138 LARRY LAlJDAN

'Scientific Progress.' Synthese. 45: 42762.

7he Logic of Scientific Discovery. New York: Basic Books.
nmf'Clures and Refurarions. London: Routledge and Kegan Paul.
'Correspondence. Invariance and Heuristics: In Praise of Conser_
vative Induction.' Srudies in lhe Hisfory and Philosophy of Science, 2. VII
Putnam, H. (1975). Malhematics, Matter, and Method, 2 vols. Cambridge: Cam
bridge University Press.
--(1976). 'What is Realism:' Proceedings of the Aristotelian Society, 76: 177-94. STRUCTURAL REALISM:
--(1978). Meaning and the Moral Sciences. London: Routledge and Kcgan Paul.
Sellars, W. (1963). Science, Perception and Reality. New York: Humanities Press. THE BEST OF BOTH WORLDS?
Sklar, L. (1967). 'Types of Inter-Theoretic Reductions.' British Journal for Philoso.
phy of Science. 18: 190-224. JOHN WORRALL
Szumilewicz. I. (1977). 'Incommensurability and the Rationality of the Develop_
ment of Science: Brilish Journal for Philosophy of Science, 28: 345-50.
Watkins. J. (1978). 'Corroboration and the Problem of Content-Comparison.' In G,
Radnitzky and G. Anderson (cds.), Progress and RalionafilY in Science, 339-78. Presently accepted physical theories postulate a curved space-time struc
Dordrecht: Reidel.
ture, fundamental particles, and forces of various sorts. What we can know
for sure on the basis of observation, at most. are only facts about the
motions of macrosapic bodies, the tracks that appear in cloud chambers in
certain circumstances, and so on. Most of the content of the basic theories
in physics goes 'beyond' the 'directly observational'-no matter how lib
eral a conception of the 'directly observational' is adopted. What is the
status of the genuinely theoreticaL observation-transcendent content of
our presently accepted theories? Most of us unreflectingly take it that the
statements in this observation-transcendent part of the theory are at
tempted descriptions of a reality lying 'behind' the observable phenomena:
that those theories really do straightforwardly assert that space-time is
curved in the presence of matter, that electrons, neutrinos, and the rest
exist and do various funny things. Furthermore, most of us unreflectingly
take it that the enormous empirical success of these theories legitimizes the
assumption that these descriptions of all underlying reality are accurate, or
at any rate 'essentially' or 'approximately' accurate. The main problem of
scientific realism, as I understand it, is that of whether or not there are,
after reflection, good reasons for holding this view that most of us
unrcflectingly
There are, of course, several anti-realist alternatives on offer. The most
widcly canvassed is some version of the pragmatic or instrumentalist view
that the observation-transcendent content of our theorics is not in fact,
and despite its apparent logical form. descriptive at all, but instead simply
'scaffolding' for the experimental laws. Theories are codification schemes;
theoretical terms like 'electron' or 'weak force' or whatever should not
be taken as even intended to refer to real entities, but instead as
fictional namcs introduced simply to order our experimental laws into a
Reprinted from Dialrcrica, 4311-2 (l'lg'l): 99-124. by permission of S()ci~te Dialectica.

