation had attained by 1860 (§5.4, note 64).
To account for such facts philosophers of an older persuasion tried
to develop a concept of approximate truth. If a physical theory is a
system of statements that are not exactly true and yet are not com-
mitted to the trash can of science, it must be because those statements
are true approximately, to a degree that scientific methodology ought
be capable of measuring. However, all attempts to calibrate approxi-
mate truth have ended in failure, essentially, I dare say, because, unless
you twist the meaning of English words beyond recognition, the truth
of a statement cannot be a matter of degree. Structuralism need
not distort the grammar of 'truth'. From its standpoint, a theory is -
or is not - exactly true of its purported models, which, in turn, are
good or bad representatives of the physical phenomena for which they
 7.2 Laws and Patterns 417

are supposed to stand. And, of course, a particular model of a partic-

ular process may be good enough for one purpose and not for another
one.
A major strength of the structuralist view is its ability to elucidate,
from this standpoint, the relations between the different and often
incompatible physical theories that vie but also cooperate with one
another. On the "received view" two theories can vie for our accep-
tance only if they contradict each other on one or more issues; but then
they cannot cooperate, for how could two mutually inconsistent sets
of statements jointly contribute to the solution of a scientific problem?
Structuralism nimbly avoids this dilemma. We shall see in §7.3 how it
dissolves the difficulty that the "received view" had to face after Kuhn's
analysis of the succession of rival theories. Here I shall describe in
structuralist terms two well-known examples of collaboration between
theories.21
My first example is Einstein's work on Mercury's perihelion-
advance, mentioned in §5.4 and alluded to above. In November 1915,
Einstein (1915h) showed that an approximate solution of eqns. (5.28)
- the second set of gravitational field equations published by him in
that month - accounted for the difference of about 43" per century
between the observed precession of Mercury's perihelion and the value
predicted by astronomers using the Newtonian theory of gravity. The
difference is also accounted for by Schwarzschild's exact solution of
eqns. (5.29) - Einstein's third and final set of November 1915 -, which
agree with eqns. (5.28) in empty space. Using more recent figures
(Weinberg 1972, p. 199) the said difference arises as follows: The
secular precession actually measured is
Aoc = 5,600.73 ± 0.41" (7.1)
The observed perihelion of Mercury must advance some 5,025" per
century due to the fact that the astronomical coordinate system is
affixed to the moving Earth. Of the remaining 575.73", some 532" can

21
Another example that deserves being studied from this point of view is the remark-
able assortment of theories involved in the analysis of experimental results in con-
temporary particle physics. For a lucid description of this conceptual pot-pourri, see
Falkenburg (1995, Ch. 5). As she aptly notes, the "theory of measurement" stan-
dardly employed in ascertaining the measurable attributes of the particle tracks, scat-
tering events and resonances recorded in high-energy experiments, is "anything but
a strict deductive system of special cases of one-and-the-same theory" (p. 183).
418 Perspectives and Reflections

be attributed, in accordance with the Newtonian theory, to gravita-

tional perturbations caused by the other planets, chiefly Venus, Earth,
and Jupiter. So, according to pre-relativistic astronomy, the secular pre-
cession of Mercury's perihelion should not be larger than
A0 = 5,557.62 ± 0.20" (7.2)
The difference Aa - A9 = 43.11 ± 0.45" agrees beautifully with the
secular precession of 43.03" predicted by General Relativity for a spin-
less test particle of negligible mass simulating Mercury's motion in an
otherwise empty spherically symmetric spacetime such as would cor-
respond to the presence of a solar mass at the axis of symmetry.22 Now,
neither Einstein nor Schwarzschild figured out the precession of
Mercury's perihelion in a relativistic model of the solar system, com-
plete with all the planets. That was far beyond their mathematical
ability. So they worked in fact with three models, namely, (i) a vacuum
solution of Einstein's eqns. (5.28) and (5.29); (ii) the standard New-
tonian model of the solar system, perfected by Laplace and Poincare;
and (iii) a model that we may call Ptolemaic or, more exactly, Eudox-
ian, consisting of an imaginary sphere affixed to the Earth and homo-
centric with it, on which the observed positions of Mercury and the
Sun are recorded. Since the question at issue does not depend on the
accurate measurement of very short time intervals, all three models
share the standard (Newtonian?) time of astronomy. It is in model (iii)
that Mercury's perihelion - that is, the point on the sphere where it
comes closest to the Sun - advances by Aa in 100 years. This model
is purely kinematic and can be converted to either (i) or (ii) by a suit-
able coordinate transformation. In both cases, the transformation
wipes out approximately 5,025" from the perihelion advance. The
rest, amounting to some 575.5", must be explained dynamically. As we
saw, model (i) takes care of some 43" and model (ii) of the remaining
532". Seeing these figures, the reader may wonder why this was
regarded as a triumph of Einstein's theory, instantiated by model (i),
over Newton's, instantiated by model (ii). Now Einstein and his fol-
lowers were certainly unable to build a relativistic model of the solar

22
Apart from the given numerical values, this prediction is significant because in the
Newtonian theory a particle with Mercury's mass and simulating its motion around
an otherwise lonely Sun would describe an ellipse with the center of gravity of the
system at one focus and its perihelion would not precess at all. Thus, Einstein's dis-
crepancy with Kepler's picture of the Solar System is steeper than Newton's.
 7.2 Laws and Patterns 419

system that accounted for the full dynamically conditioned precession

of 575.5". But still, it was clear to everybody that in such a model -
given the formal relation between eqns. (5.29) and the Poisson equa-
tion of Newtonian theory (cf. point (D) in §5.4) - the precession
increase over 43" due to the presence of massive planets would not
differ significantly from the 532" attributed to their influence in the
Newtonian model.
My second example is more speculative. I mentioned earlier the sin-
gularity theorems proved in the 1960s by Penrose, Hawking, and
Geroch (§5.5, note 73). They imply that in a typical GR spacetime
there are black holes, that is, spacetime regions that absorb all the
energy they receive without ever releasing any. Although the evidence
for the actual existence of such regions is not entirely beyond question
even now, the theoretical study of the properties of black holes soon
became a flourishing academic industry. Bekenstein suggested that
certain properties of black holes might be connected with entropy, and
other important links to thermodynamics were proposed. However,
it did not seem possible to assign a finite entropy to a black hole,
since that would imply that it had a finite temperature and there-
fore, if surrounded by thermal radiation at the same temperature,
would remain in equilibrium with it. This seemed absurd, for the black
hole would inevitably absorb some of the radiation but could not emit
anything in return. To overcome this difficulty, Hawking (1974, 1975)
constructed a QM model of a black hole. According to QM, energy
trapped inside a potential well has a finite chance of tunneling through
the barrier surrounding it. This is, for example, what happens in
radioactivity, which QM conceives as the sudden, uncaused ejection of
certain quanta of matter and radiation from the atomic nucleus to
which they are normally bound. Hawking applied the same general
idea to black holes and calculated the time that it would take for a
black hole of given mass m to evaporate completely. (If m is the mass
of the Sun, the time in years is of the order of 1063.) On the "received
view" of theories, Hawking's procedure plainly defies logic. If GR and
QM are regarded as two sets of statements, their union is an incon-
sistent set that therefore cannot be satisfied. Worse still, according to
GR there are no black holes in the flat Minkowski spacetime underly-
ing Quantum Field Theory, let alone in the Newtonian spacetime
underlying nonrelativistic QM. However, from the structuralist stand-
point, both theories can very well cooperate on this issue, as on many
others. If there is a real black hole in the world, it can be represented
420 Perspectives and Reflections

both in a GR model and in a QM model. A difficulty would arise only

if the degrees of approximation claimed for the models cannot be
jointly attained because the incompatibility between the theories
forbids it.

7.3 Rupture and Continuity

In the Preface to the second edition of the Critique of Pure Reason,
Kant speaks of the moment when, in the full light of history, physics
entered at last "upon the secure path of science" after centuries of
"random groping" (1787, p. xiv). That was in the first half of the sev-
enteenth century, when Galileo and Torricelli carried out their experi-
ments. They understood "that reason has insight only into that which
she produces after a plan of her own, and that she must not allow
herself to be kept, as it were, in nature's leading-strings, but must
herself show the way with principles of judgment based upon fixed
laws, constraining nature to give answer to questions of reason's own
determining. [. . .] Reason, holding in one hand her principles [. . .] and
in the other hand the experiment which she has devised in conformity
with them, must approach nature in order to learn from it, but not as
a pupil who must listen to everything the teacher chooses to say, but
as an appointed judge who compels the witnesses to answer the ques-
tions which he puts to them" (p. xiii). Kant was persuaded that this
"revolution in the manner of thinking" had infallibly secured progress
"in endless expansion throughout all time" (p. xi). There is no indica-
tion that he ever considered the following question, which his judicial
metaphor immediately suggests: What would happen to the secure
progress of science if reason, the judge, on her own initiative or at her
sovereign's behest, would change the terms of her questions or the rules
for evaluating the answers? Although reason began to wield her prin-
ciples at a given time and place, Kant regarded them as being above
history. The interrogation of nature by reason between Arno and
Clyde, Elbe and Ebro, in the seventeenth and eighteenth centuries of
the Christian era, was, in his eyes, only an expression of reason's time-
less nature.
The unchangeable principles by which, according to Kant, the
human understanding must "spell out appearances in order to read
them as experience" (1781, p. 314) include Euclidian geometry and
Newtonian chronometry, mass conservation, strict determinism, and
instantaneous interaction at a distance (§§3.3 and 3.4). The advent of
 7.3 Rupture and Continuity 421

Relativity and quantum physics brought these principles to nought.

Quite reasonably, many philosophers saw here a practical refutation of
Kant's philosophy. Former Kantians, like Carnap and Reichenbach,
and other advocates of logical empiricism sought to separate the infor-
mative contents of science, provided by the senses alone, from its con-
ceptual scaffolding. Neglecting Kant's warning that, without concepts,
the senses are blind, they viewed scientific concepts as a mere instru-
ment, of arbitrary design, for codifying sense data in a way that is eco-
nomical and easy to handle. Having failed in his attempt to reduce
scientific discourse to the language of sense impressions, Carnap
adopted Neurath's physicalistic language (§7.1) and tried to show that
its "observational" vocabulary sufficed for conveying the full cognitive
contents of physical theories. The permanence of that vocabulary,
despite the drastic changes of theories, secured a connection between
the latter and facilitated their comparison. When Carnap and his
friends regarded everyday laboratory talk as a factor of continuity in
physics they certainly were on the right track. But their distinction
between observational terms - that are applicable to real life things and
processes - and theoretical terms - to be used only as chips in calcu-
lation - drove them into a blind alley, for the observations of greatest
interest for today's physics can only be precisely described and
profitably interpreted in a language that is loaded with theoretical
terms (and assumptions).
As I mentioned in §7.1, the "theory-ladenness" of observation was
rediscovered by Hanson (1958) and Feyerabend (1958, 1962) and
played a key role in the devastating criticism of logical empiricism that
they and other authors undertook c. 1960. By that same time, Thomas
Kuhn developed a vision of the history of every science as a succession
of ruptures or "revolutions" separated by periods of continuity or
"normal science". According to Kuhn (1962), scientific revolutions
break the continuity of "normal science" because they bring about a
change in the manner of thinking by virtue of which any postrevolu-
tionary theory is "incommensurable" with the pre-revolutionary
theory that it displaces. This radical conclusion is inevitable if the rel-
evant facts of observation can be articulated as such only in the context
and within the perspective of one or the other theory. In this case, evi-
dently, there is no independent data base with which the theories could
be compared. On the other hand, as Shapere aptly remarked (1966; cf.
1964), if successive theories are incommensurable for the stated reason,
one cannot even claim that they concern the same subject matter, for
422 Perspectives and Reflections

the breakdown of the earlier conceptual system entails the loss of all
references.
To counter this shocking result, Hilary Putnam (1970, 1973, 1975b)
put forward his doctrine of reference without sense. In contrast with
classical Fregean semantics, in which the concept one thinks fixes the
kind of things one means, Putnam's semantics anchors the reference of
each term to an original act of name-giving - in which, for instance,
Adam points to a river in Paradise and says: "that is water" - linked
to the word's present use by a causal chain of communications. If this
were so, scientific revolutions could not alter the meaning of the main
terms of science, and successive theories could easily be compared with
one another and with experience. But the doctrine of senseless refer-
ence, although perhaps admissible in the case of the proper names of
individuals, is inapplicable to general terms, because the recognition of
different individuals as instances of the same class depends on the
concept by which one grasps them. Thus, we could not regard the fall
of heavy bodies toward the center of the earth, the circulation of the
planets about the sun, and the universal recession of distant galaxies
as phenomena of the same kind if our concept of gravity were the same
as Aristotle's.
I do not intend here to fight a theory of meaning that is no longer
upheld by its author.23 I refer to it because it shows how seriously this
distinguished philosopher took the alleged incommensurability of
scientific theories and to what extremes he was ready to go to avoid
it. In my view, however, the thesis of incommensurability rests on a
misunderstanding. Certainly, there are ruptures in the history of
physics, as indeed there must be if there is genuine intellectual novelty;
but the ruptures heal because the same factors that promote them and
make them possible contribute to restore the continuity. I shall clarify
this view with examples, but I must warn first against a common
mistake. In the literature prompted by Kuhn's book (1962), the epony-
mous scientific revolution of the seventeenth century and the great
innovations in twentieth-century physics are often treated on par with
the minirevolutions that, it is said, are happening all the time in every
puny scientific specialty. This leveling approach threw light on the
similarities between major and minor sciences, but it also generated
confusion concerning the significance of their respective revolutions.

I did so, with apologies to Putnam, in Torretti (1990, §2.6).

 7.3 Rupture and Continuity 423

Surely it is an exaggeration to say that "after a revolution the scien-

tists respond to a different world" (Kuhn 1962, p. 110) if the revolu-
tion in question concerns, for example, the origin of heart disease or
the displacement of continental plates. Indeed, when the geologists, dis-
carding the very idea of terra firma, see continents gliding over magma
like skating boards they radically alter the interpretation of countless
phenomena in their area of study, but they can go on describing them
with the same terms of physics, chemistry, and common sense used by
their predecessors; the continental plates move in the same world in
which America, Eurasia, and Africa were previously thought to rest on
the earth.
A true "revolution in the manner of thinking" in Kant's sense makes
no allowance for a general - scientific or nonscientific - form of dis-
course in which the rival theories are embedded and through which
they can communicate. If there is a single coherent global way of orga-
nizing our experience and this suffers mutation, the theories that
precede and follow such change will not even acknowledge one another
as theories. A system of reason like the one proposed by Kant is
absolute and self-contained and leaves no room for a "manner of think-
ing" that is different from its own (cf. Davidson 1974). The statement
that scientific theories separated by a scientific revolution are incom-
mensurable rests on the following misunderstanding: Its supporters
continue to see human - or, at any rate, scientific - thought in the guise
of all-encompassing unitary reason a la Kant, although they regard it
as a product of history and liable to mutations a la Darwin. The fact
is that, if - as Kant taught - there is an architecture of human reason,
it is not all of one piece, like that of Escorial or Versailles, but more
like the cathedral of Santiago de Compostela, combining many styles
that do not spoil but magnify the beauty of the whole. A more illumi-
nating architectural metaphor was proposed by Wittgenstein, who
compared "language" to a city with a labyrinth of short streets with
little squares and houses of diverse ages at the center - which he assim-
ilated to everyday speech - , surrounded by modern suburbs with
straight and regular avenues and monotonous housing - resembling the
idiolects of science and technology (1958, I, §18). Note that this
metaphor does not imply that scientific languages must be translatable
to ordinary language nor that they draw their meanings from it; but
rather that prescientific language is older and more stable and remains
afoot when one of the special jargons is remodeled or removed, and
so can maintain a connexion that would otherwise be interrupted. In
424 Perspectives and Reflections

particular, physics, while its theories can be adequately expressed only

in a language teeming with mathematical terms and formulas, must
continually resort to ordinary language in its practice, whenever
instructions are given to set up or to handle an experiment.
Take an example. In §5.1 I mentioned Michelson's interferometer
(see Fig. 12), an instrument designed to measure the velocity of the
earth relative to aether, the massless carrier of light waves postulated
by nineteenth-century physics. The instrument was built, in agreement
with the aether theory, so that the pattern of fringes generated by
mixing light traveling from a common source along two different paths
would manifest the path-dependence of the speed of light. However,
no change in the fringe pattern was detected when the paths were inter-
changed. To explain this null result FitzGerald (1889) and Lorentz
(1892) assumed that solid bodies contract, due to some action of the
aether, in the direction in which they move through it. On the other
hand, Einstein's Special Relativity (1905r) implied a completely differ-
ent reading of the interferometer results. According to it, Michelson's
instrument cannot measure the speed of the earth in the aether, for the
new theory denies the very existence of an aether. So what the inter-
ferometer experiment compares is the speed of light along a path at
rest on a laboratory that, at least for this purpose, may be regarded as
an inertial system, and its speed along another path, perpendicular to
the former, but equally at rest on the same laboratory. It is no wonder
that the experiment fails to disclose any differences. Einstein's revolu-
tion turned Michelson and Morley's perplexing experiment into an
innocuous banality. This was paid for with a drastic change in all the
fundamental concepts of physics: time, distance, mass, force. But if the
sense of these terms has changed, how can they refer to the same phe-
nomena? If physicists talked only of what they grasp with their theo-
ries, the jump between two conceptually irreconcilable theories would
sever their references. Fortunately it is not so. Physical theories grasp
only an abstract and idealized aspect of life, but physicists often think
of other more concrete aspects of it, for example, when they buy cheese
and wine at the supermarket. When they ask the janitor to dust the
interferometer they will refer to the apparatus in ways that do not in
the least depend on their understanding of what goes on when it is put
to work. Such modes of reference are shared, of course, by relativists
and physicists of the Lorentz-FitzGerald persuasion.
This state of affairs fits well into the structuralist scheme, according
to which a physical theory consists of a mathematical structure and a
 7.3 Rupture and Continuity 425

collection of intended applications, that is, of aspects or fragments of

the world that are modeled by instances of that structure. If the mod-
eling is successful - within the admissible margin of imprecision - the
fragments or aspects in question are grasped and understood by the
theory, and its terms refer in effect to them.24 But there must be a way
of referring to each one of those fragments and aspects without the
theory, for otherwise one could not designate them as candidates for
modeling. Such forms of reference are probably confused and do not
properly delineate their referents or contribute in any way to under-
stand them; but they provide a semantic handle by which to keep a
hold on the referents even if the modeling fails and the theory's con-
cepts merely glide over them. A physicist must be able to speak of that
which he does not understand; otherwise, a baffling observation or an
experimental result contrary to prediction would fall outside the reach
of discourse and could not act as a catalyzer of conceptual novelty. The
insufficiencies of accepted theory, the anomalies that will lead to its
breakdown, can only be signaled in a paratheoretic language that acts
as a factor of continuity.25

24
For instance, NASA engineers who speak of a spaceship "entering the gravitational
field of Mars" grasp both ship and planet as massive bodies that are subject to mutual
gravitational attraction in accordance with Newtonian theory.
25
D. G. Mayo's philosophy of experiment provides further evidence that physical
theories must communicate among themselves to do their jobs and should therefore
be embedded in a common discourse. Taking her cue from Suppes's 1962 criticism
of "philosophers of science [who] overly simplify the structure of science", Mayo
argues convincingly that every experiment involves the interplay of a host of models
(1996, Ch. 5 and 7). Although the breakdown and classification of the models used
in an experimental inquiry are not "a cut and dried affair" (p. 129), she groups them
for elucidation in three levels: (a) the primary model of the physical theory being
tested; (b) the experimental model that, on the one hand, specifies the key features
of the experiment and states the primary question or questions with respect to it and,
on the other hand, specifies analytical techniques for linking the data to that ques-
tion or questions; (c) the data models that articulate and "massage" the raw data of
observation in the manner required for the application of the experimental model's
analytical tools. Take for example the confrontation of General Relativity and other
modern theories of gravity with experimental data drawn from the Solar System.
Solar system experiments are modeled after a theory known as the parametrized post-
Newtonian (PPN) formalism. "The PPN framework takes the slow motion, weak
field, or post-Newtonian limit of metric theories of gravity, and characterizes that
limit by a set of 10 real-valued parameters. Each metric theory of gravity has par-
ticular values for the PPN parameters" (Will 1981, p. 10). Suitably massaged results
of astronomical observations, organized into appropriate data models, supply the
426 Perspectives and Reflections

There is another factor of continuity in the history of physics that I

believe is responsible, more than anything else, for the unity displayed
by modern mathematical physics since Galileo, despite all the changes
that it has gone through. It lies at the core of every physical theory,
which, as we have seen, can be explicated as a mathematical species of
structure. As such, it has its definite place in the ideal universe of struc-
tures and is clearly and distinctly related to other structures, including
those at the core of other physical theories. For example, the cosmo-
logical models of General Relativity are smooth four-dimensional man-
ifolds with a semi-Riemannian metric satisfying the Einstein field
equations (5.31) and possibly other conditions - such as stable causa-
tion - that are often prescribed to exclude monsters. Thus, they form a
subspecies of the much broader species of ^-dimensional Riemannian
manifolds, which is a part of the smooth manifolds, the full class of
which may be regarded in turn as a distinctive kind of topological
spaces. This example is particularly clear, because the said species of
structure, of which the relativistic worlds are very peculiar instances, are
by themselves the subject of important branches of mathematics. In
most cases, however, the conceptual core of a physical theory
must be explicated by a highly idiosyncratic species of structure, com-
bining subspecies of several familiar species in a tangled web.26 In
all cases, however, the rich available stock of structures that pure

measured values of those parameters. The measured value of this or that parameter
can then be compared with the different values assigned to it by the diverse theories
of gravity. In this way, in each particular Solar System experiment the same PPN
model of experiment mediates between the data and several alternative primary
models, based on General Relativity and its rivals. Mayo's approach undercuts
Kuhnian incommensurability. The theory instantiated by any given model is a crea-
ture of human thought and therefore, at some point, has come out of the blue.
However, the theories that experimental physics teems with do not stand in isolation,
but must be so conceived that they can be linked with other such theories via the
several model levels. Nor are these theories closed worldviews. They range from the
humble rules for drawing histograms and for calculating averages to the lofty con-
ceptual frameworks of classical dynamics and electromagnetism, relativity and
quantum mechanics. But even the latter, for all their weltanschaulich posturing, are
made to work only in a piecemeal fashion, through the testing of detachable hypothe-
ses in separate experiments. Two theories corresponding to diverse primary models
may differ deeply in their basic concepts; but, if any one of them is meant to take the
place of the other, they must both be linked to the same experimental models, or, at
any rate, they must have application to the same data models. Thus, through such
links, they are brought under a common standard.
26
The rational reconstruction of such systems is a favorite subject of research in Sneed's
school. See Moulines (1975) and Bartelborth (1987).
 7.3 Rupture and Continuity 427

mathematics studies with great generality not only furnishes physicists

with a practically inexhaustible source of revolutionary concepts, but it
also facilitates the comparison between new theories and their prede-
cessors. Thus the very factor that contributes the means for innovative
rupture makes it possible to determine the conceptual link between the
new and the old theories and to perceive the continuity that binds them.
I shall illustrate this idea with two examples. The first one is quite
modest and therefore can be explained in detail. I shall compare the
central notion of Lorentz's prerelativistic electrodynamics with the
homologous notion in Minkowski's relativistic electrodynamics.
Lorentz's theory is a modified version of Maxwell's (§4.2). In it, the
whole universe is filled with a massless aether, none of whose parts can
suffer displacement relative to its other parts. This substance penetrates
the innermost recesses of massive matter. The aether is the seat of
forces, subject to Newton's Second Law, that act on the electrically
charged bodies that move or rest in it. The forces are represented by
two time-dependent vector fields, the electric field E and the magnetic
field B. The total force F exercised at a particular time on a particle
with charge q moving through the aether with velocity v is given by
F = 4(E + vxB) (7.3)
the values of E and B being taken at the point where the particle is
located in the aether at that time. The said local, momentary values of
the fields depend - with delay - on the positions and velocities of all
the electric charges in the world. The relations between the electric and
magnetic fields and the distribution of electric charges and currents are
governed by the Maxwell equations (eqns. I-IV in Chapter Four, note
42). Let x, y, z be Cartesian coordinates for a reference frame at rest
in the aether. The vectors E and B can be analyzed, as usual, into their
components relative to the said coordinates (cf. eqn. 4.11):
E = (£x,£y,£j B = (B x ,B y ,Bj (7.4)
The numerical value of the components can be obtained by measuring
various electromagnetic effects in the laboratory.
In relativistic electrodynamics there is no aether. The Maxwell equa-
tions and the Lorentz force (7.3) take the same form relative to any
inertial frame of reference. However, a charged particle of mass m evi-
dently cannot experience the same acceleration F/ra with respect to an
inertial frame relative to which it moves with constant velocity v and
with respect to the inertial frame in which it is at rest. Neither can the
field vectors E and B be the same in two inertial systems in motion rel-
428 Perspectives and Reflections

ative to each other, for the distribution of electric currents cannot be

the same in both. Therefore, the electric and magnetic fields depend on
the arbitrary election of a reference frame and do not represent self-
subsisting physical realities. Nevertheless, as Minkowski showed, the
components of E and B relative to a Cartesian coordinate system (x,y,z)
anchored to an inertial frame happen to be the components of a geo-
metric object on Minkowski spacetime relative to the spacetime chart
(t,x,y,z) adapted to the same inertial frame and consisting of Einstein
time t and the said Cartesian coordinates x, y, z. A geometric object
is, as such, independent of the chosen reference frame, but its analysis
into components relative to a particular coordinate system varies, with
the latter, according to definite rules characteristic of each type of
object. The object in question here can therefore be regarded as the
appropriate mathematical representation of a single physical reality, the
electromagnetic field, which from a human standpoint, bound to this
or that inertial frame, is decomposed in a natural way into an electric
field and a magnetic field. Minkowski conceived the electromagnetic
field as a field of so-called polar vectors, which, on a 4-manifold, has
precisely 6 components relative to any particular chart. Today we con-
ceive it, in a mathematically equivalent way, as a field of antisymmet-
ric tensors of type (0,2). On a 4-manifold, a geometric object of this
type has 16 components relative to a given chart, but, due to anti-
symmetry, no more than 6 of them will be significantly different.
Specifically, if we denote the electromagnetic field by 3F and its 16 com-
ponents relative to (t,x,y,z) by 3% cFtX9 and so on, we have that - with
the sign conventions used by Feynman et al. (1964) -
CJTp __ Q2P —— ___ Z-7 (Jip __ Q2p —— J7 O/^ ._ CJTp __ E1
<r tx — ^xt — ^x ^ty — ^yt — ^y ^ tz ~ ~^xt — *-Jz

(7.5)

Therefore, the 16 components of 3F relative to the Lorentz chart

(x,y,z,t) are given by the following matrix:

0 -Ex ~Ey -Ez

Ex 0 -Bz By
(7.6)
Ey Bz 0 -Bx
E, B. 0
 7.3 Rupture and Continuity 429

Thus, although the relativist and prerelativist conceptions of the

physical world in general and of electrodynamics in particular are rad-
ically different, the numbers that determine electrodynamic phenom-
ena in accordance with the Maxwell equations have a definite role to
play in each theory. Measured by the same methods in the same labo-
ratories, they will be the same numbers even though they stand for
physical quantities conceived in a completely different way. This
example suggests that, among the several adjectives available for
describing the relation between a revolutionary physical theory and the
theory displaced by it, 'incommensurable' was not a remarkably
thoughtful choice.
My second example is more ambitious, and I shall barely sketch it.27
I have just recalled that models of General Relativity are Riemannian
manifolds. This concept was introduced by Riemann in order to furnish
physicists with a generalization of Euclidian space, to which they could
resort if, as Riemann anticipated, physical data obtained with methods
of ever greater precision would not fit into the Procrustean framework
of traditional geometry. Riemann was convinced that metrical relations
in a physical continuum depend on the natural forces that act in it. He
therefore thought it unlikely that the geometry that has served us so well
at the human scale could also be adequate at the cosmic and the mole-
cular scales. As a first step beyond Euclid, Riemann proposed a family
of metrics that agree optimally with Euclidian metrics on a neighbor-
hood of each point of space (see §4.1.3). The semi-Riemannian metrics
of General Relativity are designed to agree optimally on a neighborhood
of each point of spacetime with the Minkowski metric (see §§5.2 and
5.4). General Relativity is a geometrodynamic theory of gravity. A par-
ticle moving under the sole influence of gravity describes a spacetime
geodesic: Its worldline depends on geometry alone. But geometry, in
turn, depends on the forces of nature: The metric is determined by the
distribution of matter. This theory of gravity has precious little to do
with Newton's and is apparently incomparable with it. Nevertheless, a
few years after Einstein published General Relativity (1915i, 1916e),
Elie Cartan (1923) put forward a geometrodynamic formulation of
Newton's theory that is equivalent to its classical formulation (by
Laplace and Poisson; see §4.2). Cartan's formulation was further elab-
orated by Friedrichs (1927) and Havas (1964). In it, as in Einstein's
theory, a material particle influenced solely by gravity describes a space-

27
For details, see Earman and Friedman (1973).
430 Perspectives and Reflections

time geodesic, and the metric that fixes what spacetime curves are geo-
desic depends on the distribution of matter. The differential equations
governing this dependence lead to precisely the same predictions as the
classical Newtonian theory. In this way, the versatile mathematical
structure that Riemann invented on purpose to facilitate a finer adjust-
ment of physical theories to experience has made it possible to refor-
mulate an old theory in terms that allow a comparison with its
revolutionary successor. The new formulation is of course anachronis-
tic. (It is also needlessly clumsy, and nobody would dream of using it to
calculate the orbit of a communications satellite.) However, - in stark
contrast, say, with renderings of Greek poetry in modern languages - it
exactly matches the original, because only neat mathematical concepts
are at play.28
Thus, the tradition of physics has continued across all ruptures
because the same factors that contribute to innovation also facilitate
comparison between its successive stages. However, the historical con-
tinuity of physics does not imply that it advances on a "secure path"
toward a preestablished goal, nor that it is impossible to develop a dif-
ferent physics that is mathematical and experimental like ours, but is
incompatible and even incomparable with it. To prove that such an
alternative physics is possible one usually invokes the so-called
Duhem-Quine thesis, viz., that a given stock of empirical information
can be approximated by an infinite number of different theories. This
thesis is questionable, for a stock of data can hardly exist apart from
the "manner of thinking" that articulates it. However, "manners of
thinking" are not born, armed to the teeth, out of Zeus's brain, but
grow contingently in the course of history. An alternative physics could
arise elsewhere in the universe or issue from an unorthodox deviation
of ours (see Cushing 1994). After some generations and a few turns of
the road two intellectual traditions can become mutually incompre-
hensible even though they share the same roots.

