Russells Logical Atomism David Bostock download

https://ebookbell.com/product/russells-logical-atomism-david-
bostock-4671366

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

The Philosophy Of Logical Atomism 1st Edition Russell Bertrand

https://ebookbell.com/product/the-philosophy-of-logical-atomism-1st-
edition-russell-bertrand-60480114

Russells Philosophy Of Logical Analysis 18971905 Jolen Galaugher Auth

https://ebookbell.com/product/russells-philosophy-of-logical-
analysis-18971905-jolen-galaugher-auth-5375944

Barriers To Entailment Humes Law And Other Limits On Logical

Consequence 1st Edition Gillian K Russell

https://ebookbell.com/product/barriers-to-entailment-humes-law-and-
other-limits-on-logical-consequence-1st-edition-gillian-k-
russell-58194806

Daily Warmups Reading Grade 3 Russell S

https://ebookbell.com/product/daily-warmups-reading-
grade-3-russell-s-6340636
War Like The Thunderbolt The Battle And Burning Of Atlanta 1st Russell
S Bonds

https://ebookbell.com/product/war-like-the-thunderbolt-the-battle-and-
burning-of-atlanta-1st-russell-s-bonds-54878280

Russells Unknown Logicism A Study In The History And Philosophy Of

Mathematics 1st Edition Sbastien Gandon Auth

https://ebookbell.com/product/russells-unknown-logicism-a-study-in-
the-history-and-philosophy-of-mathematics-1st-edition-sbastien-gandon-
auth-5889756

Learning American Sign Language In High School Motivation Strategies

And Achievement 1st Edition Russell S Rosen

https://ebookbell.com/product/learning-american-sign-language-in-high-
school-motivation-strategies-and-achievement-1st-edition-russell-s-
rosen-7142576

Russells Unknown Logicism A Study In The History And Philosophy Of

Mathematics Sebastien Gandon

https://ebookbell.com/product/russells-unknown-logicism-a-study-in-
the-history-and-philosophy-of-mathematics-sebastien-gandon-10204892

Stealing The General The Great Locomotive Chase And The First Medal Of
Honor Russell S Bonds

https://ebookbell.com/product/stealing-the-general-the-great-
locomotive-chase-and-the-first-medal-of-honor-russell-s-bonds-2487766
Russell’s Logical Atomism
Drawing of Bertrand Russell by Augustus John, c. 1913. Image courtesy of William Ready
Division of Archives and Research Collections, McMaster University Library, Hamilton,
Canada # Bridgeman Art Library
Russell’s Logical
Atomism

David Bostock

1
3
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
# David Bostock 2012
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2012
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose the same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
Typeset by SPI Publisher Services, Pondicherry, India
Printed in Great Britain
on acid-free paper by
MPG Boooks Group

ISBN 978–0–19–965144–3

1 3 5 7 9 10 8 6 4 2
Contents

Preface vi
Abbreviations ix

Part I: Logic and Mathematics

1. Elementary Logic 5
2. The Contradiction (i): The Problem 16
3. On Denoting 31
4. The Contradiction (ii): A Simple Solution 47
5. The Contradiction (iii): A Ramified Solution 74
6. Problems 97

Part II: Knowledge

7. Acquaintance 113
8. Knowledge and its Foundations 135
9. Logical Constructions (i): Matter 151
10. Logical Constructions (ii): Minds 169
11. The Demise of Neutral Monism 190

Part III: Metaphysics

12. Propositions 201
13. Universals 236
14. Facts 252

Epilogue 279
Concordance: Reprints of Journal Articles 284
References 286
Index 297
Preface

Bertrand Russell would sometimes describe his own basic philosophy as ‘the philoso-
phy of logical atomism’. He first used this description in a little-known lecture that he
delivered in Paris in March 1911, entitled ‘Analytic Realism’, which is a different title
that he also accepted at that time as describing his basic position.1 It was a ‘realism’
insofar as it accepted that there are many things which exist quite independently of
human minds, and was ‘analytic’ because it was concerned with the analysis of these
things. The things in question Russell regarded as ‘complexes’ or ‘facts’, and the
analysis aimed to discover the simple ingredients of these facts, which count as
‘atoms’ because they are simple, and not open to any further analysis. They are called
‘logical atoms’ because they are to be reached by a process which Russell always called
‘logical analysis’, though it is not quite clear why it deserves the adjective ‘logical’. As
we might say, it is counted as logical because it is in the first place an analysis of
propositions, and logic is traditionally a study of propositions. But Russell would rather
say that propositions are used to state facts, and the analysis that he is concerned with is
really an analysis of facts, so that the logical atoms that it aims to reveal are the simple
ingredients of facts. That is to say, they are what the world is ultimately made of. That is
a brisk and bald outline of what he has in mind when he speaks of ‘logical atomism’.
This book aims to fill in many of the details.
Russell described his position as ‘the philosophy of logical atomism’ not only in that
lecture of 1911, but also in two better known works of 1914, namely in the program-
matic article ‘On Scientific Method in Philosophy’, and in the preface to his book Our
Knowledge of the External World.2 Then it is used as the title of his set of lectures on The
Philosophy of Logical Atomism of 1918; and finally it occurs once more in his own
contribution to a collection of essays by various contemporary philosophers in 1924,
which he called simply ‘Logical Atomism’. So one might say that Russell saw himself as
believing in ‘logical atomism’ during the years 1911–24. Looking back on his life from
the much later perspective of My Philosophical Development [1959], Russell actually said
There is one major division in my philosophical work: in the years 1899–1900 I adopted the
philosophy of logical atomism and the technique of Peano in mathematical logic. . . . The change
in these years was a revolution; subsequent changes have been of the nature of an evolution.
(p. 11)

Certainly, these years in Russell’s life saw an important change in his philosophical
attitude, and one might well say that they were the beginning of his development

1
Russell [1911]. The phrase ‘logical atomism’ is used on p. 135.
2
Russell [1914d: 111, 1914c: 14].
 P R E FAC E vii

towards what is now regarded as his ‘philosophy of logical atomism’. But the doctrine
did take some time to evolve. Similarly, it did not end abruptly in 1924, and his later
book The Analysis of Matter [1927a] may fairly be counted as yet a further development
of this philosophical attitude. But we may reasonably take the years 1911–24 as central.
Meanwhile, let us go back to the beginnings.
The emphasis on logic stems from Russell’s first brush with modern logic, which was
with Peano in 1900. This subject was of central importance in his analysis of mathe-
matics, and he never lost his enthusiasm for it. But in later years he employed it in other
areas too. His atomism might be traced back even earlier, to his revolt from the Hegelian
‘monism’ that was so widespread when he was an undergraduate at Cambridge.
Strongly influenced by G. E. Moore he came to reject this monism in about 1898,
and his first important publication as a ‘pluralist’ was his book A Critical Exposition of the
Philosophy of Leibniz [1900]. Of course, if a pluralist is simply one who believes that
there really are many things, and not just one, then a pluralist is not as such committed
to any form of atomism. But it is quite a natural development to suppose that the ‘many
things’ are ultimately composed of ‘many atoms’, and then it is natural to seek for
these ‘atoms’. Much of the progress in Russell’s thought, during our period, was in
his successive discoveries that this or that could be analysed, and so did not have to be
accepted as an atom. The position that he ended with was extremely economical, as we
shall see.

The book is divided into three main parts, namely on Russell’s views: I on logic and
mathematics, II on knowledge, and III on metaphysics. These parts are roughly
chronological. Russell worked mainly on logic from 1900 to 1910; he turned his
attention to the theory of knowledge in many writings from 1910 on; it is always
intertwined with his metaphysics, which has its centrepiece in the lectures on The
Philosophy of Logical Atomism of 1918. These lectures bring together many of the themes
developed independently in earlier writings. But there is one development, which
should surely be counted as part of Russell’s logical atomism, but which belongs to the
years after 1918, namely the views that he puts forward in his book The Analysis of Mind
[1921a]. This work is treated at the end of part II, where it fits very naturally, although
the chronological order is thereby disrupted.
The only other philosopher who is ever called a logical atomist is Ludwig Wittgen-
stein, in his early period, which culminated in his Tractatus Logico-Philosophicus of 1921.
Wittgenstein became a pupil of Russell’s in October 1911, and after only a year Russell
was thinking of him as an equal, and hoping that he might be his successor at Cam-
bridge. It is clear that they saw a lot of one another, and talked a great deal. But in
October 1913 Wittgenstein left Cambridge, in order to work alone (in Norway), and
he and Russell did not meet again until after the war, in December 1919. By that time
Wittgenstein had written his Tractatus, and Russell agreed to write an introduction for
it. We shall never know how much of Russell’s thought, subsequent to their meeting
in October 1911, comes from his talks with Wittgenstein. In his writings Russell very
viii PR EFAC E

clearly acknowledges a debt,3 but in my own opinion he exaggerates it. Some, and
perhaps most, of what he credits to Wittgenstein clearly has its seeds in his own
thought, from times before they met. But on several points all that we can say is that
we have no direct evidence, either way, on who influenced whom.
Wittgenstein’s version of logical atomism is clearly rather different from Russell’s,
but that is a topic for another book, and there has not been the space to include it in this
one. I regard it as a possible project for the future.

Some of the material in my Part I, especially in Chapter 4, is taken from an article of

mine entitled ‘Russell on “the” in the plural’. This was delivered at a conference at
McMaster University in 2005, and later published in the collection Russell vs. Meinong:
The Legacy of On Denoting, edited by Nicholas Griffin and Dale Jacquette, and
published by Routledge in 2008. I thank the editors and the publisher for permission
to reuse this material.
I should like to acknowledge help with this book from Peter Sullivan, who gave me
an extended criticism of all of Part I, and from two anonymous readers for Oxford
University Press, who have done the same for the whole book, and especially for
chapters in Parts II and III. The book has surely benefited from all of their suggestions.
In a quite different way I should like to acknowledge help from what I shall always
think of as ‘The Philosophy Library’ of Oxford University. (It is now officially a part of
the overarching Bodleian Library of this University, but for much of the time that
I have known it, it was an independent institution.) The staff of this library have
continually been helpful to me in all kinds of ways, and have very much facilitated my
efforts to write. Finally, I should like to acknowledge help from Merton College,
particularly in the person of Judith Kirby, who has done all my secretarial work for me
for many years. I do not know how I could have managed without her; she has been
absolutely essential.
David Bostock
Merton College, Oxford
July 2011

3
Russell [1914c: 9, 1918b: 177].
Abbreviations

References are generally given in an abbreviated form, by citing just the author and the date of
publication, e.g. ‘Strawson [1950]’. Full details are then given in the list of references at the end of
the book. In some cases I have used writings by Russell which were not published at the time,
and these are given by the date of their composition. But some important works by Russell,
which are frequently cited, are also referred to by their abbreviated titles, as follows:

PoM The Principles of Mathematics [1903]

OD ‘On Denoting’ [1905b]
ML ‘Mathematical Logic as based on the Theory of Types’ [1908]
PM Principia Mathematica, with A. N. Whitehead, vol. 1 [1910d]
KAD ‘Knowledge by Acquaintance and Knowledge by Description’ [1910/11]
PP The Problems of Philosophy [1912a]
RUP ‘On the Relation of Universals and Particulars’ [1911/12]
NA ‘On the Nature of Acquaintance’ [1914a]
RSP ‘The Relation of Sense-Data to Physics’ [1914b]
KEW Our Knowledge of the External World [1914c]
PLA ‘The Philosophy of Logical Atomism’ [1918b]
IMP Introduction to Mathematical Philosophy [1919a]
Propns ‘On Propositions: What they Are and How they Mean’ [1919b]
AMind The Analysis of Mind [1921a]
LA ‘Logical Atomism’ [1924]
AMatter The Analysis of Matter [1927a]
MPD My Philosophical Development [1959]

I add one work by Wittgenstein, which is also frequently cited:

TLP Tractatus Logico-Philosophicus [1921]
This page intentionally left blank
 PART I
Logic and Mathematics

As an undergraduate at Cambridge (1890–4) Russell had studied first mathematics and

then philosophy. As he tells us when looking back on his life in the book My
Philosophical Development [1959], an important reason for his change from the one
subject to the other was his search for certainty (p. 29). Some time earlier he had lost a
belief in the certainty of religious truth, and he then felt that if anything could be
certain it must be mathematics. However it soon became clear to his keen intellect that
the mathematics then taught at Cambridge contained a number of fallacies, and could
by no means be called ‘certain truth’. So he hoped that philosophy would be better.
At first he had embraced the idealist philosophy that was then prevalent at Cambridge,
and had been particularly interested in its application to science and mathematics. (His
doctoral dissertation was on the foundations of geometry.) But by 1898 he and his
younger contemporary G. E. Moore had come to think that idealism was a great
mistake, and had revolted against it. However he retained his previous interest in
mathematics, and in fact had already made a draft of what was to become his Principles of
Mathematics before the important meeting with Peano in 1900. He was once more
hoping that a revised view of mathematics would fulfil the yearning for certainty that
had never left him.
This was the background situation when, in 1900, he attended a mathematical
conference in Paris. There he was extremely impressed by the exactitude that
Giuseppe Peano and his school had brought to the subject, and at once read all of
Peano’s works. This is where he first met what we now think of as modern logic, for
Peano was a pioneer in this area. Russell soon mastered Peano’s logical system, and
made an important contribution to it (i.e. his paper of 1901 on ‘The Logic of
Relations’). He also felt that he had learnt from Peano how to overcome various
problems that had stood in the way of a proper understanding of the concept of
number, for example the problem of how and in what sense there could be such a
thing as a class (a collection?) of no members, and how a class (a collection?) of just
one member could itself be something different from that one member. As these
problems became clearer, he was led to think that the concept of a number could
after all be analysed in purely logical terms, and so to embrace what has since become
2 L O G I C A N D M AT H E M AT I C S

known as the ‘logicist’ view of the nature of mathematics. This can be summed up as
the two claims (a) that the concepts of mathematics can be analysed in terms of the
concepts of pure logic, and (b) that the truths of mathematics then turn out to be no
more than truths of logic. The aim of deducing mathematics lay behind almost all of
Russell’s work in logic, i.e. of the great bulk of all his efforts during the ten years
from his first introduction to Peano’s ideas in 1900 to the publication of Principia
Mathematica in 1910. Part I of this book is entirely concerned with the work that
received its final form in Principia Mathematica.
I add here that I shall regularly speak of Principia Mathematica as if it were all Russell’s
own work. This is in one way unfair, since the book was written in collaboration with
his former tutor, A. N. Whitehead, who is joint author of the final publication, and
whose collaboration dates back almost to the beginning of the project in 1900. But on
the other hand Russell himself says, in describing this work, that ‘broadly speaking,
Whitehead left the philosophical problems to me’,1 and that is what will matter for us.
With very few exceptions, we shall be concerned with the introductory portions of the
work, and it is clear that these were written just by Russell. He did indeed publish all
the important ideas separately, under his own name only.2
Chapter 1 will contain a brief exposition of what is nowadays called ‘elementary
logic’. I expect that most readers will be quite familiar with this topic already, and so
can skip all of the first section of this chapter, but I add that it was not familiar to
Russell, and he may certainly be counted as one of those who helped to develop it.
Chapter 2 turns to non-elementary matters, first giving some details on how Russell
hoped to deduce mathematics from logic, and then turning to the big problem that was
posed for this project by his discovery of what he always called ‘the contradiction’,
namely what we now call ‘Russell’s paradox’. The chapter describes Russell’s first
efforts to resolve this problem, and why he found them unsatisfactory. Chapter 3 is
then something of a digression, for it is concerned with the theory of definite descrip-
tions in his well-known article ‘On Denoting’ [1905], and this is (on the surface) just a
modification in the theory of how elementary logic should be applied to our ordinary
ways of thinking. But it is a modification with some deeper morals, and Russell himself
describes it as opening the way to the true solution to his problem. Certainly it does
have an importance in the development of that solution, even if it is mostly discussed
today in a different aspect. Then Chapters 4 and 5 describe how Russell reached his

1
This is how Russell described the work when looking back on it from the much later perspective of
My Philosophical Development [1959: 74]. There is a useful assessment of Whitehead’s contribution to the work
in Alasdair Urquhart’s introduction to volume 4 of Russell’s Collected Papers [i.e. Russell 1994], particularly
pp. xxxviii–xli.
2
Russell [1910a] is a French version of almost all of chapter 2 of the Introduction to Principia Mathematica,
together with some themes from chapter 3 of that Introduction. Much of the discussion is also anticipated in
Russell [1908].
 L O G I C A N D M AT H E M AT I C S 3

final theory, with Chapter 4 devoted to what is now called his ‘simple’ theory of types,
and Chapter 5 to the more complicated ‘ramified’ theory that he finally recommends.
In each case there is some discussion both of how Russell himself conceived this theory
and of what more recent developments have made of it. Finally Chapter 6 notes the
main problems that arise, especially for the desired deduction of mathematics, and
makes some suggestions on how they might perhaps be met.
This page intentionally left blank
 1
Elementary Logic

Russell’s meeting with Peano in 1900 was the spur to his developing what we now
think of as ‘modern logic’. Peano had introduced much of today’s notation (which
Russell took over from him), and had set out some basic principles, but done little to
systematize them. Russell began to reflect on ways of improving the basis of the system,
and parts of his Principles of Mathematics [1903], henceforth PoM, are concerned with
what is needed in logic by way of fundamental concepts, in terms of which all others
are defined, and with what he always called ‘primitive propositions’, which are the
basic premises or axioms from which all other truths of logic are to be deduced. Early in
1901 he discovered Frege’s work in the subject, which preceded Peano’s and was
certainly much better, though Russell did not at once realize that. (His first reaction to
Frege was like everyone else’s: the symbolism is entirely unfamiliar, and not at all easy
to read, so why bother?1) But when he did pay serious attention to Frege in 1902 he
began to see the value of the work. So this chapter begins with a brief account of the
‘elementary logic’ that is shared between Frege and Peano and Russell. It is a needed
background. Indeed, it is a needed background to almost all of today’s philosophy, and
I imagine that most readers will already be quite familiar with it. In that case, they may
conveniently skip the first section, which will only summarize what they already
know. But the second section may contain some new information, and the third
describes Russell’s own understanding of this elementary logic, which is not quite the
same as what is nowadays usual.