1
 STRUCTURAL REALISM 141
140 JOliN WORRALL

quantum theory's failing to be a fundamentally correct description of

system.' A more recent anti-realist position---that of van Fraassen--holds
that theortical terms do, at any rate purportedly, refer to real entities (and reality.
Notice, by the way, that the argument requires the empirical success of
are not, e.g., simply shorthand for complex observational terms), but that
a theory to be understood in a particular way. Not every empirical conse
there is no reason to assume that even our best theories are true nor even
quence that a theory has and which happens to be correct will give intuitive
'approximately' true, nor even that the aim of science is to produce true
support for the idea that the theory must somehow or other have latched
theories; instead, acceptance of a theory should be taken to involve only
on to the 'universal blueprint'. Specifically, any empirical consequence
the claim that the theory is 'empirically adequate', that it 'saves the
which was written into the theory post hoc must be excluded. Clearly it is no
phenomena' J
miracle if a theory gets right a fact which was already known to hold and
I can find no essentially new arguments in the recent discussions
which the theory had been engineered to yield. If the fact concerned was
Worrall 1982). What seem to me the two most persuasive arguments arc
used in the construction of the theory-for example, to fix the value of
very old-both are certainly to be found in Poincare and in Duhem. The
some initially free parameter-- then the theory was bOllluj to get that fact
main interest in the problem of scientific realism lies, I think, in the
(On the other hand, if the experimental result concerned was not
fact that these two persuasive arguments appear to pull in opposite
written into the theory, then the support it lends to the idea that the theory
directions: one seems to speak for realism and the other against it: yet
is 'essentially correct' is surely independent of whether or not the result
a really satisfactory position would need to have both arguments on
its side. The concern of the present paper is to investigate this tension was already known when the theory was
This inlUitive 'no miracles' argument can he made more precise in
between the two arguments and to sug&cst (no more) that an old and
various ways-all of them problematiC and some of them more prohlem
hitherto mostly neglected position may offer the best hope of reconciling
atic than others, It is. for instance, often run as a form of an 'inference to
the two.
the best explanation' or Peircian 'abduction'.1 But, as Laudan (1981) and
The main argument (perhaps 'consideration' would be more accurate)
Fine (ch. L this volume) have both pointed ouL since the anti-realist is
to incline someone towards realism I shall call the 'no miracles'
in the business of denying the validity of inference to the hest
argument (although a version of it is nowadays sometimes called the
in science. he is hardly likely to allow it in philosophy as a
'ultimate argument' for realism-see Musgrave 198R). Very roughly, this
means of arguing for realism. Perhaps more importantly, and despite the
argument goes as follows. It would be a miracle, a coincidence on a near
attempts of some philosophers to claim scientilic status for realism itself on
cosmic scale, if a theory made as many correct empirical predictions as,
the basis of its explanatory power.' there is surely a crucial. pragmatic
say, the general theory of relativity or the photon theory of light without
difference between a good scientific explanation and the 'explanation'
what that theory says about the fundamental structure of the universe
afforded by the thesis of realism for the success of our present theories. A
correct or 'essentially' or 'basically' correct. But we shouldn't
requirement for a convincing scientific explanation is independent testabil
miracles, not at any rate if there is a non-miraculous alternative. If what
ity-Newton's explanation of the planetary orbits is such a good one
these theories say is going on 'behind' the phenomena is indeed true or
because the theory yields so much else that is testable besides the orbits:
'approximately true', then it is no wonder that they get the phenomena
the ohlateness of the earth. return of Hallev's comet, and so on. Yet in the
right. So it is plausible to conclude that presently accepted theories are
indeed 'essentially' correct. After all, quantum theory gets certain
for this notion of empirical
phenomena, like the Lamb shift, correct to, whatever it is. 6 or 7 decimal I <i~) pap'"
in the view of some scientists, only a philosopher, overly impressed ramous. 'white
support for the 'novel fa~t' count mor,,'
hy merely logical possibilities. could believe that this is compatible with the ~llrnent is strongly cnticizcd by l.arry Louda" (l')Hl). Strong and cogent
reservations ahout the allegc.:d explanation that [c(llism supplies of Sc1cnCt;'S success were also
in Howard Stem's pap"r ddivered to the Ncuchiltd confer"l1ce.
I According to a f3ITIOU<.; rt:mark of Quine's, for ins1ance, the theoretical entities involved in
osition seems to have been h"ld bv I:loyct, l'iiiniluoio. anct others. It is ctisown"d by
current scicn~e (like ckctrons) are epiSctenH}logicalh' on a par with the Greek rfntk__ hoth are
Putnam (197;';): 'I think that realism is like an empirical hypothesi, in that il could he false, and
convenient Iklions jntroduc~d in lhe
that facts are relevant to supporl (.or critIClzing it): but that dot:so'( mean that n:alism is
s'cientihc (in any Slandard scn~c of "sciCI11ific"). ur that realism 1s a hypothesls:
Van Fraasscll 1tJHO. Van Fraasscn calls
see my 1'!H3 review of hi, hook).