28
A historically more significant example is mentioned by Falkenburg: In a very
influential paper, Eugen Wigner (1939) classed all the solutions of conceivable rela-
tivistic field equations - including, among others, the Maxwell equations and the
Dirac equation - according to the irreducible representations of the Poincare group.
"This classification is not linked to a particular formulation of quantum theory
[...], but embraces all relativistic theories of free fields, that is, it extends to a whole
class of physical theories which are characterized by a certain mathematical structure
of the theoretical entity 'field'" (Falkenburg 1995, pp. 229-30).
 7.4 Grasping the Facts 431

7.4 Grasping the Facts

Particular facts cannot be grasped as such unless general concepts are
brought to bear on them. Neglect of this truism has been the source of
endless blunders in philosophy. On the other hand, it borders paradox,
for how can you pinpoint a fact to bring a concept to bear on it unless
you have somehow grasped it? The classical philosophical solution of
this difficulty, due to Aristotle, is that the individual things themselves,
present to the senses, convey to any intelligent observer the concepts
under which they, their properties, and relations properly belong. From
the commonsense standpoint of someone steeped in the routines of
civilized life this seems obvious. Don't tables signal us to eat or write
on them, running water to quench our thirst, red traffic lights to stop?
It is true that we often make mistakes, but those who consciously or
unconsciously endorse the Aristotelian solution are not disturbed by
them (although mistakes are facts too, which naturally do not
announce themselves as such). The average adult members of a stable
community, in firm possession of their mental faculties, discount errors
in the classification of familiar objects as transient inconveniences,
which are easily overcome by increased attention and a few simple
tests. After all, how could such errors ever be corrected except by giving
a closer look to the state of affairs and letting it speak for itself? As
for the unfamiliar, there is a common human tendency - abundantly
documented in the annals of exploration - to assimilate it to the famil-
iar, and then, little by little, to make allowance for the differences. This
suggests that novel facts do not sport the concepts by which we grasp
them, but that we must draw these concepts from our stock-in-trade,
refurbishing them as needs be for their new jobs. It is therefore unlikely
that Aristotelianism could have risen as an explicit philosophical doc-
trine in a fast-changing social environment, such as that associated with
modernity. Indeed, the unconscious Aristotelians we still meet every-
where tend to be the same people who methodically avoid the unfa-
miliar or refuse to acknowledge it as such.
As we saw in Chapter One, modern physics began in a thoroughly
anti-Aristotelian mood, motivated by the perceived inadequacy of tra-
ditional concepts for dealing with some common phenomena, notably
with the motion of projectiles. While Aristotelian science favored
loving attention to detail, through which alone one could succeed in
conceiving the real in its full concreteness, Galileo and his followers
432 Perspectives and Reflections

conducted their research with scissors and blinkers: Scissors to cut off
not just the subjective (§7.1) but also the irrelevant; blinkers to fix the
sight on what mattered and prevent distraction by what was being -
provisionally? - left aside. The natural processes and states of affairs
under study were represented by simplified models, manageable
instances of definite mathematical structures. The inevitable discrep-
ancies between the predicted behavior of such models and the observed
behavior of the objects they stood for were ascribed to "perturbations"
and observation errors.
Galileo proclaimed that the book of nature is written in mathemat-
ical language (1623, §6; quoted in §1.3). However, natural phenom-
ena do not wear the script on their face. Kepler did not read off the
sky the concept and the properties of the ellipse to which he fitted the
observed positions of Mars; he found them in Apollonius's Conica. It
was all right for Kepler to repeat that "God is always doing geome-
try".29 But how do we mortals get to know the divine geometry of
things? Philosophers of mathematics are still bedeviled by this ques-
tion. However, in the seventeenth century the following reply, given by
Kepler, was not immediately ruled out of court:

Geometry, being part of the divine mind from time immemorial, from
before the origin of things, being God Himself (for what is in God that
was not God Himself), has supplied God with the models for the cre-
ation of the world and has been transferred to man together with the
image of God.
(Kepler GW, VI, 223; quoted in Caspar 1993, p. 271)

This is not too far from Descartes's thesis that we know bodies through
our inborn, "clear and distinct" idea of extension; after all, innate ideas
are supposedly implanted in us in the act of creation directly by God
Himself. And Malebranche's contention that we "see" all corporeal
things "in God" and are thus directly acquainted with their mathe-
matical patterning may well pass for a more precise - and more daring
- restatement of Kepler's.
Today few would countenance basing physics on our knowledge of
God. However, one implication of seventeenth-century theological
commitments has lingered on as a source of confusion. Like the smile

29
xov 9E6V del 7£CO|i£Tp£Tv. The Greek phrase, fondly quoted by Kepler (GW, VI, 298),
was attributed in Antiquity to Plato; see Plutarch (Qu. Conv. 8.2).
 7.4 Grasping the Facts 433

of the Cheshire cat, the idea of a ready-made world continues to haunt

a philosophical tradition from which the idea of its Maker has long
vanished. Although everyone is aware that physicists work on selected
aspects of reality using idealized models that cannot claim perfect ade-
quacy, the dream persists of a final theory of everything representing
the true mathematical structure of the universe.
Kepler believed that he had discovered that planets really describe
ellipses with the Sun at one focus. He ascribed to observational inac-
curacy the discrepancies between the elliptical path that he traced for
Mars and the positions recorded by Tycho Brahe. With the introduc-
tion of the telescope - in Kepler's own lifetime - the quality of astro-
nomic observation improved substantially and the said excuse was no
longer viable. However, the theory of universal gravitation put forward
in Newton's Principia (1687) satisfactorily explained both the successes
and the failings of Kepler's "laws" (§2.3). In the course of the two fol-
lowing centuries Newtonian celestial mechanics was able to account,
within the very narrow margins of error admissible in telescopic astron-
omy, for every motion tracked in the Solar System (except a small frac-
tion of Mercury's perihelion advance). So physicists and astronomers
got into the habit of thinking that Newton's law of gravity was exactly
obeyed by every bit of matter, although all its applications to particu-
lar celestial bodies involved simplified models of the objects in
question.
This understanding of physical theory weighed heavily on Kant's
decision "to deny knowledge so as to make room for faith" (1787,
p. xxx). He accepted the exact validity of Newtonian mechanics for
phenomena, and even argued that some form of instantaneous action
at a distance was a necessary prerequisite of human experience. But he
denied that any branch of human science could apply to the things-in-
themselves on which - he assumed - the phenomena are grounded. In
this way, our rational will could have free play in the realm of the really
real, even if it was excluded from everything we can see and touch by
the universal determinism that Kant and his contemporaries believed
was entailed by the exact validity of mechanics. What solace a serious
moral thinker can get from this strange doctrine is not a subject I can
discuss here. What matters to us is Kant's philosophy of phenomena.
Although I already dealt with it at some length in Chapter Three, there
are a few points that I wish to emphasize now.
First, Kant detheologized physics: No matter what its accuracy, its
sole concern are the objects of human experience, subject to the specific
434 Perspectives and Reflections

conditions of human sensibility (space and time), and therefore it

cannot in any way approach a God's eye view of reality.
Second, there is some tension between Kant's doctrine of the antin-
omies (§3.5) and his understanding of "nature" as "the connection of
phenomena determining one another with necessity according to uni-
versal laws" (1787, p. 479). For him, the "thoroughgoing connection"
of events according to unchangeable laws is a principle "from which
no deviation is allowed and no phenomenon can be exempted" (1787,
pp. 564, 570). And yet the antinomies entail that the whole of nature
is only an Idea (in Kant's sense) that is incapable of realization although
it is indispensable as a guiding principle of science (cf. 1783, §§40 and
56). So the connection of mutually determined phenomena cannot be
thoroughgoing, but only on the way of becoming so. Indeed, Kant's
solution to the First and Second Antinomies rests on the premise that
phenomena are not fully determinate in every respect (omnimode deter -
minata) like things-in-themselves, but instead they gradually approach
determinacy in the course of the progressive construction of experience
by the human understanding.
Finally, and most significantly for us, Kant offered a novel, power-
ful account of the use of mathematical concepts in grasping physical
facts that does not depend on God's geometrizing nor on our being
made privy to God's plan of creation. Concepts are needed to "spell
out appearances in order to read them as experience" (1781, p. 314),
in other words, to articulate the fugitive flickerings of sensory aware-
ness into a steady display of objects, connected in space and lasting
through time. As I noted in §3.4.1, this Kantian doctrine of the role of
concepts in the constitution of objectivity has two sides. On the one
hand, all concepts carry necessity with them and therefore, on Kant's
account, bring it into the facts that are constituted and grasped by
means of them. On the other hand, according to Kant, the very general
concepts listed in his table of categories are themselves specifically nec-
essary for there to be any facts. This second side of Kant's doctrine
mattered most to him but is hard to defend now that physics has dis-
carded the concepts and principles that Kant incorporated into his
"metaphysics of experience". Philosophical attempts to salvage Kant's
conception of a changeless human reason by watering it down and
moving to weaker principles and more abstract categories than his do
not look promising. One cannot easily forget the case of Bertrand
Russell (1897) who, renouncing Kant's fixation with Euclidian geom-
etry, equated the "form of externality" with the nonmetric projective
 7.4 Grasping the Facts 435

space that, according to Felix Klein, supported equally well the para-
bolic (i.e., Euclidian), hyperbolic (Lobachevskian), and elliptic metric
geometries (see §4.1.2). It then turned out that when physical geome-
try was actually revolutionized by Einstein the new metric adopted was
none of the above and would not fit into Russell's scheme (or Klein's).
A more likely way of preserving the gist of Kant's teaching was tried
by Strawson. His "descriptive metaphysics" concerns the "massive
central core of human thinking which has no history - or none
recorded in histories of thought". This core comprises "categories and
concepts which, in their most fundamental character, change not at
all. Obviously these are not the specialties of the most refined think-
ing. They are the commonplaces of the least refined thinking; and
are yet the indispensable core of the conceptual equipment of the
most sophisticated human beings" (Strawson 1959, p. 10). Strawson's
"massive core" may perhaps be equated with the commonsense back-
ground from which physical theories and other such intellectual
systems grow and across which they communicate. But then such a
background core is not so frightfully stable as Strawson would have it,
and it does not enjoy any privilege over physical theories when it comes
to setting the terms in which natural phenomena are to be described
and understood.
Kant's changeless scheme of forms, categories, and Ideas suggests
that he still believed that our species was created at one stroke by God
(although he would not indulge in asserting this as a scientific or philo-
sophical truth). After Darwin, we conceive human reason as a gradu-
ally evolving complex of various, possibly colliding, currents, that must
be investigated by the hermeneutic methods of cultural anthropology
and intellectual history, not by phenomenological introspection and
transcendental deduction. The idea that reason is open-ended encour-
ages the exercise of scientific freedom. The loss of certainty is more
than compensated for by improved efficacy, as anyone can readily
verify by comparing the last 400 years of physics with the preceding
2,000. Scientists remain bound by tradition but are free to radically
reshape it. In this sense, they face the surprises of experience with unfet-
tered minds. No longer constrained by the purportedly unconditional
validity of their principles, they may turn to their genius or muse for
the concepts wanted to grasp the facts.
Einstein was probably the first physicist to become fully aware of
this freedom. As he put it, the "concepts and fundamental laws" of
physics are "free inventions of the human mind (freie Erfindungen des
436 Perspectives and Reflections

menschlichen Geistes), which cannot be justified a priori by the nature

of the human mind nor in any other way" (1933, p. 180). Physics "is
in a state of evolution", and its "basis cannot be obtained through dis-
tillation by any inductive method from the experiences lived through,
but [...] can only be attained by free invention" (1936, p. 96). Ein-
stein noted that "Newton, the first creator of a comprehensive, oper-
ative system of theoretical physics, still believed that the fundamental
concepts and fundamental laws of his system could be derived from
experience" (1933, p. 181). Newton surely felt uncomfortable about
the concept of absolute space, for he must have seen that nothing in
experience appeared to correspond to it. He also felt uneasy about
forces acting at a distance. "But the enormous practical success of his
theory may well have prevented him and the physicists of the eigh-
teenth and nineteenth centuries from perceiving the fictional character
of the foundations of his system" (Ibid.). They were persuaded that
"the fundamental concepts and fundamental laws of physics are not in
a logical sense free inventions of the human mind, but can be derived
from experiments by 'abstraction', i.e., by a logical procedure" (p.
182).

Indeed, clear knowledge of the incorrectness of this view was first pro-
cured by the general theory of relativity. It showed that one could do
justice to all the relevant empirical facts on a footing (Fundament) vastly
different from Newton's, in an even more satisfactory and complete way
than on his. But quite apart from the question of superiority, the fictional
character of foundations (Grundlagen) was made perfectly evident by
the fact that two essentially different foundations can be exhibited, both
of which amply agree with experience; this proves at any rate that every
attempt at a logical derivation of the fundamental concepts and funda-
mental laws of mechanics from elementary experiences is doomed to
failure.
(Einstein 1933, p. 182)

Like genetic mutations, the concepts of science come out of the blue.
While mutations have been traced - "in principle" - to chemical
changes attributable to random quantum interactions, we have no
inkling of the origin of concepts. To suggest that the advent of Special
Relativity - that is, the switch from Galilei invariance to Lorentz invari-
ance - is related to modern aesthetic and moral relativism, or that
Heisenberg's "uncertainty" relations reflect the monetary and political
instability of the Weimar Republic, makes for entertaining science jour-
 7.4 Grasping the Facts 437

nalism but throws no light on the creation of scientific ideas. But, of

course, what matters most is not the source of our thinking, but its
aptness. So philosophers of science traditionally pay great attention to
this question: How can we tell that our concepts fit the facts that we
intend to grasp by means of them?
For Kepler and his contemporaries this amounted to: How can we
know that the concepts of science are the same that God had in mind
when he created the world? Using the metaphor that modern science
unwittingly inherited from Callicles, via the Church fathers, the same
question is often stated thus: How can we establish that a general state-
ment of fact expresses a true law of nature? Having realized that we
plainly cannot do it, philosophers set out to find clear rules and firm
criteria by which to evaluate the probability that such general state-
ments might express true laws, given the available evidence.
Research in this area - known as inductive logic, but also, more
modestly, as confirmation theory - is one of the blindest alleys in twen-
tieth-century philosophy. Until the 1950s it mainly turned around the
logical notion of probability. Keynes (1921) conceived probability as a
quantitative relation between statements based on their meanings. A
similar idea was adumbrated by Wittgenstein (1922, 5.15-5.156).
From this standpoint, given any two statements h and e, the probabil-
ity that h is true if e is true is a real number p(h,e) in the closed inter-
val [0,1], which depends solely on what h and e say. Thus, a statement
such as 'p(fc,e) = 0.75' is a logical truth. If h is a scientific hypothesis
and e is the conjunction of all the available evidence, p(h,e) can be
regarded as a measure of confirmation of h by e. Clearly, a relation of
this sort can only be made precise for statements couched in a formal
language, with fixed, explicit syntactic and semantic rules. Carnap
(1945a, 1945b, 1950, 1952) struggled steadfastly to meet this require-
ment. He was able to define logical probability only for languages too
simple to be of much use in science, under fairly arbitrary conventions
regarding the probability of basic alternative states of affairs. Most
frustratingly, every function p in Carnap's "continuum of inductive
methods" takes the value p(h,e) = 0 if h states a hypothetical law of
nature and e is a finite conjunction of particular statements of fact.
In the second half of the twentieth century a different approach to
the confirmation of scientific hypotheses by empirical evidence has won
increasing support among philosophers. Known as Bayesianism - after
Thomas Bayes, an eighteenth-century English cleric and statistician -,
it actually turns on the conception of subjective probability indepen-
438 Perspectives and Reflections

dently introduced by Frank Ramsey (1926) and Bruno de Finetti (1930,

1931, 1937). This can be roughly summarized as follows. A real
number br(s) is assigned to the "degree of belief" of a person r in a
statement s. This quantitative assignment reflects the odds that r is
willing to give or take in a bet concerning s. Specifically, r will be ready
to pay, if s comes true, br(not-s)/br(s) times the amount she is to receive
if s comes false. Ramsey and de Finetti proved that such betting behav-
ior would lead to a loss, come what may, unless the function br com-
plies with the following conditions:
Bl For any statement s, 0 < br(s) < 1.
B2 For any two mutually incompatible statements s and t, br{s v t) =
br(s) + br(s).30
B3 In particular, br(s v not-s) = 1.
Therefore, our authors concluded, a rational person must adjust its
betting rates or "degrees of belief" so that they agree with conditions
B1-B3.31 Suppose now that br is restricted to statements that assert the
occurrence of some event or are truth-functions or first-order general-
izations formed from such statements. Let S denote the set of events
that occur if and only if statement s is true. If br complies with B1-B3,
the function p, defined by p(S) = br(s) for any such statement s, is a
probability function on events, namely, r's personal or subjective prob-
ability function. Bruno de Finetti argued that this is the only concept
of probability that makes sense and is not sheer fantasy (1974,1, 3-4).
He was careful to insist that probability thus conceived applies to
events only (Ibid., I, 139), and indeed only to uncertain - that is, future
or unknown - events. Run-of-the-mill Bayesians are much less fastid-
ious, and they are ready to speak about the subjective probability of
major physical theories, such as General Relativity, and to evaluate
how it changes in the light of incoming data. Since my aim here is not
to defend Bayesianism, but just to cursorily inform about it, I take for
granted that this manner of speaking is not plain rubbish, and I shall

sv tis the statement that is true unless both s and t are false. To express in English the
exact force of the connective v, bureaucrats have created the conjunction 'and/or'.
Let me note, in passing, that if this linkage between rationality and betting behavior
holds good, a rational person must know every logical and mathematical truth. For
if s is such a truth and r is uncertain about s, so that br(s) < 1, then, under the usual
assumptions of the Ramsey-de Finetti argument, r can be cornered into making bets
she must lose come what may.
 7.4 Grasping the Facts 439

spend no effort in trying to justify it. Therefore, like the Bayesians, I

write p(s) for the subjective probability of statement s even if s is not
restricted in the way I indicated above. Following the usual practice in
the theory of probability, if p(s) > 0, one defines the subjective proba-
bility p(t\s) of statement t conditional on (the truth of) s by:

^JJ^- (7.7) 32
If p(t) > 0,

p(s\t) = P ^ A s ) ( 7- g)

Therefore,

p(*) = ^ E W (7>9)

This result is the simplest form of Bayes's Theorem, an elementary

proposition of the mathematical theory of probability on which
Bayesians build.33 Although Bayesianism has grown ever more
complex, as its advocates tease their brains to overcome the difficulties
that beset it, its basic ideas are easy - not to say facile - and can be
explained briefly. I shall focus on three. Let p stand for the subjective
probability function of a particular person at a particular time; if it is
necessary to distinguish successive times, I use numerical subscripts,
thus: pi, p2, and so on.
(i) Let h be a scientific hypothesis and e a statement of a particular
fact such that h entails e. Let p(h) and p(e) be both greater than 0 and
less than 1. Clearly, p(e\h) = 1. Therefore, p(h\e) = p(h)/p(e). Thus, the
probability of h is necessarily less than the probability of h given e, if
e is a logical consequence of h and 0 < p{e) < 1. If these conditions are

32
t A s is the statement that is false unless both s and t are true. The connective A is
well represented by the English conjunction 'and'. To avoid the restriction of condi-
tional probabilities to cases in which the condition has non-zero probability, some
authors favor taking conditional probability as a primitive concept, to be character-
ized by axioms. These axioms then replace the standard axiomatic characterization
of absolute probability.
33
Bayes's Theorem takes its name from Thomas Bayes not because he discovered it, but
because of the way he wielded it in his famous paper of 1763. Earman (1992, Ch.
1) gives a sharp and most instructive assessment of the relevance of Bayes's work to
contemporary Bayesianism.
440 Perspectives and Reflections

met, we say that e is confirmatory of h. This result, and others like it,
are touted by Bayesians as providing a satisfactory philosophical
justification of the commonplace view that confirmation by empirical
evidence increases the credibility of hypotheses in the eyes of rational
agents. However, the preceding argument shows that p{h\e) > p(h) only
if e is still uncertain, that is, if the evidence in question is not yet avail-
able; for, if e were known to the subject of the probability function p,
p(e) would equal 1, in which case p(h\e) = p(h).
(ii) Thus, the Bayesian proof that confirmatory evidence improves
the credibility of scientific hypotheses cannot simply rest on a result
like (i), which is solidly grounded on the mathematical theory of prob-
ability, but depends also on the following principle of conditionaliza-
tion: If pi and p2 are the probability functions of a rational agent right
before and right after the acquisition of new knowledge e, then p2(h)
= pi(h\e) for every hypothesis h such that 0 < p\(h) < I.34 There have
been clever attempts to prove that agents who do not comply with the
principle of conditionalization can be constrained to make bets that
they will lose come what may (Teller 1973; Skyrms 1987). If this is so,
conditionalization is, arguably, a necessary ingredient of rationality.
However, if the new knowledge e lies outside the domain of p! or, worse
still, if it provokes a change in the very concepts from which statements
are built, the principle of conditionalization cannot be applied.
(iii) The principle of conditionalization implies that rational agents
will bet with increasing confidence for hypotheses that are persistently
and consistently supported by incoming evidence, but not that they will
achieve certainty about such hypotheses, nor that they will eventually
agree on the betting rates. So, if (i) and (ii) were all there is to it,
Bayesianism would cut a poor figure as a philosophical theory of the
growth of knowledge, even if one turns a blind eye on the essential
restriction to conditionalization mentioned at the end of (ii). Therefore,
Bayesians take special pride in the mathematical theorems which estab-
lish that, under certain conditions, pn(h) converges to 1 as n grows
beyond all bounds if, for each w, the transition from pn to pn+i satisfies
the principle of conditionalization as applied to new evidence

34
Of course, in real life a global revision of one's subjective probability function may
be brought along not by the acquisition of some new certainty, but by a local change
in the said function. This possibility is taken care of by Jeffrey conditionalization (so
called after Richard Jeffrey). See Earman (1992, p. 34).
 7.4 Grasping the Facts 441

confirmatory of h. (See in particular Gaifman and Snir 1982, and the

comments in Earman 1992, Chapter 6.) This implies, in turn, that all
rational agents will ultimately agree. Still, the rate of convergence need
not be the same for different agents, and the theorems provide no hint
that might enable one to estimate this rate. They assure us only that
the subjective beliefs of rational agents converge in the long run to
practical certainty and unanimity. But then, as Keynes dryly remarked,
in the long run we shall all be dead. Moreover, as I noted at the end
of (ii), the Bayesian school, for all its mathematical sophistication,
remains committed to the feckless assumption that concepts and mean-
ings are fixed and that a rational agent will not be moved by empiri-
cal evidence to see things in a fundamentally different way.
Fortunately, the advancement of physics does not depend on the con-
vergent evolution of the subjective probability functions entertained by
rational human beings. It is enough that physics continues to produce
ever more comprehensive theories for successfully conceiving the great
families of phenomena we are wont to discern in nature. We cannot
judge a physical theory by the way it apes God's view of its subject
matter, for we have no idea how well it does it, or if it does it at all.
But this does not mean that we have to measure a theory's success by
how close to 1 is its average subjective probability, or by how fast this
average is increasing via conditionalization on new evidence.35 From
the standpoint reached in §7.2, it makes no sense to ask about the
probability of, say, Poisson's equation. For a Newtonian gravitational
field, the equation holds by definition. A meaningful question is
whether this body here - say, a comet, or a Saturnian moon - moves
through such a field, that is, whether it is part of a Newtonian gravi-
tational system or, more precisely, whether it, together with the whole
relevant fragment of reality, can be suitably modeled by a Newtonian
system. (Here 'relevant' means 'that must be taken into account in
order to understand and predict those aspects of the body's behavior
that interest us at present'; and 'suitably' means 'to within the admis-
sible margin of inaccuracy'.) Analogous questions are being asked all
the time in everyday life. Kant postulated a special faculty for answer-

35
With these remarks I do not intend to belittle the importance of statistical inference
in physics. For a powerful and mind-expanding statement of how it works in real
science I refer the reader to Mayo (1996). Mayo's "error statistics" continues the tra-
dition of C. S. Peirce and Egon Pearson, and keeps clear of the "Bayesian Way".
442 Perspectives and Reflections

ing them, which he called 'judgment' (Urteilskraft).36 In the enterprise

of grasping the facts, it relies on whatever concept or combination of
concepts looks most promising, yet it stands ready to replace them by
other, more befitting ones if and when they become available. Every-
day concepts, however, are flexible in a way that physical theories are
not. According to the conventional wisdom, "if it walks like a duck,
it swims like a duck, it quacks like a duck, then it is a duck". And a
duck's deviant behavior will not easily lead one to reclassify it as some-
thing else. (More probably, one will pronounce it "mentally ill", or
"acting under stress".) On the other hand, the unexplained perihelion
advance of Mercury by a mere 43" per century - less than eight-
thousandths of the amount observed - is enough to conclude that
neither the Sun-Mercury system, nor the entire Solar System, nor, for
that matter, any gravitational system in the world can be represented
to our full satisfaction by a Newtonian model.37 Due to their great
rigidity physical theories cannot supply us with resilient articles of
belief that could vie in endurance with myths and superstitions. But
this feature of mathematical physics is also a great source of power for
those who wield it with circumspection to find their way about the
world.

36
"Lack of judgment is properly that, which is called stupidity (Dummheit), and for
such infirmity there is no remedy" (Kant 1781, p. 133n). Certainly no algorithm to
"wash it out".
37
To be fair, I should recall that (i) the 43" secular perihelion advance was not felt to
be critical to the Newtonian theory of gravity until a viable replacement was avail-
able, viz., GR; and (ii) although Einstein had the Mercury anomaly in mind since he
began working on a new theory of gravity (see note 64 in Chapter Five), his main
motive for replacing Newton's theory was that it is not Lorentz invariant.
 Supplements

I. On Vectors
Set-theoretic concepts. Let S be a set. We write x e S to indicate that x
is an element of S. A set R is said to be a subset of S (symbolically: R c
5) if every element of R is an element of S. The set of all subsets of a set
S is called the power set of S and is denoted by &S. Set-theoreticians
assume that, given any set S, the set &S is also given. If A and B are sets,
the Cartesian product A x B is the set of all ordered pairs (<?,£>), such
that a belongs to A and b belongs to B. We write A1 for A x A, A3 for
{Ax A) x A, and A" for A""1 x A. If A and £ are sets, a mapping f of A
into B is a correspondence that assigns to each element x e A a unique
element /*(#) of £. I usually refer to the correspondence thus described
as 'the mapping f: A —> B by x •—> f(x)\ fis said to be a one-one or m/ec-
fr'fe mapping if, for any x and y, /"(#) = f(y) entails that x = y. f is said
to be surjective if it maps A owfo B, that is, if every element of B is
assigned by f to some element of A. /"is bijective if it is both injective
and surjective (i.e., if it is one-one and onto). 1

Groups and fields. A group is a set G that contains a distinguished

neutral element e and on which two operations are defined, viz., (I) the
group product, which is a mapping of G x G into G by (a,b) >-> ab, such

For those who have difficulty understanding the intuitive notions of 'assignment' and
'correspondence', mathematicians have contrived a definition of 'mapping' that is
based solely on the concept of 'set-membership'. Here it goes. We begin with the purely
set-theoretical definition of an ordered pair, proposed by Kuratowski (1921): The
ordered pair (x,y) is the set {{*},{*,)/}}. With this in hand, we define the ordered n-
tuple (xu . . ., xn), for any integer n > 2, to be the ordered pair ({xu .. ., xn_x), xn).
Let A and B be sets. The mapping / : A -> JB by x ^ f(x) is the ordered triple {A,B,C\
where C is a set of ordered pairs (x,f(x)), such that x e A, f(x) e B, and each element
of A occurs in one and only one ordered pair in C.

443
444 Supplements

that (Li) for any a in G, ae- ea = a, and (I.ii) for any a, b, c in G, a(bc)
= (ab)c; and (II) the inverse operation, a mapping of G into G that assigns
to each a an element a'1 (the inverse of <z), such thattfdT1= a~la = e. A
group is said to be abelian if, for any group elements a and b, ab = ba.
For further clarifications and examples, see §4.1.2.
A field can be readily described as a set F that incorporates two
abelian groups. (I) The additive group consists of F itself, with a group
product that is called addition and is symbolized by + ; the neutral
element is called 'zero' and denoted by 0; the inverse of an element a is
denoted by -a. (II) The multiplicative group consists of A{0) (i.e., what
remains of F when one removes 0); the group product is called multi-
plication and symbolized by x, or by •, or by mere juxtaposition (a x b
= a-b = ab); the neutral element is called 'one' and is denoted by 1; the
inverse of an element a is denoted by a'1 or by IIa. Multiplication is
extended to the whole of F by the rule: <zx0 = 0x<z = 0for any a in F.
Addition and multiplication are meshed together by the following law
of distribution: For any a, b, c in F, a(b + c) = ab + ac. The most fre-
quently encountered fields are the field Q of rationals, the real number
field R, and the complex number field C.

3 Vector spaces. The concept of a vector space is one of the most impor-
tant in modern mathematical physics. Let T be an abelian group, and
let F be a field (in physics usually either U or C). The elements of T are
called vectors; in this supplement, I represent them with boldfaced char-
acters (in particular, 0 denotes the neutral element). The group product
of T is called 'addition' and is symbolized by + . The inverse of v e T
is denoted by -v. The elements of F are called scalars. Arbitrary scalars
will be represented with lowercase italics. We define a mapping of F x
T into Y, called scalar multiplication, that assigns to each scalar a and
vector v a vector a\. T is a vector space over F - a real vector space if
F = [R, and a complex vector space if F = C - if the addition of vectors
meshes with the multiplication by scalars according to the following
rules: (i) a(v + w) = ay + aw; (ii) (a + b)\ = av + bv; (iii) a{b\) = (ab)v;
(iv) lv = v (where 1 denotes the one of F). Let ¥ b e a nonempty subset
of T, closed under vector addition and scalar multiplication (i.e., such
that, for any v and w in W and any a in F, v + w and av are in °W). Then
°W - with the operations of T restricted to W - is a vector space over F
on its own right and is said to be a subspace of T. (Exercise: Prove that
0 belongs to °W.)
Given n distinct vectors Vi, v 2 ,. . ., vw and any scalars au <32,. . . , am
the vector v = a^v^ + #2v2 + .. . + an\n is said to be a linear combination
of the Vi, v 2 ,.. ., vn. The vectors Vi, v 2 ,. . ., vn span the subset of T
formed by all their linear combinations. vl5 v 2 ,.. ., wn are linearly inde-
 I. On Vectors 445

pendent if a\\\ + <32v2 + .. . + an\n = 0 only if ax = a2 = . . . = an = 0. The

phrase 'linearly independent' conveys the fact that no vector belonging
to such a set is a linear combination of the others. (Exercise: Infer this
fact from the foregoing definition.) A finite set B of linearly independent
vectors is a basis of Y if Y is spanned by B. If B = {vu v 2 ,. . ., vM} is a
basis and v = EJl^v,, the scalars at are the components of v relative to
B. Suppose that Y has a basis formed by n vectors; this implies that
no set of linearly independent vectors contains more than n vectors.
Thus, every basis of Y has exactly n vectors. We say then that Y is an
w-dimensional vector space. Given any integer w, all w-dimensional vector
spaces over the same field F are isomorphic, that is, there are bijective
mappings of each onto the others that preserve vector addition and mul-
tiplication by scalars. If no finite collection of vectors is a basis, Y is said
to be infinite-dimensional. In 1.9 we shall see how the concept of basis
can be extended to infinite-dimensional spaces.

Inner product spaces. A real or complex vector space T is an inner

product space if it is endowed with an inner product, that is, a mapping
that assigns to each pair of vectors v and w a scalar subject to the con-
ditions IP listed below, and which I shall denote by (v|w). Let the field
of scalars be C, and let x* stand for the complex conjugate of x (in other
words, if x = a + ib, where a and b are real numbers, then x* = a - ib).
The inner product on Y obeys the following rules, for any vectors u, v
and w and any scalars a and b:

IP1 <v|w> = <w|v)*;

IP2 (vjtfw + bxx) = a(y\w) + fc(v|u);
IP3 (v|v) > 0, equality holding if and only if v = 0.