1 A modern account of elementary logic

Elementary logic, as now understood, contains two parts. First there is what is variously
called ‘propositional’ or ‘sentential’ logic, but which I think is better called the logic
of truthfunctors. Next there is what is often called ‘first-order predicate logic’, but which
I think is better called the logic of quantifiers (of the first level). Here is a brief outline
of what are now taken to be the central ideas.
Logic is concerned with truth and falsehood. (These are called the ‘truthvalues’, and
in classical logic—which is what we are here concerned with—it is assumed that these

1
‘I had seen his [i.e. Frege’s] Grundgesetze der Arithmetik, but, owing to the great difficulty of his symbolism,
I had failed to grasp its importance or to understand its contents’ (Russell, preface to PoM [1903: xvi]).
6 L O G I C A N D M AT H E M AT I C S

two are the only truthvalues.) What is true or false is called a ‘proposition’. Propositions
can be thought of as what a sentence or utterance may say, or what a person may
believe, or in other ways. (Russell’s thoughts on what propositions are will occupy us
later, in Chapter 12.) But so far as pure logic is concerned, a proposition is simply
whatever has a truthvalue: every proposition is either true or false, and no proposition
is both true and false. This book will use the letters ‘P ’, ‘Q’, ‘R’, . . . . as schematic letters
to stand in for any proposition.
Some propositions contain others as parts, with such words as ‘not’, ‘and’, ‘or’, ‘if ’
applied to the parts and (in many cases) joining them together. This book will use the
symbols ‘~’, ‘&’, ‘∨’, ‘!’ to abbreviate these words, or the ideas that they stand
for. We assume (in classical logic) that these symbols are truthfunctors, i.e. signs for
truthfunctions, which is to say that they may be explained by saying how they affect
the truthvalues of the propositions that contain them. To give the explanations in
English, they are:
‘~P ’ is true if and only if ‘P ’ is not true
‘P&Q’ is true if and only if both ‘P ’ and ‘Q’ are true
‘P∨Q’ is true if and only if either ‘P ’ or ‘Q’ or both are true
‘P!Q’ is true if and only if either ‘P ’ is false or ‘Q’ is true2
To these we may add a symbol for ‘if and only if ’, which is an ‘if ’ that works both ways,
and so is abbreviated to ‘$’. The definition is

‘P$Q’ is true if and only if both ‘P!Q’ and ‘Q!P ’ are true, i.e. if and
only if ‘P’ and ‘Q’ both have the same truthvalue.
These explanations may conveniently be summed up in ‘truthtables’, which use ‘T’ for
‘true’ and ‘F’ for ‘false’, and surely need no other explanation:

P ~P P Q P &Q P ∨Q P!Q P$Q

T F T T T T T T
F T T F F T F F
F T F T T F
F F F F T T

The five truthfunctors explained here are the ones that will be used in this book, and
are the ones that are now standard. Naturally one may if desired use other truthtables
to introduce other truthfunctors, but for ordinary purposes little is gained thereby.

2
This is the truthfunctional interpretation of ‘!’, i.e. of ‘if ’. The question of how well it represents the
ordinary English ‘if ’ is controversial. Of course the same question may be raised about the ordinary English
‘not’ and ‘and’ and ‘or’: is the truthfunctional interpretation adequate for these?
 E L E M E N TA RY L O G I C 7

I remark that Russell’s notation differs from what is here introduced, but that is a trivial
point of no importance (so long as you do also know how to read his notation).3 This
introduces the schematic language for the logic of truthfunctors. Let us now turn to the
quantifiers.
Some propositions are ‘subject-predicate’ propositions in the modern sense, i.e.
propositions in which some singular subject is referred to. What is said about
that subject then forms the predicate of the proposition.4 This book will use the
schematic letters ‘a’, ‘b’, ‘c’, . . . to stand in for any reference to a subject, and the letters
‘F ’, ‘G’, ‘H ’, . . . to stand in for what is predicated of a subject. By convention, we
write the predicate letter before the subject letter, as in ‘Fa’ or ‘Gb’. Some propositions
contain a reference to two or more subjects, and what is predicated of them is then
a relation of two or more places. We represent these by the schematic formulae ‘Fab’
or ‘Gabc’, and so on.5 These schematic formulae can then be combined into more
complex schemata by using the truthfunctors just explained. We can now introduce
the quantifiers.
These are expressed in English by ‘every’ and ‘some’, for which we use ‘8’ and ‘9’
respectively. The quantifier sign is always followed by what is called a ‘variable’, which
is then repeated in what follows it, in order to show just where the ‘every’ or the ‘some’
is applied. In a first-level logic, which is all that presently concerns us, the variable
will be one that takes the place of a subject letter, and for these we use the letters ‘x’,
‘y’, ‘z’, . . . Thus if we begin with a simple subject-predicate proposition ‘Fa’, we may
then put a variable, say ‘x’ in place of its subject letter ‘a’, and add at the front
a quantifier ‘8x’ or ‘9x’ which is said to ‘bind’ that variable. The result is ‘8xFx’
or ‘9xFx’ and these represent the propositions which claim that ‘Fa’ remains true for
everything in place of ‘a’, or that it is at least true for something in place of ‘a’. To give
a concrete example,
If ‘Fa’ is ‘Socrates is mortal’
then ‘8xFx’ is ‘Everything is mortal’
and ‘9xFx’ is ‘Something is mortal’
But in order to say that every man is mortal we need another predicate, say ‘G . . . ’, to
represent ‘ . . . is a man’. Then the sentence that we want is ‘8x(Gx!Fx)’. In a more

3
In place of ‘P &Q’, ‘P !Q’, and ‘P $Q’ Russell himself uses ‘P.Q’, ‘P ⊃Q’, ‘P Q’. Others have used
–
‘¬P ’ or ‘P ’ in place of ‘~P ’, and ‘P∧Q’ in place of ‘P &Q’. So far as I know, the use of ‘∨’ for ‘or’ is universal.
The symbol stands for the Latin ‘vel’, i.e. for the inclusive rather than the exclusive ‘or’. The symbol ‘∧’ is a
deliberately contrived contrast to ‘∨’. (I do not give Frege’s notation, which is very different.)
4
Contrast the traditional usage, which hails from Aristotle, by which a proposition of the form ‘All A’s are
B’s’ is counted as a subject-predicate proposition, even though ‘All A’s’ evidently does not refer to a singular
subject.
5
In place of ‘Fab’ Russell uses ‘aRb’, with ‘R’ to suggest a relation. So long as only two-place relations are
concerned, this is a very natural symbolism. It is usual to express particular two-place relations in this form,
e.g. ‘x¼y’, ‘x<y’, ‘x2y’. We may abbreviate by placing a slash through the relation-sign, as in ‘x6¼y’, to
express its negation.
8 L O G I C A N D M AT H E M AT I C S

complex schematic formula there may be several quantifiers. For example the formula
‘8x9yFxy’ says that for everything in place of ‘a’ the schema ‘Fab’ comes out true for at
least one thing in place of ‘b’. By contrast, the formula ‘9y8xFxy’ says that there is at
least one thing which you can put in place of ‘b’ that will make ‘Fab’ come out true for
everything whatever in place of ‘a’. You should think about this example for a bit, in
order to see clearly that there is a difference between these two formulae. (The first
does not imply the second, though the second does imply the first.) Generally,
however complicated the schema ‘(—x—)’ may be, ‘8x(—x—)’ may be read as
‘Everything is such that (—it—)’, and ‘9x(—x—)’ may be read as ‘Something is such
that (—it—)’. But where there are several quantifiers we need to know which of them
each occurrence of ‘it’ goes back to.6
I have to assume that this is an adequate explanation of the schematic language to be
used in setting out the logic of the truthfunctors and the quantifiers. I should also say
that in order to come to a good understanding of this language one needs practice in
translating between it and one’s ordinary language. This brings one to appreciate how
much of what we ordinarily say and think can be schematically represented in the
language. But I do not provide the material for such practice. There are any number of
books on elementary logic that do provide it. Instead I move on at once to what
elementary logic can do with this language.
Traditionally, logic is concerned with the validity of arguments, i.e. with what
follows from what. In logic as it is now conceived we are concerned with what follows
formally, where this is understood in terms of the formal language just introduced,
i.e. one which uses ‘P ’, ‘Q’, . . . as schematic letters for any propositions, ‘a’, ‘b’, . . . as
schematic letters for any reference to a singular subject, and ‘F’, ‘G’, . . . as schematic
letters for any predicate. So we first explain validity for such schemata. An interpretation
for the language assigns some particular propositions, or subjects, or predicates to the
schematic letters involved. It also assigns some domain for the quantifiers to range over,
i.e. some definite objects to be those of which ‘all’ or ‘some’ are concerned. Then a
single schematic formula counts as valid if it always comes out true, however its
schematic letters are interpreted, and whatever the domain of quantification is taken
to be.7 A series of such formulae representing an argument, with several formulae
(to the left) representing the premises, and one (to the right) representing the conclu-
sion, counts as a valid sequent if in all interpretations it is truth-preserving, i.e. if all

6
Note that a variable ‘x’ or ‘y’ is not a schematic letter; it is part of the notation that represents a quantifier.
I add that in place of ‘8x(—x—)’ Russell’s notation is ‘(x)(—x—)’. Other notations that have been used in
place of ‘8’ and ‘9’, often for a special purpose, are ‘∧’ and ‘∨’, and ‘П’ and ‘S’.
7
It is usual to exclude the ‘empty’ domain, i.e. the case where there is no object that qualifies as ‘something’, so
that no 9-proposition is interpreted as true. Whether this is a good idea may be debated, and is debated in many
places (including my book [1997: ch 8]). But for present purposes I follow the usual procedure, without further
debate. This suits Russell’s practice. (But in his later Introduction to Mathematical Philosophy [1919a: 203–4], he does
admit that he is no longer satisfied with this practice. Cf. p. 69, n. 42.)
 E L E M E N TA RY L O G I C 9

interpretations which make all the premises true also make the conclusion true. For
schemata, validity is a matter of truth in all interpretations.
We can here make a simplification. So far as the interpretation of propositional
letters is concerned, all that is needed is to assign them a truthvalue, for that is all that is
relevant to the validity of the formulae (or sequents) in which they occur. Similarly, for
the interpretation of a subject letter the only thing that is relevant is which object in the
selected domain it refers to; how it does so does not matter. And again for predicate
letters the only thing that is relevant is which objects in the domain they are to count as
being true of. The ordinary notion of ‘interpreting’ one of these schematic letters would
no doubt take it to be representing an ordinary sentence or subject expression or
predicate expression (as the case may be), and these expressions would have a genuine
meaning. Moreover, it is because of that meaning that the expression would have
whatever truthvalue it does, or would refer to whatever object it does refer to, or
would be true of whatever objects it is true of. But in elementary logic it turns out that
these ‘meanings’ do not matter. To put this in traditional terms, the intensions of what
counts as an interpretation are irrelevant, and what matters is just the extensions assigned,
where extensions are understood as above.8 That is enough to determine whether a
given schematic formula is ‘true under all interpretations’, or whether a given sche-
matic sequent is ‘truth-preserving under all interpretations’, and hence whether it
counts as valid.
I should make it clear that for this purpose one does not think of the truthfunctors
involved as themselves open to various interpretations. They are each to be held to a
constant interpretation, as given by their truthtables. Similarly, in any permitted
interpretation the subject letters must be taken to be referring to an object in the
domain selected, and the predicate letters must be taken to be true or false of those
objects. Finally, the quantifiers 8 and 9 range over just those objects, and are always
taken to mean every or some one of them. That also is held constant. So there are
constraints on what is to count as a ‘permitted’ interpretation, and validity is a matter of
truth (or truth preservation) in all permitted interpretations. But in future I will take this
qualification to go without saying.
This explains what it is for a schematic formula to count as valid, or a sequent of such
schematic formulae. We may now add that an actual proposition counts as ‘formally
valid’ if and only if it has a valid form, i.e. is an instance of some schematic formula that
is valid.
Similarly, an actual argument is ‘formally valid’ if and only if it has a valid form, i.e. it
is an instance of some schematic sequent that is valid. Rather than ‘formally valid’ it
would be more accurate to say ‘valid just in virtue of the truthfunctors and first-level
quantifiers that it contains’. This begs no question about what is to count as the ‘logical

8
For more on ‘intensions’ and ‘extensions’ see pp. 69–71 below.
10 L O G I C A N D M AT H E M AT I C S

form’ of a proposition or an argument, but it does indicate just which ‘forms’ are
considered in elementary logic.
Finally, the task of logic as nowadays conceived is the task of finding explicit rules of
inference which allow one to discover which formulae (or sequents) are the valid ones. In
the case of the truthfunctors this task is extremely simple. Nowadays one might say
that all that is needed is the rule for testing a whole formula (or sequent) by drawing
up a truthtable for it. But there are many other techniques, including the one
pursued by Russell (and by Frege), which I shall illustrate in a moment. Nowadays it
will seem to most people that there is no great philosophical significance in which
particular method is adopted. What is required is just a set of rules which is both (i)
‘sound’ and (ii) ‘complete’, in the sense (i) that the rules prove only formulae (or
sequents) that are valid, and (ii) that they can prove all such formulae (or sequents).
From his earliest consideration of the topic, Russell does provide such a set of rules for
the truthfunctors. He took longer to find a suitable set for the (first-level) quantifiers, but
his treatment in Principia Mathematica [vol. 1, 1910] would certainly qualify.9

2 A Russellian presentation of elementary logic

Both Russell and Frege set out their logical theories by using what is now called ‘the
axiomatic method’. Since that method is likely to be less familiar in these days than it
once was, I shall now give a detailed example. I do not follow exactly the definitions
and axioms that Russell himself adopted, for they have since been shown to be both
inelegant and inconvenient, but I do follow his general approach. (I also add a number
of details which he passes over in silence.) The system for elementary logic that is
presented here will be extended in Chapter 4, section 5, to give us a system for more
advanced logic.
First, for the logic of truthfunctors we may take just ~ and ! as our basic and
undefined notions, and define the other truthfunctors in terms of them. The formulae
of the system are then defined by these three rules of formation:
(a) Any one of the sentence letters ‘P ’, ‘Q’, . . . standing alone is a formula;
(b) If A is any formula, then ~A is also a formula;10

A &B for ~(A!~B),

A∨B for ~A!B,
A$B for (A!B) & (B!A).