~
 STRUCTURAL REALISM
143
142 JOHN WORRALL

some binary stars should be in accordance with the theory-it would be a

case of realism's 'explanation' of the success of our current theories there miracle if this were true but the theory is not. However. as we all know,
can of course be no question of any independent tests. Scientific realism Newton'S theory was rejected in favour of Einstein's in the earlv twentieth
can surely not be illferred in any interesting sense from science's success.
The 'no miracles' argument cannot establish scientific realism: the claim is century.
This would pose no problem if Einstein'S theory were simply an exten
only that, other things being equal. a theory's predictive success supplies a sion of Newton's; that is, if it simply incorporated Newton's theory as a
prima-facie plausibility argument in favour of its somehow or other special case, and then went on to say more. In general, if the development
latched on to the truth. of science were cumulative, then scientific change would pose no problem
Certainly the psychological force of the argument was sharply felt even either for the realist or for his 'no miracles' argument. The reason why
the philosophers who are usually (though, as we shall see, mistakenly) Newton's theory got so many of the phenomena correct could still be that
regarded as the great champions of anti-realism or instrumentalism: Pierre
Duhem and Henri Poincare. Here, for example, is Duhem: it was true, just not the whole truth.

Unfortunately Einstein'S theory is not simply an extension of Newton's.

The highest test. therefore of [a th.:ory] is to ask it to indicate in advance things The two theories are logically inconsistent: if Einstein's theory is true, then

which the future alone will reveal. And when the experiment is made and confirms Newton's has to be false." This is of course accepted by all present-day

the predictions obtained from our theory. we feel strengthened in our conviction
realists. The recognition that scientific progress, even in the 'successful',

that the relations established by our reason among abstract notions only correspond
to the relations among Ihings. (190h: 28) 'mature' sciences, is not strictly cumulative at the theoretical level, but

instead involves at \east an element of modification and revision is

And here Poincare: the reason why no present-day realist would claim that we have grounds

Have we any right, for instance. to enunciate Newton's law? No doubt numerous for holding that presently accepted theories arc true. Instead, the claim

observations are in agreement with it. bill is not that a simple fact of chance'> And is only that we have grounds for holding that those theories are 'approxi

how do we know hesidt.!s. that this law wlllch has bcen true for so many generations,
mately' or 'essentially' true. This last claim might be called 'modified

will not be untrue in the next') To this objection the only answ;;r you can give is: It
is very improbable. (J 90S: 186) realism'. I shall, for convenience, drop the 'modified' in what follows,

but it should be understoOd that my realists claim only that we

So the 'no miracles' argument is likely, I think, to incline a common have grounds for holding that our present theories in mature science are

sensical sort of person towards some sort of scientific realist view. But he
is likely to feel those realist sentiments evaporating if he takes a close look approximately true.
at the history of science and particularly at the phenomenon of ,,r.mt;!;,> , Professor Agazzl m ntS pap.." at Ncuchatel took the view thai Newtonian physics remains
true of objects in its intended domain and that quantum and relativislic physics are true of
revolutions. s in quite different domams. But this position is surely untenable. Newton's theory was not
Newton's theory of gravitation had a stunning range of predictive suc (its 'intended referent' was not) macroscopic objects mOYing with yeiocilies small com
cess: the perturbations of the planetary orbits away from strict Keplerian with thai of light. [t waS about all material objects moving with any velocity you like. And
theory is wrong (or so we now think). gloriously wrong, o[ course. but wrong. Moreover, it
ellipses. the variation of gravity over the earth's surface, the return of isn't even, strictly speaking, right about certain bodies and certain motions and 'only' wrong
Halley's comet, precession of the equinoxes, and so on. Newtonians even when we are dealing with microscopic objects or bodies mOYing at very high velocities. If
and quantum theory are correct, then Newton's theory's predictions about the motion
turned empirical difficulties (like the initially anomalous motion of of any body. even the most macroscopic and slowest-moving. are strictly false, It's just that their
Uranus) into major slIccesses (in this case the prediction of a hitherto falsity lies well within experimental error. That is. what is true is that Newton's theory is an
unknown trans-lJranian planet suhsequently christened Neptune). emoiricallv fauiliess approximation for a whole range of cases. Its also true, as Agazzi claimed.
's slill often see themselves as applying classical physics in a whole
icists were wont to bemoan their fate at having been born after Newton c1ear-sighled account of what they are doing is, I think, that Ihey are
there was only one truth to he discovered ahout the 'system of the world', .lpported theories available to them-viz. quantum mechanics and
It's jusl that they know that these theories themselves entail the meta-result
and Newton had discovered it. Certainly an apparenlly hugely convincing
that. for their purposes (of sending rockets to the moon or whatever). it will make no
'no miracles' argument could be-and was--constructed on behalf of difference to act as iflhey were applying classical physics, and indeed that il would be
Newton's theory. It would be a miracle if Newton's theory got the empirical point of view a waste of effort to apply the mathematically more demanding newel
theories onlv for that sophistication to become entirely irrelevant when it comes to empirical
etary motions so precisely right, that it should be right about Neptune and
about Halley's comet, that the molion of incredibly distant objects like application.