IP1 and IP2 jointly entail that (<av|w) = #*(v|w). If the field of scalars is
R, rule (i) naturally becomes (v|w) = (w|v). Other notations often used
for the inner product, instead of (v|w), are v w and (v,w).
The length of a vector v in Y is defined by ||v|| = V(v|v). Note that ||v||
is necessarily a real number (by IP1). The distance between two vectors
v and w is then given by ||v — w|| (I write v — w for v + -w). The mapping
v— • > ||v|| evidently satisfies conditions N1-N4 in §4.1.3 and is therefore
a norm in Y. It is assumed that Y has the topology induced by this norm,
so that, for each v in Y and p in 1R the set {u: u e Y and ||v - u|| < p} is
an open neighborhood of u. The inner product satisfies Schwarz's
inequality: |(v|w)| < ||v|| • ||w||.
By using the inner product, one defines the angle between two vectors
a and b as

a|b>
446 Supplements

This definition has a natural motivation in the space of directed segments

or "arrows" in Euclidian space, which is of course the prototype of all
vector spaces: Let a and b be two such arrows issuing from a point P;
assume that ||a|| measures Euclidian length. Then, by the trigonometric
"law of cosines", ||a - b|p = ||a|p + ||b|p - 2||a||-||b||cos4(a,b); so, to obtain
||a - b|p = ||a||2 - 2(a|b) + ||b|p we must put

The definition of angle entails that, for any scalar m and vector v,
cos4_(a,raa) = 1; therefore, 4_(a,raa) = 0. Thus, it is fair to say that mul-
tiplication by a scalar preserves the "direction" of vectors. If cos4_(a,b)
= 0, 4_(a,b) = nil and vectors a and b are said to be mutually perpen-
dicular or orthogonal. This happens if and only if (a|b) = 0. A set of
vectors (v,), indexed by a set of indices Q, is said to be orthonormal if,
for all indices i, / in Q, ||v/|| = 1 and (v/|v7) = 0 whenever i ^ j . An ortho-
normal set is said to be complete if it is not contained in any other such
set. {Exercise: Prove that every orthonormal set is a linearly independent
set.)
Linear and multilinear functions. Let T be a vector space over a field F.
A linear function on T is a mapping (p: Y —> F such that, for any v and
w in Y and any a and b in F, (p(<zv + bw) = <zcp(v) +fccp(w).If (p and \\f
are two linear functions on Y, their sum cp + \|/ is defined by the condi-
tion: (9 + V|/)(v) = (p(v) + \j/(v) for every v in Y. Clearly, (p + \\f is a linear
function on T, for ((p + \|/)(<zv + bw) = y(av + bw) + \\f(av + bw) = #<p(v)
+fccp(w)+ a\\f(\) + b\\f(w) = a(q>(\) + v|/(v)) +fc(cp(w)+ i|/(w)) = a(y + \j/)(v)
+ b((p + V|/)(w). Let the product ay of the linear function cp by the scalar
a be defined by (tf(p)(v) = ^cp(v) for every v in Y. ay is also a linear func-
tion. It is an easy exercise to show that the set of linear functions on
T, with these operations, is a vector space over F. We denote it by Y*
and call it the dual of Y. (Show, in particular, that the function which
assigns the 0 of F to every vector in T is linear and performs as the 0
of°r\)
Let T be an inner product space. If v is any vector in Y, the mapping
of T into F by u >-» (v|u) is a linear function (by IP2) and so belongs to
Y*. On the other hand, it can be proved that, for each (p in T*, there is
a unique v in T such that, for any u in T, (p(u) = (v|u). Thus, there is a
one-one correspondence between the vectors in T and the linear func-
tions in Y*. This is the origin of the notation that I use for the inner
product. It was introduced by the physicist Dirac, who denoted an arbi-
trary vector of a vector space Y by |u) and called it a ket, an arbitrary
vector of the dual space Y* by (v| and called it a bra, and let the bra[c]ket
 I. On Vectors 447

(v|u) stand for the value of (v| at |u). If (v| is precisely the bra that the
said one-one correspondence assigns to the ket |v), it turns out that (v|u)
is identical with the inner product of |v) and |u).
An w-linear function on T is a mapping of Yn into F that is linear in
each argument. For example, a bilinear function cp on T is a real-valued
mapping defined on Y2 that satisfies the following twofold condition of
linearity: For every a and b in F and every u, v, and w in T, (p(<2u + bv,w)
= <3(p(u,w) + £>(p(v,w) and (p(u,<rv + bw) = #(p(u,v) + £>(p(u,w). This
definition can be readily extended to scalar-valued mappings defined on
Y* x y*9 on y x Y\ and on y* x y. The last two would illustrate the
notion of a mixed bilinear function on Y and Y*.
Linear operators and matrices. Let Y and °W be vector spaces over a
field F. A linear operator Y —> °W is a mapping A of Y into W such that,
for any v and w in Y and any a and b in F, A(av + bw) = A(av) + A(bw).
This implies, in particular, that A(0) = 0. Linear operators can be added
together and multiplied by scalars, much like linear functions. Thus, if
A and B are linear operators T —> °W and a is any scalar, aA and A + B
are the mappings of T into T such that, for any v in T, (aA)(v) = A(av)
and (A + B)(v) = A(v) + B(v), respectively. Both aA and A + B are linear
operators T —> W, as the reader will readily show. Hereafter I write Av
for A(v).
In the remainder of this section, A, B, C . . . denote linear operators
y —> Y. In particular, I denotes the identity operator, defined by Iv = v
for each v in T. If there is a v * 0 in T such that Av = av, that is, if A
merely multiplies the vector v by the scalar a, we say that a is a proper
value (or eigenvalue) of A and that v is the corresponding proper vector
(or eigenvector). Note that if v is a proper vector of A and b is any scalar,
bv is also a proper vector corresponding to the same proper value.2 If
two or more linearly independent vectors are proper vectors of A cor-
responding to the same proper value X, A is said to be degenerate. Any
nonzero linear combination of the said proper vectors is again a proper
vector corresponding to X. (Prove it!) The number of linearly indepen-
dent proper vectors corresponding to X is called the multiplicity or the
degree of degeneracy of X.
Assume now that T is finite-dimensional and let V = {vl5 v2, . . . , vn]
be a basis of T. Obviously, if A is a linear operator and we know Av/
for each v,- in V, we can determine Av for any vector v in T, for we can
calculate the components of Av relative to V from the components of v
and of Av1?. . . Avm relative to V. Let v = Ef^tf/V/ and Av* = S"=iA;7vy (/ =
! , . . . , « ) . Clearly,

2
Let Av = as. Then A(bv) = bA\ = bay = a(bv).
448 Supplements

Av =

so the components of A(v) relative to V are # / = E-LiAi^-, a{ = E"=1A2A

. . ., an' = I?=iAniaj. (Verify this for « = 2.) Thus, just like the vector v is
represented, relative to the basis V, by the scalar w-tuple (au . . . , an) of
its components, so the linear operator A can represented by the n x n
scalar array

n Au . . . A\n

A i Anl ...Ann

This is the matrix of A in the basis V.

Generally speaking, an n x m matrix is an array of nm scalars
arranged in n rows and m columns. Each scalar element of a matrix is
labeled with two integers, the first of which indicates the element's row
and the second its column. I designate a matrix by its typical element
enclosed in parentheses, say, (A/;). If Aif = 0 whenever / ^ j9 the matrix
(Aij) is said to be diagonal. The sum of two matrices (Ay) and (Bif) is the
matrix (A,7- + B^) that is formed by adding each pair of elements with the
same indices. Evidently, two matrices can be added only if both have
the same number of rows and the same number of columns. We multi-
ply a matrix by a number r (of the same kind as its elements) by multi-
plying each element by r: r(Aij) = (rAjj). The product of (Aij) and (Bif) -
in that order - is the matrix

(Ati)x(Bti) = (ZkAtkBkj) (S2)

whose ij-th. element is formed by multiplying the ith row of (A/;), element-
by-element, by the ;th column of (£z/) and adding together the products.
Evidently, two matrices can be multiplied only if the first has as many
columns as the second has rows. The n x n square matrix (8/;) - where
8/, = 1 if / = / and 0 otherwise - acts as the "one" of matrix multiplica-
tion. We call it the unit matrix and denote it by 1M, or simply by 1 (omit-
ting the subscript does not usually create confusion). Note that an n x
m matrix (Aif) can be multiplied by 1W ow the left and by 1W on the right•;
in either case, the product is (A{j). Note further that 1 is the matrix of
the identity operator I in every basis.
The rules of matrix addition and multiplication by a scalar sound
natural enough. The peculiar rule of matrix multiplication can be readily
justified in the light of what was said above about linear operators.
Regard the components of v relative to V as an n x 1 matrix (afl). Then,
the components of Av relative to V form the n x 1 matrix
 I. On Vectors 449

which is obtained by multiplying (an) on the left by (A/V), the matrix of

A relative to V. Moreover, if B is another linear operator on Y whose
matrix relative to the basis V is (B/7), the composite linear operator C =
AB, which sends v to ABv, is represented, relative to V, by the matrix
(Cfi) = (Ifk=iAjkBki) = (Aji) x (Bji). (Again the reader should work this out
for the case n = 2.)
I must mention two important scalar-valued functions that are defined
on square matrices, viz., the trace and the determinant. The trace of the
n x n matrix (Az/) is the sum of its diagonal elements:

Tr(A,,) = Z A (S3)

It is worth noting that, if the linear operator A is represented, relative

to different bases of T , by the matrices (Aif) and (A/), Tr(Az/) = Tr(A/).
We are therefore entitled to speak of the trace Tr(A) of the operator A.
The reader has probably met 3 x 3 and even 4 x 4 determinants in
high-school algebra. The determinant det(A2/) of our n x n matrix (A/;)
is constructed as follows: Let a be a permutation of ( 1 , 2 , . . . ,n); form
the product Alc(1)A2a(2) • • • Ana(n); prefix to it the sign + or - , depending
on whether a is odd or even; do this with each of the n\ permutations
of ( 1 , 2 , . . . ,w); det(Ay) is the sum of the n\ terms thus obtained. Again,
if A is represented, relative to different bases of T , by the matrices (Aif)
and (A/), we have that det(A/;) = det(A/).
Now look again at the system of n linear equations Z £ = 1 A ^ = a[
printed right after eqn. (SI). Clearly, the linear operator A is an injec-
tive mapping if and only if this system has unique solutions ak (1 < i <
n). The reader may recall that this is so if and only if det(A?/) * 0. This
fact provides a useful way of characterizing the set of proper values of
A. Let X be any scalar. Consider the linear operator (A — A,l). X is a proper
value of A if and only if there is a v in T such that (A - AJ)v = 0. If (A
- X\) is not injective, there are vectors ua and u 2 such that (A - X\)ui =
(A - X\)\i2 and Ui — u2 = v =£ 0. Thus v is a proper vector of A, corre-
sponding to the proper value X. As we know, if there is one such proper
vector v, there are many more (viz., a\, for any scalar a). Thus X is a
proper value of A if and only if (A - X\) is not injective, that is, if and
only if det(Ajj - X§ij) = 0. Since det(A z/ - ^8?/) is an nth degree polynomial
in X, there can be at most n different scalars X satisfying the equation
det(A/; - Xbjj) = 0. They are the proper values of A. The set of proper
vectors that correspond to them is, of course, infinite; it includes a set
of n linearly independent ones. If T is an inner product space, this set
can be so chosen that it constitutes an orthonormal set, in effect an
orthonormal basis of T .
If T is an arbitrary vector space and A is a linear operator T —> T ,
the set of scalars X such that (A — X\) is not injective is known as the
450 Supplements

spectrum of A. (This name is motivated by physical applications.) We

have just shown that, for finite-dimensional T, the spectrum of A is pre-
cisely the set of its proper values.
7 Hilbert spaces. I shall now define a class of vector spaces of central
importance in Quantum Mechanics (see §6.2.5). A Hilbert space is a
complete inner product space. Inner product spaces were defined in 1.4.
An inner product space %C is said to be complete if every Cauchy sequence
of vectors in $£ converges to a vector in %C. An infinite sequence Vi, v2,
. . . of vectors in 5^ is a Cauchy sequence if for every positive real number
5 there is an integer N such that, for every m, n > N, ||vm — vw|| < 8. The
sequence Vi, v2, . . . is said to converge to the vector v if for every posi-
tive real number 5 there is an integer N such that, for every n > N,
||v — vj| < 5.3 In the rest of this section, 3€, %C\ and so on, denote complex
Hilbert spaces.
Given our definitions, any finite-dimensional inner product space is a
Hilbert space. But, of course, it is the infinite-dimensional ones that
demand a special study. If %£ is infinite-dimensional and we are given any
list vl5 v 2 ,. . . , vw, no matter how long, of linearly independent vectors,
there is always in %C some other vector vn + x that is linearly independent
of them. Still, due to the completeness of W, it makes sense to speak of
linear combinations of infinitely many vectors in "K. Consider a sequence
Vi, v 2 ,. . . of vectors, none of which is a linear combination of those pre-
ceding it. Suppose that a^yi, axvx + <32v2,. . ., £JU<z*v*,... is a Cauchy
sequence converging to a vector v. We then regard the infinite series
T%=lakVk = v as a linear combination - in an expanded sense - of the
vectors Vi, v2, V3,. . . By using the Axiom of Choice one can readily prove
that any Hilbert space 1& contains a complete orthonormal set of vectors
{v,j / e Q] such that, for any v in $£, v = 5^>EQtf^v^,with at — (v;|v) for
each i e Q. Or, in Dirac notation,
(S4)

Applying the bra (v| on the left- and right-hand sides of eqn. (S4)
we obtain at once the Theorem of Pythagoras in arbitrarily many
dimensions:
IMI2 = Hv> = 5>h>< v *l v > = I>*|v>*<v*|v> = 5>*|v>| 2 (S5)
keQ

3
Convergence as defined here is sometimes qualified as strong. In contrast with it, weak
convergence is defined as follows: The sequence vl5 v 2 ,. . . is said to converge to the
vector v if, for every positive real number 5, there is an integer N such that, for every
n > N, |(v|v) - (v,,| v,,)| < 5. Strong convergence implies but is not implied by weak
convergence.
 I. On Vectors 451

If the orthonormal set {v,-| i e Q] spans ^t in this way, we are surely

justified to call it a basis of 3€. The cardinality of Q provides a "dimen-
sion-number" for 3€. In particular, eqns. (S4) and (S5) can be proved for
countable Q, without resorting to the Axiom of Choice, if %£ is separa-
ble, that is, if *K contains a countable subset £f that is dense in W. (The
subset Sf is dewse in ^ if, for any v in $£, there is an element of Sf in
every neighborhood of v.) Indeed, %C is separable if and only if it con-
tains a countable complete orthonormal set of vectors.
Let 3{ be a nonempty subset of a Hilbert space %C. If 3£ contains every
linear combination of vectors in 3{, 3£ is said to be a /me#r submanifold
of 3€. 3£ is then a vector space (with the operations of %C restricted to %).
However, 3£ need not be a Hilbert space (with the said operations), for
there might be a Cauchy sequence of vectors in % that does not converge
to a vector in X If % is a Hilbert space in its own right when the oper-
ations that characterize W are restricted to %, we say that % is a sub-
space of %t.
Linear operators Y —> %£ (where T is any vector space) are defined as
in 1.6. If T is a linear submanifold of %C, I speak of linear operators in
M {on W, if T = VC). Let A be a linear operator in %C. Suppose that there
exists a linear operator Af such that, for all v and w in H£\ (v|Aw) =
(Afv|w); Af is called the adjoint of A. A is a bounded operator if there
is a real number C > 0 such that, for every u at which A is defined, ||Au||
< C ||u||. The least such number is called the norm of A and is denoted
by ||A||. If A is bounded, its adjoint Af exists and is also a bounded oper-
ator, with the same norm as A. If A and Af exist and are defined on all
of ^C, they are both bounded. Continuous linear operators on Hi are
bounded. If the linear operator A: T —> %C is bounded and T is dense in
3^, A can be extended uniquely to a bounded operator on %C.
A is said to be unitary if its adjoint Af is the inverse operator A"1, in
other words, if A+A = AA1^ = I. Unitary operators preserve the inner
product; for, if A is unitary, then, for any vectors v and w in 3€, (Av|Aw)
= (AfAv|w) = (v|w).
A is said to be self-adjoint if A = Af. If A is self-adjoint, all its proper
values are real. For suppose that for some v in T, Av = av; then a(\\\)
= (v\a\) = (v|Av) = (Av|v) = (av\v) = a*(\\v); so a = a*; therefore, a is real.
If A is self-adjoint, any two proper vectors of A that correspond to dif-
ferent proper values are mutually orthogonal. Suppose that Av = av, Aw
=fow,and a * b; then a(w\v) = (w|Av) = (Aw|v) = b *(w|v) = b(w\v); thus
(a - b)(w\\) = 0; since (a - b) * 0, we have that (w|v) = 0; therefore, w
and v are orthogonal. These properties of self-adjoint operators moti-
vate their use in Quantum Mechanics.
The study of linear operators in an infinite-dimensional Hilbert %C
must face two difficulties that do not arise in thefinite-dimensionalcase:
452 Supplements

(i) The spectrum of such an operator does not consist simply of proper
values corresponding to its proper vectors, and (ii) the operator can be
unbounded. "It is a fact of life that many of the most important opera-
tors which occur in mathematical physics are not bounded" (Reed
and Simon 1980, p. 249). Because of this one must pay attention to the
operator's precise domain of definition (whereas, if the operator is
bounded, one can always deal simply with its unique extension defined
on the whole %£). With regard to (i), I shall consider only the case of a
self-adjoint operator A in 3€. Its spectrum s(A) typically comprises a dis-
crete collection of real numbers that are proper values of A in the usual
sense, and its complement, the continuous spectrum, which cannot be
associated simply with proper vectors. (Indeed, some such operators -
like the position operator in Quantum Mechanics - only have a contin-
uous spectrum.) The continuous spectrum was dealt with differently
by Dirac and von Neumann. Dirac's approach amounts to a bold
generalization of the concepts of proper vector and proper value that
makes the latter applicable to the continuous spectrum. For this purpose,
it uses the notorious Dirac "delta function", which was spurned as non-
sense until Schwartz's theory of distributions (1957/1959) made it
respectable. Von Neumann's approach attains rigor within the bounds
of classical analysis. Mathematicians and philosophers prefer it to this
day. Neither approach can be properly explained within the bounds of
this book.4
8 Tensor product of two Hilbert spaces. Let 3^ and ^ 2 be complex Hilbert
spaces. It can be shown that there is a Hilbert space %C and a mapping
®: (v,w) ^ v 0 w of 1 ] x ^ into 3€ such that, for every scalar a, b and
suitable vectors u, v, w, t,
(i) [an + by) ® w = a(u <8> w) + b(y <8> w);
(ii) u ® (ay + bw) = a(u 0 v ) + b(u <g> w);
(iii) (u ® t|v 0 w) = (u|v)(t |w);
(iv) every neighborhood of every vector in %C contains a linear combi-
nation of vectors in the range of ® (i.e., a linear combination of
vectors of the form Vi ® v2, with v, in %Ch i = 1,2).
Moreover, if the pairs (2C,®) and <2C',®') both satisfy the above require-
ments, there is a unique isomorphism fiffl-^W such that v 0 ' w = f(y
® w) for every pair (v,w) in 2£a x 2€2. By the tensor product 3^® ^ 2 of
spaces ^ i and <K1 I mean the structure that is realized by any such pair
(^, ®) by dint of conditions (i)-(iv). We call v 0 w the tensor product

4
Brief descriptions are given by Redhead (1987, pp. 6-16).
 II. On Lattices 453

of vectors v and w. If A1 and A2 are linear operators on W1 and 3€2,

respectively, the linear operator Ai <E> A2 on Htx 0 $£2 is defined by: Ai
<8> A2(v 0 w) = AiV 0 A2w, for every suitable v and w.

II. On Lattices

Partial ordering. A relation R is said to be binary (or dyadic) if it is apt

to hold between the elements of an ordered pair. I write aRb for 'object
a has relation R to object b' (or, synonymously: 'R holds between the
first and the second element of the ordered pair (<?,£>)'). A set S is par-
tially ordered by the binary relation < if the following conditions are
satisfied for any elements a, b, and c of S:

PO1. a < a (the relation < is reflexive).

PO2. \ia<b and a * b, it is not the case that b < a (< is antisymmetric).
PO3. If a < b and b < c, then a < c (< is transitive).

If S is partially ordered by <, we say that (S,<) is a poset (for 'partially

ordered set'). Depending on the nature of the objects in 5, we read 'a <
V* as 'a is less than or equal to b" or as 'a precedes or is equal to 6'; alter-
natively, we read '<? < b' as lb is greater than or equal to a9 or as 'b
follows or is equal to a9 Consider any subset {at e S: i e 3}. Suppose
that there is an element b in S such that (i) b < a{ for every i s i , and
(ii) c < b for every c in S such that, for every i e $, c < a{. We say then
that b is the infimum of {au . . . , # „ } . Suppose that there is an element b
in S such that (i) a{ < b for every i e $, and (ii) b < c for every c in S
such that, for every i e $, a;< c. We say then that b is the supremum of
{*,- G S:ie $}.

Lattices. The poset (L,<) is a lattice if the following conditions are

fulfilled:
LI. L contains a minimal element 0 that is less than or equal to every
element of L.
L2. L contains a maximal element 1 that is greater than or equal to
every element of L.
L3. If a and b are any two elements of L, L contains an element that is
the infimum of {a, b] and an element that is the supremum of {a, b}.
When speaking of lattices, the supremum of a and b is called their join
and is denoted by a v fo, and the infimum of a and b is called their meet
and is denoted by a Afc.Our definitions imply that, for any a, b e L, a
< b if and only if # = a A 6 and b = a v b. From PO1-PO3 and L1-L3
the following statements can be readily inferred, for every a, b, c e L:
454 Supplements

L4 A . a A b = b A a. L4V. a v b = b v a.
L5A. a A (b A c) = (a A b) A c. L5V. a v (b v c) = {a v b) v c.
L6. a = a v (b A a) = a A (b v a).
An atom in a lattice (L,<,0,1) is an element a e L such that <3 ^ 0
and, for any x e L, if 0 < x < a, then either 0 = x or x - a. In other
words, tf is an atom if there are no elements in the lattice between a and
0. The lattice (L,<,0,1) is said to be atomic if every element of L, except
0, is either greater than or equal to an atom.
3 Orthocomplementation. Let a and b belong to the lattice (L,<,0,1). If a
vfc = 1 and a A b - 0, b is said to be a complement of a. (L,<,0,1) is a
complemented lattice if every element of L has a complement. Consider
now an injective mapping of L into L that assigns to each a G L an
element a' e L such that
OC1. <z' is a complement of a,
OC2. (a')' = a, and
OC3. V < a' for any b e L such that a < b.
The mapping meeting these requirements can be defined on a lattice in
only one way (or not at all). The object a' is called the ortho complement
of a. A lattice (L,<,0,1,'} on which such a mapping is defined is said to
be ortho complemented. Note that an element a of an orthocomple-
mented lattice may have several complements, although only one of them
will jointly satisfy OC2 and OC3.
Orthocomplementation can obviously be defined in any poset that is
not a lattice, provided that it contains a maximal and a minimal element
(i.e., provided that it satisfies conditions LI and L2, even if it does not
satisfy L3). We speak then of an ortho complemented poset.
4 Boolean lattices. A Boolean lattice is a lattice (L,<,0,1) in which the two
following distributive laws hold for any three a, b, c e L:
Dv/A. a v (b A c) = (a v b) A (a v c);
DA/v. a A (b v c) = [a A b) v (a A c).

In a Boolean lattice every element a has one and only one complement,
which I denote by a\ because it satisfies the rules of orthocomplementa-
tion OC1-OC3. a' is the maximum of the lattice elements whose meet with
a is 0. In other words, x = af if and only if y A a = 0 for every y < x. On the
other hand, any orthocomplemented lattice in which the orthocomple-
ment of an element is its only complement is necessarily a Boolean lattice.5

5
There is no essential difference between a Boolean lattice and what is known as
Boolean algebra. Take any set B, endowed with two binary algebraic operations v and
 III. Terms from Topology 455

Given any set E, its power set 2P£ is partially ordered by the relation
of inclusion c. (SPEjCi) is a lattice with maximal element £ and minimal
element 0 . Take the union of two sets for their join and the intersection
of two sets for their meet. Define the complement of any A c I by A' =
E\A (the set of all elements of S that do not belong to A). The lattice
(SPZ,c,Z,0/) is Boolean; we call it the Boolean lattice of subsets of Z.
Every Boolean lattice (L,<,0,1/) can be represented by the Boolean lattice
of subsets of some fundamental set ZL by virtue of a mapping /: L —>
S^IL which is such that (i) /"(0) = 0 , (ii) /"(I) = E, and, for any a9 b s L,
(iii) f(a') = I\f(a) and (iv) a < b if and only if f(a) c f(b). These condi-
tions imply, of course, that for any a, b e L, f(a A b) = f(a) n f(b), and
f(a vb)= f(a) u f(b).

5 Modularity and orthomodularity. An orthocomplemented lattice

(L,<,0,l/) is modular if it satisfies the following condition:

M. For any a, b, c e L, such that a < b, a v (c A b) = (a v c) A b.

In a modular lattice, any permutation of a, b, and c satisfies the distrib-

utive laws Dv/A and DA/v, provided that a < b.
An orthocomplemented lattice (L,<,0,1/) is orthomodular if it
satisfies the following condition:

OM. For any a, b e L, such that a < b, a v (b A a') = b.

In an orthomodular lattice, any permutation of a, a\ and b satisfies the
distributive laws Dv/A and DA/v, provided that a < b.
Conditions M and OM can hold also for an orthocomplemented
poset (L,<,0,l/) that is not a lattice.

III. Terms from Topology

Topological spaces. Let X be any set of objects, which we call points,
and T a collection of subsets of X, which we call open sets. T is a topol-
ogy on X and (X,T) a topological space if and only if the following three
conditions are met:

A (i.e., two mappings of B x B into B) that satisfy conditions L4-L6 and D. If both
operations have "neutral elements" 0 and 1 such that for any a e B,avO = a = aA
1, and there is a one-one mapping of B into B by x •—> x' such that for any a e B, a
v a' - 1 and a A a' = 0, ( B , V , A , 0 , 1 / ) is a Boolean algebra. B is partially ordered by
the relation < defined by: a < b if an only if a A b = a and avb = b. The reader should
verify that, pursuant to the definitions in the main text, (B,<,0,1/> is indeed a Boolean
lattice.
456 Supplements

Tl. The entire set X and the empty set 0 are open sets.
T2. The union of any collection of open sets is an open set
T3. The intersection of any two open sets is an open set.
If A is an open set in (X,T), its complement X\A (the set of all points in
X but not in A) is said to be closed. If A is any subset of X, the inter-
section of all closed sets that include A is clearly a closed set,6 called the
closure of A. If p is a point of X, a neighborhood of p is any subset of
X that includes an open set that contains p. (This idea of neighborhood
is central to topology, which is naturally regarded as the theory of neigh-
borhood structures; a definition of topology that is equivalent to the one
above can be given purely in terms of neighborhood systems attached to
the points of X.) Let (X,T) be a topological space, let Y be a subset of
X, and consider the collection 17 of all sets that are the intersection of a
member of T with Y. (Y,U) is a topological space. (Exercise: Prove this.)
U is the subset topology induced in Y by T. Topologies are often defined
by fixing a base, that is, a collection of easily characterized sets. The
topology generated from a given base is the smallest topology that
includes the base plus all other sets required by conditions Tl, T2, and
T3. The standard topology of Euclidian space is generated from the open
balls (a set of points in Euclidian space is an open ball if it contains all
the points that lie at less than a certain distance from a certain point,
and only them). The concept of compactness provides an exact charac-
terization of finiteness in purely topological terms, without resorting to
metric concepts (such as distance or volume). An open covering of the
topological space (X,T) is a collection C of sets in T such that each point
in X belongs to one member of C (at least). An open covering C of (X,T)
is said to have a sub covering C if C is also an open covering of (X,T)
and every member of C is a member of C. A topological space (X,T) is
said to be compact if and only if every open covering of (X,T) has a
finite subcovering (i.e., a subcovering consisting of finitely many sets).
Consider an ordinary circle S and a straight line R; let Ts and TR be the
subset topologies induced, respectively, in S and R by the standard
Euclidian topology. Then, (5,TS) is compact and (R,TR) is not. (Exercise:
Prove it.) By using the concept of topological space we can achieve a
precise general definition of continuity. Suppose that (X,T) and (Y,U)
are two topological spaces. The mapping f:X —> Y is said to be contin-
uous if and only if every open set of (Y,U) is the image under f of an
open set of (X,T). If f is a continuous bijective mapping with continu-

Each closed set that includes A has an open complement that excludes A. The union
of these open sets is open by T2. The intersection of the closed sets that include A is
precisely the complement of this union, so it is a closed set.
 III. Terms from Topology 457

ous inverse f'1, X and Y obviously share the same topological structure
that is faithfully portraited by f (and by f'1). We then say that f is an
homeomorphism and that X and Y are homeomorphic.

2 Topological manifolds. Let a and b be real numbers such that a < b. The
sets {x e U: a < x < b], {x e (R: a < x], and {x e U: x < b] are open inter-
vals^ and every open interval falls under one of these three descriptions
for some pair of real numbers such as a and b. The standard topology
of the real line R is generated from the base formed by the open inter-
vals.7 The standard topology of Un can be generated from the base
formed by the Cartesian products of n open intervals. A topological
space (X,T) is said to be a topological n-dimensional manifold if every
point of X has an open neighborhood that is homeomorphic to Un (with
the standard topology). Comparing this definition with the notion of
an ^-manifold (or real w-dimensional smooth manifold) explained in
§4.1.3, we see that every w-manifold is also a topological ^-dimensional
manifold.

3 Borel sets of the real line. Consider any set 5. Consider a collection 2ft (S)
of subsets of S such that (i) S G 2&(S); (ii) if A e Sft(S), its complement
S\A e 2ft(S); and (iii) if Au A 2 , . . . is any countable list of subsets in 28(S),
their union uJLiA* e SB(S). Then S&(S) is a Borel field of S. If M is any
nonempty collection of subsets of S, the smallest Borel field of S that
contains M is said to be the Borel field generated by M. Let 2ft(R) denote
the Borel field of U generated by the open intervals. The sets in 28(IR) are
called the Borel sets of U. They are precisely the Lebesgue measurable
subsets of the real line. $k(U) is important in the theory of probability
because, by dint of the Axiom of Choice - which most mathematicians
accept, as one told me, "in order to survive" - a probability function
satisfying the standard Kolmogorov axioms cannot be consistently
defined on the entire power set SPIR; so, normally, a function that
expresses the probability that, say, the value of a physical quantity lies
in a particular subset of IR is not defined for all the subsets of [R but only
for the Borel sets.