9
There is no adequate treatment of the logic of the (first-level) quantifiers in Peano, nor in Russell’s PoM
[1903]. There is an adequate treatment in Frege’s Begriffsschrift [1879], but in 1903 Russell had not yet
assimilated this.
10
It should be obvious that if A is any formula then ~A is the result of writing that formula preceded by a
negation sign ‘~’. Similarly for other uses of A, B, C in this paragraph.
 E L E M E N TA RY L O G I C 11

(c) If A and B are formulae, then (A!B) is also a formula.

The definitions of other truthfunctors are: for any formulae A, B we abbreviate

We now turn from rules of formation to rules of proof, and begin by laying down
three axiom schemas: for any formulae in place of A, B, C, the following formulae are
axioms

Ax 1: ‘ A ! (B!A),
Ax 2: ‘ (A!(B!C)) ! ((A!B) ! (A!C)),
Ax 3: ‘ (~A!~B) ! (B!A).
The sign ‘‘’ says that what follows it may be proved. So this rule is that any axiom may
be proved, and we add a further rule of proof, which is often called Modus Ponens,
but is better called the rule of Detachment, namely

Det: If ‘A and ‘(A!B), then ‘B.

A proof is then a (finite) series of formulae, each one of which is either an axiom or a
consequence by the rule of Detachment of two formulae that are earlier in the series.
A proof is said to be a proof of the last formula of the series. This completes the system
for the logic of the truthfunctors. It is both sound and complete for that logic, in the
sense explained earlier.
Let us now turn to the logic of the (first-level) quantifiers. This uses the letters ‘a’,
‘b’, . . . as subject letters, the letters ‘F’, ‘G ’, . . . as predicate letters, and the letters ‘x’,
‘y’, . . . as (subject) variables. So we start by adding a further rule of formation:
(d) A predicate letter immediately followed by a series of one or more subject letters
is a formula.
(The formulae given by this rule are called the ‘atomic’ formulae of the system.) We
may use 8 as our only quantifier symbol, and define 9 in terms of it, but before we can
formulate its formation rule we shall need some preliminary explanations. The quanti-
fier 8 will be immediately followed by a variable, say v, to form a quantifier prefix 8v.
This will then be followed by an expression A(v) containing the variable v, and
surrounded by brackets, so as to form the whole expression 8v(A(v)). This whole
expression is said to be the scope of its initial quantifier prefix 8v. In a simple case, there
will be no further occurrence of the same prefix 8v, using the same variable v, within
the expression A(v). In that case we say that the occurrences of v within A(v) are free in
A(v), i.e. are free to be bound by a quantifier, and the occurrences of v within 8v(A(v))
are bound in 8v(A(v)), i.e. are bound by its initial quantifier prefix.11 In a more
complex case, one occurrence of the prefix 8v may occur within the scope of another,
and in that case an occurrence of v is bound by the quantifier prefix which has the

11
Note that an occurrence of v within a quantifier prefix 8v counts as bound by that prefix itself.
12 L O G I C A N D M AT H E M AT I C S

smallest scope that includes it. An occurrence of v is free in an expression A(v) if

and only if it is not bound by any quantifier prefix that occurs within the expression.
We can now formulate our final rule of formation:
(e) If A(t) is any formula containing a subject letter t,12 and if A(v) results by
substituting for each occurrence of t in A(t) an occurrence of a variable v that
is free within A(v), then 8v(A(v)) is a formula.
This completes the formation rules.13 That is, the formulae of the system are just those
expressions that can be shown to be formulae by using the five rules (a)–(e) now given.
We add the obvious definition of the quantifier 9, namely (for any variable n)

9v(A(n)) for ¬8v(¬A(v))

Let us now turn to rules of proof for the quantifier 8. Where A(v) contains
occurrences of v that are free in A(v), and A(t) results by substituting a subject letter
t for all such occurrences of v in A(v), we lay down as axioms all formulae

Ax 4: ‘ 8v(A(v)) ! A(t)
Where A(v) is as before, and where B contains no free occurrence of the variable v,
we also lay down as axioms

Ax 5: ‘ 8v(B!A(v)) ! (B!8v(A(v)))
Finally, we add a further rule of inference which is called the rule of Generalization:
if A(v) contains free occurrences of v just where A(t) contains occurrences of t, then

Gen: If ‘ A(t), then ‘ 8v(A(v))

A proof is defined as before, i.e. as a series of formulae, each of which is either one of
the axioms 1–5 or a consequence by Detachment or Generalization of preceding
formulae. This completes the system for elementary logic. Just as the quantifier-free
part of this system is both sound and complete for the logic of truthfunctors, so too the
whole system is both sound and complete for the logic of the quantifiers of first level.
In each case, the soundness of the system is very easily proved, for one only has to
observe that all the axioms are valid and that the two rules of Detachment and
Generalization preserve validity. But the proof of completeness is not trivial, and no
such proof was discovered until Gödel [1930].
I add a couple of comments on this presentation of elementary logic. First the rules
of formation have been framed so that (i) no variable occurs free in what is counted as a
formula, but only in expressions which are parts of formulae but not themselves

12
The letter ‘t’ stands for ‘term’. In elementary logic our only terms are the subject letters. In the more
advanced logic of Chapter 4 there will be further terms.
13
A usual abbreviation allows a quantifier to bind two variables at once. Thus ‘8xy(A(x,y))’ abbreviates
‘8x(8y(A(x,y)))’, and similarly ‘9xy(A(x,y))’.
 E L E M E N TA RY L O G I C 13

formulae; and (ii) no quantifier is vacuous, i.e. every quantifier prefix does bind one or
more occurrences of a variable in what follows it. These features of the system keep it
closer to what in our ordinary language are familiar sentences expressing propositions.
However they may be lacking in other treatments of elementary logic, for rules which
ignore them are in fact simpler. If vacuous quantifiers are permitted, the rules will
ensure that they add nothing, i.e. that if v does not occur free in A then 8v(A) is
everywhere interchangeable with the plain A. If formulae are allowed to contain free
variables, then those variables will function just as schematic letters do in our treatment,
and in fact there will be no need for separate alphabets to distinguish the two. I shall
exploit this point when I come to formulate the extended logical system of Chapter 4.
Second, it is obvious that there can be different choices of which symbols to take as
defined, of which formulae to take as axioms, and of which further rules of inference to
adopt. But all choices should lead to the same results, in the sense that they should
provide proofs of just the same theorems. The axioms that are adopted here have been
chosen mainly because they lead quickly and easily to other methods of proof, which
are actually much more convenient than are the axiomatic proofs that Russell himself
employs. (The crucial result for this purpose is what is called ‘the deduction theorem’,
and this comes without difficulty from axioms 1, 2, and 5.) But I do not here offer any
further development of this logic; I have merely chosen a basis that can quite easily
be developed.14
Let us now turn back from the formal details of elementary logic, as that subject is
understood nowadays, in order to say something of how Russell himself understood
the situation, so many years ago (i.e. in 1910). The main point of difference is that he
never attained our conception of what counts as a valid formula, i.e. one that is true
under all interpretations, so he had nothing against which to check his (perfectly
correct) rules of proof.

3 Russell and validity

Like Frege, Russell never thought of the quantifiers in his logic as being open
to different interpretations by a different choice of domain. While he does of
course recognize that in practice the use of the quantifiers may be restricted to a specific
domain—e.g. all numbers, or all material objects, or whatever—still he thinks that this
is not their ‘proper’ use. For in logic the quantifiers should always be construed as
ranging over everything whatever, and any intended restriction should be explicitly
expressed, e.g. as ‘for every x, if x is a material object, then . . . ’. Indeed, Russell goes
further than Frege on this point. For Frege always thought of his first-level quantifiers
as ranging over the total domain of ‘all objects’, but at the same time he also made use
of second-level quantifiers, governing variables which take the place of predicate letters

14
Such a development may be found in many textbooks, including my [1997: ch. 5].
14 L O G I C A N D M AT H E M AT I C S

rather than subject letters. And he insisted that these quantifiers have a different
interpretation: they range over ‘all concepts’ rather than ‘all objects’, and—in his
view—concepts are not objects. But in his early days (e.g. in his PoM and for some
time thereafter) Russell could not accept this distinction. His position, put in Fregean
language, was that every concept is an object, and so is anything else that we can talk
about. Everything whatever must be what Frege would call an object and what Russell
calls a ‘term’.15 In consequence, the quantifiers used in logic should always have the
same domain, namely the all-inclusive and maximum domain of everything whatever.
Quite apart from this matter of domains for the quantifiers, Russell did not think of
his logical formulae as containing schematic letters, open to any number of different
interpretations. On the contrary, his view is that in logic one makes genuine assertions,
that no genuine assertion can contain what is merely a schematic letter, and conse-
quently that these letters should always be understood as variables bound by the tacit
occurrence of an initial universal quantifier. (In Russell’s own terminology they are
merely ‘apparent’ variables, i.e. variables which are governed by a quantifier.) It is true
that in working out proofs it is convenient to treat them as ‘free’ variables, i.e. variables
that are not bound by any quantifier (which Russell calls ‘real’ variables). That is, they
are treated as if they stood for some definite proposition or subject or predicate, as the
case might be, but one which is left unspecified. However, when they are treated in
this way, the formulae of logic do not make any genuine assertions, which Russell
thinks that they should do. So his ‘language’ for logic is not thought of as a merely
schematic language, using schemata which propositions of a genuine language may or
may not exemplify. What looks like a schema must really be a universally quantified
assertion, with the quantifier left unexpressed.16
Because he does not think of the formulae of logic as open to different interpreta-
tions, he does not have our concept of validity as ‘truth in all interpretations’. He
simply has the concept of what counts as a truth of logic, and his investigations are
designed to provide a ‘foundation’ for such truths by providing the basic axioms and
rules of inference from which they can all be deduced. In modern terms, Russell was
quite familiar with the idea that logic provides proofs (which we now symbolize by ‘‘’),
but had only an intuitive and unworked-out idea of validity (which we symbolize by
‘⊨’). He welcomed Wittgenstein’s claim that the truths of logic should all be ‘tautol-
ogies’, but then found that he did not know how to define a tautology, and this was a
problem that he never resolved.17 His initial attitude to logic was that one could just
‘see’ that its basic rules and axioms are correct, and so could be confident in whatever
could be deduced from them. But, as we shall see, this attitude changed as a result of his

15
See e.g. PoM, }}47–9.
16
This makes it difficult to be clear about what is happening in a proof. Russell’s problems on this point
are discussed in Kemp [1998].
17
I return to it on pp. 69–72 below.
 E L E M E N TA RY L O G I C 15

continued wrestling with ‘the contradiction’, which from now on will be his major
problem in thinking about logic.
Before I come to this, I add just one further remark about the elementary logic that
has been the subject of this chapter. Neither Russell, nor anyone else who has been
interested in the ‘logicist’ claim about mathematics, can believe that the elementary logic
here described is all of logic. There must be more to the subject than this, if there is to
be any chance that it will imply all of mathematics. Frege had assumed a second-level
logic, in which quantifiers are allowed to bind variables which take the place of
predicate expressions, as well as variables which take the place of subject expressions.
He had also made use of the notion of a class, in a way that turned out to be a disaster.
In consequence, Russell was wary of just what more would be needed as logic, beside
what is now generally called ‘elementary’ logic. But he was always aware that more
would be needed.
 2
The Contradiction (i): The Problem

Mathematicians are used to thinking in terms of classes, or sets. At the time that we are
now concerned with, i.e. around 1900, there was no recognition that one might usefully
distinguish classes from sets, because two distinct notions are involved. Roughly
speaking, sets are to be conceived as ‘built out of their members’, along the lines first
suggested by Zermelo,1 whereas classes are conceived rather as the extensions of pre-
dicates, and therefore as given by our independent understanding of those predicates.
This distinction is a post-Russell development, and when we are thinking of Russell’s
problems we must try to set it aside. I shall therefore use the words ‘class’ and ‘set’
interchangeably in this book, and I shall use the usual current notation for sets, i.e.2
{a, b, c, . . . } for the set which has as members just a, b, c, . . .
{x: —x—} for the set which has as members just those
things x such that —x—.
Anyway, as I say, mathematicians are used to thinking in these terms, and at the time
when Russell was beginning to think about how mathematics might be ‘reduced’ to
logic this tendency had quite recently been reinforced. Aside from Peano, and Frege,
two mathematicians who had each been making important contributions were Dede-
kind and Cantor, and both of them thought in terms of classes (or sets). It will be useful
to begin this chapter with a brief description of their work, for Russell was much
influenced by it.

1 The development of mathematics

Dedekind had proposed in his [1872] a way of thinking of the real numbers, i.e. those
numbers that are expressed in our usual decimal notation only by unending decimals.
His basic thought was that a real number is determined by a ‘cut’ in the rational
numbers, i.e. a cut which separates all the rational numbers into two non-empty classes,

1
I shall describe Zermelo’s system, and the way that his thoughts have since been developed, in }5 of this
chapter.
2
Russell’s own notation for the second is ‘^x (—x—)’.
 THE CONTRADICTION ( I ): T H E P RO B L E M 17

those less than it and those greater than it.3 He demonstrated that the usual arithmetic
of the real numbers could be deduced, via this conception, from that of the rational
numbers, and—more importantly—that this conception also allowed one to prove, for
the first time, the existence of suitable real numbers to be the limits of various familiar
series of numbers. Russell took over this idea, but modified it. Instead of saying that
a real number is something (invented by us) that is ‘determined’ by a cut in the rational
numbers, why not just identify the real number and the cut? Or, to simplify things still
further, let us say that the real number is the lower half of the cut, i.e. the class of all the
rational numbers that are ordinarily thought of as less than it. That is: a real number just
is a class of rational numbers such that (i) at least one rational number is a member of it,
and (ii) at least one is not, and (iii) any rational less than some member of the class is also
a member of the class, and (iv) there is no greatest member of the class. Any class of
rational numbers that satisfies these four conditions will be the lower half of a Dedekind
cut, and if we simply identify the real numbers with such classes then a perfectly good
theory of real numbers will result. It was Russell himself who proposed the identifica-
tion, which may at first seem somewhat artificial, but which clearly does suit the logicist
approach. For the real numbers are then ‘constructed’, by purely logical means, from
the rational numbers, which are presumed to be already familiar. But it was Dedekind
who had shown that the resulting theory was entirely satisfactory from the mathemati-
cal point of view.4
For the next step we need the notion of an ordered pair, which we write as ‘<a,b>’
for the pair which has a as first member and b as second member. It was discovered by
Wiener [1912–14] how to represent an ordered pair in the vocabulary of ordinary set
theory, and his idea, as modified by Kuratowski [1920], has been used ever since.5 This
allows one to regard two-place relations simply as sets of ordered pairs. But Russell
took the opposite approach, with relations as primitive, and defined the ordered pair
<a,b> as the particular relation R such that, for all x,y, ‘xRy’ holds if and only if x ¼ a &
y ¼ b. In any case, once ordered pairs are available, it is easy to see how each rational
number n/m is determined by an ordered pair of integers <n,m>, and so may be
identified either with one such pair (e.g. the fraction ‘in its lowest terms’), or with
the set of all such ordered pairs that determine the same rational number. Similarly,
the signed numbers can be identified either with one particular pair of unsigned

3
There may also be a rational number which is neither less than it nor greater than it, but equal to it. This
is a trivial complication, which will very shortly fall away.
4
Of course, there are also other identifications which are equally satisfactory from this point of view. In
particular, I should mention that Cantor had offered an approach to the real numbers that was rather different
from Dedekind’s, but would have provided an equally effective identification. However Russell chose to
work from Dedekind’s basic idea.
5
What is now called the Wiener/Kuratowski definition is
<a,b> for {{a}, {a,b}}.
Many other definitions are available which have the same required consequence, namely
<a,b> ¼ <c,d> $ a ¼ c & b ¼ d
18 L O G I C A N D M AT H E M AT I C S

numbers—say ‘+n’ for ‘<0,n>‘, and ‘–n’ for ‘<n,0>’—or again with a class of them.
So, in general, all the familiar kinds of number can be ‘constructed’ as classes built up
in several stages from the natural numbers.6 What remains, then, is to show how the
natural numbers may in their turn be identified with classes of some other and familiar
things.
Here another work of Dedekind’s, i.e. his [1888], was helpful, but not yet sufficient
by itself. Dedekind had analysed the structure of the series of natural numbers: they
form a series of the kind nowadays called a ‘progression’, which means that there is a
first number, and for each number a next number, and no more than this requires. The
last condition can also be put by saying that each number can be reached by a
succession of steps, starting with the first number and going on step by step, with
each step taking one from the number last reached to the one next after it. Dedekind’s
important discovery was how to put this in purely logical terms, namely: whatever
is true of the first number, and is such that if it is true of any number then it is also true
of the next, must be true of all numbers.7 This is called the principle of mathematical
induction. It allows one to set down a simple set of axioms for the natural numbers, and
Dedekind in effect showed that these axioms form a sufficient basis for the deduction of
all the familiar properties of the natural numbers. Russell first learnt of these axioms
from Peano, and he therefore called them ‘Peano’s postulates’, which is how they have
been known ever since. But in fact Peano took them (with acknowledgement) from
Dedekind’s work. From Russell’s point of view, the important contribution made by
these postulates was that they showed just what conditions a suitable way of ‘con-
structing’ the natural numbers would have to satisfy, if it was to be adequate for all
mathematical purposes. But they did not by themselves show how to find any items
that did in fact satisfy them.
Here is where Cantor’s work was helpful. Cantor’s principal contribution was his
theory of infinite numbers [1895, 1897], which we shall come to shortly, but as a
preliminary he proposed a criterion of identity for all cardinal numbers,8 whether finite
or infinite, namely this: two classes have the same number of members, i.e. are
equinumerous, if and only if there exists a relation which correlates their members
one-to-one. (That is: each member of the first class bears the relation to one and only
one member of the second, and for each member of the second class there is one and
only one member of the first that bears the relation to it.) As with Dedekind on the real
numbers, Cantor did not himself suggest that this criterion provided an identification of
the cardinal numbers with any class of more familiar items, but Russell saw that it could
be used as such. His idea was that two classes share the same number if they share

6
Russell’s chosen way of doing this is most conveniently set out in his Introduction to Mathematical
Philosophy [1919a], henceforth IMP, chapter 7.
7
Frege’s Begriffsschrift [1879] contained the same discovery, so he has priority. But Dedekind’s work was
independent of Frege’s, and at the time was much better known.
8
The cardinal numbers are those that answer the question ‘how many?’ (A little thought will show that not
all the things that we call numbers do answer this question.)
 THE CONTRADICTION ( I ): T H E P RO B L E M 19

membership of the class of all classes with that number, so the number may simply be
identified with the class of all classes that have that number.9 That is to say: a number is
a class of classes, which (i) has at least one class as a member, and (ii) has as its other
members all and only the classes which are equinumerous with that one. Moreover,
equinumerousness can then be explained in terms of one-one correlations, as Cantor
did, and this is an explanation in purely logical terms. We therefore have a construction
of the cardinal numbers which apparently provides just what logicism needs.
The construction so far applies to all cardinal numbers, whether finite or infinite, but
the natural numbers are just the finite cardinal numbers, and so we now need a way of
distinguishing these. Dedekind’s work shows us how to do so: we need only define the
first number 0, and for each number n the next number n+1, and then his axioms will
do the rest. We have shown how to define 0: it is the class of all classes with 0 members,
i.e. with no members, and that is easily described.10 Moreover, since the number n is
the class of all n-membered classes, it is easy to see that the number n+1 will be the class
of all those classes that have one more member than do the classes that are members of
n. And a class a, has one more member than a class b if and only if, upon removing one
member of a, what is left is a class equinumerous with the class b. So we now have
definitions of 0 and of n+1, and we can define the finite cardinal numbers as those that
can be reached from 0 by a succession of steps from n to n+1, which is to say: those such
that whatever is true of 0, and is true of n+1 whenever it is true of n, must be true of
them. Then all that remains to be done is to show that the finite cardinals, so defined,
do satisfy Peano’s postulates. This requires us to prove just two more, quite simple,
theses: (i) that 0 is a number which does not come next after any (cardinal) number, and
(ii) that for every finite cardinal number there is one and only one (cardinal) number
that comes next after it. The first is in fact an immediate consequence of the definition
of a cardinal number, and the second looks as if it will follow from the theory of classes
which has been presupposed in all our thinking.11
This outlines the position as Russell saw it in 1900. It looked as if the theory of
classes would provide, first a construction of the natural numbers, and a demonstration
that they satisfy Peano’s postulates, and then further constructions from them of the
signed integers, the rational numbers, the real numbers, and any other entities of
traditional mathematics. This would establish the logicist claim that all of mathematics
could be reached from a starting point which comprised just pure logic and nothing
else. But of course the whole scheme, as here outlined, depends upon our being able to

9
Looking back on things in his much later work My Philosophical Development [1959], henceforth MPD,
Russell seems to claim on p. 70 to have discovered this definition for himself. But if he did mean to claim this,
then it is suggested by Rodrı́guez Consuegra [1987] that his memory was not entirely accurate. We need not
enter into this question. (As in note 7, Frege [1884] has the priority.)
10
Since classes are the same when their members are the same, there is just one class which has no
members, namely the null class. So 0 is the class which has just this class as a member and no other members.
11
In fact the second of these theses will give trouble. But we shall not come to that until Chapter 6, when
Russell’s final theory has been reached.
20 L O G I C A N D M AT H E M AT I C S

say that the theory of classes is itself to be counted simply as logic. In June 1901 Russell
discovered the paradox which is called after him, and which made this assumption
about the theory of classes extremely doubtful.

2 How the contradiction was discovered

He made his discovery while thinking of Cantor’s theory of infinite cardinal numbers, so
let us say a little more of this. Apparently there are classes with an infinite number of
members, for example the class of all the natural numbers. According to Cantor’s criterion,
any class which can be correlated one-to-one with this class will also have the same infinite
number of members as it does. Such a class is said to be ‘denumerable’, and both the finite
and the denumerable classes are said to be ‘countable’, for their members can be matched
one-to-one with some or all of the natural numbers, which is exactly what one does when
counting. Now it turns out that the class of all pairs of natural numbers is also denumera-
ble, and since this is a slightly surprising result I briefly indicate a proof of it. Consider the
pairs as written out in a two-dimensional and doubly infinite array thus:

<0,0> <0,1> <0,2> <0,3> <0,4> <0,5> ….

<1,0> <1,1> <1,2> <1,3> <1,4> ….

<2,0> <2,1> <2,2> <2,3> <2,4> ….

<3,0> <3,1> <3,2> <3,3> <3,4> ….

<4,0> <4,1> <4,2> <4,3> <4,4> ….

….

….

….

….

….

We reorder this as a singly infinite array, by taking it in the order indicated by the
arrows, namely
<0,0>, <0,1>, <1,0>, <2,0>, <1,1>, <0,2>, <0,3>, <1,2>, <2,1> . . . .

It is obvious that this array can be matched one-to-one with the natural numbers in
their natural order, and is therefore denumerable. It follows that the class of all signed
integers is denumerable, and so—more surprisingly—is the class of all rational numbers.
But the class of all real numbers is not denumerable, and this is a crucial theorem of
Cantor’s theory.
For simplicity, let us just consider the real numbers between 0 and 1, each repre-
sented as an unending decimal. To secure a unique representation, let us discard all
those which end with an infinite string of occurrences of the digit ‘0’. (That is, we
retain ‘0.4999 . . . ’ but discard ‘0.5000 . . . ’, for these are two different ways of writing
the same real number.) Then, if there were a correlation between these real numbers
and the natural numbers, we could think of it in this way. We list on the left each of the
 THE CONTRADICTION ( I ): T H E P RO B L E M 21

natural numbers, in their natural order, and opposite each we list on the right the real
number with which it is correlated, e.g. thus:
1 012345 ....
2 022345 ....
3 023345 ....
4 013445 ....
5. 034556
.. ....
.. .
Now, go down the diagonal of the array of digits on the right, after the decimal point,
and form a new decimal between 0 and 1 that is different from all those listed, e.g.
by writing a ‘2’ in its nth place if the nth digit of the nth real number is odd, and a ‘3’ if
it is even. So, with the present example, we form the decimal
023233 . . . .

This differs from each one in the list, because it differs from the nth one listed at its nth
place. So the supposed list of all the real numbers between 0 and 1 was not a complete
list, for this one was left out, and the argument shows that there could be no such
complete list. That is, there cannot be a one-to-one correlation between the natural
numbers and all the real numbers between 0 and 1.
This ‘diagonal’ method of argument can be generalized in various ways. Cantor
himself used it to prove the more general result that every class has more subclasses than
it has members. We may begin by picturing the proof in the same way as before.
Imagine that there is a correlation between the members of the class and its subclasses,
and think of the members as listed on the left and the correlated subclasses as listed on
the right. Write the subclass by first writing the whole class, and then crossing out the
members that are not in that subclass, e.g. thus:
a {a,6 b, 6 c, d, . . . .}
b {6 a, b, c, d, . . . .}
c {6 a,6 b, 6 c, d, . . . .}
..d ..{a, b, 6 c, 6 d, . . . .}
. .
Then form a new subclass, omitted from the correlation, by once more going down
the diagonal, and putting a member in the new subclass if and only if it is not in the
subclass with which it is correlated. So, in the present illustration, one forms the new
subclass
{6 a, 6 b, c, d, . . . }

This shows the general idea behind Cantor’s argument, but at the cost of supposing that
there is such a thing as a list of all the members of the class, which apparently assumes
that the class is countable. However this is needed only to help us to picture the
22 L O G I C A N D M AT H E M AT I C S

argument, and we can state the argument in quite general terms without any picturing.
In fact we can show that, however many members our class has, it must have more
subclasses than it has members, for the two cannot be correlated one-to-one. Assuming
any correlation between all the members and some of the subclasses, one finds that
there must be a subclass omitted from it, namely the subclass which has as members
all and only those members of the whole class that are not members of the subclass
with which they are correlated. This result, which is given in Cantor [1895] is often
called ‘Cantor’s theorem’. It shows that his infinite numbers never give out, i.e. that
for each one there is another that is greater than it. But it also led Russell to the
discovery of the contradiction that is called after him.
Russell began with the idea that there ought to be a greatest infinite number, namely
the number of all the things that there are. To put this in terms of classes, it would be
the number of the universal class, i.e. the class which has everything whatever as a
member. If there is such a class, then all of its subclasses should be members of it. So
there is a correlation which correlates all the subclasses with some of the members, and
the correlating relation is simply the one-to-one relation of identity. But Cantor’s
argument deduces a contradiction from this supposition, by showing that there is a
subclass which is omitted from the correlation. What is it? Well, the argument shows
that it is the class of all those classes that are not members of themselves. It follows that if
there does exist such a thing as the universal class then this supposed subclass of it
cannot exist. But once the point is noted one sees that there is a very simple argument
that goes directly to this conclusion, without the roundabout route through Cantor’s
diagonal argument. To suppose that there is a class w which contains as members all and
only those classes that are not members of themselves is to suppose12
9 w 8 x(x2w $ x 2
= x)

From this supposition, taking the particular case in which x is w, we deduce

9 w(w 2 w $ w 2
= w)

This is a contradiction. One can only infer that there is no such class as w is supposed to
be. So either there is something wrong with what is predicated of x in the formula
‘x 2
= x’, or this is a predicate for which there is no corresponding class. When Russell
paid serious attention to Frege’s logic in 1902, he saw that Frege’s system did yield this
contradiction, and wrote to him about it.13 At the time both of them thought that
there must be quite a simple correction that would avoid the contradiction but still
allow for almost all of the ‘new logic’ that they were each committed to. However
Russell quite soon came to see that the problem was serious.

12
‘2’ symbolizes the relation of membership, so that ‘x 2 y’ represents ‘x is a member of y’ (and ‘x 2 = y’
represents ‘x is not a member of y’).
13
Russell’s letter, and Frege’s reply, are conveniently translated in van Heijenoort [1967: 124–8]. One can
only assume that Frege himself had not read Cantor [1895] with sufficient attention.
 THE CONTRADICTION ( I ): T H E P RO B L E M 23

3 Some first reactions

His first and rather hasty reaction was a ‘theory of types’ that he sketched in an
appendix to PoM. The basic idea was to take as a starting point the things which are
not classes, which he here calls ‘individuals’. These are entities of type 0. Then there
are classes which have these individuals as members, and no others, and these are classes
of type 1. Then there are classes which have these classes of type 1 as members, and
these are classes of type 2. And so on upwards. This much is already part of his final
theory of types, which we shall come to later. But the early version in PoM also
contains two features which later disappeared. (i) The hierarchy of classes was not in
fact the strict hierarchy just described but rather a cumulative hierarchy, since Russell adds
that any two types have a sum, which is also a type, and which includes the classes of
both of those two types summed together. (ii) The PoM theory includes an infinite
type, above all the finite types (but, in contrast to the ZF theory of sets, which I shall
describe shortly, Russell has only one infinite type). His main reason at the time was the
case of the numbers. For, as we have seen, he wished to construe the number n as the
class of all classes which have just n members, and such classes may be of any type
whatever. So it appears that the number must be a class of a type that includes them all,
and this can only be an infinite type.
Russell quite soon abandoned this theory, though he never tells us why. The
difficulty over an infinite type is clear: although ‘x 2 x’ will never be true in any finite
type, still it may well be true in the infinite type, and so may ‘x 2 = x’, so the original
contradiction breaks out once more. One can block this by refusing to contemplate an
infinite type, and I imagine that Russell quite soon came to realize this, and so
abandoned infinite types for ever after. But then there is a problem. If we confine
our attention to the finite types, then the theory puts a severe restriction on what classes
there are, and (at present) there seems to be no good rationale for such a restriction. For
example, what is wrong with a class of all n-membered classes? It is obvious that the
theory outlined in PoM contains no answer. A further difficulty is that this theory is only
a theory of classes built from individuals, and has nothing to say about other kinds of
classes. But Russell has already seen that there is also an analogous problem over classes
of propositions. His discussion in PoM ends with the thought that for each class of
propositions there should be a further proposition, which states that all the members
of that class are true. But by Cantor’s theorem there must be more classes of proposi-
tions than there are propositions, and so this too is a contradiction.
In his further thought on this contradiction Russell quite soon came to see that it was
not specially concerned with classes. Just as there is a problem over the (supposed) class
of all classes that are not members of themselves, so there is exactly the same problem
over the (supposed) property of being a property which does not apply to itself. The
same problem even affects such ordinary entities as words and phrases. The linguistic
predicate ‘ . . . is not true of itself ’ (often abbreviated to ‘ . . . is heterological’) is easily
argued to be true of itself if and only if it is not true of itself. One might perhaps infer
24 L O G I C A N D M AT H E M AT I C S

that there is no such class as the one just mentioned, and that there is no analogous
property, but is it possible to conclude that the expression ‘ . . . is not true of itself ’ does
not exist? Have we not just shown, by quoting it, that it does exist? Nevertheless,
despite his awareness of these similar cases, Russell continued for a while to think of his
problem as one that concerned classes. That is why his next thoughts are presented as
thoughts on classes. He tells us of these in his article ‘On some difficulties in the Theory
of Transfinite Numbers and Order Types’ [1906a], where he first outlines what he calls
‘the zigzag theory’ and ‘the theory of limitation of size’. These are both theories about
what classes exist, which introduce no type distinctions between classes. But they do
claim that certain predicates do not define a corresponding class.