7
Exercise: Prove that if U is endowed with the said standard topology, a function
f:U —> R is continuous according to the familiar definition of a continuous real-valued
function of a real variable if and only if it is also continuous in accordance with the
topological definition of continuity given at the end of III.l.
 References

Alexander, H. G., ed. (1956). The Leibniz-Clarke Correspondence. Manchester:

Manchester University Press, 1956.
Alpher, R. A. and R. C. Herman (1948). "Evolution of the universe". Nature.
162:774-75.
Alpher, R. A., H. Bethe, and G. Gamow (1948). "The origin of chemical
elements". Physical Review. 73:803-804.
Ampere, A. M. (1827). "Memoire sur la theorie des phenomenes electrody-
namiques uniquement deduite de Pexperience, dans lequel se trouvent reunis
les Memoires que M. Ampere a communiques a l'Academie royale des Sci-
ences dans les seances des 4 et 26 decembre 1820, 10 juin 1822, 22 decembre
1823, 12 septembre et 21 novembre 1825". Memoires de l'Academie Royale
des Sciences. 6:175-388 (1823). (Volume 6 of the Memoires de VAcademie des
Sciences corresponds to 1823, but was actually issued in 1827; both dates are
given on the title page.).
Anderson, C. D. (1932). "The apparent existence of easily deflectable positives".
Science. 76:238-39.
Anscombe, G. E. M. (1971). Causality and Determination: An Inaugural
Lecture. Cambridge: Cambridge University Press.
Appell, P. (1893). Traite de mecanique rationelle. Tome I. Paris: Gauthier-Villars.
Arabatzis, T. (1996). "Rethinking the 'discovery' of the electron". Studies in
History and Philosophy of Modern Physics. 27B:405-35.
Arana, J. (1990). Apariencia y verdad: Estudio sobre la filosofia de P. L. M. de
Maupertuis (1698-1759). Buenos Aires: Charcas.
Aristotle. Aristotelis Opera. Ex recognitione I. Bekkeris edidit Academia Regia
Borussica. Berlin: G. Reimer, 1831. 2 vols. (My quotations refer, as is cus-
tomary, to the pages, columns, and lines of this edition; however, they are taken
from various more recent editions and may not always agree with Bekker's
text.)
Aristotle. De Anima. Books II and III. Translated with Introduction and Notes
by D. W. Hamlyn. Oxford: Clarendon Press.
Aspect, A., J. Dalibard, and G. Roger (1982). "Experimental tests of Bell's
inequalities using time-varying analyzers". Physical Review Letters.
49:1804-7.

458
 References 459

Aspect, A., P. Grangier, and G. Roger (1982). "Experimental realization of

Einstein-Podolsky-Rosen Gedankenexperiment: A new violation of Bell's
inequalities". Physical Review Letters. 48:91-94.
Auyang, S. Y. (1995). How Is Quantum Field Theory Possible? New York:
Oxford University Press.
Ayer, A. J., ed. (1959). Logical Positivism. New York: Free Press.
Baade, W. (1956). "The period-luminosity relation of the cepheids". Publications
of the Astronomical Society of the Pacific. 28:5-16.
Baldwin, F. J. (1901/1902). Dictionary of Philosophy and Psychology. New York:
Macmillan. 2 vols.
Balzer, W., C. U. Moulines, and J. D. Sneed (1983). "The structure of empirical
science: Local and global". In R. Barcan Marcus et al., eds., Proceedings of
the 7th International Congress of Logic, Methodology and Philosophy of
Science, 1983. Amsterdam: North-Holland, 1986, pp. 291-306.
Barbour, J. B. (1989). Absolute or Relative Motion? A Study from a Machian
Point of View of the Discovery and the Structure of Dynamical Theories. Part
1: The Discovery of Dynamics. Cambridge: Cambridge University Press.
Barrow, J. D. and F. J. Tipler (1986). The Anthropic Cosmological Principle.
Oxford: Clarendon Press.
Bartelborth, T. (1988). Eine logische Rekonstruktion der klassischen Elektrody-
namik. Frankfurt a. M.: Peter Lang.
Bartels, A. (1994). Bedeutung und Begriffsgeschichte: Die Erzeugung wis-
senschaftlichen Verstehens. Paderborn: Ferdinand Schoningh.
Basilius (St. Basil). Epistulae. Edited by Y. Courtonne. Paris: Belles Lettres,
1957-1966.
Bayes, T. (1763). "An essay towards solving a problem in the doctrine of
chances". Royal Society of London Philosophical Transactions. 53:370-418.
Bechler, Z. (1991). Newton's Physics and the Conceptual Structure of the
Scientific Revolution. Dordrecht: Kluwer.
Belinfante, E J. (1973). A Survey of Hidden-Variables Theories. Oxford:
Pergamon.
Bell, J. and M. Hallett (1982). "Logic, quantum logic and empiricism". Philos-
ophy of Science. 49:375-79.
Bell, J. S. (1964). "On the Einstein-Podolsky-Rosen paradox". Physics.
1:195-200 (reprinted in Bell 1987, pp. 14-21).
(1966). "On the problem of hidden variables in Quantum
Mechanics". Review of Modern Physics. 38:447-52 (reprinted in Bell 1987,
pp. 1-13).
-(1975). "On wave packet reduction in the Colemann-Hepp model".
Helvetica Physica Ada. 48:93-98 (reprinted in Bell 1987, pp. 44-51).
-(1987). Speakable and Unspeakable in Quantum Mechanics. Cambridge-
Cambridge University Press.
-(1990). "Against 'measurement'". In A. I. Miller, 1990, pp. 17-31.
Beltrametti, E. G. and G. Cassinelli (1981). The Logic of Quantum Mechanics.
Reading, MA: Addison-Wesley.
Berkeley, G. (1710). A Treatise Concerning the Principles of Human Knowlege
[sic]. Dublin: Printed by A. Rhames, for J. Pepyat.
460 References

(1744). Siris: A Chain of Philosophical Reflexions and Inquiries Con-

cerning the Virtues of Tar-water, and Divers other Subjects Connected
Together and Arising One from Another. Dublin: Printed by M. Rhames, for
R. Gunne.
-(1752). Alciphron: or, the Minute Philosopher. In Seven Dialogues
containing an Apology for the Christian Religion, against those who are
called Free-thinkers. 3rd Ed. London: J. and R. Tonson and S. Draper. (First
edition, 1732).
Bernays, P. (1967). "Hilbert, David". In Edwards 1967, vol. 3, pp. 496-504.
Bernoulli, D. (1738). Hydrodynamica. Argentorati [Strasbourg]: Dulsecker.
(English translation by T. Carmodus and H. Cobus, Hydrodynamics, New
York: Dover, 1968.)
Bernoulli, J. (1713). Ars conjectandi: Opus posthumum. Basilea [Basle]:
Impensis Thurnisiorum, fratrum.
Birkhoff, G. and J. von Neumann (1936). "The Logic of Quantum Mechanics".
Annals of Mathematics. 37:823-43 (reprinted in C. A. Hooker 1975).
Birkhoff, G. D. (1923). Relativity and Modern Physics. Cambridge, MA:
Harvard University Press.
Blau, S. K. and A. H. Guth (1987). "Inflationary cosmology". In S. Hawking
and W. Israel 1987, pp. 524-603.
Bohm, D. (1951). Quantum Theory. Englewood Cliffs, NJ: Prentice-Hall.
(1952). "A suggested interpretation of the Quantum Theory in terms of
'hidden variables'". Physical Review. 85:166-79, 180-93.
-(1980). Wholeness and the Implicate Order. London: Routledge.
Bohm, D. and B. J. Hiley (1993). The Undivided Universe: An ontological
interpretation of quantum theory. London: Routledge.
Bohr, N. (CW). Collected Works. Amsterdam: North-Holland, 1972- .
(1913). "On the constitution of atoms and molecules". Philosophical
Magazine. 26:1-25, 476-502, 857-75.
-(1918). "On the quantum theory of line-spectra. Part I. On the general
theory". Det Kongelige Danske Videnskabernes Selskab, Matematisk-fysiske
Meddelser. (8) 4 1:1-36 (reprinted in B. L. Van der Waerden 1967, pp.
95-137).
-(1922). The Theory of Spectra and Atomic Constitution. Cambridge:
Cambridge University Press.
-(1924). On the Application of the Quantum Theory to Atomic Structure,
Part I: The Fundamental Postulates. Supplement to the Cambridge Philo-
sophical Society Proceedings.
-(1928). "The quantum postulate and the recent development of atomic
theory". Nature. 121:580-90 (printed version of Bohr's Como lecture of 1927.
Reprinted in Bohr 1934, pp. 51-91).
-(1934). Atomic Theory and the Description of Nature. Cambridge:
Cambridge University Press.
-(1935). "Can quantum-mechanical description of physical reality be con-
sidered complete?". Physical Review. 48:696-702 (reprinted in J. A. Wheeler
and W. H. Zurek 1983, pp. 145-51; the text printed on p. 148 should go on
p. 149, and vice versa).
 References 461

(1949). "Discussion with Einstein on epistemological problems in atomic

physics". In P. A. Schilpp 1949, pp. 199-241.
(1958). Atomic Physics and Human Knowledge. New York: Wiley.
-(1963). Essays 1958-1962 on Atomic Physics and Human Knowledge.
New York: Wiley.
Bohr, N., H. Kramers, and J. C. Slater (1924). "The quantum theory of radia-
tion". Philosophical Magazine. 47:785-822 (reprinted in B. L. Van der
Waerden 1967, pp. 159-76).
Boltzmann, L. (WA). Wissenschaftliche Abhandlungen. Herausgegeben von F.
Hasenohrl. Leipzig: Barth, 1901. 3 vols. (reprint: New York, Chelsea, 1968).
(1871a). "Einige allgemeine Satze iiber Warmegleichgewicht". Sitzungs-
berichte der Akademie der Wissenschaften zu Wien. 63:679-711 (reprinted in
Boltzmann WA, I, 260-87).
-(1871b). "Analytischer Beweis des zweiten Hauptsatzes der mechanischen
Warmetheorie aus den Satzen iiber das Gleichgewicht der lebendigen Kraft".
Sitzungsberichte der Akademie der Wissenschaften zu Wien. 63:712-32
(reprinted in Boltzmann WA, I, 288-308).
-(1872). "Weitere Studien iiber das Warmegleichgewicht unter Gas-
molekiilen". Sitzungsberichte der Akademie der Wissenschaften zu Wien.
66:275-370 (reprinted in Boltzmann WA, I, 316-402).
(1896a). Vorlesungen uber Gastheorie. Teil I. Leipzig: Barth.
-(1896b). "Entgegnung auf die warmetheoretischen Betrachtungen des
Hrn. E. Zermelo". Annalen der Physik. (3) 57:773-84 (reprinted in Boltzmann
WA, III, 567-78).
-(1897). "Zu Hrn. Zermelos Abhandlung 'Uber die mechanische Erk-
larung irreversibler Vorgange'". Annalen der Physik. (3) 60:392-98 (reprinted
in Boltzmann WA, III, 579-86).
(1898). Vorlesungen iiber Gastheorie. Teil II. Leipzig: Barth.
(1905). Populare Schriften. Leipzig: Barth.
-(LGT). Lectures on Gas Theory. Translated by S. G. Brush. Berkeley:
University of California Press, 1964 (unaltered reprint: New York, Dover,
1995).
Bondi, H. and T. Gold (1948). "The steady-state theory of the expanding uni-
verse". Monthly Notices of the Royal Astronomical Society. 108:252-70.
Born, M. (1924). "Uber Quantenmechanik". Zeitschrift fur Physik. 26:379-95.
(1926a). "Zur Quantenmechanik der Stossvorgange. (Vorlauflge Mit-
teilung)". Zeitschrift fur Physik. 37:863-67.
-(1926b). "Quantenmechanik der Stossvorgange". Zeitschrift fiir Physik.
38:803-27.
(1927a). "Das Adiabatenprinzip in der Quantenmechanik". Zeitschrift
fur Physik. 40:167-92.
-(1927b). "Physical aspects of quantum mechanics". In M. Born 1969, pp.
6-12. (First published in Nature 119:354-57 (1927).)
-(1962). Atomic Physics. 7th Ed. From the original translation of J.
Dougall revised by the author in collaboration with R. J. Blin-Stoyle. London:
Blackie &c Son.
(1968). My life and my views. New York: Scribner's.
462 References

(1969). Physics in my generation. 2nd rev. ed. New York: Springer.

Born, M. and P. Jordan (1925a). "Zur Quantentheorie aperiodischer Vorgange".
Zeitschrift fur Physik. 33:479-505.
(1925b). "Zur Quantenmechanik". Zeitschrift fur Physik. 34:858-78.
(English translation qf §§1-3 in B. L. Van der Waerden 1967, pp. 277-306.)
Born, M., W. Heisenberg, and P. Jordan (1926). "Zur Quantenmechanik II".
Zeitschrift fiir Physik. 35:557-615. (English translation in B. L. Van der
Waerden 1967, pp. 321-85.)
Boskovic, R. J. (1758). Philosophiae naturalis theoria redacta ad unicam legem
virium in natura existentium. Vindobonae [Vienna].
(1763) Theoria philosophiae naturalis redacta ad unicam legem virium in
natura existentium. Venetiis [Venice]: ex Typographia Remonidniana.
Bose, S. N. (1924). "Plancks Gesetz und Lichtquantenhypothese". Zeitschrift fiir
Physik. 26:178-81.
Bourbaki, N. (1970). Theorie des ensembles. Paris: Hermann.
Boyle, R. (1666). The origine of formes and qualities (according to the corpus-
cular philosophy) illustrated by considerations and experiments. Oxford:
Printed by H. Hall printer to the University, for Ric. Davis.
(1674). De hypothesis mechanics excellentia et fundamentis considera-
tiones qucedam amico propositce. London: Typis T.N., impensis Henrici
Herringman.
-. (SPP). Selected Philosophical Papers. Edited with an introduction by
M. A. Stewart. Indianapolis: Hackett, 1991.
Bradley, J. (1728). "A Letter from the Reverend Mr. James Bradley Savilian Pro-
fessor of Astronomy at Oxford and F.R.S. to Dr. Edmund Halley Astronom.
Roy. etc. giving an Account of a new discovered Motion of the Fix'd Stars".
Royal Society of London Philosophical Transactions. 35:637-61.
Braginsky, V B. and E Y. Khalili (1992). Quantum Measurement. Edited by
K. S. Thorne. Cambridge: Cambridge University Press.
Braginsky, V. B. and V I. Panov (1972). "Verification of the equivalence of iner-
tial and gravitational mass". Soviet Physics JETP. 44:463-66 (English trans-
lation; the Russian original was published in Zh. Eksp. Teor. Fiz., 71:873-79
(1971).)
Braithwaite, R. B. (1953). Scientific Explanation: A Study of the Function of
Theory, Probability and Law in Science. Based Upon the Tamer Lectures,
1946. Cambridge: Cambridge University Press.
Brown, H. R. and R. Harre, eds. (1988). Philosophical Foundations of Quantum
Field Theory. Oxford: Clarendon Press.
Brown, R. (1828). "A brief account of observations made in the months of June,
July, and August 1827, on the particles contained in the pollen of plants; and
on the general existence of active molecules in organic and inorganic bodies".
Edinburgh New Philosophical Journal. 5:358-71 (reprinted in Philosophical
Magazine 4:161-73 (1828)).
Bub, J. (1997). Interpreting the Quantum World. Cambridge: Cambridge
University Press.
Buchwald, J. Z. (1985). From Maxwell to Microphysics: Aspects of Electro-
 References 463

magnetic Theory in the Last Quarter of the Nineteenth Century. Chicago: Uni-
versity of Chicago Press.
-(1994). The Creation of Scientific Effects: Heinrich Hertz and Electric
Waves. Chicago: University of Chicago Press.
-, ed. (1995). Scientific Practice: Theories and Stories of Doing Physics.
Chicago: University of Chicago Press.
Buijs-Ballot, C. H. D. (1858). "Uber die Art der Bewegung, welche wir Warme
und Elektrizitat nennen". Annalen der Physik. (2) 103:240-59.
Bunge, M. (1967). Foundations of Physics. Berlin: Springer.
Burbury, S. H. (1894). "Boltzmann's minimum function". Nature. 51:78.
Busch, P., P. J. Lahti, and P. Mittelstaedt (1996). The Quantum Theory of
Measurement. 2nd rev. ed. Heidelberg: Springer.
Butts, R. E. (1968). William Whewell's Theory of Scientific Method. Pittsburgh,
PA: University of Pittsburgh Press (reprint: Indianapolis, Hackett, 1989).
Campbell, L. and W. Garnett (1884). The Life of James Clerk Maxwell. 2nd Ed.
London: Macmillan.
Campbell, N. R. (1920). Physics: The Elements. Cambridge: Cambridge
University Press (reprinted as Foundations of Science: The Philosophy of
Theory and Experiment•, New York, Dover, 1957).
Cao, Tian Yun (1997). Conceptual Developments of 20th Century Field
Theories. Cambridge: Cambridge University Press.
Carnap, R. (1928). Der logische Aufbau der Welt. Berlin: Weltkreis-Verlag.
. (1936/37). "Testability and meaning". Philosophy of Science. 3:419-71;
4:1-40.
(1945a). "On inductive logic". Philosophy of Science. 12:72-97.
-(1945b). "The two concepts of probability". Philosophy and Phenome-
nological Research. 5:513-32.
-(1950). Logical Foundations of Probability. Chicago: University of
Chicago Press.
-(1952). The Continuum of Inductive Methods. Chicago: University of
Chicago Press.
-(1956). "The methodological character of theoretical concepts".
Minnesota Studies in the Philosophy of Science. 1:38-76.
Carnot, S. (1824). Reflexions sur la puissance motrice du feu et sur les machines
propres a developper cette puissance. Paris: Chez Bachelier (facsimile reprint:
Paris, Jacques Gabay, 1990).
Cartan, E. (1923). "Sur les varietes a connexion affine et la theorie de la rela-
tivite generalised (premiere partie)". Annales de VEcole Normale Superieure.
40:325-412.
Caspar, M. (1993). Kepler. Translated and edited by C. D. Hellman. New York:
Dover.
Chandrasekhar, S. (1995). Newton's Principia for the Common Reader. Oxford:
Clarendon Press.
Charlton, W. (1970). Aristotle's Physics, Books I and II. Translated with intro-
duction and notes. Oxford: Clarendon Press.
Chrysippus (FM). "Fragmenta Moralia". In von Arnim, SVE Vol. 3, pp. 3-191.
464 References

Cini, M. (1983). "Quantum theory of measurement without wave packet

collapse". // Nuovo Cimento. 73B:27-55.
Ciufolini, I. and J. A. Wheeler (1995). Gravitation and Inertia. Princeton, NJ:
Princeton University Press.
Claggett, M. (1959a). The Science of Mechanics in the Middle Ages. Madison,
WI: University of Wisconsin Press.
, ed. (1959b). Critical Problems in the History of Science. Madison, WI:
University of Wisconsin Press.
Clausius, R. (1857). "Uber die Art der Bewegung, welche wir Warme nennen".
Annalen der Physik. (2) 100:353-80.
(1858). "Uber die mittlere Lange der Wege, welche bei der Molecularbe-
wegung gasformigen Korper von den einzelnen Moleciilen zuriickgelegt
werden; nebst einigen anderen Bemerkungen uber die mechanische Warmethe-
orie". Annalen der Physik. (2) 105:239-58.
Clifton, R. K., M. L. G. Redhead, and J. N. Butterfield (1991a). "Generalization
of the Greenberger-Horne-Zeilinger algebraic proof of nonlocality". Founda-
tions of Physics. 21:149-84.
(1991b). "A second look at a recent algebraic proof of nonlocality".
Foundations of Physics Letters. 4:395-403.
Cohen, I. B. (1980). The Newtonian Revolution: With illustrations of the
transformation of Scientific Ideas. Cambridge: Cambridge University
Press.
Cohen-Tannoudji, C , B. Diu, and F. Laloe (1977). Quantum Mechanics. Trans-
lated from the French by S. R. Hemley, N. Ostrowsky, and D. Ostrowsky.
Paris: Hermann. 2 vols.
Colodny, R. G., ed. (1965). Beyond the Edge of Certainty. Englewood Cliffs, NJ:
Prentice-Hall.
, ed. (1966). Mind and Cosmos. Pittsburgh, PA: University of Pittsburgh
Press.
, ed. (1972). Paradigms and Paradoxes: The Philosophical Challenge of
the Quantum Domain. Pittsburgh, PA: University of Pittsburgh Press.
ed. (1986). From Quarks to Quasars: Philosophical Problems of
Modern Physics. Pittsburgh, PA: University of Pittsburgh Press.
Cornford, F. M. (1936). "The invention of space". In Essays in Honour of
Gilbert Murray. London: Allen 6c Unwin, pp. 215-235 (reprinted in
M. Capek, ed., The Concepts of Space and Time, Dordrecht: Reidel, 1976,
pp. 3-16).
Costabel, P. (1960). Leibniz et la dynamique: Les textes de 1692. Paris:
Hermann. (Leibniz's Essay de dynamique is printed on pp. 97-106.)
Courant, R. (1936). Differential and Integral Calculus. Volume II. Translated by
E. J. McShane. London: Blackie 8c Son.
Cushing, J. T. (1994). Quantum Mechanics: Historical Contingency and the
Copenhagen Hegemony. Chicago: University of Chicago Press.
d'Alembert, J. (1758). Traite de dynamique dans lequel les loix de Vequilibre et
du mouvement des corps sont reduites au plus petit nombre possible, et demon-
trees d'une maniere nouvelle, et oil Von donne un principe general pour trouver
le mouvement de plusieurs corps qui agissent les uns sur les autres d'une
 References 465

maniere quelconque. Nouvelle Edition, revue et fort augmentee par l'Auteur.

Paris: Chez David (facsimile reprint: Paris, Jacques Gabay, 1990).
Damerow, P., G. Freudenthal, P. McLaughlin, and J. Renn (1992). Exploring the
Limits of Freclassical Mechanics. A Study of Conceptual Development in Early
Modern Science: Free Fall and Compounded Motion in the Work of Descartes,
Galileo, and Beeckman. New York: Springer.
Daneri, A., A. Loinger, and G. M. Prosperi (1962). "Quantum theory of mea-
surement and ergodicity conditions". Nuclear Physics. 33:297-319.
Darrigol, O. (1992). From c-Numbers to q-Numbers: The Classical Analogy in
the History of Quantum Theory. Berkeley, CA: University of California Press.
Davidson, D. (1974). "On the very idea of a conceptual scheme". Proceedings
and Addresses of the American Philosophical Society. 47:5-20.
de Broglie, L. (1923a). "Ondes et quanta". Comptes rendus de YAcademie des
Sciences. 177:507-10.
(1923b). "Quanta de lumiere, diffraction et interferences". Comptes
rendus de YAcademie des Sciences. 177:548-50.
-(1925). "Recherches sur la theorie des quanta". Annales de physique.
3:22-128.
Debs, T. A. and M. L. G. Redhead (1996). "The twin 'paradox' and the con-
ventionality of simultaneity". American Journal of Physics. 64:384-92.
de Finetti, B. (1930). "Funzione caratteristica di un fenomeno aleatorio".
Memorie della Reale Accademia dei Lincei. TV 5:86-133.
(1931). "Sul significato soggettivo della probability". Fundamenta Math-
ematicae. 17:298-329.
(1937). "La prevision: ses lois logiques, ses sources subjectives". Annales
de ITnstitut Henri Poincare. 7:1-68 (translated as "Foresight: Its logical laws,
its subjective sources" in Kyburg and Smokier 1980, pp. 53-118).
(1970). Teoria della probability. Torino: Einaudi.
de Gandt, F. (1995). Force and Geometry in Newton's Principia. Translated by
C. Wilson. Princeton, NJ: Princeton University Press.
de Maria, M. and A. Russo (1985). "The discovery of the positron". Rivista di
Storia delle Scienze. 2:237-86.
Descartes, R. (AT). CEuvres. Edited by C. Adam and P. Tannery. Paris: Cerf,
1897-1912. 12 vols.
(1637). Discours de la methode pour bien conduire sa raison, et chercher
la verite dans les sciences. Plus La dioptrique, Les meteores, et La geometrie,
qui sont des essais de cette methode. Leyde: de l'imprimerie de I. Maire.
-(1641). Meditationes de prima philosophia, in quibus Dei existentia &
animae a corpore distinctio demonstratur. Amstelodami: apud Ludovicum
Elzevirium.
-(1644). Principia philosophiae. Amstelodami: apud Ludovicum
Elzevirium.
de Sitter, W. (1917a). "On the relativity of inertia: Remarks concerning Einstein's
latest hypothesis". K. Nederlandse Akademie van Wetenschappen. Proceed-
ings. 19:1217-25.
(1917b). "On the curvature of space". K. Nederlandse Akademie van
Wetenschappen. Proceedings. 20:229-43.
466 References

(1917c). "On Einstein's theory of gravitation, and its astronomical con-

sequences. III". Monthly Notices of the Royal Astronomical Society. 78:3-28.
-(1933). "On the expanding universe and the time-scale". Monthly
Notices of the Royal Astronomical Society. 93:628-34.
d'Espagnat, B., ed. (1971). Foundations of Quantum Mechanics. Proceedings of
the International School of Physics "Enrico Fermi", Course IL. New York:
Academic Press.
Destouches-Fevrier, P. (1951). La structure des theories physiques. Paris: Presses
Universitaires de France.
DeWitt, B. S. (1970). "Quantum mechanics and reality". In B. S. DeWitt and
N. Graham 1973, pp. 155-65. (Originally published in Physics Today, 23
9:30-35 (September 1970).)
(1971). "The many-universes interpretation of quantum mechanics". In
B. d'Espagnat 1971, pp. 211-62 (reprinted in B. S. DeWitt and N. Graham
1973, pp. 167-218).
DeWitt, B. S. and N. Graham, eds. (1973). The Many-Worlds Interpretation of
Quantum Mechanics. A Fundamental Exposition by Hugh Everett, III, with
Papers by J. A. Wheeler, B. S. DeWitt, L. N. Cooper, D. van Vechten, and N.
Graham. Princeton, NJ: Princeton University Press.
Dicke, R. H., P. J. E. Peebles, P. G. Roll, and D. T. Wilkinson (1965). "Cosmic
black-body radiation". Astrophysical Journal. 142:414-19.
Diels, H. and W. Kranz (DK). Die Fragmente der Vorsokratiker. Siebente
Auflage. Berlin: Weidmannsche Verlagsbuchhandlung. 3 vols. (items are
identified by a number designating an author or school, followed by a letter
- A for testimonies, B for direct quotations, C for spuria -, followed by a
number).
Dirac, P. A. M. (1926a). "On the theory of quantum mechanics". Royal Society
of London Proceedings. A112:661-77.
(1926b). "The physical interpretation of the quantum dynamics". Royal
Society of London Proceedings. A113:621-41.
-(1927). "The quantum theory of the emission and absorption of radia-
tion". Royal Society of London Proceedings. Al 14:243-65 (reprinted in
J. Schwinger 1958, pp. 1-23).
-(1928). "The quantum theory of the electron". Royal Society of London
Proceedings. A117:610-24.
(1930). The Principles of Quantum Mechanics. Oxford: Clarendon Press.
-(1931). "Quantized singularities in the electromagnetic field". Royal
Society of London Proceedings. A133:60-72.
-(1958). The Principles of Quantum Mechanics. 4th Ed. (revised).
Oxford: Clarendon Press (first edition 1930, second edition 1935, third edition
1947).
-(1984). "The requirements of a fundamental physical theory". European
Journal of Physics. 5:65-67.
Drake, S. (1978). Galileo at Work: His Scientific Biography. Chicago: University
of Chicago Press.
Dugas, R. (1955). A History of Mechanics. Translated by J. R. Maddox.
Neuchatel: Editions du Griffon (the original French was published in 1950,
also by Editions du Griffon).
 References 467

Duhem, P. (1905-1906). Les origines de la statique. Paris: Hermann. 2 vols.