4 The zigzag theory

We can deal quite briefly with the first, the so-called ‘zigzag’ theory. This depends
upon a generalization which is intended to capture all cases where we know how to
produce a contradiction which shows that some described class cannot exist.14 Russell
suggests that in all of them there is a kind of ‘zigzagging’ feature, which he describes in
this way: we get a contradiction when we suppose that there is a class u such that (i) all
the members of u have a certain property , and (ii) there is a function f whose value for
that class as argument is again an object with the property , but one which is not a
member of u. For then either there will be no such thing as the class of everything which
has the property , or—if there is such a class—then the function f will have no value
for that class as argument (pp. 141–2). The idea is that in the construction of that class
there is a kind of ‘zigzag’ between the members of the class and the values of the
function f. But, to put the idea in much more general terms, it is this: a class will fail to
exist when the predicate which defines it has a certain kind of complexity, and nice
simple predicates will always succeed in defining classes. For example, there will be a
universal class of all objects, defined by a predicate such as ‘ . . . is identical with itself ’ or
‘ . . . is either red or not red’. But, unsurprisingly, Russell has to admit that he has not
found what seems to him to be a satisfactory way of distinguishing between the simple
predicates that do define classes and the ‘zigzaggy’ ones that do not.
However, there is a general objection to the idea that we can distinguish the
apparent classes that really do exist from the others that do not just by considering
the kind of predicates that define them. First, there will surely be infinitely many
(‘simple’) class descriptions that do succeed in defining classes. (For example, start with
any object a that you like and consider the unending series a, {a}, {{a}}, {{{a}}}, . . . .)
Hence, if every class is an individual there will be infinitely many individuals. By
Cantor’s theorem one then expects there to be uncountably many classes of indivi-
duals. But second, in any learnable language there cannot be more than countably

14
The generalization is reached by reflecting on the argument that leads to the Burali-Forti paradox, as
Russell says at [1906a: 141]. I give a brief statement of that paradox on p. 26 below.
 THE CONTRADICTION ( I ): T H E P RO B L E M 25

many predicates of any kind, whether those that do or those that do not define classes,
as Russell himself well knew [1906b: 184–5]. This is because, if the language is to be
learnable, it must start from a finite basis, e.g. of letters, or phonemes, or words, or
what you will. Further, any expression of the language will have only a finite length.
It follows that the expressions of that language cannot be more than countably many.
Hence only countably many of the candidates for being classes can have their case
determined in this way, i.e. by the kind of predicate that purports to define them, and
this leaves uncountably many still undecided. Some supplementary criterion is there-
fore needed, and has yet to be supplied. Russell’s discussion seems to suggest that all
classes without a definition should be ruled out. At any rate he says that every existing
class must be defined by a ‘norm’, i.e. a propositional function, and that a propositional
function is an expression of a certain kind.15 This is a problem that will come up once
more in Chapter 6, for Russell’s final theory seems to be liable to a similar problem.
But, by the time that we reach that theory, the idea of zigzagginess will have been
dropped and replaced by something altogether simpler.
With that we may leave the ‘zigzag’ theory. It is sometimes said that this theory has
some resemblance to the theory which Quine proposed in his ‘New Foundations for
Mathematical Logic’ [1937]. I make no comment on that, except to remark—as is now
well known—that Quine’s theory would seem to be a very peculiar theory, and one
which appears to have no natural model.16 Before I can treat adequately of Russell’s
second suggestion, a theory of ‘limitation of size’, I must introduce more background
information on Cantor’s theory of infinite numbers, and in particular his theory of the
infinite ordinal numbers.

5 Digression: infinite ordinals and ZF set theory

The natural numbers 1, 2, 3, 4, and so on, may be used both for counting and for
numbering off the members of an ordered series as the first, second, third, and so on of
that series. In the first use they are called cardinal numbers, and in the second use ordinal
numbers. A series which matches the series of all natural numbers, in their natural order,
so that each member can be labelled as ‘the nth member’ for some natural number n, is
the shortest kind of infinite series. It is said to have length o. But, given any such series,
we can always form a longer one by adding a further object at the end, and this new
series is said to have length o+1. A series of length o has no last member, but a series of
length o+1 does have a last member, and this is a member with no immediate
predecessor. Then we can add another object to give us a series of length o+2, and
so on. By continuing such additions one forms a series of length o+o, which has a first

15
[1906a: 136]. Russell takes over the word ‘norm’ from the person with whom he is disputing in that
article (i.e. Dr E. W. Hobson). It is not his usual word, and he does not employ it elsewhere.
16
For useful surveys of investigations into Quine’s NF, and its descendant ML, see Quine [1963: ch. 13],
and Hatcher [1968: ch. 7].
26 L O G I C A N D M AT H E M AT I C S

part of length o followed by a second part of length o. An example would be the series
of all the even numbers followed by the series of all the odd numbers. This length is also
called o2. Clearly we can progress in the same way to o3, o4, and so on, and then to
oo, which is also called o2. These numbers which are used to number off the members
of a series, and to measure its length, are the infinite ordinal numbers, and a sketch of
how the series of all infinite ordinal numbers begins is this:
o, o+1, o+2 . . . . o+o (=o2)
o2, o3, o4 . . . . oo (=o2)
o2, o3, o4, . . . . oo
ω
ω
ω ωω
, .... ∈0
17
ωω, ω , ω

All of these give the lengths of countable series, i.e. series with no more members than
there are natural numbers, so examples can be found just by rearranging the natural
numbers in an unfamiliar order. Cantor’s theory also introduces uncountable ordinal
numbers, but we shall not need to go into any detail about these, save to mention one
problem that concerns the series of all the ordinal numbers. This is the paradox
discovered by Burali-Forti [1897], which antedates Russell’s discovery of his paradox,
and was the first indication of the problems lurking in Cantor’s theory.
The series which are measured by the ordinal numbers are called ‘well-ordered’
series. The distinctive character of such series is that although they may contain several
infinite ‘ascents’ they cannot contain an infinite ‘descent’. That is, if you start from any
member of the series and go back from there to an earlier member, and then back again
to a still earlier member, and so on, then your ‘descent’ must end, by reaching the
bottom, after only finitely many steps. Another way of putting the same condition is
this: if some member of a well-ordered series has a certain property then there must be
an earliest member that has that property, i.e. there must be a point in the series where
that property first occurs.18 Now the idea is that any well-ordered series will have a
length that is given by an ordinal number, for it will be possible to match the members
of the series one by one with the ordinal numbers, and then the length of the whole
series will be given by the ordinal that is next greater than all of those used in the
matching. But this idea apparently leads to a contradiction, for the series of all the
ordinal numbers is itself a well-ordered series, and so should have an ordinal number.
But that ordinal will have to be greater than all the ordinals in the series, and so (inter
alia) greater than itself, which is evidently impossible. So one must apparently infer that
there is no ordinal number which measures the length of this series. Why not? Well,
quite a plausible suggestion is that the series of all the ordinal numbers is too long to be

17
20 is the first ordinal number 2 such that o2 = 2.
18
Notice, incidentally, that the familiar series of the signed integers is not in this sense well ordered. Nor is
any stretch of the series of rational numbers.
 THE CONTRADICTION ( I ): T H E P RO B L E M 27

measured by an ordinal number. As we know from his correspondence, Cantor himself

had pondered a somewhat similar suggestion about infinite classes, namely that some of
them might be too large for their size to be measured by a cardinal number. Neither he nor
anyone else had put such a suggestion into print at this time, but it was quite a natural one,
and it is not particularly surprising that Russell should have explored it, as his next
attempt to meet the problem posed by his contradiction. What is perhaps surprising is
that he should have come so near to the basic ideas that we now see as underlying the set
theory that is widely accepted today, namely the ZF theory, which can certainly be
described as a theory of ‘limitation of size’. It will be helpful if I give a brief outline.
The theory was first presented in Zermelo [1908], i.e. two years after Russell’s paper
of 1906, and initially it did not seem to have much to do with the idea of size
limitation. But its important connection with the ordinal numbers became explicit in
Mirimanoff [1917], and the consequence for size was added as a basic axiom by
Fraenkel [1922]. As we now understand it, the theory is best seen as resulting from a
conception of sets as constructed in stages, with the sets at each stage having as members
only the sets (or other entities19) that already existed at earlier stages. The stages are well
ordered, and so can be indexed by the ordinal numbers, and there are to be as many of
them as there are ordinal numbers. The important constraint is that every set has to
belong to some stage of construction, i.e. the stage at which it is first constructed, and
the ordinal of that stage is also called the ‘rank’ of the set. To illustrate, no set is a
member of itself, because the members of a set have to be of lower ranks than is the set
itself. So if one tried to form a set of all those sets that are not members of themselves
this would have to be a universal set, which had all sets as members. But there cannot be
such a set, because it could not have a definite rank, i.e. there could be no stage at
which it was constructed. For every stage is followed by a further stage, which contains
new sets that cannot be members of sets formed at earlier stages. One must add that this
talk of sets being ‘constructed in stages’ is of course a metaphor, and we are not to
suppose that sets are literally created at earlier or later times. But it is a metaphor which
is very helpful when trying to picture the situation.
Finally, let us come explicitly to the idea of limitation of size. In his [1922] Fraenkel
proposed a new axiom, called the axiom of replacement, which directly enforces this
idea. It stipulates that, if the members of a collection can be correlated one-to-one with
some or all of the members of a known set, so that the collection is no bigger than the
known set, then that collection is itself a set. The rationale is that the correlation
between the members of the known set and the proposed one would yield a correla-
tion between the ordinals that are the ranks of those members, and then since the ranks
of the members of the known set are all less than some ordinal, namely the rank of the

19
One may begin with some individuals which are not sets, and then the first stage of set construction has
sets which contain them as members. But we get a simpler theory if we begin with nothing at all, so that at the
first stage we can form only one set, i.e. the null set, which has no members. (This is known as the theory of
pure sets, i.e. sets built up from nothing but sets.)
28 L O G I C A N D M AT H E M AT I C S

set itself, the correlated ordinals will also be less than some ordinal, which will be
the rank of the proposed set. With Fraenkel’s axiom added, we have what is now called
the Zermelo-Fraenkel (i.e. ZF) theory of sets, which is widely accepted today. To
elaborate Fraenkel’s idea further, one may add that von Neumann in his [1925] and his
[1929] made a further proposal, namely that there is just one size that is too big for there
to be a set of that size, namely the size of the collection of all sets altogether. He offered
various arguments (which I shall not discuss) to show that this is a reasonable way of
making Fraenkel’s proposed axiom more precise, and that it conforms very nicely to
the idea that sets are to be thought of as constructed in a well-ordered series of stages.
Let us now return to Russell in 1906.

6 Russell on limitation of size

When Russell comes to describe the idea that some purported classes would be too big to
exist, he at once connects this with the theory of ordinals. As he says, the Burali-Forti
paradox can certainly be taken as suggesting that a class of all ordinal numbers would be
too big to exist, and this is the starting point of the idea. He adds that any other purported
class, which would have as many members as there are ordinal numbers, would also be
too big. Taking on board Zermelo’s axiom of choice,20 and its consequence that every
class can be well ordered, he therefore proposes as the foundation of the theory that a
class exists if and only if its members can be correlated one-to-one with some proper
initial segment of the series of all ordinal numbers, i.e. with the ordinals less than some
given ordinal. This idea is only slightly different from today’s thought that a class exists if
and only if it has some ordinal number as its rank, to mark the stage at which it is first
constructed. In each case it is the ordinal numbers that are being invoked to determine
what is to count as ‘too big’, and in fact the two suggestions are provably equivalent in
the presence of the axiom of choice. So, since Russell’s proposal in this case is so near to
the theory that is generally accepted these days, one asks why he feels able to reject it in
little more than a page. I think that there are two things to be said in explanation, though
neither of them is actually said in Russell’s own discussion.
The only point that he makes himself is that the proposal is not very helpful until we
know
how far up the series of ordinals it is legitimate to go. It might happen that o was already
illegitimate . . . or it might happen that o2 was illegitimate, or oo or o1 or any other ordinal
having no immediate predecessor. We need further axioms before we can tell where the series
begins to be illegitimate (p. 153, my emphasis).21

20
Zermelo [1904]. I shall discuss this axiom, and its significance, in section 2 of Chapter 6.
21
By ‘illegitimate’ Russell presumably means ‘non-existent’. Compare an earlier statement: ‘At present it
is not easy to see where this series [sc. of ordinals] begins to be non-existent, if such a bull may be permitted’
(p. 144).
 THE CONTRADICTION ( I ): T H E P RO B L E M 29

From today’s perspective one replies that that is exactly what the task is, namely to
provide further axioms. Zermelo began this task by introducing axioms of pairing and
union, which provide the finite ordinals, an axiom of infinity which posits the first
infinite ordinal, and a power set axiom that then takes us yet further. Fraenkel added to
these his axiom of replacement, which certainly ‘legitimized’ o2, oo, and many more.
In more recent times we have become familiar with yet more axioms of the same kind,
postulating the existence of hitherto ‘inaccessible’ ordinals. But it would appear that
Russell himself never gave any attention to this task of providing suitable existence
axioms, even though he had himself suggested exactly that project. Why so?
One reason that was probably influential is that he was still pursuing his goal of
showing that all mathematics is really just logic. But logic is by tradition supposed to be
a completely general subject, making no particular assumptions about the existence of
this or that particular kind of thing. As Russell was aware, it cannot altogether avoid all
existence claims. In the third theory sketched in this paper of 1906 it assumes the
existence of propositions, and in Russell’s final theory it assumes the existence of
propositional functions, but these assumptions are supposed to be of an entirely general
nature. However the various special assumptions of set existence that are made in the
ZF axioms are quite different in style, and in fact, as soon as it became accepted that a
good set theory would have to have such axioms, the idea that set theory counted as
a part of ‘logic’ was very generally rejected. Since what Russell wanted was an
acceptable logic, it is not altogether surprising that he did not pursue this line of thought.
I think that there was probably another reason at work, though again it is not
avowed in the article in question. Russell was aware that the contradiction which
affected the naı̈ve view of classes was equally a contradiction for the naı̈ve view of
properties and relations, of propositional functions, and even of predicates taken as
linguistic expressions. Precisely the same contradiction turns up in each case.22 Now
with classes the ‘limitation of size’ approach is in a way attractive, just because we do
quite naturally think of classes as ‘constructed out of ’ their members, and so the idea
that a class cannot exist until a stage ‘after’ all its members exist does have some appeal.
But there is no such appeal in the other cases. Properties are not ‘constructed from’ the
objects that have them; nor are propositional functions ‘constructed from’ those of
their arguments that happen to yield true propositions as values; nor does a linguistic
predicate depend for its existence on the items that it happens to be true of. It strikes us
as weird to hold that such linguistic predicates as ‘ . . . is the same as itself ’ or ‘ . . . is
either red or not red’ cannot exist, because if they did they would be true of too many
things. In these cases the idea of ‘limitation of size’ seems entirely inappropriate. But
Russell always wanted some general solution to these vexing paradoxes, so this is again a
good reason for him not taking this proposal very seriously.

22
This point is made explicitly at several places in his [1906a], namely pp. 137, 140, 145n., 154n.
30 L O G I C A N D M AT H E M AT I C S

This same paper [1906a] also offers a third theory, which Russell calls the ‘no-classes’
theory. He indicates in this paper that it is his preferred option, and in a paper written
soon after [1906b] he offers a fuller explanation and an even more enthusiastic
endorsement. But to understand this we must first digress to consider his well-
known article ‘On Denoting’ [1905b]. For he says himself that it was this article on
denoting that provided the breakthrough for his hitherto frustrated attempts on the
contradiction.23

23
MPD, p. 79. Cf. letter to Jourdain of March 1906, published in Grattan-Guinness [1977: 79].
 3
On Denoting

1 The early theory of denoting

To appreciate the value which Russell himself placed upon this well-known article, we
must first go back to his earlier theory of denoting, which was published as chapter V of
his PoM. The theory concerns what he calls ‘denoting phrases’, and these consist of a
common noun or noun phrase (such as ‘man’ or ‘man who broke the bank at Monte
Carlo’) preceded by what is now called a ‘determiner’. Russell mentions in particular
phrases of the form ‘all A ’s’, ‘every A ’, ‘any A ’, ‘some A ’, ‘an A ’, ‘the A ’; and his
theory is a theory of how these phrases function.1 The main feature of the theory is, as
the name implies, that they function by denoting, but why should he think this?
The discussion in PoM is preceded by this general principle:
In the present chapter, certain questions are to be discussed belonging to what may be called
philosophical grammar. The study of grammar, in my opinion, is capable of throwing far more
light on philosophical questions than is commonly supposed by philosophers. On the whole,
grammar seems to me to bring us much nearer to a correct logic than the current opinions of
philosophers; and in what follows, grammar, though not our master, will yet be taken as our
guide. (PoM, }46)
It seems to me very likely that Russell is being guided by ordinary English grammar
when he thinks that the role of these phrases is to denote. On the usual understanding,
the paradigmatic kind of denoting expression is a proper name. No one would say that
the phrases which Russell calls ‘denoting phrases’ are proper names, but English
grammar does treat them in much the same way. Setting aside some idioms which
clearly have a special explanation, it is very largely true that wherever grammar allows
you to write a proper name you can also write any of these phrases, and vice versa.2

1
I shall continue to speak neutrally of denoting phrases. Russell usually speaks of the denoting concepts
which (in PoM ) he takes these phrases to express.
2
Russell himself later suggests that it is grammar that has led those whom he calls ‘traditional logicians’ to
treat ‘I met Jones’ and ‘I met a man’ as having the same form (IMP, p. 168). A number of apparent exceptions
to the supposed interchangeability of Russell’s denoting phrases with one another, and with proper names,
are given in Oliver [1999], but in most cases they are quite easily explained away. One point that Russell
notes himself is that ‘all A ’s’ is a plural expression, taking a plural verb, whereas his other phrases are (usually)
singular. Nowadays we all think of this as a trivial point of no significance, but in PoM Russell did not.
32 L O G I C A N D M AT H E M AT I C S

That is surely why Russell thinks of them as denoting. But his theory of how they
denote does distinguish them from proper names, since his view (in PoM ) is that a
name denotes directly, so that the object which is named is itself a constituent of the
proposition that the name is used to express, whereas the other ‘denoting phrases’
denote indirectly, by expressing a denoting concept which then denotes some further
object or objects. So in this case the propositions expressed contain as constituents not
the objects denoted but the concepts that denote them.3
What, then, are the objects that these phrases allegedly denote? Russell thinks of ‘all
A ’s’ as denoting a class, construed as a plural object, and I postpone discussion of this
until the next chapter. He thinks of ‘every A ’, ‘any A ’, ‘some A ’, and ‘an A ’ as in a way
denoting all the things that are A, but each in a different way, for it is clear that
substituting one of these phrases for another can make a difference to what is said.
For example, these two sentences do not mean the same as one another:
I did not kiss every girl there
I did not kiss any girl there
In each of them the denoting phrase refers somehow to all the girls there, but there
must be some difference between ‘every’ and ‘any’ which explains why the two
sentences do not mean the same. Russell’s first thought is that each of them denotes
a combination of all the girls—in fact, a combination formed by ‘and’—but it must be a
different combination in each case. His way of distinguishing between these suppo-
sedly different ‘combinations’ is something that we need not investigate here.4
Unfortunately there is an aspect of this early theory in PoM, which is relevant to the
later theory of ‘On Denoting’ [1905b] (hereafter OD ), but is open to serious doubt:
was Russell assuming in PoM that each of his denoting phrases did succeed in denoting
something? Much later, when looking back on his early days in MPD [1959], he
implies that this was his assumption at the time:
If you say that the golden mountain does not exist, it is obvious that there is something that you
are saying does not exist—namely the golden mountain; therefore the golden mountain must
subsist in some shadowy Platonic world of being, for otherwise your statement that the golden
mountain does not exist would have no meaning. I confess that, until I hit upon the theory of
descriptions, this argument seemed to me convincing. (p. 84)

The position described here is in effect Meinong’s position. I note that it is in the first
place a position concerning denoting phrases which begin with the definite article
‘the’, i.e. those which are nowadays called definite descriptions; it is not obvious

3
This is offered as an explanation of how we can understand a proposition which is about infinitely many
objects: we do not need to grasp each of those many objects, but only the one denoting concept that denotes
them all (}60, }72, }141).
4
The detail of Russell’s theory is frequently obscure. It is discussed in my [2009b]. What is said here
represents what Dau [1986] fairly calls the ‘official’ version of Russell’s theory, but there are also other
versions to be found in the PoM account.
 ON DENOTING 33

whether Russell intended it to apply to other denoting phrases too. I add that when
Russell refers to the time when he ‘hit upon the theory of descriptions’, he need not be
intending to refer precisely to the date (October 1905) when the paper OD was
published. We do now have a working paper of his (called ‘On Fundamentals’5 )
which is dated ‘begun 7th June 1905’, and which shows him struggling towards the
theory that was published in OD. It is therefore not at all surprising that in a short
review which he published in July 1905 6 we find several features which anticipate
the OD theory, including both a Fregean treatment of ‘all’ and ‘some’, and a clear
statement that a definite description beginning with ‘the’ may not denote anything—
neither anything that exists nor anything that subsists (i.e. ‘has being’).7 These anticipa-
tions of OD do not seriously conflict with Russell’s statement in MPD that ‘until he
hit upon the theory of descriptions’ he found Meinong’s argument convincing. But
is he misremembering?8
There are several passages in PoM which do seem to show an agreement with
Meinong, for example this one:
Being is that which belongs to every conceivable term, to every possible object of thought—in
short to everything that can possibly occur in any proposition, true or false . . . “A is not” must
always be either false or meaningless. For if A were nothing it could not be said not to be; “A is
not” implies that there is a term A whose being is denied, and hence that A is . . . Numbers, the
Homeric gods, relations, chimeras and four-dimensional spaces all have being, for if they were
not entities of a kind, we could make no propositions about them. (}427. Cf. }}47–8.)

On the other hand the view that denoting phrases may not succeed in denoting
anything is quite compatible with the overall position of PoM, since PoM does assign
a meaning to such phrases, which is put together from the meanings of the noun phrase
in question and the preceding determiner. They would retain this meaning even if they
failed to denote anything. Moreover, a proposition that is expressed by means of a
denoting phrase contains only the meaning of this phrase as a constituent, and not the
object(s) that it denotes. That would apparently allow us to make propositions about
non-existent things without assuming that they must have some kind of being. And we
do also have passages in PoM that seem to endorse this possible lack of denotation, e.g.

5
This is now published in his Collected Papers, vol. 4 [1994: 359–413].
6
[1905a]. This is a review of a paper by H. MacColl, and is entitled ‘The Existential Import of
Propositions’.
7
In Russell’s vocabulary, being is equated with subsistence, which is a more general notion than
existence. Concrete objects, existing in space and time, are said to exist; abstract objects (such as numbers)
do not exist but they do subsist. That is, both kinds of objects have being. It is not clear how the golden
mountain, which would be a concrete object and not an abstract object, can be said to subsist, though it is
clear that it does not in fact exist.
8
The question here at issue was first raised in a helpful way in Hylton [1990: ch. 6]. Hylton holds to the
traditional view that in PoM Russell was still an adherent of Meinong’s doctrine, and that he changed his
mind somewhere between PoM and OD. It has been discussed by others since, notably Griffin [1996], Makin
[2000: ch. 3] and more recently [2009], and Levine [2001]. Makin argues that Russell was never a Meinongian
over denoting phrases, though he was over proper names, until he came to see that—in doubtful cases—they
could be viewed as covert definite descriptions. Stevens [2010] takes issue with Makin.
34 L O G I C A N D M AT H E M AT I C S

It is necessary to realise that a concept may denote although it does not denote anything . . . All
denoting concepts, as we saw, are derived from class-concepts; and A is a class-concept when
‘x is an A ’ is a propositional function. The denoting concepts associated with A will not denote
anything when and only when ‘x is an A ’ is false for all values of x. (}73)

But Russell also recognizes in this section of PoM that there is a problem over the
supposed propositions that can be expressed by means of these empty denoting
concepts. He has begun by saying that all such propositions seem to be false, but at
once qualifies this idea by calling it ‘only a first step’, and not by itself ‘an adequate
explanation’. The ensuing discussion uses the example ‘chimaeras are animals’ and it
assumes (contrary to }427, cited above) that chimaeras neither exist nor subsist. But this
leads it to conclude that ‘it seems most correct to reject the proposition altogether’, and
this apparently means that the sentence (if taken literally, and not rewritten in some
more tractable way) should be regarded as not expressing a proposition at all.9 One can
only conclude that if, as here, the Meinongian position on non-existents is put aside,
still PoM can hardly offer a satisfactory view of sentences which appear to be about
them.10 However, the position of OD on this question is perfectly clear, so let us now
turn to this.

2 The theory of 1905

For the overall development of Russell’s thought the main importance of OD is that it
is where Russell decisively breaks away from his original respect for English surface
grammar. Wittgenstein is evidently thinking of OD when he says in his Tractatus Logico-
Philosophicus [1921], henceforth TLP,
It was Russell who performed the service of showing that the apparent logical form of a
proposition need not be its real one. (TLP 4.0031)

For OD rejects all of Russell’s early theory of denoting in PoM. It is true that the only
case which it discusses in any detail is that of ‘the A ’, but the opening pages show that
Russell is now putting forward a quite different theory of all his other denoting phrases
too. Thus ‘every man’ and ‘all men’ are now identified, and both are taken to be short
for ‘8x (if x is human, then . . . )’. Similarly ‘a man’ and ‘some man’ are now identified,
and both are taken to be short for ‘9x (x is human, and . . . )’. I add that ‘no man’ is

9
An unpublished mss of 1903, i.e. written shortly after PoM was completed, suggests that ‘we shall have
to say that “the present king of France is bald” is neither true nor false’ (Collected Papers, vol. 4 [1994: 286]).
But this mss apparently accepts that it is a proposition, though it lacks a truthvalue, whereas }73 of PoM claims
that it is not a proposition at all.
10
As the passage continues we find ‘Nothing is a denoting concept, which denotes nothing’. Neither the
word ‘nothing’ nor the similar phrases ‘no A ’ were originally listed as denoting phrases. It is surely very odd to
suppose that they do denote certain special combinations, either of all things or of all A ’s, formed presumably
by the combining notion ‘neither . . . nor . . . nor . . . ’. But this is what the theory of chapter V would seem to
require. However it is even worse to say that, since these are denoting phrases which denote nothing, the
sentences that contain them express no propositions at all. This point is made by Oliver [1999: 264].
 ON DENOTING 35

now included as a similar phrase, and this is no longer any kind of problem, as it was
in PoM.11
What one misses in OD is any kind of argument for the new position on these
phrases, and any account of what was wrong with the old one. There is not even an
explanation of how it can make a difference whether we say ‘every’ or ‘any’, though
this was a central feature of the PoM theory. We can be sure that Russell would now
explain the difference as we do, i.e. as a difference in the scopes of what is actually the
same quantifier. That is, on the new theory ‘I did not kiss any girl there’ and ‘I did not
kiss every girl there’ are respectively to be analysed as
8x (x was a girl there ! ~(I kissed x ))
~8x (x was a girl there ! I kissed x )
What changes from one sentence to the other are the relative scopes of ‘8’ and ‘~’, but
there is no difference in what is ‘denoted’. However it must be admitted that in OD
Russell never says this, and never even hints at it. Presumably the explanation is that
this theory of the quantifiers has been so much a part and parcel of Russell’s own
thinking, ever since his detailed study of Frege in 1903, that he has now forgotten that
it still needs to be explained to others. But in any case, whatever the reason is, one has
to say that OD contains no proper account of what the new theory of the quantifiers
now is, or of what was wrong with the old one.
But it is a radically new theory, and more so than Russell’s own terminology
suggests.12 He opens the article by saying that it is a theory of ‘denoting phrases’, and
he explains what these phrases are by giving the same sort of examples as before.
He adds that ‘a phrase is denoting solely in virtue of its form’, and evidently means by
this its grammatical form. One supposes that he intends this to allow for the fact stated
in his next sentence, that some denoting phrases do not denote anything, e.g. ‘the
present king of France’. For he goes on to state that in other cases the phrases do denote.
But this is not what his theory now is. The theory is much better described as the
theory that these so-called denoting phrases never denote, for that is not what their role
is. Unfortunately, he does not say this. What he says instead is that these phrases ‘do not
have meaning in isolation’, and the intended contrast is with a word that genuinely
does denote, such as a proper name. This does ‘have meaning in isolation’, for its
meaning is just the object that it denotes.
Let us now come to the subject with which OD is mainly concerned, namely
definite descriptions of the form ‘the A ’. In PoM he had treated all his so-called
denoting phrases as having the same role, namely to denote. In OD he still treats

11
See the previous note.
12
In PoM Russell is aware that English sentences containing his denoting phrases are equivalent to those
containing unrestricted quantifiers and bound variables that in OD he gives as their analyses. But he there
claims that they do not mean the same, and so the one cannot be analysed as the other. See in particular
PoM, }89.
36 L O G I C A N D M AT H E M AT I C S

all these phrases in the same way, and the new theory is that none of them work
by denoting; instead they are all complex quantifying expressions. The obvious
opponent here is Frege. For Russell has now adopted Frege’s way of handling ‘all’,
‘every’, ‘any’, ‘a’, and ‘some’, but Frege had treated ‘the’ differently. He had regarded
‘the A ’ as a (complex) name, and not a quantifying phrase. Russell’s main claim in
OD is that the same treatment does extend to ‘the’. Just as ‘An A is B ’ is now to be
analysed as
9x(x is A & x is B )

so also ‘The A is B ’ is to be analysed as

9x(x is A & 8y(y is A ! y=x ) & x is B )

The definite article ‘the’ is thus treated in the same way as the indefinite article ‘a’,
except that it contains a uniqueness claim as well as an existential claim. But in this case
Russell does argue quite extensively for his position. Most of his arguments, but not
all, concern the case when the definite description is not (uniquely) satisfied. This
applies to both of his opening arguments, first against Meinong (p. 45), and then against
Frege (pp. 45–7).