(1906). La theorie physique: son objet, sa structure. Paris: Chevalier et
Riviere.
-(1906-1913). Etudes sur Leonard de Vinci, ceux qu'il a lus, ceux qui Vont
lu. Paris: Hermann. 3 vols.
-(1908). "La valeur de la theorie physique, a propos d'un livre recent".
Revue generale des Sciences pures et appliquees. 19:7-19 (reprinted in Duhem
1914, pp. 473-509).
-(1913). "Examen logique de la theorie physique". Revue scientifique.
51:713-40. (English translation in Duhem HPS, pp. 232-38.)
-(1913-58). Le systeme du monde: Histoire des doctrines cosmologiques
de Platon a Copernic. Paris: Hermann. 10 vols. (Vols. 1-5 appeared in
1913-1916; vols. 6-10 were posthumously published between 1954 and
1958).
-(1914). La theorie physique: son objet, sa structure. Deuxieme edition,
revue et augmentee. Paris: Marcel Riviere (reprint of first edition of 1906, sup-
plemented with two appendices).
-(HPS). Essays in the History and Philosophy of Science. Translated and
edited, with introduction, by R. Arew and P. Barker. Indianapolis: Hackett,
1996.
Earman, J. (1992). Bayes or Bust? A Critical Examination of Bayesian Confir-
mation Theory. Cambridge, MA: The MIT Press.
(1995). Bangs, Crunches, Whimpers, and Shrieks: Singularities and
Acausalities in Relativistic Spacetimes. New York: Oxford University Press.
Earman, J. and M. Friedman (1973). "The meaning and status of Newton's Law
of Inertia and the nature of gravitational forces". Philosophy of Science.
40:329-59.
Earman, J. and J. Mosterin (1999). "A critical analysis of inflationary cosmol-
ogy". Philosophy of Science. 66 (forthcoming).
Earman, J. and J. Norton (1987). "What price spacetime substantivalism? The
hole story". British Journal for the Philosophy of Science. 38:515-25.
Earman, J., A. I. Janis, G. J. Massey, and N. Rescher, eds. (1993). Philosophical
Problems of Internal and External Worlds: Essays on the Philosophy of Adolf
Grunbaum. Pittsburgh, PA: University of Pittsburgh Press.
Eddington, A. S. (1924). The Mathematical Theory of Relativity. 2nd Ed.
Cambridge: Cambridge University Press.
Edwards, P., ed. (1967). The Encyclopedia of Philosophy. New York: Macmil-
lan. 8 vols.
Ehlers, J. E, A. E. Pirani, and A. Schild (1972). "The geometry of free fall and
light propagation". In L. O'Raifeartaigh 1972, pp. 63-84.
Ehrenfest, P. (CSP). Collected Scientific Papers. Edited by M. J. Klein. Amster-
dam: North-Holland, 1959.
Ehrenfest, P. and T. Ehrenfest (1906). "Uber eine Aufgabe aus der Wahrschein-
lichkeitsrechnung, die mit der kinetischen Deutung der Entropievermehrung
zusammenhangt". In P. Ehrenfest CSP, pp. 128-30 (originally published in
Mathematisch-Naturwissenschaftliche Blatter, Nr. 11-12, 1906).
Einstein, A. (CP). The Collected Papers of Albert Einstein. Princeton, NJ: Prince-
ton University Press, 1987- .
468 References

(1902b). "Kinetische Theorie der Warmegleichgewichtes und des zweiten

Hauptsatzes der Thermodynamik". Annalen der Physik. (4) 9:417-33
(reprinted with editorial notes in Einstein CP 2, 56-75).
-(1903). "Eine Theorie der Grundlagen der Thermodynamik". Annalen
der Physik. (4) 11:170-87 (reprinted with editorial notes in Einstein CP 2,
76-97).
-(1904). "Zur allgemeinen molekularen Theorie der Warme". Annalen der
Physik. (4) 14:354-62 (reprinted with editorial notes in Einstein CP 2,
98-108).
-(1905i). "Uber einen die Erzeugung und Verwandlung des Lichtes
betreffenden heuristischen Gesichtspunkt". Annalen der Physik. (4) 17:132-48
(reprinted with editorial notes in Einstein CP 2, 149-69).
-(1905k). "liber die von der molekularkinetischen Theorie der Warme
geforderte Bewegung von in ruhenden Fliissigkeiten suspendierten Teilchen*
Annalen der Physik. (4) 17:549-60 (reprinted with editorial notes in Einstein
CP 2, 223-36).
-(1905r). "Zur Elektrodynamik bewegter Korper". Annalen der Physik.
(4) 17:891-921 (reprinted with editorial notes in Einstein CP 2, 275-310).
-(1905s). "1st die Tragheit eines Korpers von seinem Energiegehalt
abhangig". Annalen der Physik. (4) 18:639-41 (reprinted with editorial notes
in Einstein CP 2, 311-15).
-(1906b). "Zur Theorie der Brownschen Bewegung". Annalen der Physik.
(4) 19:371-81 (reprinted with editorial notes in Einstein CP 2, 333-45).
-(1906e). "Das Prinzip von der Erhaltung der Schwerpunktsbewegung und
die Tragheit der Energie". Annalen der Physik. (4) 20:627-33 (reprinted with
editorial notes in Einstein CP 2, 359-66).
-(1907j). "Uber das Relativitatsprinzip und die aus demselben gezogenen
Folgerungen". Jahrbuch der Radioaktivitdt und Elektronik. 4:411-62
(reprinted with editorial notes in Einstein CP 2, 433-8
-(1915g). "Zur allgemeinen Relativitatstheorie (Nachtrag)". K. Preuss-
ische Akademie der Wissenschaften Sitzungsberichte, pp. 799-801 (reprinted
with editorial notes in Einstein CP 6, 225-29).
-(1915h). "Erklarung der Perihelbewegung des Merkurs aus der allge-
meinen Relativitatstheorie". K. Preussische Akademie der Wissenschaften
Sitzungsberichte, pp. 831-39 (reprinted with editorial notes in Einstein CP 6,
233-43).
-(1915i). "Die Feldgleichungen der Gravitation". K. Preussische Akademie
der Wissenschaften Sitzungsberichte, pp. 844-47 (reprinted with editorial
notes in Einstein CP 6, 244-49).
-(1916e). "Die Grundlagen der allgemeinen Relativitatstheorie". Annalen
der Physik. (4) 49:769-822 (reprinted with editorial notes in Einstein CP 6,
283-339).
-(1916j). "Strahlungs-Emission und -Absorption nach der Quantentheo-
rie". Verhandlungen der Deutschen Physikalischen Gesellschaft. 18:318-23
(reprinted with editorial notes in Einstein CP 6, 363-70).
-(1916n). "Zur Quantentheorie der Strahlung". Mitteilungen der
Physikalische Gesellschaft Zurich. 18:47-62 (reprinted with editorial notes in
 References 469

Einstein CP 6, 381-98; the same paper appeared, with a minor correction, in

Physikalische Zeitschrift 18:121-28 (1917)).
-(1917a). liber die spezielle und die allgemeine Relativitdtstheorie.
(Gemeinverstdndlich). Braunschweig: Vieweg (reprinted with editorial notes in
Einstein, CP 6, 521-39).
-(1917b). "Kosmologische Betrachtungen zur allgemeinen Relativitatsthe-
orie". K. Preussische Akademie der Wissenschaften Sitzungsberichte, pp.
142-52 (reprinted with editorial notes in Einstein CP 6, 540-52).
-(1922). "Bemerkung zu der Arbeit von A. Friedmann, 'Uber die Kriim-
mung des Raumes'". Zeitschrift fur Physik. 11:326.
-(1923). "Notiz zu der Arbeit von A. Friedmann, 'Uber die Krummung
des Raumes'". Zeitschrift fiir Physik. 16:228.
-(1924). "Quantentheorie des einatomigen idealen Gases". Preussische
Akademie der Wissenschaften. Physisch-mathematische Klasse Sitzungs-
berichte, pp. 261-67.
-(1925a). "Quantentheorie des einatomigen idealen Gases. 2. Abhand-
lung". Preussische Akademie der Wissenschaften. Physisch-mathematische
Klasse Sitzungsberichte, pp. 3-14.
-(1925b). "Quantentheorie des idealen Gases". Preussische Akademie der
Wissenschaften. Physisch-mathematische Klasse Sitzungsberichte, pp. 18-
25.
-(1933). "Zur Methode der theoretischen Physik". In Einstein 1934, pp.
176-187. (The Herbert Spencer Lecture, delivered at Oxford, 10 June 1933.)
(1934). Mein Weltbild. Zweite Auflage. Amsterdam: Querido Verlag.
-(1935). "Elementary derivation of the equivalence of mass and energy".
Bulletin of the American Mathematical Society. 41:223-30.
-(1936). "Physics and Reality". In Einstein, Out of my later years. Revised
reprint edition. Secaucus, NJ: Citadel Press, 1956, pp. 59-97. (Originally
published in The Journal of the Franklin Institute, vol. 221, no. 3, March
1936.)
Einstein, A. and M. Besso (1972). Correspondance 1903-1955. Traduction,
notes et introduction de P. Speziali. Paris: Hermann.
Einstein, A., H. Born, and M. Born (1969). Briefwechsel 1916-1955. Miinchen:
Nymphenburger Verlagshandlung.
Einstein, A. and M. Grossmann (1913). "Entwurf einer verallgemeinerten Rela-
tivitatstheorie und einer Theorie der Gravitation". Zeitschrift fiir Mathematik
und Physik. 62:225-61.
Einstein, A. and W. de Sitter (1932). "On the relation between the expansion
and the mean density of the universe". Proceedings of the National Academy
of Sciences. 18:213-14.
Einstein, A. and L. Infeld (1949). "On the motion of particles in general rela-
tivity theory". Canadian Journal of Mathematics. 1:209-41.
Einstein, A., B. Podolsky, and N. Rosen (1935). "Can quantum-mechanical
description of physical reality be considered complete?". Physical Review.
47:777-80 (reprinted in J. A. Wheeler and W. H. Zurek 1983, pp. 138-41).
Einstein, A., L. Infeld, and B. Hoffmann (1938). "The gravitational equations
and the problem of motion". Annals of Mathematics. 39:65-100.
470 References

Ellis, G. F. R. and R. M. Williams (1988). Flat and Curved Space-Times. Oxford:

Clarendon Press.
Ellis, R. L. (1849). "On the foundations of the theory of probabilities". Trans-
actions of the Cambridge Philosophical Society. 8:1-6.
Euler, L. (1736). Mechanica, sive motus scientia analytica exposita. Petropoli [St.
Petersburg]: Academia scientiarum.
(1748). "Reflexions sur l'espace et le terns". Histore de VAcademie Royale
des Sciences et Belles Lettres (Berlin). 4:324-33.
-(1765). Theoria motus corporum solidorum seu rigidorum ex primis
nostrae cognitionis principiis stabilita et ad omnes motus qui in huiusmodi
corpora cadere possunt, accommodata. Rostock: A. F. Rose.
-(1768-74). Lettres a une princesse d'Allemagne sur divers sujets de
physique et de philosophie. St. Petersbourg: Academie Imperiale des Sciences.
3 vols.
Everett (III), H. (1957). "'Relative state' formulation of quantum mechanics".
Reviews of Modern Physics. 29:454-62. (Everett's Princeton doctoral
dissertation; reprinted in B. S. DeWitt and N. Graham 1973, pp. 141-
49.)
(1973). "The theory of the universal wave function". In B. S. DeWitt and
N. Graham 1973, pp. 3-140 (this detailed presentation of Everett's ideas was
written about the same time as his brief Ph.D. thesis of 1957, but it remained
unpublished until 1973).
Falkenburg, B. (1995). Teilchenmetaphysik: Zur Realitdtsauffassung in Wis-
senschaftsphilosophie und Microphysik. 2. Auflage. Heidelberg: Spektrum.
Fano, U. (1957). "Description of states in Quantum Mechanics by density matrix
and operator techniques". Reviews of Modern Physics. 29:74-93.
Faraday, M. Diary. London: G. Bell & Sons, 1932-1936. 7 vols.
(EREM). Experimental Researches in Electricity and Magnetism.
London: Richard & John E. Taylor, 1839-1855. 3 vols.
Fermi, E. (1926). "Zur Quantelung des idealen einatomigen Gases". Zeitschrift
fur Physik. 36:902-12.
(1937). Thermodynamics. Englewood Cliffs, NJ: Prentice-Hall.
Fevrier, P. (1937a). "Les relations d'incertitude de Heisenberg et la logique".
Comptes rendus de VAcademie des Sciences. 204:481-83.
(1937b). "Sur une forme generale de la definition d'une logique".
Comptes rendus de VAcademie des Sciences. 204:958-59.
-(1951). See Destouches-Fevrier, P.
Feyerabend, P. K. (1958). "An attempt at a realistic interpretation of experience".
Proceedings of the Aristotelean Society. 58:143ff (reprinted in Feyerabend
1981, I, 17-36).
(1962). "Explanation, reduction and empiricism". Minnesota Studies
in the Philosophy of Science. 3:28-97 (reprinted in Feyerabend 1981, I,
44-96).
-(1970). "Against Method: Outline of an Anarchistic Theory of Knowl-
edge". Minnesota Studies in the Philosophy of Science. 4:17-130 (enlarged
version in book form: London, New Left Books, 1975; revised edition:
London, Verso, 1988).
 References 471

(1981). Philosophical Papers. Cambridge: Cambridge University Press. 2

vols.
Feynman, R. P., R. B. Leighton, and M. Sands (1964). The Feynman Lectures
on Physics. Volume II: Mainly Electromagnetism and Matter. Reading, MA:
Addison-Wesley.
Finkelstein, D. (1962/63). "The logic of quantum physics". Transactions of the
New York Academy of Sciences. 25:621-37.
(1969). "Matter, space and logic". In R. S. Cohen and M. Wartofsky, eds.,
Boston Studies in the Philosophy of Science. Dordrecht: Reidel, Vol. 5, pp.
199-215.
-(1972). "The physics of logic". In R. G. Colodny 1972, pp. 47-66.
FitzGerald, G. F. (1885). "On the structure of mechanical models illustrating
some properties of the aether". In G. F. FitzGerald 1902, pp. 157-62 (origi-
nally published in Philosophical Magazine, 19:438-43 (1885)).
(1889). "The ether and the earth's atmosphere". Science. 12:390.
(1902). Scientific Writings of the Late George Francis FitzGerald. Edited
by J. Larmor. Dublin: Hodges and Figgis.
Fourier, J. (1822). Theorie analytique de la chaleur. Paris: Firmin Didot (facsimile
reprint: Paris, Jacques Gabay, 1988).
Fox, R. (1971). The Caloric Theory of Gases from Lavoisier to Regnault.
Oxford: Clarendon Press.
Frank, P. (1932). Das Kausalgesetz und seine Grenzen. Wien: Julius Springer.
(1935). "Philosophische Deutungen und Missdeutungen der Quanten-
theorie". Erkenntnis. 6:303-17.
Frege, G. (KS). Kleine Schriften. Herausgegeben von I. Angelelli. Darmstadt:
Wissenschaftliche Buchgesellschaft, 1967.
(1879). Begriffschrift, eine der arithmetischen nachgebildete Formel-
sprache. Halle a.S.: Louis Nebert.
-(1884). Die Grundlagen der Arithmetik: Eine logisch mathematische
Untersuchung uber den Begriff der Zahl. Breslau: Wilhelm Koebner.
Friedman, M. (1983). Foundations of Space-Time Theories: Relativistic Physics
and the Philosophy of Science. Princeton, NJ: Princeton University Press.
(1992). Kant and the Exact Sciences. Cambridge, MA: Harvard
University Press.
Friedmann, A. (1922). "Uber die Krummung des Raumes". Zeitschrift fur
Physik. 10:377-86.
(1924). "Uber die Moglichkeit einer Welt mit konstanter negativer
Krummung des Raumes". Zeitschrift ftir Physik. 21:326-32.
Friedrichs, K. (1927). "Eine invariante Formulierung des Newtonschen Gravita-
tionsgesetzes und des Grenziiberganges vom Einsteinschen zum Newtonschen
Gesetz". Mathematische Annalen. 98:566-75.
Funkenstein, A. (1986). Theology and the Scientific Imagination from the Middle
Ages to the Seventeenth Century. Princeton, NJ: Princeton University Press.
Gaifman, H. and M. Snir (1982). "Probabilities over rich languages". Journal of
Symbolic Logic. 47:495-548.
Galileo Galilei (EN). Le Opere. Nuova ristampa della Edizione Nazionale.
Firenze: G. Barbera, 1964-1966. 20 vols.
472 References

(1623). II saggiatore nel quale con bilancia esquisita e giusta si ponder-

ano le cose contenute nella Libra astronomica e fdosofica di Lotario Sarsi
Sigensano, scritto in forma di lettera all' ULmo et rever.mo mons.re d. Virginio
Cesarini. Roma: Appresso Giacomo Mascardi.
-(1638). Discorsi e dimostrazioni matematiche intorno a due nuove scienze
attenenti alia Mecanica & i Movimenti Locali. Leida: Appresso gli Elsevirii.
-(1974). Two New Sciences, including Centers of Gravity and Force of
Percussion. Translated by S. Drake. Madison, WI: University of Wisconsin
Press.
Gamow, G. (1948). "The evolution of the universe". Nature. 162:680-82.
Gauss, C. F. (WW). Werke. Achter Band. Herausgegeben von der Koniglichen
Gesellschaft der Wissenschaften zu Gottingen. Leipzig: Teubner, 1900.
Geiger, H. and E. Marsden (1909). "On a diffuse reflection of the a-particles".
Royal Society of London Proceedings. 82:495-99.
Gell-Mann, M. and J. B. Hartle (1993). "Classical equations for quantum
systems". Physical Review D. 47:3345-82.
Gentzen, G. (1934). "Untersuchungen iiber das logische Schliessen". Mathema-
tische Zeitschrift. 39:176-210, 405-31.
Gibbs, J. W. (1902). Elementary Principles in Statistical Mechanics developed
with especial reference to the Rational Foundations of Thermodynamics. New
York: Scribner's (reprint: Woodbridge, CN, Ox Bow Press, 1981).
Gilson, E. (1930). Etude sur le role de la pensee medievale dans la formation du
systeme cartesien. Paris: Vrin.
Gleason, A. M. (1957). "Measures on the closed subspaces of a Hilbert space".
Journal of Mathematics and Mechanics. 6:885-93 (reprinted in C. A. Hooker
1975, pp. 123-33).
Godel, K. (1949). "An example of a new type of cosmological solutions of Ein-
stein's field equations of gravitation". Reviews of Modern Physics. 21:447-50.
Goldstein, H. (1950). Classical Mechanics. Reading, MA: Addison-Wesley.
Gooding, D. (1981). "Final steps to the field theory: Faraday's study of magnetic
phenomena, 1845-1850". Historical Studies in the Physical Sciences.
11:231-75.
Gouy, L. G. (1888). "Note sur le mouvement brownien". Journal de physique
theorique et appliquee. 7:561-64.
Greenberger, D. M., M. A. Home, and A. Zeilinger (1989). "Going beyond Bell's
Theorem". In M. Kafatos, ed., Bell's Theorem, Quantum Theory, and Con-
ceptions of the Universe. Dordrecht: Kluwer; pp. 73-76.
Greenberger, D. M., M. A. Home, A. Shimony, and A. Zeilinger (1990). "Bell's
Theorem without inequalities". American Journal of Physics. 58:1131-43.
Guth, A. H. (1982). "Inflationary universe: A possible solution of the horizon
and flatness problems". Physical Review D. 23:347-56.
(1997). The Inflationary Universe: The Quest for a New Theory of
Cosmic Origins. Reading, MA: Addison-Wesley.
Hacking, I. (1965). Logic of Statistical Inference. Cambridge: Cambridge
University Press.
Hafele, J. C. and R. E. Keating (1972). "Around-the-world atomic clocks".
Science. 177:166-70.
 References 473

Hall, A. R. and M. B. Hall, eds. (1962). Unpublished Scientific Papers of Isaac

Newton. A Selection from the Portsmouth Collection in the University Library,
Cambridge. Cambridge: Cambridge University Press.
Hamilton, W. R. (1835). "Second Essay on a General Method in Dynamics".
Royal Society of London Philosophical Transactions. 125:95ff. (reprinted
together with the - first - essay "On a General Method in Dynamics" in Hamil-
ton, Collected Papers, Vol. II. Cambridge: Cambridge University Press, 1940,
pp. 103-211).
Hannabuss, K. (1997). An Introduction to Quantum Theory. Oxford:
Clarendon Press.
Hanson, N. R. (1958). Patterns of Discovery. Cambridge: Cambridge University
Press.
Harman, P. M. (1982a). Metaphysics and Natural Philosophy: The Problem of
Substance in Classical Physics. Sussex: Harvester.
(1982b). Energy, Force, and Matter: The Conceptual Development of
Nineteenth-Century Physics. Cambridge: Cambridge University Press.
-(1987). "Mathematics and reality in Maxwell's dynamical physics". In R.
Kargon and P. Achinstein 1987, pp. 267-97.
Havas, P. (1964). "Four-dimensional formulations of Newtonian mechanics and
their relation to the special and the general theory of relativity". Reviews of
Modern Physics. 36:938-65.
Hawking, S. W. (1974). "Black hole explosions?". Nature. 248:30-31.
(1975). "Particle creation by black holes". Communications of Mathe-
matical Physics. 43:199-220.
Hawking, S. W. and G. F. R. Ellis (1973). The Large Scale Structure of Space-
Time. Cambridge: Cambridge University Press.
Hawking, S. and W. Israel, eds. (1987). Three Hundred Years of Gravitation.
Cambridge: Cambridge University Press.
Hawking, S. W. and R. Penrose (1970). "The singularities of gravitational col-
lapse and cosmology". Royal Society of London Proceedings. A 314:529-48.
Heisenberg, W. (1925). "Uber quantentheoretische Umdeutung kinematischer
und mechanischer Beziehungen". Zeitschrift fur Physik. 33:879-93 (English
translation in B. L. Van der Waerden 1967, pp. 261-76).
Heisenberg, W. (1927). "Uber den anschaulichen Inhalt der quantentheoretis-
chen Kinematik und Mechanik". Zeitschrift fiir Physik. 43:172-98.
(1930). The Physical Principles of the Quantum Theory. Translated into
English by C. Eckart and F. C. Hoyt. Chicago: University of Chicago Press.
-(1969). Der Teil und das Game. Miinchen: Piper.
Helmholtz, H. (WA). Wissenschaftliche Abhandlungen. Leipzig: Barth,
1882-1895.
(1847). Uber die Erhaltung der Kraft. Berlin: Reimer (reprinted, with
addenda, in Helmholtz WA, I, 12-75).
-(1866). "Uber die tatsachlichen Grundlagen der Geometrie". Verhand-
lungen des naturhistorisch-medicinischen Vereins zu Heidelberg. 4:197-202.
-(1887). "Zahlen und Messen, erkenntnistheoretisch betrachtet". In
Philosophische Aufsdtze Eduard Zeller gewidmet. Leipzig: Verlag Fues
(reprinted in H. von Helmholtz, Die Tatsachen in der Wahrnehmung/Zdhlen
474 References

und Messen. Darmstadt: Wissenschaftliche Buchgesellschaft, 1959; pp.

77-112).
Hempel, C. G. and P. Oppenheim (1948). "Studies in the logic of explanation".
Philosophy of Science. 15:135-75.
Hepp, K. (1972). "Quantum theory of measurement and macroscopic observ-
ables". Helvetica Physica Ada. 45:237-48.
Herapath, J. (1821). "A mathematical inquiry into the causes, laws and princi-
pal phenomenae of heat, gases, gravitation, etc.". Annals of Philosophy. (2)
1:273-93, 340-51, 401-06.
Herivel, J. (1965). The Background to Newton's Principia: A Study of Newton's
Dynamical Researches in the Years 1664-1684. Oxford: Clarendon Press.
Herschel, J. F. W. (1830). A Preliminary Discourse on the Study of Natural
Philosophy. London: Longman, Rees, Orme, Brown, 6t Green, and John
Taylor (facsimile reprint: Chicago, University of Chicago Press, 1987).
Hertz, H. (1894). Die Prinzipien der Mechanik in neuem Zusammenhang
dargestellt. Leipzig: Barth (facsimile reprint: Darmstadt, Wissenschaftliche
Buchgesellschaft, 1963).
Hilbert, D. (1899). "Grundlagen der Geometrie". In Festschrift zur Feier der
Enthullung des Gauss-Weber-Denkmals in Gottingen. Leipzig: Teubner, pp.
3-92.
Hintikka, J. (1967). "Kant on the mathematical method". The Monist.
51:352-75 (reprinted in L. W. Beck, ed., Kant Studies Today. LaSalle, IL: Open
Court, 1959, pp. 117-40).
Hooke, R. (PW). Posthumous Works. London: Frank Cass & Co., 1974.
Hooker, C. A., ed. (1975). The Logico-Algebraic Approach to Quantum
Mechanics. Volume I. Dordrecht: Reidel.
Hopf, L. (1948). Introduction to the Differential Equations of Physics. Trans-
lated by W. Nef. New York: Dover.
Hoyle, F. (1948). "A new model for the expanding universe". Monthly Notices
of the Royal Astronomical Society. 108:372-82.
Hubble, E. P. (1929). "A relation between distance and radial velocity among
extragalactic nebulae". Proceedings of the National Academy of Sciences.
15:168-73.
Hughes, R. I. G. (1989). The Structure and Interpretation of Quantum Mechan-
ics. Cambridge, MA: Harvard University Press.
Hume, D. (THN). A Treatise of Human Nature. Edited by L. A. Selby-Bigge.
Oxford: Clarendon Press, 1888.
Hunt, B. J. (1987). "'How my model was right': G. F. FitzGerald and the
reform of Maxwell's theory". In R. Kargon and P. Achinstein 1987, pp.
299-321.
Huygens, C. (OC). (Euvres completes. Publiees par la Societe Hollandaise des
Sciences. La Haye: Martinus Nijhoff, 1888-1950. 22 vols.
(1690). Traite de la lumiere. Avec un discours de la cause de la pesanteur.
Leide.
Ignatowsky, W. (1910). "Einige allgemeine Bemerkungen zum Relativitat-
sprinzip". Verhandlungen der Deutschen Physikalischen Gesellschaft.
12:788-96.
 References 475

(1911). "Das Relativitatsprinzip". Archiv fiir Mathematik und Physik. (3)

17:1-24; 18:17-41.
Jammer, M. (1961). Concepts of Mass in Classical and Modern Physics.
Cambridge, MA: Harvard University Press.
(1966). The Conceptual Development of Quantum Mechanics. New
York: McGraw-Hill.
-(1974). The Philosophy of Quantum Mechanics: The Interpretations of
Quantum Mechanics in Historical Perspective. New York: Wiley.
Jaskowski, S. (1934). "On the rules of suppositions in formal logic". Studia
Logica. 1:5-32 (reprinted in S. McCall, ed., Polish Logic 1920-1939, Oxford:
Clarendon Press, 1967, pp. 231-58).
Jauch, J. and C. Piron (1963). "Can hidden variables be excluded in Quantum
Mechanics?". Helvetica Physica Acta. 38:827-37.
Jordan, P. (1926). "Uber kanonische Transformationen in der Quanten-
mechanik". Zeitschrift fur Physik. 37:383-86; 38:513-17.
(1927). "Uber eine neue Begriindung der Quantenmechanik". Zeitschrift
fiir Physik. 40:809-38.
Joule, J. P. (1843). "On the heat evolved during the electrolysis of water". In
J. P. Joule 1884, pp. 109-23 (read to the Manchester Literary and Philosoph-
ical Society on 24 January 1843).
(1844). "On the changes of temperature produced by the rarefaction and
condensation of air". In J. P. Joule 1884, pp. 171-89 (communicated on 20
June 1844 to the Royal Society of London, which only published an abstract).
-(1849). "On the mechanical equivalent of heat". In J. P. Joule 1884, pp.
298-328 (read to the Royal Society of London on 21 June 1849; published in
Royal Society of London Philosophical Transactions, 140:61-82 (1850)).
-(1884). The Scientific Papers of James Prescott Joule. London: The Phys-
ical Society (an additional volume appeared in 1886).
Kant, I. (Ak.). Gesammelte Schriften. Herausgegeben von der K. Preussischen,
bzw. Deutschen Akademie der Wissenschaften. Berlin, 1902ff.
(1746). Gedanken von der wahren Schdtzung der lebendigen Krdfte und
Beurtheilung der Beweise derer sich Herr von Leibniz und andere Mechaniker
in dieser Streitsache bedienet haben, nebst einigen vorhergehenden Betrach-
tungen welche die Kraft der Korper uberhaupt betreffen. Konigsberg: Martin
Eberhard Dorn (reprinted in Kant, Ak. I, 1-181).
-(1755). Allgemeine Naturgeschichte und Theorie des Himmels, oder
Versuch von der Verfassung und dem mechanischen Ursprunge des ganzen
Weltgebdudes nach Newtonischen Grundsdtzen abgehandelt. Konigsberg und
Leipzig: Johann Friedrich Petersen (reprinted in Kant, Ak. I, 215-368).
-(1756). Metaphysica cum Geometria lunctae Usus in Philosophia
Naturalis, cuius Specimen I. continet Monadologiam Physicam. Konigsberg
(reprinted in Kant, Ak. I, 473-87).
-(1766). Trdume eines Geistersehers, erldutert durch Trdume der Meta-
physik. Konigsberg: Kanter (reprinted in Kant, Ak. II, 315-73).
-(1768). "Von dem ersten Grunde des Unterschiedes der Gegenden im
Raume". Wochentliche Konigsbergsche Frag- und Anzeigungsnachrichten. N°
6-8, 13. und 20. Februar 1768 (reprinted in Kant, Ak. II, 375-83).
476 References

(1770). De mundi sensibilis atque intelligibilis forma et principiis.

Regiomonti: Io. lac. Kanter (reprinted in Kant, Ak. II, 384-419).
-(1781). Critik der reinen Vernunft. Riga: Johann Friedrich Hartknoch (the
first half, which was substantially modified in the second edition, is reprinted
in Kant, Ak. IV, 1-252).
-(1783). Prolegomena zu einer jeden kiinftigen Metaphysik, die als Wis-
senschaft wird auftreten konnen. Riga: Johann Friedrich Hartknoch (reprinted
in Kant, Ak. IV, 252-383).
-(1786). Metaphysische Anfangsgriinde der Naturwissenscbaft. Riga:
Johann Friedrich Hartknoch (reprinted in Kant, Ak. IV, 465-565).
-(1787). Critik der reinen Vernunft. Zweyte hin und wieder verbesserte
Auflage. Riga: Johann Friedrich Hartknoch (reprinted in Kant, Ak. Ill, pp.
1-552).
-(1933). Critique of Pure Reason. Second edition with corrections. Trans-
lated by N. Kemp Smith. London: Macmillan.
Kargon, R. and P. Achinstein, eds. (1987). Kelvin's Baltimore Lectures and
Modern Theoretical Physics: Historical and Philosophical Perspectives. Cam-
bridge, MA: MIT Press.
Kelvin, Lord. See Thomson, W.
Kepler, J. (GW). Gesammelte Werke. Herausgegeben von M. Caspar et al.
Miinchen: C. H. Beck, 1939- .
Keynes, J. M. (1921). A Treatise on Probability. London: Macmillan.
King, H. R. (1956). "Aristotle without Prima Materia". Journal for the History
of Ideas. 17:370-89.
Kirchhoff, G. (1860). "Uber das Verhaltnis zwischen dem Emissionsvermogen
und dem Absorptionsvermogen der Korper fur Warme und Licht". Annalen
der Physik. (2) 109:275-301.
Klein, F. (GA). Gesammelte mathematische Abhandlungen. Berlin: Springer,
1921-1923. 3 vols.
(1871). "Uber die sogenannte Nicht-Euklidische Geometrie". Mathema-
tische Annalen. 4:573-625.
-(1872). Vergleichende Betrachtungen uber neueregeometrische Forschun-
gen. Erlangen: A. Duchert.
-(1873). "Uber die sogenannte Nicht-Euklidische Geometrie (Zweiter
Aufsatz)". Mathematische Annalen. 6:112-45.
-(1874). "Nachtrag zu dem 'zweiten Aufsatz' uber Nicht-Euklidische
Geometrie". Mathematische Annalen. 7:531-37.
(1890). "Zur Nicht-Euklidischen Geometrie". Mathematische Annalen.
37:544-72.
(1893). "Vergleichende Betrachtungen uber neuere geometrische
Forschungen". Mathematische Annalen. 43:63-100 (revised version of Klein
1872).
-(1911). "Uber die geometrischen Grundlagen der Lorentz-Gruppe".
Physikalische Zeitschrift. 12:17-27.
Kochen, S. and E. P. Specker (1967). "The problem of hidden variables in
Quantum Mechanics". Journal of Mathematics and Mechanics. 17:59-87
(reprinted in C. A. Hooker 1975, pp. 293-328).
 References 477

Koyre, A. (1923). Descartes und die Scholastik. Bonn: Bouvier.