3. Negative arguments: Against Meinong and Frege

Meinong had held that all definite descriptions do refer, and refer to an object that
does satisfy the description in question. However, this object may be one that does
not exist, for existing (sein ) is one thing, and satisfying a description (sosein ) is
another. An object may lack the first while still possessing the second. (I make no
attempt to explain why Meinong held this view.13 ) Russell very fairly objects that
this theory infringes the law of non-contradiction. According to Meinong ‘the
golden mountain’ refers to an object that is golden, and is a mountain, but does
not exist. Similarly ‘the round square’ refers to an object that is round and is square,
and so (of necessity) does not exist. But we have only to build existence into the
description and we have an explicit contradiction. For the description ‘the golden
mountain that does exist’ will now refer to an object that is golden, and is a
mountain, and does exist, but also does not exist. Russell protests that a theory
according to which there are things that both do and do not exist should be rejected,
unless there is absolutely no alternative. In my opinion this is a very strong objection,
and it is not worth trying to rescue Meinong from it.14

13
It is not irrelevant that Meinong studied under Brentano, and that Brentano claimed that it was peculiar
to thoughts that they always have an ‘intentionality’. This means that they are always directed towards some
object, though that object may be one that does not exist.
14
Meinong’s theory was declared dead and buried by Ryle [1972: 7], but by that time serious attempts to
resurrect it had already begun, for example in Routley’s work of the early 1960s. But since Routley’s book
 ON DENOTING 37

Frege’s position is more complicated. For mathematical purposes he recommends a

language in which a definite description ‘the A ’ always has a reference, but not of the
Meinongian sort. If there is one and only one thing that is A, then it refers to that thing;
but if there is not then we must supply something else for it to refer to, and Frege’s
suggestion is that we take it to refer to the class of all A ’s. If there are no A ’s, this will be
the empty class; if there are several A ’s then Frege assumes that they will always form a
non-empty class, and Russell does not here question that assumption.15 But he does
very fairly state that this idea is ‘plainly artificial’, and surely Frege would not deny this.
It is not a proposal for how to construe our existing language, but a proposal to move
away from it towards a better language.
As for our present language, Frege thinks that names, predicates, and sentences will
normally have both a sense and a reference (and, as I have noted, he counts a definite
description as a kind of name). Normally, a name will refer to an object, a predicate will
refer to what Frege calls a concept, and a sentence—according to Frege—will refer to
whichever truthvalue it has. But one of Frege’s leading principles is that the reference
of a complex expression depends upon the references of its simpler components.
It follows that if a definite description ‘the A ’ lacks a reference, then any sentence
that contains it will also lack a reference, i.e. it will have no truthvalue. And this is
what he thinks does happen in our ordinary language. Even though all the expressions
in question may have a perfectly good sense, still some of them may turn out to lack
a reference, with the consequence just described.
There is a very obvious and simple objection to this theory, and it is surprising that
Russell does not at once propose it when discussing Frege’s theory. Consider explicitly
existential statements of the forms ‘the A exists’ and ‘the A does not exist’. (In more
idiomatic language these are ‘there is (or: is not) such a thing as the A ’.) When the
description ‘the A ’ is empty, e.g. when nothing is A, then the first of these sentences
should be false and the second should be true. But on Frege’s theory both of them must
lack a truthvalue, and that is surely an absurd result.16 Although Russell does not
actually say this, he does say something more complex which has the same moral.
Consider the sentence ‘If one and only one thing is A, then the A is A ’.17 Russell
claims that this should always be true, both when there is one and only one thing that is
A, and when there is not. He then objects that on Frege’s theory it cannot be true in
the second case, and this objection is similar to the one that I have just given. But (a) the

[1980] was not published until some time later, priority must be assigned to Findlay [1963]. The most recent
extended defence of Meinong that I know is Jacquette [2009]. (But, against him, see also Griffin [2009].)
15
The assumption must appear very doubtful when one sees how it leads straight to Russell’s paradox.
But that point is hardly important here.
16
Perhaps Russell deliberately avoids this objection here because he is reserving the case of explicitly
existential statements for his puzzle (iii), discussed below.
17
For simplicity, this makes a quite unimportant alteration to Russell’s own wording.
38 L O G I C A N D M AT H E M AT I C S

example is more complex than is needed, which leaves it open to a possible response,18
and (b) Russell carelessly mis-states it. He claims that when there is nothing that is A
then the sentence must for Frege be ‘nonsense’, and Frege should reply that that is not
so; on his theory it has a perfectly good sense, but is lacking in truthvalue. Similarly
Russell complains that ‘The King of France is bald’ is in fact ‘plainly false’, though on
the theory that he is considering it ‘ought to be nonsense’. Frege has the same reply.19
But those mis-statements can be corrected, and then the objections to Frege are
entirely effective. I do not see how he could adequately reply, especially to the
objection that was put first, concerning explicitly existential statements. Perhaps the
explanation is that he was always more concerned to improve on our ordinary
language than to explain how it actually functions. But still, his untenable claim that
an empty name must always lead to a lack of truthvalue does stem from a general
principle which he accepts and regards as important. Besides, he does attempt to defend
this claim as a correct account of our ordinary language when discussing the name
‘Odysseus’ [1892: 62–3]. It is not a claim that he can easily abandon, though Russell is
right to say that it must be abandoned.
We may conclude that Russell does have cogent objections both to Meinong’s
theory and to Frege’s theory. But there are, of course, yet other theories which are
rivals to his own and which have not been considered here. A fairly obvious one, that
seems to me to be defensible, is this modification of Frege’s theory: when a definite
description is not (uniquely) satisfied, then the simple (i.e. ‘atomic’) sentences that
contain it are just false, and the truthvalues of more complex sentences that contain
it are then to be calculated from this starting point. (On this theory ‘the A exists’ is
atomic, and so true or false according as there is or is not a unique A; then the
truthvalue of ‘~ (the A exists)’ is calculated from that of ‘the A exists’ in accordance
with the truthtable for ‘~’.) This rival theory is defended in some detail in part III of my
[1997], and that defence is not repeated here. But one can say that it is to a large extent in
agreement with Russell’s own thoughts about what ought to be the truthvalues of
sentences containing empty descriptions. Let us move on.
After the initial objections to Meinong and to Frege, Russell proceeds to state three
puzzles which a theory of denoting ought to be able to resolve (pp. 47–8). Then he
digresses to develop what he first calls ‘a rather curious difficulty’, and later ‘an
inextricable tangle’, for theories which aim to distinguish between the meaning
(or sense) and the denotation (or reference) of a definite description. This includes
both Frege’s theory and his own past theory of PoM, and it is generally agreed that
he is mainly thinking of the latter. The central example is the phrase ‘the first line of
Gray’s Elegy’ and the passage is generally referred to as the ‘Gray’s Elegy’ argument

18
Russell is assuming, as one does in classical logic, that if ‘P ’ is false then ‘P!Q ’ is true. But someone
might reply that that is not how conditional statements work in our ordinary language.
19
As mentioned earlier (note 9) it is Russell’s own previous theory, in PoM, }73, and not Frege’s theory,
that is forced to dismiss such sentences as ‘nonsense’.
 ON DENOTING 39

(pp. 48–50). After this digression he then returns to his three puzzles concerning
denoting, and shows how his own theory does resolve them (pp. 51–5). I will take
this digression first, though Russell himself might well object to calling it a ‘digression’,
because in his eyes it was the most important argument in OD. That is because it is
the most general. But its interpretation is certainly controversial.20 I will not add to the
many accounts that have been offered, save for one quick comment.
The theory in question holds that a definite description will (usually) have both a
meaning (or sense) and a denotation (or reference). It is the meaning that determines
the denotation, and not vice versa, because ‘there is no backward road from denota-
tions to meanings’ (p. 50). But Russell begins by claiming that the relation between
meaning and denotation ‘is not merely linguistic through the phrase: there must be a
logical relation involved, which we express by saying that the meaning denotes the
denotation’ (p. 49). Compare what he had said earlier in PoM:
There is a sense in which we denote, when we point or describe, or employ words as symbols for
concepts; this, however, is not the sense that I wish to discuss. But the fact that description is
possible—that we are able, by the employment of concepts, to designate a thing which is not a
concept—is due to a logical relation between some concepts and some terms, in virtue of which
such concepts inherently and logically denote such terms. It is this sense of denoting which is here
in question. (}56)
Then the argument which follows (in OD ) can only be interpreted as aiming to show
that the supposed meaning (or sense) of the phrase in question cannot be denoted
except ‘purely linguistically through the phrase’. Of course one can denote the meaning
of the phrase ‘the A ’ by using the definite description ‘the meaning of the phrase “the
A” ’. But Russell’s ensuing argument never mentions this possibility, presumably
because he thinks of it as already ruled out on the ground that it is a ‘merely linguistic’
way of denoting that meaning. The argument that he offers is concerned with other
ways of trying to denote the meaning, and it concludes that none of them work. For
the sake of argument21 one might agree that his reasoning on this is correct, and that
a reference to the meaning of the phrase must go via a reference to the phrase itself.
But what would be wrong with that?
Perhaps the most plausible suggestion is one advanced by Kremer [1994] and by
Noonan [1996]. In later writings Russell explicitly affirms his ‘principle of acquain-
tance’, according to which one must be acquainted with every constituent of a
proposition that one understands. (I shall discuss this principle in Chapter 7.) This
idea was certainly present in his thought at the time when he was writing OD, as is clear
from the fact that it is mentioned as highly relevant both near the beginning and near

20
The latest discussion of the passage known to me is that of Salmon [2009]. In his note 2 he cites no less
than twenty-four previous discussions.
21
In fact the reasoning is just muddled. There is a clear and destructive commentary in Urquhart [2005],
who correctly observes that all those who have attempted to find a defensible argument behind what Russell
actually says have had to import further things that he did not say, or to introduce qualifications to what he
did say, or both.
 Another Random Document on
Scribd Without Any Related Topics
The Project Gutenberg eBook of Around the World
in Seven Months
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with
this ebook or online at www.gutenberg.org. If you are not located
in the United States, you will have to check the laws of the country
where you are located before using this eBook.

Title: Around the World in Seven Months

Author: Charles J. Gillis

Release date: August 18, 2013 [eBook #43495]

Most recently updated: October 23, 2024

Language: English

Credits: Produced by Greg Bergquist and the Online Distributed

Proofreading Team at http://www.pgdp.net (This file was
produced from images generously made available by The
Internet Archive/American Libraries.)

*** START OF THE PROJECT GUTENBERG EBOOK AROUND THE

WORLD IN SEVEN MONTHS ***
TRAVELLING IN JAPAN.
AROUND THE WORLD
IN SEVEN MONTHS

BY

CHARLES J. GILLIS

Printed for

Private Distribution

COPYRIGHT, 1891
BY
CHARLES J. GILLIS

The Knickerbocker Press, New York

Printed and Bound by

G. P. Putnam's Sons
 With the Compliments
of the Author

CONTENTS

CHAPTER PAGE

I.— Yokohama 1
II.— Yokohama 8
III.— Tokio 14
IV.— Nikko 19
V.— Miynoskita 22
VI.— Kobé 27
VII.— Osaka 31
VIII.— Hong-Kong 33
IX.— Canton 39
X.— Hong-Kong 46
XI.— Singapore 51
XII.— On Board the "Kaisar-I-Hand" 57
XIII.— Colombo 61
XIV.— Newava Eliya 66
XV.— On Board the "Rohilla" 70
XVI.— Calcutta 76
XVII.— Darjeeling 81
XVIII.— Benares 85
XIX.— Lucknow 90
XX.— Cawnpore 93
XXI.— Agra 95
XXII.— Delhi 99
XXIII.— Jeypore 105
 XXIV.— Bombay 111
XXV.— On Board the "Khedive" 115
XXVI.— Through the Suez Canal 119
XXVII.— Cairo and the Pyramids 123
XXVIII.— Jaffa 132
XXIX.— Ramleh 134
XXX.— Jerusalem 136
XXXI.— Jaffa 140
XXXII.— On Board the "Poccir" 142
XXXIII.— Constantinople 147
XXXIV.— Conclusion 154
Distances Travelled 158
 AROUND THE WORLD IN SEVEN
MONTHS.

CHAPTER I.

YOKOHAMA.

Y o k o h a m a , J a p a n , O c t . 1 0 , 1 8 8 9 .

A T 9.50 A.M., on the morning of the 8th of September I went

aboard the vestibule train of the New York Central &
Hudson River Railroad, at Forty-second Street, New York;
and having travelled on the principal railroads around the
world, I can truly say that no train which has ever carried
me has approached this one in luxurious ease, comfort,
and safety. The train rolled into the Chicago depot at 9.50 the next
morning—exactly twenty-four hours. I was detained in Chicago for two
days, and then left by the Rock Island route for San Francisco.
At Ogden, we were detained two days by the burning of a bridge
built over a ravine—one hundred feet long and about the same height.
The fire destroyed the massive snow-sheds and great trees for a long
distance. The fire took place Friday. Telegrams were sent to
Sacramento, and the next day word came that twenty-one car loads of
material had been sent with mechanics to erect a new bridge.
 The new bridge was erected in four days. Our train was the first to
pass over it, and I remarked how substantially the new erection had
been constructed. We reached the summit at noon, and the Palace
Hotel, San Francisco, at midnight.
On the 21st of September we went aboard the steamer Rio-de-
Janeiro, built for the southern trade—370 feet long, 38 feet wide, 3,500
tons—six tubular boilers, each 13 feet diameter, 10½ feet long. I
remarked what heavy consumers of coal such shaped boilers must be,
and the engineer said there was no room to put in any other kind.
I found myself the sole occupant of a large and well ventilated state-
room. At 3 P.M., Captain Ward, standing on the bridge, gave the signal,
and the voice of an officer sang out, "All ashore that's going." Several
hundreds of Chinese men and a dozen women, in showy dresses,
crowded the wharf. The friends of the missionaries on the wharf sang a
parting hymn. The big propeller started. A tug pulled the ship's bow
around, and away we went on our voyage of 4,700 miles across the
Pacific. We passed the Golden Gate and the Seal Islands—covered with
huge seals—and then on towards our destination.
I soon made the acquaintance of most of the passengers, forty-five
in number—including fourteen missionaries of the Presbyterian Board,
nice young people going out to their duties in China and Japan. I took
my seat at the dining-table, and found that I had at my right an
agreeable companion, a captain in the German army, and at the left a
charming miss of ten, Bessie, daughter of J. De Romero, secretary of
the Spanish Legation to China.
The first week out was a rough one. The weather was bad, and the
ship rolled fearfully, so that we could not walk on deck. The waves
were immense, and consequently nearly every one was sick. I felt a
little nausea for a couple of days, but soon did duty regularly at the fine
feast placed before us three times a day, the specialty being splendid
California fruits—peaches, plums, grapes, and oranges, any of which
would bring a prize in an agricultural show.
Day after day the tireless engine drove the propeller. The splendid
ship rushed on and on, not a moment's stop the entire distance. Not a
sail or a steamer seen from port to port, and not even a whale. Once
some porpoises and flying-fish, and once, when a thousand miles from
land, a land-hawk lighted on the cross-trees, and proceeded leisurely to
feast on a captured bird, and during the night flew away.
The never-ending water was very impressive in its desolation. Better
weather came, the ship was steady, and we could walk on deck. My
little friend and I romped along the deck from end to end in safety, but
once a rude wave threw us down, and dashed us against the sides of
the vessel, taking off some inches of skin from me, but the child was
unhurt, and I did not mind a little thing like that.
I had early made the acquaintance of Mr. Mathews, the chief
engineer, and once went into the hold and inspected the boilers and
machinery of the huge ship. I spent a good deal of time in the chief
engineer's room, listening to strange tales of ship and shore.
On the 9th inst., as we were approaching our destination, I was
shown an engrossed resolution complimenting the captain, beautifully
illustrated with a pen-and-ink sketch of the ship by Señor Romero. After
dinner, one of the passengers was selected to make the presentation
address. He said:
"Ladies and gentlemen, fellow-passengers by the good ship Rio-de-
Janeiro: I act with pleasure as chairman on this auspicious occasion,
and congratulate you on the near termination of our long trip across
the great Pacific Ocean, rendered safe by the skill of the navigators and
pleasant by the efforts of the officers, one and all. I have been many
times across many seas, sometimes in magnificent floating palaces, but
never on one so neat and clean, and where every detail has been so
carefully attended to.
"'O'er the glad waters of the dark blue sea,
Our thoughts as boundless and our souls as free,
Far as the breeze can bear, the billows foam,
Survey our empire and behold our home.'
"Our only home indeed for a brief period of time. But who can fail to
remember the pleasant acquaintances made, even if we go around the
world? For 'they that go down to the sea in ships; that do business in
great waters; these see the works of the Lord, and his wonders in the
deep.' Even if we look upon the lofty Himalayas, the Alps, the
Apennines, and the Juras, and linger about gigantic Mont Blanc and her
white-robed sisters, or the beautiful Jungfrau, or sail along the classic
shores of the blue Mediterranean,—wherever we go, and whatever we
see, the scenes on this good ship will be photographed, as it were, on
our memories as long as we live—the romps on the deck, perchance
with a charming miss; or the tramp, tramp with military regularity with
those of mature age; the hours of looking upon the moonlit sea,
listening to the song and music of our missionary friends. God bless
them and their cause!
"The temporary annoyance of sea-sickness will be forgotten. And
now, fellow-passengers and friends, let us resolve that, like the
passion-flower of the wilderness, which always bears within its bosom
the true cross, we will bear within our bosom the true cross of 'enmity
towards none, charity and goodwill for all,' and thus we shall be an
honor to ourselves, the dear ones at home, the country we came from,
and our God.
"Captain Ward, by directions of the passengers on this ship, permit
me to present to you an engrossed resolution, signed by all of us, and
beautifully illustrated by Señor Romero, and expressing the hope—
which has been so often said before on like occasions—that your
voyage through life may be as safe and pleasant as you have made
ours. I bid you farewell."
The captain made a suitable reply, and the company all stood up and
drank his health.
One more night on the ship; and the next morning we sighted land
and passed along near it for forty miles. It was a rough country,
evidently of volcanic formation, and not so thickly populated as I
expected to see, considering that there are thirty-eight million people in
Japan. At last we cast anchor in the splendid harbor of Yokohama, one
of the most commodious and beautiful in the world, where a tug took
us off the ship. We were detained an hour or two at the custom-house,
and then each took a jinrickisha, a low, two-wheeled chaise with a man
between the shafts, who trotted up to the "Grand," the most perfect of
hotels. We went directly to our rooms, which had been previously
engaged.