(1965). Newtonian Studies. London: Chapman £>c Hall.
Kragh, H. (1996). Cosmology and Controversy: The Historical Development
of Two Theories of the Universe. Princeton, NJ: Princeton University
Press.
Kramers, H. (1924a). "The law of dispersion and Bohr's theory of spectra".
Nature. 113:673-674 (reprinted in B. L. Van der Waerden 1967, pp. 177-80).
(1924b). "The quantum theory of dispersion". Nature. 114:310-11
(reprinted in B. L. Van der Waerden 1967, pp. 199-201).
Kramers, H. and W. Heisenberg (1925). "Uber die Streuung von Strahlung durch
Atome". Zeitschrift fur Physik. 31:681-708 (English translation in B. L. Van
der Waerden 1967, pp. 223-52).
Kronig, A. K. (1856). "Grundziige einer Theorie der Gase". Annalen der Physik.
(2) 99:315-22.
Kuhn, T. S. (1959). "Energy conservation as an example of simultaneous dis-
covery". In M. Claggett 1959b, pp. 321-56 (reprinted in Kuhn 1977, pp.
66-104).
(1962). The Structure of Scientific Revolutions. Chicago: University of
Chicago Press.
-(1970). The Structure of Scientific Revolutions. 2nd Ed., enlarged.
Chicago: University of Chicago Press.
-(1977). The Essential Tension: Selected Studies in Scientific Tradition and
Change. Chicago: University of Chicago Press.
-(1978). Black-Body Theory and the Quantum Discontinuity: 1894-1912.
Oxford: Clarendon Press.
Kuratowski, K. (1921). "Sur la notion d'ordre dans la theorie des ensembles".
Fundamenta Mathematicae. 2:161-171.
Kyburg Jr., H. E. and H. E. Smokier, eds. (1980). Studies in Subjective Proba-
bility. Huntington, NY: Krieger.
Lagrange, J. L. (1788). Mecanique analitique. Paris: Chez la Veuve Desaint (fac-
simile reprint: Paris, Jacques Gabay, 1989).
Lambert, J. H. (1786). "Theorie der Parallellinien". Magazin fur die reine und
angewandte Mathematik. 2:137-64; 3:325-58 (written about 1766; posthu-
mously published by J. Bernoulli).
Landau, L. D. and E. M. Lifshitz (1960). Course of Theoretical Physics. Vol. I,
Mechanics. Translated from the Russian by J. B. Sykes and J. S. Bell. Oxford:
Pergamon Press.
Lange, L. (1885). "Uber das Beharrungsgesetz". K. Sdchsische Gesellschaft der
Wissenschaften zu Leipzig; math.-phys. Cl. Berichte. 37:333-51.
Laplace, P. S. (OC). (Euvres completes. Paris: Gauthier-Villars, 1878-1912. 14
vols.
(1795). Essai philosophique sur les probabilites. Paris (later republished
as the Introduction to Laplace, Traite analytique des probabilites, Paris 1814,
reprinted in Laplace, OC, VIII).
Larmor, J. (1900). Aether and Matter: A Development of the Dynamical Rela-
tions of the Aether to Material Systems on the basis of the Atomic Constitu-
tion of Matter, including a discussion of the influence of the Earth's motion
478 References

on optical phenomena. Being an Adams Prize Essay in the University of Cam-

bridge. Cambridge: Cambridge University Press.
-(1904). "Lord Kelvin on optical and molecular dynamics". Nature.
70(Supplement):iii-v.
Leibniz, G. W. (GP). Die philosophischen Schriften. Herausgegeben von C. J.
Gerhardt. Hildesheim: Olms, 1965. 7 vols. (unaltered reprint of the original
edition published in Berlin, 1875-1890).
(OFI). Opuscules et fragments inedits. Extraits des manuscrits de la
Bibliotheque royale de Hanovre par L. Couturat. Hildesheim: Georg Olms,
1961 (facsimile reprint of the first edition: Paris, Alcan, 1903).
Lemaitre, G. (1927). "Un univers homogene de masse constante et de rayon
croissant, rendant compte de la vitesse radiale des nebuleuses extra-
galactiques". Annales de la Societe Scientifique de Bruxelles. A 47:49-59.
Linde, A. (1983). "Chaotic inflation". Physical Letters. B 129:177-81.
(1986). "Eternal chaotic inflation". Modern Physics Letters. A 1:81.
(1987). "Inflation and quantum cosmology". In S. Hawking and W. Israel
1987, pp. 604-30.
Lindsay, R. B. and H. Margenau (1957). Foundations of Physics. New York:
Dover (corrected reprint of first edition, published in 1936).
Lobachevsky, N. I. (ZGA). Zwei geometrische Abhandlungen. Aus dem Russis-
chen iibersetzt von F. Engel. Leipzig: Teubner, 1898-99. 2 vols.
[Locke, J.] (1690). An Essay concerning Humane Understanding. In four books.
London: Printed for Thomas Basset and sold by Edward Mory. (Published
anonymously; the author's name was added in the second edition.)
Locke, J. (1699). Mr. Locke's Reply to the Right Reverend Lord Bishop of
Worcester's Answer to his Second Letter. London: Printed for A. and J.
Churchil; and E. Castle.
London, F. (1926a). "Uber die Jacobischen Transformationen der Quanten-
mechanik". Zeitschrift fur Physik. 37:915-25.
(1926b). "Winkelvariable und kanonische Transformationen in der
Undulationsmechanik". Zeitschrift fur Physik. 40:193-219.
London, F. and E. Bauer (1939). La theorie de Vobservation en mecanique quan-
tique. Paris: Hermann. Actualites scientifiques e industrielles, N° 775 (English
translation, including a new paragraph by Fritz London, in J. A. Wheeler and
W. H. Zurek 1983, pp. 217-59).
Lorentz, H. A. (CP). Collected Papers. Edited by P. Zeeman and A. D. Fokker.
The Hague: Nijhoff, 1935-1939. 9 vols.
(1892). "The relative motion of the earth and the ether". In Lorentz CP,
vol. 4, pp. 219-23 (originally published in Dutch in Versl. K. Nederlandse
Akademie van Wetenschappen 1:74-79 (1892)).
-(1895). Versuch einer Theorie der electrischen und optischen Erschein-
ungen in bewegten Korpern. Leiden: Brill (reprinted in Lorentz CP, vol. 5, pp.
1-137).
-(1897). "Uber den Einfluss magnetischer Krafte auf die Emission des
Lichtes". Annalen der Physik. (3) 63:278-84.
-(1899). "Theorie simplified des phenomenes electriques et optiques dans
les corps en mouvement". In Lorentz CP, vol. 5, pp. 139-55 (originally pub-
 References 479

lished in Dutch in Versl. K. Nederlandse Akademie van Wetenschappen,

7:507-22 (1899)).
-(1904). "Electromagnetic phenomena in a system moving with any veloc-
ity smaller than that of light". K. Nederlandse Akademie van Wetenschappen.
Proceedings. 6:809-31 (reprinted in Lorentz, CP, vol. 5, pp. 172-97).
Lorenzen, P. and K. Lorenz (1978). Dialogische Logik. Darmstadt: Wis-
senschaftliche Buchgesellschaft.
Loschmidt, J. (1876). "Uber den Zustand des Warmegleichgewichtes eines
Systemes von Korpern mit Riicksicht auf die Schwerkraft". Sitzungsberichte
der Akademie der Wissenschaften zu Wien, math.-naturwiss. Klasse. 73:128-
42.
Lovelock, D. and H. Rund (1975). Tensors, Differential Forms, and Variational
Principles. New York: Wiley.
Lucas, J. R. (1984). Space, Time and Causality: An Essay in Natural Philoso-
phy. Oxford: Clarendon Press.
Liiders, G. (1951). "Uber die Zustandsanderung durch den Messprozess".
Annalen der Physik. (5) 8:322-28.
Ludwig, G. (1953). "Der Messprozess". Zeitschrift fur Physik. 135:483-511.
(1954). Grundlagen der Quantenmechanik. Berlin: Springer.
(1979). Einfuhrung in die Grundlagen der theoretischen Physik. Band 4:
Makrosysteme, Physik und Mensch. Braunschweig: Vieweg.
-(1984). Einfiihrung in die Grundlagen der theoretischen Physik. Band 3:
Quantentheorie. 2. Auflage. Braunschweig: Vieweg.
-(1984/85). Foundations of Quantum Mechanics. Translated by C. A.
Hain. New York: Springer. 2 vols. (much revised and enlarged English edition
of Ludwig 1954).
-(1985/87). An Axiomatic Basis for Quantum Mechanics. Volume I:
Derivation of Hilbert Space Structure. Volume II: Quantum Mechanics and
Macrosystems. Berlin: Springer.
Lukasiewickz, J. (1920). "On three-valued logic". In J. Lukasiewickz, Selected
Works. Edited by L. Borkowski. Amsterdam: North-Holland, 1970; pp. 87-88
(first published in Polish in Ruch Filosoficzny, 5:169-71).
Mach, E. (1868). "Uber die Definition der Masse". Carl's Repertorium der
Experimentalphysik. 4:355-59 (reprinted in Mach 1909, pp. 50-54).
(1872). Die Geschichte und die Wurzel des Satzes von der Erhaltung der
Arbeit. Vortrag gehalten in der K. Bohm. Gesellschaft der Wissenschaften am
15. Nov. 1871. Prag: Calve.
-(1875). Grundlinien der Lehre von den Bewegungsempfindungen.
Leipzig: Engelmann.
-(1883). Die Mechanik in ihrer Entwicklung historisch-kritisch dargestellt.
Leipzig: Brockhaus.
(1886). Beitrdge zur Analyse der Empfindungen. Jena: Fischer.
-(1896). Die Principien der Wdrmelehre: Historisch-kritisch entwickelt.
Leipzig: Barth.
-(1905). Erkenntnis und Irrtum: Skizzen zur Psychologie der Forschung.
Leipzig: Barth.
(1906). Die Analyse der Empfindungen und das Verhdltnis des physischen
480 References

zum psychischen. Jena: Fischer (fifth edition, substantially enlarged, of Mach

1886).
-(1909). Die Geschichte und die Wurzel des Satzes von der Erhaltung der
Arbeit. Vortrag gehalten in der K. Bohm. Gesellschaft der Wissenschaften am
15. Nov. 1871. Leipzig: Barth (unmodified reprint of Mach 1872).
-(1921). Die Prinzipien der physikalischen Optik: Historisch und er-
kenntnispsychologisch entwickelt. Leipzig: Barth.
-(AS). The Analysis of Sensations and the Relation of the Physical to the
Psychical. Translated from the first German edition by C. M. Williams. Revised
and supplemented from the fifth German edition by S. Waterlow. New York:
Dover (English translation of Mach 1886/1906).
-(El). Erkenntnis und Irrtum: Skizzen zur Psychologie der Forschung.
Darmstadt: Wissenschaftliche Buchgesellschaft, 1968 (unmodified reprint of
the fifth edition, Leipzig, 1926).
-(M). Die Mechanik: historisch-kritisch dargestellt. Darmstadt: Wis-
senschaftliche Buchgesellschaft, 1963 (unmodified reprint of the ninth edition,
Leipzig, 1933).
Mackey, G. W. (1963). The Mathematical Foundations of Quantum Mechanics:
A lecture-note volume. Reading, MA: Benjamin/Cummings.
Mainzer, K. (1995). "Quantentheorie". In J. Mittelstrass, ed., Enzyklopddie
Philosophie und Wissenschaftstheorie. Band 3. Stuttgart: Metzler, 1995; pp.
429-36.
Maupertuis, P. L. M. (1732). Discours sur les differentes figures des astres avec
une exposition des systemes de MM. Descartes et Newton. Paris: Imprimerie
Royale.
Maxwell, J. C. (1855/56). "On Faraday's lines of force". In Maxwell 1890, vol.
I, pp. 155-229 (read on 10 December 1855 and 11 February 1856 at the Cam-
bridge Philosophical Society; published in the Society's Transactions, 10 1
(1857)).
(1860). "Illustrations of the dynamical theory of gases". In Maxwell
1890, vol. I, pp. 377-409. (Originally published in the Philosophical Maga-
zine, (4) 19:19-32; 20:21-37 (I860).)
-(1861-62). "On physical lines of force". In Maxwell 1890, vol. I, pp.
451-513. (Originally published in the Philosophical Magazine.)
-(1864). "A dynamical theory of the electromagnetic field". In Maxwell
1890, pp. 526-97 (received on 27 October and read on 8 December 1864 at
the Royal Society of London; published in the Philosophical Transactions,
155).
-(1866). "On the dynamical theory of gases". In Maxwell 1890, vol. II,
pp. 26-78 (read at the Royal Society of London on 31 May 31 1866; origi-
nally published in the Philosophical Transactions, 157:49-88 (1867)).
-(1873). A Treatise on Electricity and Magnetism. Oxford: Clarendon
Press.
-(1883). Theory of Heat. Seventh edition. London: Longmans, Green, and
Co.
-(1890). The Scientific Papers of James Clerk Maxwell. Edited by W. D.
Niven. Cambridge: Cambridge University Press. 2 vols. (unaltered reprint:
New York, Dover, 1965).
 References 481

(1891). A Treatise on Electricity and Magnetism. Third edition. Oxford:

Clarendon Press. 2 vols. (unabridged, slightly altered reprint: New York,
Dover, 1954).
Mayo, D. G. (1996). Error and the Growth of Experimental Knowledge.
Chicago: University of Chicago Press.
McMullin, E., ed. (1965). The Concept of Matter in Greek and Medieval Phi-
losophy. Notre Dame: University of Notre Dame Press (reprint of the first part
of The Concept of Matter, ed. by E. McMullin, Notre Dame: University of
Notre Dame Press, 1963).
Mehra, J. and H. Rechenberg (1982-88). The Historical Development of
Quantum Theory. New York: Springer. Five volumes bound as six.
Messiah, A. (1961). Quantum Mechanics. Translated from the French by G. M.
Temmer and J. Potter. Amsterdam: North-Holland Publishing Co. 2 vols.
Michelson, A. A. (1881). "The relative motion of the earth and the luminifer-
ous ether". American Journal of Science. 22:120-29.
(1891). "On the application of interference methods to spectroscopic
measurements". Philosophical Magazine. 31:338-46 (continued in Philosoph-
ical Magazine, 34:280-99 (1892)).
Michelson, A. A. and E. W. Morley (1887). "On the relative motion of the earth
and the luminiferous ether". American Journal of Science. 34:333-45.
Migne, J. P., ed. (PG). Patrologia graeca. Paris, 1857-1866.
Mill, J. S. (1843). A System of Logic Ratiocinative and Inductive. London: John
W. Parker.
(1874). A System of Logic Ratiocinative and Inductive, Being a Con-
nected View of the Principles of Evidence and the Method of Scientific Inves-
tigation. 8th Ed. New York: Harper & Brothers. (This edition, which appeared
shortly after Mill's death, was the last one he corrected for publication. The
University of Toronto Press published in 1973 a critical edition with all variant
readings. Unfortunately, I no longer had access to it when I wrote this book.)
Miller, A. I., ed. (1990). Sixty-Two Years of Uncertainty: Historical Philosoph-
ical and Physical Inquiries into the Foundations of Quantum Mechanics. New
York: Plenum.
(1994). Early Quantum Electrodynamics: A Source Book. Translations
from the German by W. Grant. Cambridge: Cambridge University Press.
Minkowski, H. (1909). "Raum und Zeit". Physikalische Zeitschrift. 10:104-11.
(1915). "Das Relativitatsprinzip". Jahresbericht der Deutschen
Mathematiker-Vereinigung. 24:372-82. (Lecture delivered at the Gottingen
Mathematical Society on 5 November 1907; printed also in Annalen der
Physik (4) 47:927ff. (1915).)
Mittelstaedt, P. (1978). Quantum Logic. Dordrecht: Reidel.
(1979). "Quantum Logic". In G. Toraldo di Francia 1979, pp. 264-99.
Mormann, T. (1996). "Categorical structuralism". In W. Balzer and C. U.
Moulines, eds., Structuralist Theory of Science: Focal Issues, New Results.
Berlin: de Gruyter, pp. 265-86.
Muller, F. A. (1997). "The Equivalence Myth of Quantum Mechanics". Studies
in the History and Philosophy of Modern Physics. 28:35-61, 219-47.
Nersessian, N. J. (1984). Faraday to Einstein: Constructing Meaning in Scien-
tific Theories. Dordrecht: Nijhoff.
482 References

Neumann, C. (1870). Uber die Principien der Galilei-Newton'schen Theorie.

Leipzig: Teubner.
Neurath, O. (1932/33). "Protokollsatze". Erkenntnis. 3:204-14. (Translated by
G. Schick as "Protocol sentences" in Ayer 1959, pp. 199-208.)
Newton, I. (1687). Philosophiae naturalis principia mathematica. Londini: Jussu
Societatis Regiae ac Typis Josephi Streater.
(1713). Philosophiae naturalis principia mathematica. Editio secunda
auctior et emendatior. Cantabrigias.
-(1726). Philosophiae naturalis principia mathematica. Editio tertia aucta
&C emendata. Londini: Apud Guil. & Joh. Innys, Regiae Societatis typographos.
-(1934). Mathematical Principles of Natural Philosophy and his System of
the World. Translated into English by A. Motte in 1729. Translations revised,
and supplied with a historical and explanatory appendix, by F. Cajori. Berke-
ley: University of California Press. 2 vols.
-(1972). Philosophiae naturalis principia mathematica, 3rd Ed. (1726)
with variant readings. Assembled and edited by A. Koyre and I. B. Cohen.
Cambridge, MA: Harvard University Press. 2 vols.
-(1974). Newton's Philosophy of Nature: Selections from his Writings.
Edited and arranged with notes by H. S. Thayer. New York: Hafner Press
(fourth printing; originally published by Hafner Publishing Co. in 1953).
-(Opticks). Opticks, or A Treatise on the Reflections, Refractions, Inflec-
tions & Colours of Light. Based on the Fourth Edition, London, 1730. New
York: Dover, 1979.
Noether, E. (1918). "Invariante Variationsprobleme". Gottinger Nachrichten,
pp. 235-57.
Omnes, R. (1994). The Interpretation of Quantum Mechanics. Princeton, NJ:
Princeton University Press.
O'Raifeartaigh, L., ed. (1972). General Relativity: Papers in honour of J. L.
Synge. Oxford: Clarendon Press.
The Oxford English Dictionary (OED). Second Edition on Compact Disc for
the Apple Macintosh®. Oxford: Oxford University Press, 1994.
Pais, A. (1982). 'Subtle is the Lord. . .': The Science and the Life of Albert
Einstein. Oxford: Clarendon Press.
(1986). Inward Bound: Of Matter and Forces in the Physical World.
Oxford: Clarendon Press.
-(1991). Niels Bohr's Times, in Physics, Philosophy, and Polity. Oxford:
Clarendon Press.
Pars, L. A. (1965). A Treatise on Analytical Dynamics. New York: Wiley.
Pascal, B. L'CEuvre de Pascal. Texte etabli et annote par J. Chevalier. Paris:
Gallimard.
Pasch, M. (1882). Vorlesungen uber neuere Geometrie. Leipzig: Teubner.
Pauli, W. (1925a). "Uber den Einfluss der Geschwindigkeitsabhangigkeit
der Elektronenmasse auf den Zeemaneffekt". Zeitschrift fiir Physik. 31:373-
85.
(1925b). "Uber den Zusammenhang des Abschlusses der Elektronen-
gruppen im Atom mit der Komplexstruktur der Spektren". Zeitschrift fiir
Physik. 31:765-83.
 References 483

(1926a). "Quantentheorie". In H. Geiger and K. Scheel, eds., Handbuch

der Physik. Berlin: Springer. Vol. 23, pp. 1-278.
-(1926b). "Uber das Wasserstoffspektrum vom Standpunkt der neuen
Quantenmechanik". Zeitschrift fiir Physik. 36:336-63 (English translation in
B. L. Van der Waerden 1967, pp. 387-415).
-(1940). "The connexion between spin and statistics". Physical Review.
58:716-22.
Peano, G. (1895-1908). Formulaire des Mathematiques. Torino: Bocca. 5 vols.
Peirce, C. S. (CP). Collected Papers. Edited by C. Hartshorne, P. Weiss, and
A. W. Burks. Cambridge, MA: The Belknap Press of Harvard University Press.
1931-1958. 8 vols.
(1892). "The doctrine of necessity examined". The Monist. 2:321-37
(reprinted in Peirce CP, 6.28-35).
(1982- ). Writings of Charles S. Peirce: A Chronological Edition. Max
H. Fisch, General Editor. Bloomington: Indiana University Press.
(RLT). Reasoning and the Logic of Things. The Cambridge Conferences:
Lectures of 1898. Edited by K. L. Ketner. Cambridge, MA: Harvard Univer-
sity Press, 1992.
Penzias, A. A. and R. W. Wilson (1965). "A measurement of excess antenna tem-
perature at 4080MHz". Astrophysical Journal. 142:419-21.
Perrin, J. (1913). Les Atomes. Paris: Felix Alcan.
Atoms. Translated by D. L. Hammick. Woodbridge, CT: Ox Bow Press.
Planck, M. (PAV). Physikalische Abhandlungen und Vortrdge. Braunschweig:
Vieweg, 1958. 3 vols.
(1900). "Zur Theorie des Gesetzes der Energieverteilung im Normal-
spectrum". In Planck PAV, vol. I, pp. 698-716 (originally published in Ver-
handlungen der Deutschen Physikalischen Gesellschaft, 2:237-45 (1900)).
-(1906). "Das Prinzip der Relativitat und die Grundgleichungen der
Mechanik". In Planck PAV, vol. II, pp. 115-20 (originally published in Ver-
handlungen der Deutschen Physikalischen Gesellschaft, 4:136-41 (1906)).
Platon. CEuvres Completes. Paris: Societe d'Edition Les Belles Lettres, 1920ff. 13
vols. (I quote Plato by this edition. As is customary, reference is made to the
pages of Henri Estienne's edition, Paris, 1578.)
Poincare, H. (1891). "Les geometries non-euclidiennes". Revue generale des sci-
ences pures et appliquees. 2:769-74 (reprinted with additions as Chapter III
of Poincare 1902).
(1898). "La mesure du temps". Revue de metaphysique et de morale.
6:1-13 (reprinted in Poincare, La valeur de la science, Paris: Flammarion,
1905).
(1902). La science et I'hypothese. Paris: Flammarion.
-(1905). "Sur la dynamique de l'electron". Comptes rendus de VAcademie
des Sciences. 140:1504-08.
-(1906). "Sur la dynamique de l'electron". Rendiconti del Circolo matem-
atico di Palermo. 21:129-75.
Pollard, H. (1976). Celestial Mechanics. The Mathematical Association of
America. Carus Mathematical Monographs, No. 18.
Polya, G. (1977). Mathematical Methods in Science. The Mathematical Associ-
484 References

ation of America (revised edition; the book originally appeared in 1963 as Vol.
11 of the Studies in Mathematics series sponsored by the School Mathemat-
ics Studies Group).
Price, H. (1996). Time's Arrow and Archimedes' Point: New Directions for the
Physics of Time. Oxford: Oxford University Press.
Putnam, H. (1957). "Three-valued logic". Philosophical Studies. 8:73-80
(reprinted in Putnam 1979, pp. 166-73).
(1962). "What theories are not". In Putnam 1979, pp. 215-27 (first pub-
lished in E. Nagel, P. Suppes, and A. Tarski, eds., Logic, Methodology and
Philosophy of Science, Stanford, CA: Stanford University Press, 1962).
-(1965). "A philosopher looks at quantum mechanics". In Putnam 1979,
pp. 130-58 (first published in R. G. Colodny 1965, pp. 75-101).
-(1967). "Time and physical geometry". Journal of Philosophy. 64:240-47
(reprinted in Putnam 1979, pp. 198-205).
-(1969). "The logic of quantum mechanics". In Putnam 1979, pp. 174-97.
(First published as "Is logic empirical?" in R. Cohen and M. Wartofsky, eds.,
Boston Studies in the Philosophy of Science, Dordrecht: D. Reidel, Vol. 5, pp.
216-41).
-(1970). "Is semantics possible?". In H. Putnam 1975, vol. 2, pp. 139-52
(first published in H. Kiefer and M. Munitz, eds. Language, Belief and Meta-
physics, Albany: State University of New York Press, 1970).
-(1973). "Explanation and reference". In H. Putnam 1975, vol. 2, pp.
196-214 (first published in Pearce and Maynard, Conceptual Change,
Dordrecht: D. Reidel, 1973, pp. 199-221).
(1974). "How to think quantum-logically". Synthese. 29:55-61.
-(1975a). Mind, Language and Reality. Philosophical Papers, Volume 2.
Cambridge: Cambridge University Press.
-(1975b). "The meaning of 'meaning'". In Putnam 1975, pp. 215-71 (first
published in Minnesota Studies in the Philosophy of Science, 7).
-(1979). Mathematics, Matter and Method. Philosophical Papers, Volume
1 (2nd Ed.). Cambridge: Cambridge University Press.
-(1981). "Quantum mechanics and the observer". Erkenntnis.
16:193-219 (reprinted in Putnam 1983, pp. 248-70).
-(1983). Realism and Reason. Philosophical Papers, Volume 3. Cam-
bridge: Cambridge University Press.
Ramsey, F. P. (1926). "Truth and probability". In F. P. Ramsey 1931, pp. 156-98.
(1931). The Foundations of Mathematics and other Logical Essays.
London: Routledge & Kegan Paul.
Redei, L. (1968). Foundation of Euclidean and Non-Euclidean Geometries
According to F. Klein. Oxford: Pergamon Press.
Redhead, M. (1987). Incompleteness, Nonlocality, and Realism: A Prole-
gomenon to the Philosophy of Quantum Mechanics. Oxford: Clarendon Press.
(1988). "A philosopher looks at quantum field theory". In H. R. Brown
and R. Harre 1988, pp. 9-23.
-(1993). "The conventionality of simultaneity". In J. Earman, A. I. Janis,
G. J. Massey, and N. Rescher 1993, pp. 103-28.
Reich, K. (1932). Die Vollstdndigkeit der Kantischen Urteilstafel. Berlin: Richard
Schoetz.
 References 485

Reichenbach, H. (1924). Axiomatik der relativistischen Raum-Zeit-Lehre.

Braunschweig: Vieweg.
(1928). Philosophie der Raum-Zeit-Lehre. Berlin: de Gruyter.
(1944). Philosophical Foundations of Quantum Mechanics. Berkeley:
University of California Press.
Riemann, B. (1867). "Uber die Hypothesen, welche der Geometrie zugrunde
liegen". Gottinger Abhandlungen. 13:133-52.
Rietdijk, C. W. (1966). "A rigorous proof of determinism derived from the
special theory of relativity". Philosophy of Science. 33:341-44.
(1976). "Special relativity and determinism". Philosophy of Science.
43:598-609.
Robertson, H. P. (1929). "On the foundations of relativistic cosmology".
National Academy of Sciences Proceedings. 15:822-29.
(1935/36). "Kinematics and world-structure". Astrophysical journal.
82:284-301; 83:187-201, 257-71.
Robertson, H. P. and T. W. Noonan (1969). Relativity and Cosmology. Philadel-
phia: Saunders.
Robinson, H. M. (1974). "Prime matter in Aristotle". Phronesis. 19:168-88.
Roque, X. (1997). "The manufacture of the positron". Studies in History and
Philosophy of Modern Physics. 28B-.73-129.
Rossi, B. and D. B. Hall (1941). "Variation of the rate of decay of mesotrons
with momentum". Physical Review. 59:223-28.
Ruhla, C. (1992). The Physics of Chance from Blaise Pascal to Niels Bohr. Trans-
lated from the French by G. Baron. Oxford: Oxford University Press.
Rumford, Benjamin Count (CW). Collected Works. Volume I, The Nature of
Heat. Edited by S. C. Brown. Cambridge, MA: The Belknap Press of Harvard
University Press, 1968.
(1798). "An experimental inquiry concerning the source of the heat which
is excited by friction". Royal Society of London Philosophical Transactions.
88:80-102 (reprinted in Rumford CW, I, 1-26).
Russell, B. (1897). An Essay on the Foundations of Geometry. Cambridge: Cam-
bridge University Press (unaltered reprint: New York, Dover, 1956).
(1917). Mysticism and Logic, and Other Essays. London: Allen &c Unwin.
(1959). My Philosophical Development. London: Allen &c Unwin.
Saccheri, G. (1733). Euclides ah omni naevo vindicatus, sive conatus geometri-
cus quo stabiliuntur prima ipsa universae geometriae principia. Mediolani: Ex
Typographia Pauli Antonii Montani (reprinted as Girolamo Saccheri's Euclides
vindicatus, with facing English translation by G. B. Halsted: New York,
Chelsea, 1986).
Schilpp, P. A., ed. (1949). Albert Einstein: Philosopher-Scientist. Evanston, IL:
Library of Living Philosophers.
Schneider, I., ed. (1988). Die Entwicklung der Wahrscheinlichkeitstheorie von
den Anfdngen bis 1933: Einfiihrungen und Texte. Darmstadt: Wis-
senschaftliche Buchgesellschaft.
Schrodinger, E. (WM). Collected Papers on Wave Mechanics. Together with Four
Lectures on Wave Mechanics. Translated by J. F. Shearer and W. M. Deans.
New York: Chelsea, 1982.
(1926a). "Quantisierung als Eigenwertproblem. (Erste Mitteilung)".
486 References

Annalen der Physik. (4) 79:361-76 (English translation in Schrodinger WM,

pp. 1-12).
-(1926b). "Quantisierung als Eigenwertproblem. (Zweite Mitteilung)".
Annalen der Physik. (4) 79:489-527 (English translation in Schrodinger WM,
pp. 13-40).
-(1926c). "Uber das Verhaltnis der Heisenberg-Born-Jordanschen Quan-
tenmechanik zu der meinen". Annalen der Physik. (4) 79:734-56 (English
translation in Schrodinger WM, pp. 45-61).
-(1926d). "Quantisierung als Eigenwertproblem. (Dritte Mitteilung:
Storungstheorie, mit Anwendung auf den Starkeffekt der Balmerlinien)".
Annalen der Physik. (4) 80:437-90 (English translation in Schrodinger WM,
pp. 62-101).
-(1926e). "Quantisierung als Eigenwertproblem. (Vierte Mitteilung)".
Annalen der Physik. (4) 81:109-39 (English translation in Schrodinger WM,
pp. 102-23).
-(1928). Four Lectures on Wave Mechanics. Glasgow: Blackie 6c Son
(reprinted in Schrodinger WM).
-(1935). "Die gegenwartige Situation in der Quantenmechanik". Natur-
wissenschaften. 23:807-12, 823-28, 844-49 (English translation in J. A.
Wheeler and W. H. Zurek 1983, pp. 152-67).
-(1995). The Interpretation of Quantum Mechanics. Dublin seminars
(1949-1955) and other unpublished essays. Edited and with Introduction by
M. Bitbol. Woodbridge, CN: Ox Bow Press.
Schwartz, E., ed. (ACO). Ada conciliorum oecumenicorum. Vol. 1.1.6. Berlin:
de Gruyter, 1928.
Schwartz, L. (1957/1959). Theorie des distributions. Paris: Hermann. 2 vols.
Schwarzschild, K. (1916a). "Uber das Gravitationsfeld eines Massenpunktes
nach der Einsteinsche Theorie". K. Preussische Akademie der Wissenschaften
Sitzungsberichte, pp. 189-96.
(1916b). "Uber das Gravitationsfeld einer Kugel aus inkompressibler
Fliissigkeit nach der Einsteinsche Theorie". K. Preussische Akademie der Wis-
senschaften Sitzungsberichte, pp. 424-34.
Schweber, S. S. (1994). QED and the Men Who Made It: Dyson, Feynman,
Schwinger, and Tomonaga. Princeton, NJ: Princeton University Press.
Schwinger, J. (1958). Selected Papers on Quantum Electrodynamics. New York:
Dover.
Seelig, C. (1957). Albert Einstein: A Documentary Biography. Translated by M.
Savill. London: Staples Press (English translation of Seelig, Albert Einstein:
eine dokumentarische Biographie, Zurich: Europa Verlag, 1954).
Shamos, M. H., ed. (1959). Great Experiments in Physics: First Hand Accounts
from Galileo to Einstein. New York: Holt, Rinehart and Winston.
Shapere, D. (1964). "The structure of scientific revolutions". Philosophical
Review. 73:382-94 (reprinted in Shapere 1984, pp. 37-48).
(1966). "Meaning and scientific change". In R. Colodny 1966, pp. 41-85
(reprinted in Shapere 1984, pp. 58-101).
-(1984). Reason and the Search for Knowledge: Investigations in the
Philosophy of Science. Dordrecht: D. Reidel.
 References 487

Shapiro, S. (1997). Philosophy of Mathematics: Structure and Ontology. New

York: Oxford University Press.
Shimony, A. (1963). "Role of the observer in quantum theory". In Shimony
1993, pp. 3-33 (originally published in American Journal of Physics,
31:755-73 (1963)).
(1990). "An exposition of Bell's Theorem". In A. I. Miller 1990, pp.
33-43 (reprinted, with some changes, in Shimony 1993, pp. 90-103).
-(1993). Search for a Naturalistic World View. Volume II, Natural Science
and Metaphysics. Cambridge: Cambridge University Press.
Sklar, L. (1993). Physics and Chance: Philosophical Issues in the Foundations of
Statistical Mechanics. Cambridge: Cambridge University Press.
Skyrms, B. (1987). "Dynamic coherence and probability kinematics". Philoso-
phy of Science. 54:1-20.
Sneed, J. D. (1971). The Logical Structure of Mathematical Physics. Dordrecht:
Reidel.
Solmsen, R (1958). "Aristotle and Prime Matter: A reply to Hugh R. King".
Journal for the History of Ideas. 19:243-52.
Sommerfeld, A. (1910a). "Zur Relativitatstheorie. I. Vierdimensionale Vektoral-
gebra.". Annalen der Physik. (4) 32:749-76.
(1910b). "Zur Relativitatstheorie. II. Vierdimensionales Vektoranalysis".
Annalen der Physik. (4) 33:649-89.
-(1915a). "Zur Theorie der Balmerschen Serie". Sitzungsberichte der
Bayerischen Akademie der Wissenschaften, pp. 425-58.
-(1915b). "Die Feinstruktur der Wasserstoff und wasserstoffahnlichen
Linien". Sitzungsberichte der Bayerischen Akademie der Wissenschaften, pp.
459-500.
-(1916). "Zur Quantentheorie der Spektrallinien". Annalen der Physik.
(4) 51:1-94, 125-67.
(1919). Atombau und Spektrallinien. Braunschweig: Vieweg.
-(1924). Atombau und Spektrallinien. Vierte umgearbeitete Auflage.
Braunschweig: Vieweg.
Spinoza. Opera. Im Auftrag der Heidelberger Akademie der Wissenschaften her-
ausgegeben von C. Gebhardt. Heidelberg: Carl Winters Universitatsbuch-
handlung, n.d. 4 vols.
Stachel, J. (1986). "Do quanta need a new logic?". In R. G. Colodny 1986, pp.
229-347.
Stachel, J. and R. Torretti (1982). "Einstein's first derivation of mass-energy
equivalence". American Journal of Physics. 50:760-63.
Stein, H. (1970). "On the notion of field in Newton, Maxwell, and beyond".
Minnesota Studies in the Philosophy of Science. 5:264-310 (pp. 287ff. contain
comments by Gerd Buchdahl and Mary Hesse, followed by Stein's reply).
(1991). "On Relativity Theory and the openness of the future". Philoso-
phy of Science. 58:147-67.
Stein, H. and A. Shimony (1971). "Limitations on measurement". In B. d'Es-
pagnat 1971, pp. 56-76.
Strawson, P. R (1959). Individuals: An Essay in Descriptive Metaphysics.
London: Methuen.
488 References

Sudbery, A. (1986). Quantum Mechanics and the Farticles of Nature: An Outline

for Mathematicians. Cambridge: Cambridge University Press.
Suppe, F. (1967). "The Meaning and Use of Models in Mathematics and the
Exact Sciences". Ph.D. dissertation. University of Michigan.
, ed. (1977). The Structure of Scientific Theories. Second Edition. Urbana,
IL: University of Illinois Press.
-(1989). The Semantic Conception of Theories and Scientific Realism.
Urbana, IL: University of Illinois Press.
Suppes, P. (1960). "A comparison of the meaning and uses of models in mathe-
matics and the empirical sciences". Synthese. 12:287-301.
(1962). "Models of data". In E. Nagel, P. Suppes, and A. Tarski, eds.,
Logic, Methodology and Philosophy of Science: Stanford, CA: Stanford
University Press, pp. 252-61.
-(1967). "What is a scientific theory". In S. Morgenbesser, ed., Philoso-
phy of Science Today. New York: Basic Books, pp. 55-67.
-(1969). "The structure of theories and the analysis of data". In F. Suppe
1977, pp. 266-307.
Teller, P. (1973). "Conditionalization and observation". Synthese. 26:218-
58.
(1988). "Three problems of renormalization". In H. R. Brown and R.
Harre 1988, pp. 73-89.
-(1995). An Interpretive Introduction to Quantum Field Theory. Prince-
ton, NJ: Princeton University Press.
Thomson, J. (1884). "On the Law of Inertia, the Principle of Chronometry and
the Principle of Absolute Clinural Rest, and of Absolute Rotation". Royal
Society of Edinburgh Proceedings. 12:568-78.
Thomson, W. (1874). "The kinetic theory of the dissipation of energy". Royal
Society of Edinburgh Proceedings. 8:325-34.
Toraldo di Francia, G., ed. (1979). Problems in the Foundations of Physics.
Proceedings of the International School of Physics "Enrico Fermi", Course
LXXII. Amsterdam: North-Holland.
Torretti, R. (1967). Manuel Kant: Estudio sobre los fundamentos de la filosofia
critica. Santiago de Chile: Ediciones de la Universidad de Chile.
(1978). Philosophy of Geometry from Riemann to Poincare. Dordrecht:
Reidel (corrected reprint: Dordrecht, Reidel, 1984).
-(1983). Relativity and Geometry. Oxford: Pergamon (corrected reprint:
New York, Dover, 1996).
-(1990). Creative Understanding: Philosophical Reflections on Physics.
Chicago: University of Chicago Press.
Truesdell, C. (1968). Essays in the History of Mechanics. New York: Springer.
Uhlenbeck, G. E. and S. Goudsmit (1925). "Ersetzung der Hypothese vom
unmechanischen Zwang durch eine Forderung beziiglich des inneren Verhal-
tens jedes einzelnen Elektrons". Naturwissenschaften. 13:953-54.
Van der Waerden, B. L., ed. (1967). Sources of Quantum Mechanics. Amster-
dam: North-Holland (unaltered reprint: New York, Dover, 1968).
van Fraassen, B. C. (1972). "A formal approach to the philosophy of science".
In R. G. Colodny 1972, pp. 303-66.
 References 489

Voigt, W. (1887). "Uber das Doppler'sche Prinzip". Gottinger Nachrichten, pp.

41-51.
von Arnim, J. (SVF). Stoicorum veterum fragmenta. Leipzig: Teubner,
1903-1905. 4 vols.
von Laue, M. (1911). "Zur Dynamik der Relativitatstheorie". Annalen der
Physik. (4) 35:524-42.
(1920). "Theoretisches uber neuere optische Beobachtungen zur Relativ-
itatstheorie". Physikalische Zeitschrift. 21:659-62.
(1955). Die Relativitatstheorie. 1. Band. Die spezielle Relativitatstheorie.
6. durchgesehene Auflage. Braunschweig: Vieweg.
von Mises, R. (1964). Mathematical Theory of Probability and Statistics. New
York: Academic Press,
von Neumann, J. (1927). "Mathematische Begriindung der Quantenmechanik".
Nachrichten der Gesellschaft der Wissenschaften zu Gottingen. pp. 1-57.
(1932). Mathematische Grundlagen der Quantenmechanik. Berlin:
Springer.
-(1955). Mathematical Foundations of Quantum Physics. Translated by
E. T. Beyer. Princeton, NJ: Princeton University Press (English translation of
von Neumann 1932).
Wald, R. M. (1984). General Relativity. Chicago: University of Chicago Press.
Walker. A. G. (1935). "On the formal comparison of Milne's kinematical system
with the systems of general relativity". Monthly Notices of the Royal Astro-
nomical Society. 95:263-69.
(1937). "On Milne's theory of world-structure". London Mathematical
Society Proceedings. 42:90-127.
Warwick, A. (1995). "The sturdy Protestants of science: Larmor, Trouton, and
the Earth's motion through the ether". In J. Z. Buchwald 1995, pp. 300-
43.
Waterston, J. J. (1846). "On the physics of media that are composed of free and
perfectly elastic molecules in a state of motion". Royal Society of London
Philosophical Transactions. 183A:5-79 (only an abstract appeared in the year
of submission: Royal Society of London Proceedings 5:604 (1846)).
Watts, I. (1725). Logick: or, the right use of reason in the enquiry after truth,
with a variety of rules to guard against error, in the affairs of religion and
human life, as well as in the sciences. London. (The 20th Ed. of this influen-
tial textbook was printed in Glasgow by William Smith for John Bryce and
Robert Farie in 1779.)
Weinberg, S. (1972). Gravitation and Cosmology: Principles and Applications
of the General Theory of Relativity. New York: Wiley.
(1995/96). The Quantum Theory of Fields. Cambridge: Cambridge
University Press. 2 vols.
Westfall, R. S. (1971). Force in Newton's Physics: The Science of Dynamics in
the Seventeenth Century. London: Macdonald.
(1980). Never at Rest: A Biography of Isaac Newton. Cambridge: Cam-
bridge University Press.
Wheeler, J. A. and W. H. Zurek, eds. (1983). Quantum Theory and Measure-
ment. Princeton, NJ: Princeton University Press.
490 References

Whewell, W. (1837). History of the Inductive Sciences, from the Earliest to the
Present Times. London: J. W. Parker. 3 vols.
(1840). The Philosophy of the Inductive Sciences, founded upon their
History. London: J. W. Parker.
-(1844). "On the fundamental antithesis of philosophy". Cambridge
Philosophical Society Proceedings. 8:170-81 (reprinted in Whewell 1847, II,
647-68, and in R. E. Butts 1968, pp. 54-75).
-(1847). The Philosophy of the Inductive Sciences, founded upon their
History. A new edition, with corrections and additions<*ajnd an appendix, con-
taining philosophical essays previously unpublished. London: J. W. Parker. 2
vols. (reprint: London, Frank Cass, 1967).
-(1858). Novum Organum Renovatum. Being the second part of the
Philosophy of the Inductive Sciences. 3rd Ed., with large additions. London:
John W. Parker & Son.
-(I860). On the Philosophy of Discovery. Chapters Historical and Criti-
cal, including the completion of the third edition of the Philosophy of the
Inductive Sciences. London: John W. Parker & Son (reprint: New York, Burt
Franklin, 1971).
Whitehead, A. N. and B. Russell (1910-1913). Principia Mathematica. Cam-
bridge: Cambridge University Press. 3 vols.
Whittaker, E. T. (1928). "Note on the law that light-rays are the null-geodesies
of a gravitational field". Cambridge Philosophical Society Proceedings.
24:32-34.
(1951/53). A History of the Theories of Aether and Electricity. London:
Thomas Nelson &c Sons (unaltered reprint: New York, Dover, 1989).
Wigner, E. P. (1939). "On unitary representations of the inhomogeneous Lorentz
group". Annals of Mathematics. 40:149-204.
(1952). "Die Messung quantenmechanischer Operatoren". Zeitschrift fur
Physik. 133:101-08.
-(1970). "On hidden variables and quantum mechanical probabilities".
American Journal of Physics. 38:1005-1009.
-(1983). "Interpretation of Quantum Mechanics". In J. A. Wheeler and
W. H. Zurek 1983, pp. 260-314 (lectures originally given in the Physics
Department of Princeton University during 1976; revised for publication in
1981).
Will, C. M. (1981). Theory and Experiment in Gravitational Physics. Cam-
bridge: Cambridge University Press.
Williams, C. J. W. (1982). Aristotle's De Generatione et Corruptione. Translated
with notes. Oxford: Clarendon Press.
Wilson, W. (1915). "The quantum theory of radiation and line spectra". Philo-
sophical Magazine. 29:795-802.
Wise, M. N. and C. Smith. "The practical imperative: Kelvin challenges the
Maxwellians". In R. Kargon and P. Achinstein 1987, pp. 323-48.
Wittgenstein, L. (1922). Tractatus Logico-Philosophicus. London: Routledge and
Kegan Paul.
(1953). Philosophische Untersuchungen. Edited by R. Rhees and G. E.
M. Anscombe. Oxford: Blackwell.
 References 491

Wolf, A. (1939). A History of Science, Technology, and Philosophy in the Eigh-

teenth Century. New York: Macmillan.
Wolter, A. B. (1963). "The Ockamist critique". In E. McMullin 1965, pp.
124-46.
Wroblewski, A. (1985). "De mora luminis: A spectacle in two acts with a
prologue and an epilogue". American Journal of Physics. 53:620-30.
Young, T. (1807). A Course of Lectures on Natural Philosophy and the Mechan-
ical Arts. London: J. Johnson.
Zawirski, Z. (1932). "Les logiques nouvelles et le champ de leur application".
Revue de metaphysique et de morale. 39:503-19 (Zawirski's proposal for a
three-valued quantum logic was first put forward in Polish, in a Poznan
journal, in 1931).
Zeeman, P. (1897). "On the influence of magnetism on the nature of the light
emitted by a substance". Philosophical Magazine. 43:226-39 (English trans-
lation of two articles published in Dutch in 1896).
Zermelo, E. (1896). "Uber einen Satz der Dynamik und die mechanische
Warmetheorie". Annalen der Physik. (3) 57:485-94.
 Index

a-particles, 309, 311 analytic statements, 98, 11 On

abduction, 228-30 analytical mechanics, 84-96, 176
aberration of starlight, 25In and/or, 43 8n
absolute space idle in Newtonian Anderson, Carl, 394
mechanics, 53-54 angle between two vectors, 445
accident, 54 angular momentum of earth, 5In
action: by contact, 75, 80; faster- Anscombe, G. E. M., 128n
than-light, 354 anthropic principle, 400-401
Aeschylus, 13 On anticipations of perception (Kant),
aether, 181n, 251-52; Aristotle's, 11, 120,369
16, 175n; Descartes's, 175n; antinomies (Kantian), 107, 109,
Lorentz's, 180, 252, 427; 140-45, 434
Maxwell's, 173-80; motion of antiperistasis, 11
earth in, 251; partial drag, 251, Antiphon, 405
260n antisymmetric, 133; tensor, 428
aim of physical theory: Berkeley, Apollonius of Perga, 12n, 27, 432
102, 103; Duhem, 243; Appell, Paul, 239n
Helmholtz, 184; Mach, 237, 238; approximation, 74, 231, 246
Maxwell, 199; Newton, 41, 77 Aquinas, Thomas, 8
aitema, 148 Arabatzis, Theodore, 309n
aitia, 13 On Arana, Juan, 80n
Alexander, H. G., 56n, 99n Archimedes, 7, 26
Alpher, Ralph A., 304n, 305 Aristotle, 1, 2, 3, 5, 6, 8-13, 14, 36,
Ampere, Andre-Marie, 79n, 82-83, 41, 79, 130n, 140n, 175n, 188n,
171, 178 310, 379, 416, 422, 431;
analogies of experience (Kant), 120- experience and experiment, 3;
38, 232n, 369 four simple bodies, 10; two-tiered
analogy: in mathematics and universe, 8
philosophy, 134; in physics, 173, arithmetic and intuition of time,
198 118-19

493
494 Index

Arnauld, Antoine, 100 366, 375

Aspect, Alan, 350, 352n Beltrametti, Enrico G., 383
astronomy, Greek, 12, 5In; Plato's Beltrami, Eugenio, 151
view, 14 Bentley, Richard, 77, 79
atlas, 158, 262; see also Lorentz Berkeley, George, 98, 101-104, 402
atlas, Poincare atlas Bernays, Paul, 410
atoms, 15n, 127n, 141, 310; in Bernoulli, Daniel, 195n
lattice, 454; planetary model, Bernoulli, Jakob, 200n
310-11; stability, 312 Besso, Michele, 294n, 297, 313, 315
Auyang, Sunny Y., 396n Bethe, Hans, 304n
axioms of geometry, 114, 117, 408- Birkhoff, Garrett, 337n, 379-83
10 Birkhoff, George David, 300n
Ayer, A. J., 403 BKS theory, 317
black holes, 419
p-particles, 309 black-body radiation, see thermal
Baade, W., 304 radiation
Bacon, Francis, 181 Blackett, Patrick Maynard Stuart,
Baldwin, F. J., 225, 228 394
Balmer, Johann Jakob, 312, 329 Blau, Steven K., 306
Balzer, Wolfgang, 414 Bohm, David, 128, 350-55, 373,
Barbour, Julian B., 24n, 6In 375-78
Barrow, John D., 400 Bohr, Niels, 310-13, 315-18, 321,
Bartels, Andreas, 188n, 299 323, 325, 329, 333, 349, 350,
Barthelborth, Thomas, 426n 355n, 363, 364, 368-73, 393,
base (of topology), 456 395
basis (of vector space), 445 Boltzmann, Ludwig, 205-14
Bauer, Edmond, 357n, 389 Bolyai, Janos, 149, 150n, 151, 378
Bayes, Thomas, 437, 439 Bondi, Hermann, 304
Bayes's theorem, 439 Boolean algebra, 454n-55n
Bayesianism, 228, 437-41 Boolean lattice, 380, 454-55; of
Becquerel, Henri, 309, 310 propositions, 380; of subsets, 381,
being said in various ways 455
(Aristotle), 8-9 Borel field, 457
Bekenstein, J. D., 419 Borel sets, 457
belief: degrees of, 438; methods of Borel, Emile, 457
fixing it, 224; has no place in Borelli, Giovanni Alfonso, 43n
science, 224n-225n Born, Max, 281, 308, 313, 318,
believer's physics, 247 321, 322n, 324, 325, 326n, 329,
Belinfante, F. J., 374n 333-36, 368, 373, 393
Bell inequality, 352, 353n, 355, Bose, Subas Chandra, 320-21
375n Boskovic, Roger Joseph, 80, 81
Bell, J. L., 385 Bothe, Walther, 318
Bell, J. S., 337n, 350, 352, 355, Bourbaki, Nicolas, 413
 Index 495

Boyle, Robert, 14-15, 17, 180 Castelli, Benedetto, 3

bra, 446 cat: Cheshire, 433; Schrodinger's,
Bradley, James, 25In 362n
Bradwardine, Thomas, 54n categories: Aristotelian, 9; Kantian,
Braginsky, Vladimir B., 67n, 363 111-12, 222n
Brahe, Tycho, 61, 433 Cauchy, Augustin-Louis, 450
Braithwaite, R. B., 244n Cauchy sequence, 450
British Thermal Unit, 182n causality principle (as understood by
Brown, Robert, 211 physicists), 128-29, 131, 132n,
Brownian motion, 205n, 211, 213n, 133-34
242n causation and determinism, 132,
Bruno, Giordano, 54n 231, 360n
Bub, Jeffrey, 356n, 357n causation and necessary connection,
Buchwald, Jed Z., 180n 130-31
bucket experiment, Newton's, 240- cause, causal relation, 71-73, 78-
41, 290 79, 102, 103, 128-34, 142, 144,
Buijs-Ballot, C. H. D., 196n 184, 185n, 231n-232n, 369, 370,
Bunge, Mario, 413n 371
Burbury, S. H., 207n Cavendish, Henry, 82n
Busch, Paul, 356n, 358n Cay ley, Arthur, 156
Butterfield, Jeremy, 353n central forces, 80-83, 184, 250
Butts, Robert, 217n Cepheid variable stars, 304
chance, 232, 315, 337-38, 401
Calderoni, Mario, 227 Chandrasekhar, Subrahmanyan, 62n
Callicles, 405, 437 change: Aristotle's classification, 9
caloric, 181, 187 characteristic states of apparatus,
Campbell, Lewis, 172n 357
Campbell, N. R., 245n Charlton, W., 12n
Cantor, Georg, 155n, 233 chart, 158, 262; see also Lorentz
Cao, Tian Yu, 318n, 396n chart
Carnap, Rudolf, 104, 402, 403-404, Chevalier, Jacques, 98
411-12, 421, 437 Chevalley, Claude, 413n
Carnot engine, 188-91, 192, 214- Chrysippus, 282n, 405
15 Cicero, 13, 282n
Carnot, Sadi, 187-94, 214-15 Ciufolini, Ignazio, 67n, 300n
Cartan, Elie, 429 Claggett, Marshall, 2n
Cartan, Henri, 413n Clarke, Samuel, 76n, 99
Cartesian coordinate system, 5; Clausius, Rudolf, 187, 188, 192,
right-handed, 253n 193, 194, 195-98
Cartesian product, 443 Clausius's postulate, 188, 189, 190
Caspar, Max, 432 Clifton, R. K., 353n
Cassinelli, Gianni, 383 closed set, 456
Cassini, Giovanni Domenico, 39n closure, 456
496 Index

Cohen, I. B., 61n convergence: strong, 450; weak,

Cohen-Tannoudji, Claude, 336n, 450n
346, 364 coordinate system, 5; see also
collapse of the wave function, 361, Cartesian coordinate system,
362 chart, generalized coordinates,
colligation of facts, 219 Lorentz chart
collineation, 156 coordinate transformation, 158-59,
collision laws, 30-33 263
common sense, 398-99, 404, 435 Copenhagen interpretation, see
compactness, 456 under QM
complementarity, 368-73 Copernicus, Nicolas, 242
complete inner product space, 450 Coriolis, Gaspard Gustave, 35n
complete orthonormal set of vectors, Cornford, Francis M., 54n
446 correspondence principle, 313, 318
component: relative to a basis, 445; cosmological constant, 75n, 300
relative to a chart, 294 cosmological principle, 302; see also
composition of directed quantities perfect cosmological principle
(Descartes), 19-20; of forces cosmology: criticized by Kant, 140-
(Newton), 20n, 49-50; of 44; relativistic, 299-306, 426;
orthogonal vectors (Galileo), 28n; steady-state, 304
of velocities (Descartes), 20 Costabel, Pierre, 34n
Compton, Arthur Holly, 316, 318, Cotes, Roger, 42n, 76, 77
369 Coulomb, Charles Augustin de, 82,
Compton effect, 316n 135n, 170, 178
Comte, Auguste, 79, 103, 216, 229, Coulomb's law of electrostatic force,
407 82, 169, 312n
conditionalization, 440-41 Courant, Richard, 62n
configuration space, 91, 332 covariance, see general covariance,
confirmation theory, 437 invariance
conservation: of Cartesian covariant, 161, 167n, 293n, 294n
momentum, 18; of classical Crescas, Hasdai ben Abraham, 54n
momentum, 36, 126-27, 316-17; cross-ratio, 156
of energy, 88, 127, 182, 183-86, curvature, 166-67; constant, 167
187, 316-17; of "force" (Leibniz), curve, 159
33-36; of matter, 122-28 Cushing, James, 430
consilience of inductions, 221-22
constraints, 85, 86 d'Alembert, Jean, 85, 87n
contextuality, 375n, 378 d'Alembert's principle, 87; for
continuum: Peirce's definition, 233 holonomic systems, 90
contravariant, 161, 167n, 293n, Dalibard, J., 352n
294n Damerow, Peter, 20n
conventionality in physics, 51, 256, Daneri, Adriana, 356n, 364
273, 274 Dante Alighieri, 8
 Index 497

Darrigol, Olivier, 308n, 316n, 317n, differentiable manifold, see

318 manifold, smooth
Darwin, Charles, 423, 435 differential equations, 59n, 75n,
Davidson, Donald, 423 131, 132, 133-34; and
Davisson, Clinton Joseph, 326 determinism, 93-94, 131-32
Davy, Humphry, 18In, 182n Dirac, Paul Adrien Maurice, 321,
de Broglie, Louis, 128, 325-26, 327, 330n, 336, 339, 388, 393, 394,
332, 335, 374n 395, 396, 430n, 452
de Finetti, Bruno, 438 Dirac notation for inner product of
de Gandt, Francois, 43n, 61n, 67n vectors, 330n, 388, 446-47
de Maria, M., 395n Dime's 5 function, 452
de Sitter, Willem, 301, 302n, 303, distance: symmetric function, 275
305 distributive laws, 454
de Voider, Burcher, 99 Doppler effect, 40, 214; see also
Debs, Talal A., 280n redshift
decoherence, 364-67 Drake, Stillman, 24n, 398
deduction from phenomena, 61, 78; du Fay, Charles, 81
see also Newton's rules of Dugas, Rene, 84n
philosophy Duhem, Pierre, 186n, 216, 242-48,
deduction, transcendental, 111 430
deductive inference, 226-27 Duhem-Quine thesis, 430
deferent, 12n Duns Scotus, 2
degeneracy, degree of, 447 dynamical system, 84-94;
Democritus, 15n, 310 conservative, 88, 206; holonomic,
dense (in a topological space), 451 88, 90
density operator, see statistical dynamis, Greek for 'force', 84
operator Dyson, Freeman, 395
density, 42n, 124
Descartes, Rene, 1, 2, 6, 8, 10, 14, Earman, John, 297n, 305n, 306n,
17-20, 21, 30, 31, 32, 33, 34, 35, 429n, 440n, 441
36, 42, 43, 46, 65n, 71, 76, 79, effective field theories, 395
81, 113, 126, 175n, 184, 274n, efficiency of heat engine, 188
400, 402, 406, 432 Ehlers, Jiirgen, 271
determinacy, thoroughgoing, 109, Ehrenfest, Paul, 21 On, 297
141, 434 Ehrenfest, Tatiana, 210n
determinant, 449 eidos, 9
determinism, 94, 132, 142-44, eigenvalue, see proper value
230-34, 334, 337-38, 401; eigenvector, see proper vector
kinematical, 280-83; logical, 280 Einstein, Albert, 35n, 37, 42, 73n,
DeWitt, Bryce S., 368, 392 74, 75n, 105, 137, 147, 167,
Dicke, R. H., 305 180, 204, 205, 211, 234, 241,
Dieudonne, Jean, 413n 242, 246n, 248, 249-306, 307,
differentiate function, 159 308, 310, 313-16, 318, 321,
498 Index

322n, 326, 349-55, 373, 385, Einstein's, 290, 292; strong and
417, 418, 424, 429, 435-36, weak, 290n
442n Erlangen Program, 152-55, 264
Einstein field equations, 297-98, error statistics (Mayo), 228n, 441n
426; modified in 1917, 300 Euclid, 7, 147-49, 406, 408, 429
Einstein static world model, 167, Euclid's fifth postulate, 148, 151
299-300 Euclidian geometry, 114-15, 116-
Einstein tensor, 298 117, 162, 164, 169n, 378;
Einstein time, 256, 273, 283n; difference with Lobachevskian
consistency conditions, 256-57 geometry, 149
electromagnetic field, 172-73, 175- Eudoxus, 14, 61n
79, 250, 427, 428-29; energy, Euler, Leonhard, 84, 105, 106, 168,
176, 184; momentum, 176; 239
tensor, 428 event (meaning 'spacetime point'),
electromagnetic waves, 174, 178, 261
179, 180 Everett (III), Hugh, 368, 387-93
electromotive force, 176 exclusion principle, 319, 321
electron, 309; Greek for 'amber', exp(x), 196n
81 expectation value, 344
electrotonic state, 177 experiment, 3, 17, 144, 245-46,
element of reality, 349, 350 371, 401, 425n
elliptic geometry, 156 explanation, scientific, 102, 242-43,
Ellis, George F. R., 56n, 143, 401 244n
Ellis, R. L., 200 explication of conceptions, 219
Empedocles, 36
energetics, 186n, 242n facts and concepts, 431, 434, 437
energy, 35n, 175n, 182, 183-86, Falkenburg, Brigitte, 328n, 355n,
188n, 286; kinetic, 186; of 395n, 417n, 430n
electromagnetic field, 176, 184; false assumptions: use of in physics,
potential, 186; psychic, 186n; see 26
also conservation of energy Fano, U., 347
energy-momentum tensor, 295 Faraday, Michael, 83, 171, 172,
ensemble, 204, 347-48 173, 177, 178, 180n, 309
entailment, 379n Fermat, Pierre, 20
entropy, 188, 191-95, 207n, 208, Fermi, Enrico, 187, 188, 321
213 Fevrier, Paulette, 379
epicycle, 12n Feyerabend, Paul K., 220, 229,
EPR paradox, 308, 349-55, 356, 379n, 404, 421
373, 378; fallacy in EPR Feynman, Richard P., 177n, 178n,
argument, 353 179n, 428
equations of motion: Hamilton, 93, field (algebraic), 168n, 444
325, 326, 338; Lagrange, 90, 91, field (physical), 168-80; effective,
94, 95; Newton, 84 395; electric, 170;
equivalence principle: Newton's, 60; electromagnetic, 175-80;
 Index 499

gravitational, 170, 441; term 16, 20-30, 34, 36-37, 43n, 59,
introduced by Faraday, 171n; for 62, 100, 204, 230, 398, 406, 420,
mathematical concept, see field 426, 431, 432, 436; experiments,
(algebraic), scalar field, tensor 3, 24, 37; medieval antecedents,
field, tetrad field, vector field 2n; see also Galileian relativity,
Finkelstein, David, 386n inertia principle in Galileo's sense
Fischer, Ernst, 331 Galileian relativity, 53n; see also
FitzGerald, George Francis, 179n, relativity principle (Newton)
252, 424 Galileian transformation, 254, 260
Fizeau, Armand Hippolyte Louis, Gamow, George, 304
175 Garnett, William, 172n
FLRW (or FRW) worlds, 303, 306 Gauss, Carl Friedrich, 149, 150,
fluid, 168-69, 173 151, 164, 166
force 42-50, 103-104, 184-85; Gaussian curvature, 166-67
meaning 'energy', 33-36, 43, Gay-Lussac, Louis-Joseph, 184
183n; relative, 287; see also Geiger, Hans, 310, 318, 362n
central forces, forces of constraint, Gell-Mann, Murray, 364, 365
impressed forces, lines of force, general covariance, 91, 296-97
vis generalized coordinates, 89
forces of constraint, 85, 86-87 generalized force, 89
Fourier, Joseph, 79n, 207, 322, 323, generalized momentum, 92
393 Gentzen, Gerhard, 384
Fox, Robert, 18In geodesic, 166, 270; incomplete,
frame of nature, 41 305n; inextensible, 305n
Franck, James, 336 geodesic law of motion, 292, 295
Frank, Philipp, 132n, 372n geometric object, 169, 428
free-fall, 21, 62, 73n geometry: analytical, 4-5; based on
free particle, 50, 51, 273 intuition, 114, 116-18; equal
free will, 20, 143-44, 401 rights for each, 157; more than
freedom, degrees of, 89 one in physics, 150-51; most
Frege, Gottlob, 98, 119, 227, 408, general problem according to
409, 410n, 411, 422 Klein, 155; paradigm of secure
Fresnel, Augustin Jean, 175n, 251 knowledge, 113; science of space,
Freud, Sigmund, 186n 113; see also axioms of geometry,
Friedman, Michael, 56n, 116n, 429n elliptic geometry, Euclidian
Friedmann, Alexander, 292n, 301, geometry, Lobachevskian
302, 303 geometry, non-Euclidian geometry,
Friedmann world models, 301-303 projective geometry
Friedrichs, K., 429 Germer, Lester Halbert, 326
Geroch, Robert, 305, 419
y-rays, 309 Gibbs, J. Willard, 204, 205, 207n
galaxy recession, 301, 303 Gilson, Etienne, 2n
Galenus, 405n Gleason, A. M., 375
Galilei, Galileo, 1, 2, 3, 7, 8, 14, 15, God (Judaeo-Christian), 13, 17, 18,
500 Index

19, 54, 55n, 71, 72, 74, 75, 76, H-function, 207
77, 97, 98, 99, 102, 105n, 106, H-theorem, Boltzmann's, 208-11,
109n, 110, 113, 126, 138, 139, 212
141n, 142, 399, 400, 406, 432, Haack, Susan, 379n
433, 434, 441; always doing Habicht, Conrad, 298n
geometry, 432; "vindicated" by Hacking, Ian, 198n
Newton and Rousseau, 105n Hafele, J. C , 277n
gods (Aristotle's), 9, 14 Hall, A. Rupert, 52, 54, 55, 297n
Gold, Thomas, 304 Hall, D. B., 277n
Goldstein, Herbert, 87n Hall, Mary Boas, 52, 54, 55
Gooding, David, 171 Hallett, Michael, 385
Goodman, Nelson, 402 Hamilton, William Rowan, 93, 118,
Goudsmit, Samuel Abraham, 320 326, 338, 376
Gouy, Louis-Georges, 21 In Hamilton-Jacobi equation, 93, 95,
GR (general relativity), 241, 249, 327, 376
425n, 429-30, 442n; relation with Hamiltonian function, 92, 325,
SR, 249n, 292 328n; matrix, 330, 331; operator,
Graham, Neill, 392 328n, 331, 340, 365
Grangier, P., 352n Hannabuss, Keith, 336n
gravitation law, Newton's, 58, 59, Hanson, Norwood Russell, 404, 421
67, 74, 135, 137, 222, 249, 260, Harman, P. M., 172n, 175n
289, 298, 417, 418n, 419, 429, Hartle, James B., 364, 365
442n; proposed modifications of, Hartsoeker, Nikolaus 76
74n Havas, Peter, 429
gravitational "constant" variable, Hawking, Stephen W., 143, 305,
74, 75n 401,419
gravity, 49, 57-67, 75-79, 289-99, heat, 180-215; latent, 186; manifest,
429; see also Einstein field 186
equations, free-fall, geodesic law heat death of universe, 195
of motion, gravitation law, heaviness and lightness, 10
heaviness and lightness Heaviside, Oliver, 177
Greenberger, D. M., 352, 353n Heisenberg, Werner, 132n, 308,
Grossmann, Marcel, 75n, 291, 292, 316n, 318, 320, 321-25, 329,
294, 296, 298n 332, 333, 336, 340, 348, 355,
groundless ground, 142 368, 394, 395, 436
group, 152-55, 443-44; abelian, Heisenberg picture, 340
444; additive, 444; multiplicative, Heisenberg's indeterminacy relations,
444 348-49, 369, 436
group velocity, 326 Helmholtz, Hermann von, 83, 162,
Guth, Alan H., 306 178, 179, 182, 183-86, 216, 250
Hempel, Carl G., 244n
h, see Planck's constant h Hepp, K., 366n
t> (for h/2n), 320 Herapath, John, 195n
 Index 501