CHAPTER II.

YOKOHAMA.

Y o k o h a m a , O c t . 1 2 , 1 8 8 9 .

T HE Grand Hotel, where I am located, is very large and

first-class in all respects. It is two hundred feet long,
fronting the matchless bay, with an extension along a
canal of two hundred feet.
From the room I occupy, I look down upon the canal
and a fine bridge which spans it. Across this bridge goes
a constant procession of men, women, and children, some horses and
carriages, and occasionally a single ox drawing a cart. But every thing
looks so different, and is managed so differently from what one has
been accustomed to, that I am more and more impressed with the idea
that I am no longer in this world, but in some wonderland beyond the
stars.
The view of the bay from the front of the hotel is said to be, by
some, the finest in the world. The harbor is very large, and could float
all the navies of all nations. At anchor, in different directions, are iron-
clad war-ships, English, French, German, Italian, Russian, and
Japanese,—only one showing the stars and stripes, the St. Mary, an old
side-wheel boat, about as large as a Brooklyn ferry-boat, and of course
about as useless.
One of the naval vessels flying the British flag is an immense iron-
clad of six thousand tons. With the commander, Captain May, I became
acquainted. He has on board an Armstrong gun of one hundred and
twenty tons, the largest I think ever made, which will throw a bolt of
half a ton a dozen miles, and penetrate through a wrought-iron plate
twelve inches thick.
The captain expressed himself as having serious doubt of the
efficiency of such monster ships and guns in actual warfare, as smaller,
swifter crafts could run around them, and have great advantages in
that respect.
Most of the coal used in this part of the world comes from England,
and is consequently very high-priced.
The harbor is crowded with many large passenger steamers, and a
great fleet of fishing craft. Towards evening the latter presented a
beautiful appearance, coming in.
I took a short walk beyond the canal and over steep hills. There are
few horses or oxen to be seen. Most of the transportation is done by
men. A two-wheeled cart, loaded with perhaps five hundred brick, was
being pushed up a steep hill by eight men, who rested often and were
much fatigued by their exertions.
Last evening our party started for a moonlight excursion along the
smooth and finely macadam-paved streets of the city. Under the
guidance of one of our number, jinrickishas—the national cab for
transporting light packages and passengers—were called, The
translation of this word is pullman-car, and thus we have the
extraordinary coincidence of the same name for a crude vehicle, drawn
by a man, and for the splendid Pullman palace cars of America, which,
with the Wagner, so far surpass in elegance and comfort all others.
Away we went at a tremendous pace, each having a Chinese lantern,
my carriage leading. We passed through well-lighted streets, lined with
stores filled with showy goods, into the suburbs, a mile or two away,
and stopped at a tea-house, where we were received with much
bowing and ceremony by the women in charge, who spoke a little
English. We were shown up a narrow stair-way into a small hall, and a
bargain was made for a national performance by two musicians and ten
dancing girls, who presently appeared, draped in beautiful Japanese
costumes. These dancing girls were all very pretty, with their almond
eyes and dark skins, and apparently not more than twelve or fifteen
years old. They were the most jolly and laughing young women one
ever saw. They shook hands with all the company, and then danced
very nicely, to very poor music, in their swaying robes. At a signal,
these robes were thrown aside, and the girls appeared in long loose
dresses made of white and red material, much like the stripes on the
American flag. The dancing, which was very graceful, continued for
some time; but the show became somewhat monotonous. The ladies
did not care to see any thing more, and we came away. Again we took
to the road, making a very unique procession by moonlight—passing
over many bridges and boulevards, and rows upon rows of brilliantly
illuminated tea-houses,—and returned to the hotel.
The natives do not wear European clothing as much as I expected.
The upper part of their persons is covered with a loose sack, often
open in front, and the jolly and laughing children of ten years and
under are generally entirely nude.
This morning all hands took another trip through the markets. We
examined cane factories, and stores filled with silks and beautiful goods
of native manufacture. Again we looked upon the street scenes and
their never-ending wonders.
From my bedroom window I see a large factory for the making of
ice, which is sold at a penny a pound and is the greatest of comforts in
this country. Butter and cigars are so strong and bad that I have left off
the use of them, and do not expect to try them again until I get to
Austria; but soda-water of excellent quality is to be had everywhere,
and is a great comfort.
We spent the evening listening to the splendid Japanese Marine Band
which played lovely music for two hours in front of the hotel; all the
musicians were natives. It was gayly bright at the hotel, the entire front
being illuminated by paper lanterns of various colors; the big ships in
the bay shown by their own numerous lamps, and the light of the full
moon glittering upon the moving waters. I have seen no such beautiful
show since 1887, when, at the city of Florence, I witnessed great
festivities attendant upon the unveiling of a statue to an Italian patriot;
the great dome of the cathedral, the bridges, and the boats on the
river Arno, the palaces on the hills, and the whole beautiful city being
illuminated in the most splendid manner—but this is a digression and I
must write now about Japan.
I have to-day wandered about the city alone, and have seen more of
the well-to-do natives. These are better dressed—always the flowing
dressing-gown pattern, and stilted and inconvenient slipper-like shoes.
Everywhere, the babies are carried in bags, on the backs of their
mothers, or more often by the older children. There are great numbers
of babies to be seen all over the city, carried about in this way, and
they always appear very happy, well fed, and comfortable.
This morning at two o'clock I was awakened by an earthquake which
rolled my bed about the room. It lasted twenty seconds or so, but I did
not think it worth while to get up, and soon went off to sleep again. I
am told such entertainments are frequent in this country, and one must
get used to them.

CHAPTER III.

TOKIO.

T o k i o , J a p a n , O c t . 1 5 , 1 8 8 9 .

A FTER being entertained at Yokohama, on the morning of

the 14th instant, with a slight earthquake, we left for
this, the capital city of the Empire, on a finely built and
equipped, narrow-gauge (3 ft. 6 in.) railroad. Every
square yard of the country we traversed was cultivated in
the highest degree—Distance eighteen miles, time one
hour.
I noticed that the locomotives, cars, and all the equipments about
the railroad were of English manufacture from Manchester and
Birmingham. I was informed that most of the twelve hundred miles of
railroads in Japan were owned and run by the Government. The chief
in charge is a native educated in England, who scouted the idea that
any other country could produce any thing fit to be used on railroads.
We had previously engaged rooms at one of the two hotels in the
city, where foreigners are entertained, and after an excellent dinner,
took jinrickishas, of which there are 80,000 in this city, and had a long
run through the interminable streets. The city has a population of
1,600,000 and covers a space of thirty-six square miles, the streets
being very narrow and the houses mostly of wood, one and two stories
high—the stores all small.
After passing through streets for some miles, we came to others,
wider and lighted brilliantly by gas and electricity, through which
carriages were not allowed to pass. The houses, for miles, were
occupied as tea-houses, and were brilliantly illuminated, like the gin
palaces of London, or the whiskey saloons of America. Great vans were
passing along, on which dancing and theatrical performances were
going on. There were also a good many theatres in active operation.
One of the evenings that we were in the city, these streets were
occupied by an immense annual flower show, one of the features of
which was a big elephant constructed of chrysanthemum flowers of
many colors. The effect was very gorgeous.
The next day we spent going about the great city and seeing its
wonders, chief of which is the Mikado's palace and grounds. I called on
the American Minister and asked him to get me a permit to go into the
palace, but he said it was impossible, no foreigners being allowed in
the palace or the grounds. The palace and gardens looked like
immense fortifications, being surrounded by three moats, each a
hundred feet wide, and filled with water, and by three stone walls, each
thirty or forty feet high.
The palace is in the heart of the city, and I should say the grounds
were two hundred acres in extent, all, including the neighboring
streets, being lighted by the New York Edison Company. I saw the
superintendent who had charge of the construction of the plant, who
said it took them a year to do it.
 The women to be seen in the streets and tea-houses are invariably
small and very pretty, except some of the married ones, who have their
teeth colored black in accordance with an ancient custom, which makes
them look hideous.
It is very cool in this part of Japan at this season. There is not
sufficient frost to affect the crops, but one gets cold riding about—and
there is no efficient method of heating the houses. There is no coal
used for domestic purposes, and wood is very scarce and high. If you
ask for a fire, at most hotels, they bring you a copper pan containing
ignited charcoal covered with ashes, which does about as much good
as a kerosene lamp. I suffer greatly with the cold, and would be glad to
pay a large price for a pair of Arctic overshoes.
The price of newspapers, printed in the English language, at
Yokohama is twenty-five cents a copy, or thirty dollars per annum. They
have very little news, and almost none from America.
We went through the museum, and saw many extraordinary
curiosities of ancient and modern Japan. Among them was a stuffed
rooster in a glass case, whose tail feathers were ten feet long. I
thought there was some humbug about it, but I afterwards saw a live
one with tail feathers twelve feet long.
The public buildings are modern, large, and handsome, and the
people very polite and good-natured. The streets are narrow. Great
crowds are everywhere. It seems to me that I must have seen a
hundred thousand people to-day. Every thing about the city is strange,
often disagreeable and offensive. A couple of days in it is quite
sufficient, and I shall be very glad to go away to-morrow.
 CHAPTER IV.

NIKKO.

N i k k o , J a p a n , O c t . 2 0 , 1 8 8 9 .

W E left Tokio on the 17th, at 6.46 A.M., for a station called

Utsumorama, ninety-three miles. Arrived at noon, and,
after an excellent lunch, started in jinrickishas for this
place, and a most extraordinary and unique trip it proved
to be.
The road was built hundreds of years ago by a Shinto
king, and is an admirable example of engineering; well drained, and
with an excellent foundation of small stones, which needed only a top-
dressing and a steam roller to make it as good as any in Europe. It is
lined on both sides with immense pine and cedar trees. Many of these
trees are twelve feet in diameter; and often the roots are grown
together, so that four or five trees look like one. They are sixty to
eighty feet high, and afford an excellent shade.
The distance from the railway station to this place is twenty-five
miles, and we made it in four hours with two men harnessed to the
jinrickishas tandem. We made only one stop of half an hour for lunch,
which we brought with us, and ate at one of the numerous tea-houses.
We arrived at 4 P.M., delighted at the wonderful sights, but much
fatigued and very cold. Rooms had been engaged for us in an excellent
hotel, excellent in all respects except that there was no way of heating,
unless with pans of charcoal. I suffered greatly from the cold, though I
had warm clothing, including a heavy overcoat which had done me
good service the previous winter at Montreal when the thermometer
stood at thirty degrees below zero.
Near the hotel are a dozen, or more, costly and grotesque edifices,
much adorned with carved wood statues of horrible-looking beasts and
devils, covered with bronze and gold. There are temples of Buddha,
and gorgeous mausoleums of kings who died five hundred years ago,
 situated in a park of big trees; but looking at them, though interesting,
was not agreeable, and I was quite satisfied with one visit.
To-day we made an excursion to a lake among the mountains, five
thousand feet high. I was furnished with a mountain horse which
proved an ugly brute; kicked and stumbled, and put the bit between his
teeth, so that I could not control him, and he nearly trotted the life out
of me. We went up and along the winding paths, passing numerous
water-falls, one of which was 750 feet high, and at last reached the
lake, which is of great beauty. The mountains rise directly from the
water's edge. They are covered to their very tops with green trees, the
leaves of which have a singular feathery appearance.
The tea-houses where we stopped for lunch were models of
cleanliness and comfort. We brought our own provisions as usual, but
had in addition a boiled fish just taken from the lake.
We stopped in and around the tea-houses for some hours; and then
I mounted my ugly brute of a horse and rode back to the hotel, a much
used-up man. I was glad to get a bath and to retire early.
We returned to the railway station by the road we came, and again
made the distance in four hours, with only one stop of fifteen minutes.

CHAPTER V.

MIYNOSKITA.

M i y n o s k i t a , J a p a n , O c t . 2 4 , 1 8 8 9 .

ESTERDAY at 10 A.M. we left Yokohama, arrived at the railroad station at

twelve, and reached this favorite watering-place, among the
mountains, in four hours by jinrickishas. Our rooms had been engaged
in an excellent hotel, called Fujiya, and soon after our arrival a fine
dinner was served of soup, fish, roast beef, sago pudding, and other
 delicacies, to which we did ample justice. The waitresses
were all pretty native girls, dressed in their native
costumes; there were a dozen, or more, of them about
the hotel. These waitresses were pleasant, jolly, and very
polite, but very small in stature; some of them walked
under my outstretched arm, and all of them might have
done so.
I have a fine front room, and look out upon the surrounding
mountains, which are very lofty and covered with green trees. This is
an ideal mountain resort—great mountains, a roaring river winding
some hundreds of feet below the road, and numerous water-falls; the
water rushing down into the river. From one point of view I counted
seven water-falls, and found, on trial, that one of them came from a
hot spring far up among the mountains, and the water was quite warm
when it reached the road. I walked along the road for several miles and
found it wonderfully romantic everywhere. The road itself is a fine
specimen of engineering, very expensive to build, and almost as good
as the famous one built by Napoleon III., from Geneva to Chamouni.
We are here rather too late in the season to thoroughly enjoy the
place and surroundings, it being cold and the methods of heating
houses imperfect, but in summer it must be perfectly lovely.
There is another hotel being erected near the one we are in, and I
was much interested watching their method of work. They required a
lot of earth for filling in, and were transporting it in baskets from the
mountains above. Two men would fill a basket, suspend it across their
shoulders by a bamboo pole, dump it where wanted, and return for
more. I longed to present them with a wheelbarrow, and show them
how to move earth ten times faster than they were doing. It would
appear that there are no saw-mills in this country, for the men were
sawing out boards and timber by hand, to use in the construction of
this hotel. A stick of timber a foot or two in diameter was arranged with
one end resting on the ground, and the other placed on a wooden
horse four or five feet high; a man then mounted the stick and
laboriously sawed out boards with a hand-saw. The workmen had no
clothing on except a breech cloth, and were all doing constant and
faithful service for, as we were informed, ten hours a day; the pay
being ten cents per day. For similar service in our country, as every one
knows, mechanics are paid from $2.50 to $3.50 per day.
We left the hotel at nine this morning, and took a trip among the
mountains to Lake Hakone. I selected my horse this time, and he
proved an excellent animal, a small shaggy fellow, kind and easy
trotting, but much given to stumbling and letting both heels fly if
another horse came near, which little amusement of his nearly
unseated me several times. We went up six thousand feet over the
worst of mountain roads, but my animal walked carefully, often along
narrow paths, where a fall would have tumbled us down hundreds of
feet below. I enjoyed the ride very much. It took six horses and seven
chairs to accommodate our party, each horse having a man to attend to
him, and each chair carried by four men, making a large procession.
We arrived in two or three hours at an hotel on the lake, and after an
excellent lunch took boats and crossed over to near the foot of
Fusiyama, the horses and men going around to meet us.
Fusiyama is the brag mountain of Japan, the only one of much size in
the Empire, and is universally known and photographed in all possible
ways. It is fourteen thousand feet high, and is, as I write, covered with
snow, and presents a beautiful appearance from the lake.
We landed and walked over the mountains to the place where the
horses had been sent. The sun was terribly hot in some places, and in
others the only path was along the bed of dry brooks. We passed over
the crater of an active volcano, steam and smoke rushing out near the
path. The guide said it was dangerous to wander from the path, and
pointed out where two native guides had fallen through and had not
been seen since. There was no wandering after this fact was stated.
After two or three hours of dreadful fatigue, we found our horses, and I
was very glad to mount my shaggy old fellow, who carried me safely
over slippery rocks, along narrow paths, and a road (where there was
any) as bad as a road could be, arriving at the hotel at six, much
fatigued, but in good form and ready for the excellent dinner which
was waiting our arrival.
After thoroughly enjoying this delightful spot for two days, we
started down the mountain road in the morning and came along in
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com