Herivel, John, 48n, 62n hypothesis: Newton's sense, 6In,

Herman, Robert C , 305 78
Herodotus, 130n
Herschel, John, 407, 412 Ideas of Reason, 138-45, 434
Hertz, Heinrich, 83n, 88n, 179, idempotent, 343
259n, 284 identity mapping, 55n
Hertz, Marcus, 110 identity of indiscernibles, 98-101
Hicks, W. M., 179n Ignatowsky, W. von, 258
hidden variables, 374 impetus, 20-21
Hilbert, David, 91, 308, 331, 336, implication, material, 379n
338, 408-10, 412n, 413 impulse, 47
Hilbert space, 91, 308, 331, 336, incommensurability (Kuhnian), 288,
338-39, 413, 450-52; difficulties 404, 421-30, 426n, 429
of infinite-dimensional case, incongruent counterparts, 105-106
451-52; see also tensor product of independent events, 200n
Hilbert spaces indeterminacy of translation, 244-
Hiley, B. J., 378n 45
Hintikka, Jaakko, 114n indiscernibles, see identity of
Hipparchus, 12n indiscernibles
Hippocrates, 188n individuation of spatial points by
Him, G. A., 182n spatial structure, 55, 282
Hoffmann, Banesh, 296 induction (inference), 61, 73, 74, 78,
hole argument, 296-97 219-21, 227-28, 231, 247;
holonomic, 88n equated by Peirce with statistical
homeomorphic, 457 inference, 227
Homer, 13 On induction, electromagnetic, 172
Hooke, Robert, 39n inertia, 43-45; Mach's view, 240-
Hopf, L., 132n 41, 290, 291, 299-300, 301, 385;
horizon: event, 367n; particle, 367n see also inertia principle, inertial
Home, Michael A., 352, 353n frame, inertial motion, inertial
Hoyle, Fred, 304 time scale, vis inertiae
Hubble, Edmund, 303 inertia principle, 18-19, 30, 31, 45-
Hughes, R. I. G., 336n, 374n 46, 51, 53-54, 239, 240-41; in
Hume, David, 71, 120, 130, 131, Galileo's sense, 21, 25-26, 53n
132,135 inertial frame, 53, 54, 250, 253-55,
Hunt, Bruce J., 179n 258; time coordinate for, 255-56
Huygens, Christiaan, 20, 30-33, inertial motion: effectively
39n, 40, 43n, 61n, 76, 77, 175n, impossible in Newtonian world,
406 46; free-fall reclassified as, 73n;
hyle, 9; translated materia by see also inertia principle
Cicero, 13; see also prime matter inertial time scale, 51, 255, 276
hyperbolic geometry, see Infeld, Leopold, 295, 296
Lobachevskian geometry infimum, 453
502 Index

inner product, 445 60, 61-62, 63, 76, 222, 406, 433
inquiry, 224 Kepler, Johannes, 60, 61, 62, 63,
intensive magnitudes, 119-20 406, 418n, 432, 433, 437
interaction (Kantian category), 134— ket, 446
38 Keynes, John Maynard, 437, 441
interferometer, 52, 251, 424 Khalili, Farid Y., 363
intuition, 114, 116-18; axioms of, khora, 13n
117; form and formal, 116; kinesis, 9, 10, 12
intellectual, 139 kinetic theory of gases, 195-205,
invariance, 154 206, 208
inverse, 153, 154, 444 King, Hugh R., 12n
island universes, 104, 303 Kirchhoff, Gustav Robert, 307n,
isomorphism, 55n 313
Klein, Felix, 152-57, 261, 264, 435
Jacobi, Carl Gustav Jacob, 93, 95, Klein, Oskar, 394
327, 376 Knutzen, Martin, 104
Jammer, Max, 239n, 308n, 316n, Kochen, S., 375, 384n
372n, 379n Kohlrausch, Rudolph, 174
Jaskowski, Stanislaw, 384 Kolmogorov, Andrei Nikolaevich,
Jauch, Josef M., 383, 384n 457
Jeffrey, Richard, 440n Koyre, Alexandre, 2n, 42n, 70n,
Jesus Christ, 406 76n
join, 453 Kragh, Helge, 304n
Jordan, Pascual, 321, 322, 324, 329, Kramers, Hendrik, 316-18, 321
336, 393, 394n Kronig, A. K., 195
Jordanus Nemorarius, 2n Kuhn, Thomas S., 35n, 183, 288,
Joule, James Prescott, 182-83, 187 289n, 393n, 394, 404, 417, 421,
jumps, 311-12, 327, 332, 333, 341, 422, 423, 426n
342, 356n, 360-62 Kuratowski, K., 443n

Kant, Immanuel, 42, 57, 81, 97- Lagrange, Joseph-Louis, 50, 84-93,
146, 185, 216, 217n, 222, 223, 176, 178, 332
227, 230, 281, 301, 307n, 368- Lagrangian, 91, 206n
69, 370, 420-21, 423, 433-35, Lahti, Pekka J., 356n, 358n
441-42 Lambert, Johann Heinrich, 114n,
Keating, R. E., 277n 151
kelvin (degree of temperature), 191 Landau, Lev, 127n
Kelvin, William Thomson, 1st Lange, Ludwig, 53, 56, 273
Baron, 171n, 175n, 178, 179n, Laplace, Pierre-Simon de, 94, 169,
185, 187, 188, 208 429
Kelvin's postulate, 188, 189, 193, large numbers: strong law, 200;
194 weak law, 200n
Kepler's "laws" of planetary motion, Larmor, Joseph, 179n
 Index 503

lattice, 379, 453-54; atomic, 454; light-cone, see null-cone

complemented, 454; modular, Linde, Andrei, 306
455; of subspaces of Hilbert Lindsay, Robert Bruce, 87n
space, 381-82; linear: combination, 444; function,
orthocomplemented, 454; 446; independence, 444-45;
orthomodular, 455; see also submanifold, 451; subspace, 451
Boolean lattice linear operators, 447-50; adjoint,
laws of nature, 77, 102, 129, 143, 451; bounded, 451; norm of, 451;
144, 243, 246, 405-407, 412; in self-adjoint, 451; unitary, 451
Greek Church fathers, 405-406; lines of force, 171, 172, 173, 174
prescribed by human Lobachevsky, N. I., 149-52, 378
understanding according to Kant, Lobachevskian geometry, 147,
116-17: see also collision laws, 149-52; relative consistency,
Coulomb's law of electrostatic 151
force, geodesic law of motion, Locke, John, 75-76, 77, 113, 123n
gravitation law, Kepler's "laws" of logic: dialogical, 386n; imaginary,
planetary motion, Newton's Laws 378; inductive, 437; nonstandard,
of Motion, Planck's law of 378-79; quantum, 378-86
thermal radiation, Snell's law logical empiricism, 402, 404, 410-
Lebesgue, Henri, 457 12
Leibniz, Gottfried Wilhelm, 20, 21, logical positivism, see positivism
33-36, 43, 44, 47, 56, 76, 98- Loinger, A., 356n, 364
101, 104, 106, 126, 141n, 170, London, Fritz, 336, 357n, 389
184 Lorentz, Hendrik Antoon, 180, 252,
Lemaitre, Georges Henri, 301, 303 253, 257, 259n, 308, 319, 424,
length of lines, 163; in Minkowski 427, 436
spacetime, 269-70 Lorentz atlas, 263; boost, 257,
length of moving rod, 271-73 259n, 264; chart, 262; force,
length of vector, 445 179n, 284, 287n, 427; group,
Leonardo da Vinci, 242 252n, 257n, 264; transformation,
Levi-Civita, Tullio, 270, 294 252, 257n, 260
Lifschitz, E. M., 127n Lorentz-FitzGerald contraction,
light: deflection by gravity, 290; 252, 424
instantaneous propagation, 36, Lorentz invariance of the laws of
37, 39n; speed, 36-40; same as physics, 259, 283n, 289, 308,
speed of electromagnetic waves, 393, 394
174-75, 415n; numerical value set Lorenz, Kuno, 386n
by international convention, 274; Lorenzen, Paul, 386n
two-way speed, 274; see also Loschmidt, Joseph, 208, 210, 212
aberration of starlight, light Lovelock, David, 127n
principle Lucian, 123n
light principle, 253, 256, 273, 274 Lucretius, 54n
lightlike, see null Liiders, Gerhart, 36In
504 Index

Liiders rule, 361n 329-32, 336

Ludwig, Giinther, 364 matter (modern concept), 13-20, 78;
Lukasiewicz, Jan, 379 a fiction, 100, 102; an
abstraction, 184; a well-founded
Mach, Ernst, 49, 186n, 216, 234- phenomenon, 102n; the moveable
42, 290, 291, 299, 300, 301, 385, that fills space, 126
402 matter and form: Aristotle, 9, 12;
Mach number, 234n Kant, 106-107
Mackey, George W., 382 matter of phenomena (Kant), 126
Maier, Anneliese, 242n Maupertuis, Pierre Louis Moreau
Mainzer, Klaus, 328n de, 80n
Malebranche, Nicolas, 432 Maxwell, James Clerk, 5, 83, 170,
manifold, 155n; Riemannian, 165, 171-79, 180, 182n, 185, 186,
429; of constant curvature, 167; 187n, 197-205, 206, 249, 250,
smooth, 158; topological, 457 259, 260n, 295, 415n, 427, 428,
manifold topology of a smooth 429
manifold, 159n Maxwell-Boltzmann distribution of
mapping, 443; bijective, 443; velocities, 198, 201-203, 207-
continuous, 456; injective, 443; 208, 209, 210
surjective, 443 Maxwell equations, 177, 177n-
Margenau, Henry, 87n 179n, 249, 250, 259, 260n, 284,
Mariotte, Edme, 184 287n, 427, 428, 429, 430n
Marsden, Ernest, 311 Mayo, Deborah G., 205n, 228n,
mass, 42, 43, 49, 124-26, 239-40, 425n-426n, 441n
283-89; additivity, 288; inertial mean free path, 196
and gravitational, 66-67, 290n; measurement problem of QM, 308,
longitudinal, 285-86, 289; 355-67, 373
measure of energy content, 286; mechanaomai, 34n
proper, 288, 289; relativistic, 287, mechanical equivalent of heat, 182-
289; relativity of, 259; rest, 288; 83
transversal, 285-86, 289 mechanical models, 173-75, 179
mathematics: acid test of genuine mechanics, see analytical mechanics,
science, 97; language of the book matrix mechanics, QM (quantum
of nature, 15, 432; science of mechanics), wave mechanics
quantity, 4; science of structures, meet, 453
4,413 Mehra, Jagdish, 308n
matrix, 324, 448; determinant, 449; mentalism, 101-102
diagonal, 448; differentiation, Mercury's perihelion advance, 74n,
324; multiplication, 448-49; 246n, 298n, 299, 416, 417-19,
noncommutativity, 324-25; trace, 433, 442
449; unit, 448 Mersenne, Marin, 20
matrix mechanics, 308, 321-25; Merton College, 2n
equivalence with wave mechanics, Messiah, Albert, 336n
 Index 505

metaphysica specialise 110, 113, coexisting in one body, 28;

138, 140 uniform, 7, 21, 51; uniformly
meter, 262, 274 accelerated, 21-26; see also
metric, Minkowski, 267, 292, 299, equations of motion, kinesis,
303, 429; Riemannian, 165, 267; perpetual motion, phora, quantity
semi-Riemannian, 269, 426, 429 of motion,
Michell, John, 82n Motte, Andrew, 48n
Michelson, Albert Abraham, 52n, Moulines, C. Ulises, 402, 414, 426n
251, 252, 312, 424 Muller, F. A., 329n
microwave background radiation, multilinear, see ^-linear
see thermal radiation, cosmic Murphey, M. G., 223n
background
Migne, J. P., 406n ^-linear function, 447
Mill, John Stuart, 216, 232n, 407 ?z-manifold, 158
Miller, A. I., 150n, 394 w-tuple, ordered, 443n
Milne, E. A., 304 nabla operator, 95
mind and body, 20, 234, 400 nature: Aristotelian definition, 9;
Minkowski, Hermann, 109n, 257, book written in mathematical
260-71, 292n, 295, 299, 412, language, 15, 432; economy of,
419, 427, 428, 429 72, 406; frame of, 41; origin of
Mittelstaedt, Peter, 356n, 358n idea, 8
mixture of quantum states, 345, 346 necessity, 121-22, 129-31, 134
Mobius strip, 158 neighborhood, 456
model, 415, 416-17; two senses of Nernst, Walther, 194, 316n
word, 415n Nersessian, Nancy, 18 On
momentum: classical, 18, 35-36, 42; Neumann, Carl, 51, 53
Cartesian, 18; relativistic, 287; Neurath, Otto, 402-403, 404, 421
term meaning 'increment', 21; see neutral element, 153, 443
also conservation of momentum, Newton, Isaac, 1, 14, 20n, 21, 39n,
generalized momentum, quantity 41-79, 98, 100, 105, 106, 113,
of motion 162, 170, 222, 234, 239-41,
monad, 81, lOln, 104 246n, 249, 250, 282n, 283, 288,
Moore, Henry, 158 297, 322, 373, 377, 378, 385,
Morley, Edward W., 251, 252, 424 406, 418, 429, 433, 436, 442n
Mormann, Thomas, 414 Newton's Laws of Motion, 33, 45,
morphe, 9 87, 175, 180, 259-60, 283; First,
Mosterin, Jesus, 306n 45-46, 51, 53-54, 239, 240-41,
motion: absolute and relative 273; Second, 47-48, 284; Third,
distinguished by Newton, 53; 48-49, 184
Aristotle's concept, 9, 19; Newton's rules of philosophy, 61,
Descartes's concept, 17, 30-31; 69-74
forced, 11, 16; natural, 10, 16; of Neyman, Jerzy, 228
missiles, 11, 20-21, 27-30; two Nicol, William, 347n
506 Index

Noether, Emmy, 127 oscillator: anharmonic, 324; virtual,

Noether's theorem, 127 317-18
non-Euclidean geometry, 149-52, Ostwald, Wilhelm, 186n
156-57; modelled on projective
plane, 156; modelled on sphere Pais, Abraham, 289n, 309n, 333,
with imaginary radius, 114n, 151 368
nondegenerate: Hilbert space Panov, V. I., 67n
operator, 339, 447; tensor, 165 parabolic geometry, see Euclidian
nonlocality, 355n, 375n, 378 geometry
Noonan, T. W., 53n parallel, 150
norm: in vector space, 164; of linear parallel transport, 147, 270
operator, 451 parallelizable, 268
normalized vector, 338 parallellism, angle of, 150
Norton, John, 297n parallellogram rule, see composition
null: curve, 269; interval, 266; of directed quantities
vector, 269 parameter (of curve), 159
null-cone, 266; future, 266; inside, parametric curve (of coordinate
267; outside, 267; past, 266 function), 160n
Parmenides, 15n
objective, 398n, 401-402 Pars, L. A., 84n, 87n
observables, 322, 338, 356n, 357, particle, 50; see also free particle
358n, 378; collective, 365; Pascal, Blaise, 97, 204
determined by theory, 322n; Pasch, Moritz, 114n
represented in QM by self-adjoint path, 159
operators, 338; see also proper Paton, Herbert J., 11 In
state of an observable Pauli, Wolfgang, 318, 319-20, 321,
Occhialini, Giuseppe Paolo 325, 394, 395
Stanislao, 394 Peano, Giuseppe, 411
Ockam, William, 2, 12, 13 Pearson, Egon, 228, 441n
Oersted, Hans Christian, 82, 83, Peirce, Charles Sanders, 216, 222-
171, 172 34, 441n
Omnes, Roland, 364, 365-67 Penrose, Roger, 305, 419
open covering, 456 Penzias, Arno A., 305
open interval, 457 perfect cosmological principle, 304n
open rectangle, 158n permutation, 153
open set, 455 perpetual motion, physical and
operational, 255 mechanical, 34n
Oppenheim, P., 244n Perrin, Jean, 205n, 211, 242n
ordered pair, 443n phase space, 91, 204, 338
ordering, partial, 453 phase velocity, 326
Oresme, Nicolas, 54n phenomena: Newton's sense 6In,
orthogonal, 446 71; not fully determinate, 141-42,
orthonormal, 446 434; ordered by human mind
 Index 507

(Kant), 107-13, 115-17; well- 407-408, 410-12

founded, 100-101 potential, 88; electrostatic, 171;
Philoponus, John, l l n , 20 gravitational, 171; quantum, 376
philosophy, Newton's rules, see potential being, 9
Newton's rules of philosophy power set, 443
phora, 9, 10 Poynting vector, 295
photon, 254, 307-308, 314, 315, PPN (parametrized post-Newtonian)
316, 320 formalism, 425n, 426n
physical system, 337n; see also pragmatism, 223, 225
dynamical system predicates, see terms
physicalistic language, 402-403, Price, Hew, 214n
411,421 Priestley, Joseph, 82n
Pindar, 13 On prime matter, 12
Pirani, F. A. E., 271 principle, see anthropic principle,
Piron, Constantin, 384n causality principle,
Planck, Max, 210, 286, 305, 307, correspondence principle,
310, 311, 313, 314, 315, 316, 320 cosmological principle,
Planck's constant /?, 307, 311, 313, d'Alembert's principle, equivalence
369 principle, exclusion principle,
Planck's law of thermal radiation, inertia principle, light principle,
305, 307, 313-15, 320 perfect cosmological principle,
Plato, 3, 11, 13, 113,242,405, reciprocity principle, relativity
406, 414, 432n principle, Ritz combination
Playfair's axiom, 148n principle, thermodynamics -
Plutarch of Chaeroneia, 432n first principle, thermodynamics
Podolsky, Boris, 308, 349-55 - second principle
Poe, Edgar Allan, 303 probability, 180, 195, 196,
Poincare, Henri, 151, 162n, 210, 197-202, 203-204, 337n, 341,
216, 252, 255n, 257n, 259n, 383, 437-41; conditional, 439;
430n introduced into quantum physics
Poincare atlas, 263n; group, 257n, by Einstein, 314-15, 349;
263n, 430n; transformation, 257n measured by squared \|/-function,
point transformation, 263-64 333, 336n
Poisson, Denis, 169, 293, 419, 429, product, 153, 443; see also
441 Cartesian product, inner product,
Pollard, Harry, 62n tensor product
Poly a, George, 28 projection mapping, 161
Poncelet, Jean-Victor, 152 projective geometry, 152, 156
Porter, Cole, 229 projector, 342-43
poset, 454; orthocomplemented, 454 proper state of an observable, 357n,
positive definite, 339 358n
positivism, 79, 98, 101, 103-104, proper time, 270
216, 221, 238, 402-404, proper value, 325n, 447
508 Index

proper vector, 339, 358n, 447 quantum theory: old, 309-13, 319,
Prosperi, G. M., 356n, 364 321, 355; relativistic, 393-97, 419
psi (\|/) function, 327-28, 332; Quetelet, Lambert Adolphe Jacques,
collapse of, 361, 362; defined in 180
configuration space, 332; of the Quine, Willard van Orman, 430
universe, 337, 387; probabilistic
interpretation, 333-36, 373 radioactive decay, 309, 314, 333
Ptolemy, 12n radioactivity, 309
Putnam, Hilary, 281, 288, 379n, Ramsey, Frank Plumpton, 438
383-86, 407, 422 random sequence, 200n
Pythagoras's theorem, 108n, 109n, Rankine, William John, 35n, 185,
118; for arbitrarily many 188n
dimensions, 450 ratios, universal calculus of, 6-7
Rayleigh-Jeans "law", 205
QED (quantum electrodynamics), reason as guide, 138, 139
394-96 Rechenberg, Helmut, 308n
QM (quantum mechanics), 128, reciprocity principle, 258
131, 308, 396, 419; background, Redhead, Michael, 273n, 280n,
308-21; Copenhagen 336n, 353n, 357n, 375n, 396,
interpretation, 368-73, 392; 452n
genesis, 321-36; groundlines, redshift, cosmological, 249;
336-48; HV-extensions, 374, 375; gravitational, 29In
incomplete? 349-55; reference without sense, 288, 422
indeterminacy relations, 348-49, reflection, 56
369; many-worlds interpretation, Reich, Klaus, 11 In
387-93; paradoxes, 349-67; Reichenbach, Hans, 273, 274, 275-
philosophies, 367-93 77, 379, 421
quadrivium, 4 Reichenbach time, 275-76;
qualities: occult, 77, 78; primary modified, 276-77
and secondary, 15-16, 76, 398 relativistic quantum theories, 393-
quantifier, 227n 97
quantity of matter, 18, 42, 124, 288; relativity of motion: Descartes, 31;
see also mass Kant, 146
quantity of motion, 18, 34, 35, 42 relativity principle (Einstein), 253,
quantum, 254n, 307-308; 256, 259, 287n; (Newton), 53,
electrodynamics, see QED; 250, 259
mechanics, see QM relativity theory, 42, 249-306, 308;
quantum condition, 312; sharpened, alleged conflict with quantum
325 nonlocality, 355; see also GR
quantum evolution, two laws? 341- (general relativity), SR (special
42, 360, 387 relativity)
quantum logic, 378-86 Remond, Nicolas, 10In
quantum numbers, 312 renormalization, 395
 Index 509

reparametrization, 160 Schneider, Ivo, 195

Resnik, Michael, 4n Schrodinger, Ernst, 308, 325,
revolutions, scientific, 422, 423 327-31, 332, 333, 334, 335, 336,
Ricci-Curbastro, Giuseppe, 294 337, 340, 358, 360, 361, 362n,
Ricci calculus, 294 363n 373, 377, 393
Ricci tensor, 297, 298 Schrodinger equation: time-
Riemann, Bernhard, 105, 147, 157- dependent, 328n, 330, 333, 334-
68, 261, 429, 430 35, 337, 340, 341, 342, 358, 360,
Riemann tensor, 166-67 361, 377, 393; time-independent,
Riesz, Frigyes, 331 328, 393
Rietdijk, C. W., 280, 281 Schrodinger picture, 340
rigid body, 50, 162, 271-73 Schultz, Johann, 118, 135n
Ritz, Walther, 312, 324 Schwartz, E., 406n
Ritz combination principle, 312, Schwartz, Laurent, 413n, 452
324 Schwarz, Hermann Amandus, 445
Robertson, H. P., 53n, 302 Schwarz's inequality, 445
Robinson, H. M., 12n Schwarzschild, Karl, 293n, 299,
Roger, G., 352n 417,418
Romer, Ole, 36-40, 274n Schweber, Silvan S., 393n, 394, 395,
Rontgen, Wilhelm Konrad, 309 396
Roque, Xavier, 395n Schwinger, Julian, 395
Rosanes, Jacob, 324 second, 262
Rosen, Nathan, 308, 349-55 Seelig, Carl, 316n
Rossi, B., 277n sensation, 235-36
rotation, 56 sensorium, 56-57
roulette, 197-98 separable, 451
Rousseau, Jean-Jacques, 105n Shaftesbury, Anthony Ashley
Rubens, H., 316n Cooper, 3rd Earl of, 105
Ruhla, Charles, 353n Shamos, Morris H., 82n
Rumford, Benjamin Count, 181 Shapere, Dudley, 421
Rund, Hanno, 127n Shapiro, Stuart, 4n, 408n
Russell, Bertrand, 79, 162n, 410, Shimony, Abner, 338, 353n, 357n,
411,434,435 358n
Russo, A., 395n simultaneity, 51-52, 255-57, 273-
Rutherford, Ernest, 245n, 309, 311, 77; and interaction, 135-37, 281;
314,333, 394 relativity of, 258-59
singularity theorems, 305n, 419
Saccheri, Girolamo, 149 Sklar, Lawrence, 212n
Sade, Donatien Alphonse Francois, Skyrms, Brian, 440
Comte de, 105 Slater, John Clarke, 316-18
scalar, 168n, 444 Smith, Crosbie, 179n
scalar field, 168 Sneed, Joseph, 412, 414, 426n
Schild, A., 271 Snell's law, 237
510 Index

snub as a paradigmatic term of Stern, Otto, 326n

physics, 14 Stillingfleet, Edward, 75
Solmsen, Friedrich, 12n Stoney, George Johnstone, 309n
Sommerfeld, Arnold, 292n, 298n, Strawson, Peter Frederick, 13In
312, 313, 325, 393, 394 stress tensor, 295n; Maxwell's, 295
Sophocles, 5In structuralism, 412-20, 424-25
space, 50-57, 105-109, 112n, 113- structure, species of, 413n
18, 141, 145-46, 240, 370; subcovering, 456
mathematician's sense, 154; see subgroup, 153
also configuration space, Hilbert subjective, 388, 398n, 401-402
space, phase space, topological sublimity, 28
space, vector space subset, 443
spacelike: curve, 269; interval, 266; subset topology, 456
vector, 269 substance, 8n, 9, 54, 106, 122-28,
spacetime interval, 265, 267 135-36, 238
spacetime, 257; adumbrated by Sudbery, Anthony, 336n, 357n
Kant, 108; implicit in Einstein's superposition, 338; nontrivial, 357
first relativity paper, 257; Suppe, Frederick, 408n, 412n
Minkowski, 109n, 260-71, 419, Suppes, Patrick 412, 413, 414, 425n
427, 428; neo-Newtonian, 56n, supremum, 453
429-30; singularities, 305, 419 symmetric, 133
Specker, E. P., 375, 384n symmetric difference, 38In
spectrum (of linear operator), 450 synchronization of distant clocks: by
spin, 319, 394 clock transport, 255; by Einstein's
Spinoza, Baruch, 98, 106, 399-400 method, 255-56; prerequisite of
SR (special relativity), 249, 250- velocity measurement, 274
271, 292, 326, 424, 428, 436; synechism, 231, 233-34
philosophical problems, 271-289; synthetic statements, 11 On
relation with GR, 249n, 292
St. Athanasius, 406n tachyon, 266n
St. Basil, 406n Tait, Peter Guthrie, 177
St. Gregory Nazianzenus, 406n tangent bundle, 161
St. John Chrysostomus, 406n tangent space, 159-61, 169n
St. Thomas Aquinas, 8 Teller, Paul, 395, 396n, 440
Stachel, John, 286n, 297n, 386n temperature, absolute scale, 188,
Stark, Johannes, 329 191
statistical inference, 227-28 tensor field, 293n, 294n; see also
statistical operator, 345-47 Einstein tensor, electromagnetic
statistics: Bose-Einstein, 321, 396; field tensor, energy-momentum
Fermi-Dirac, 321, 396; Maxwell- tensor, Ricci tensor, Riemann
Boltzmann, 320-21; see also error tensor, stress tensor
statistics tensor product of Hilbert spaces,
Stein, Howard, 281n, 338, 396n 342, 452-53
 Index 511

terms, observational, 403, 411, 412, Tomonaga, Sin-Itiro, 395

421; theoretical, 104, 403, 411, topological space, 455-56
412, 414n, 421; theory-laden, topology, 455; generated from a
404, 421 given base, 456; [RM-standard,
tetrad, 268 155n, 457; subset, 456
tetrad field, 268 Torretti, Roberto, 56n, 11 In, 155n,
textbooks, reliability of, 38n-39n 162n, 198n, 216n, 258, 260n,
theoretical terms, see terms, 276n, 286n, 297n, 298n, 408n,
theoretical 413n, 414n, 422n
theory of everything, 246-47, 407, Torricelli, Evaristo, 420
433 torsion balance, 82n
theory, physical, 242-48; received trace, 449
view, 407, 408, 410-12, 417; see transcendental (Kant's sense), 11 In
also aim of physical theory transformation, 154; see also
thermal radiation, 205n; cosmic coordinate transformation,
background, 249, 304-305, 306 Galileian transformation, Lorentz
Thermodynamics: First Principle of, transformation, Poicare
187; Second Principle of, 187-88, transformation, point
194,207,209,211 transformation
thing-language, 403 translation, spatial, 56
things-in-themselves, 107, 109, 123, Truesdell, Clifford, 45
140, 141, 434 truth, approximate, 416
Thomson, J. J., 309, 310 twin paradox, 277-80
Thomson, James, 51, 52, 53, 255n tychism, 230, 23In, 232, 234
Thomson, William, see Kelvin
three-men paper, 324 Ueberweg, Friedrich, 162
time, 50-57, 106-109, 112n, Uhlenbeck, George E., 320
122-23, 128, 129, 141, 370; uncertainty relations, see
Archimedean standpoint outside Heisenberg's indeterminacy
it, 214; direction of, 212-14; relations
local, 252; structurally similar to unitary evolution, 340, 372
straight line, 5-6; see also proper
time values excluded from physics, 399-
time coordinate for inertial frame, 400
255-56 Van der Waerden, B. L., 308n,
time directedness of thermal 313n, 317, 321
phenomena, 207 van Fraassen, Bas C , 357n, 412
time reversal invariance, 206, 21 On variance, 344
time since the creation of the world, Vasil'ev, N. A., 378
303 vector field, 161n, 168n, 169, 173
timelike: curve, 269; interval, 266; vector space, 444-45; complex, 444;
vector, 269 dual, 446; norm, 164; real, 444;
Tipler, Frank J., 400 see also Hilbert space
512 Index

velocity (directed) versus speed, 18, Wheeler, John Archibald, 67n, 300n,
42 363n, 387
verifiable in Comte's sense, 229 Whewell, William, 216-22, 242
vibrating string, 327 Whitehead, Alfred North, 410, 411
Vienna circle, 234, 402, 403, 404 Whittaker, Edmund T., 180n, 295
Villard, Paul Ulrich, 309 Wigner, Eugen, 338, 353n, 356,
virtual displacement, 86 357n, 394n, 430n
vis impressa, 43, 44; see also Will, Clifford M., 425
impressed forces Williams, C. J. E, 12n
vis inertiae, 43-45, 77 Williams, Ruth M., 56n
vis insita, 43 Wilson, Charles Thomson Rees,
vis viva, 35, 104 312n
Voigt, Woldemar, 260n Wilson, Robert W, 305
Volta, Alessandro, 82 Wise, M. Norton, 179n
Voltaire, 34 Wittgenstein, Ludwig, 423, 437
von Laue, Max, 53n, 295, 309 Wolf, A., 182n
von Mises, Richard, 201n Wolff, Christian, 140, 141n
von Neumann, John, 329, 336, Wolter, Allan B., 13n
337n, 360-62, 363, 374-75, world (Minkowski's sense), 262
379-383, 387, 413n, 452 worldline, 269
von Weizsacker, Carl Friedrich, 368, Wroblewski, Andrzej, 38n, 39n
379n
Xerxes, 405
Wald, Robert M., 305n X-rays, 309, 316
Walker, A. G., 303
Warburg, E., 316n Young, Thomas, 35n, 175n
Warwick, Andrew, 179n
Waterston, J. J., 195n Zawirski, Zygmunt, 379
Watts, Isaac, 398n Zeeman, Pieter, 319, 320
wave mechanics, 308, 325-29; Zeeman effect, 319; anomalous,
equivalence with matrix 319-20
mechanics, 329-32, 336 Zeilinger, Anton, 352, 353n
Weber, Wilhelm Eduard, 174 Zeno of Elea, 6; "dichotomy", 6;
Weil, Andre, 413n "arrow", 6, 10
Weinberg, Steven, 395, 396, 417 Zermelo, Ernst, 210
Westfall, Richard, 41n, 43n zero, 153
Weyl, Hermann, 301, 413n Zurek, Wojciech Hubert, 363n