...J..

 145
STRUCTURAL REALISM
144 JOHN WORRALL

scienc has so far been 'essentially' cumulative at all levels-theoretical as

This realist claim involves two terms which are notoriously difficult to well ascsobservationaL It seems reasonable. therefore, to infer induetivcly
clarify. I shall propose my own rough characterization of the 'mature' that that development will continue to be 'esscntially cumulative' in the
sciences shortly. As for 'approximately true', well-known and major diffi. future. This presumably means that, even should our present theories
culties stand in the way of any attempt at precise analysis. Indeed, various be replaced, they will continue to appear 'approximately' correct in
attempted characterizations (such as Popper's in terms of 'increasing veri the light of the successor theories. Such a development is, of course,
have turned out to be formally deeply fIawed. 7 Although we do logically compatible with the genuinely theoretical assumptions. both of
often operate quite happily at the intuitive level with the notion of ap. presently accepted theories and of those destined to be accepted in the
proximate truth, it is surely not the sort of notion which can happily be left future, being entirely untrue. However. this is highly implausible, since
as a primitive. For one thing: if the notion is going to do the work that it would make the empirical success of all these theories entirely mysteri
realists need it to do. it is going to have to be transitive. Realists need to ous: while, on the other hand, the assumption that our present theories
claim that although some presently accepted theory may subsequently be are approximately true is enough to explain the empirical success as
modified and replaced, it will still look 'approximately true' in the light not
just of the next theory which supersedes it, but also in the light of the non-miraculous,
No one, I take it (reiterating the point made earlier), would claim that
theory (if any) which supersedes the theory which supersedes it. etc. But is this argument is completely watertight. Thc inductive 'inference' from
transitivity a property that the notion of approximate truth possesses even 'essential cumulativity' in the past to 'essential cumulativity' in the future
intuitively') could of course be questioned. Moreover, there is still the problem of what
But there is anyway an important prior question here: that of whether exactly is involved in approximate truth: and indeed the problem of
or not, talking intuitively. in advance of formal analysis, the history whether or not the assumption of the approximate truth of our present
of science (or some selected part of it) speaks in favour of successive theories really would explain their empirical success. It might seem plaus
scientific theories heing increasingly good 'approximations to the truth'. ible. intuitively speaking, to suppose that if a theory is 'approximately' or
This clearly depends on just how radical theory change has standardly 'essentially' true. then it is likely that most of its consequences will them
been in science. Again, of course, we are dealing in unfortunately vague selves be 'essentially' correct, To take a straightforwardly empirical ex
terms. But surely the realist claim--that we have grounds for holding that ample, say tbat I make a slight arithmetical error in totting up my bank
our present theories are approximately true-is plausible onlv to the balance and come to the strictly mistaken view that my total
extent that it seems reasonable to say that Newton's theory, for fortune is 100, when the truth is that it is 103. Will it seem 'miraculous'
"approximates' Einstein's, and that, in generaL the development of science if this strictly false theory none the less supplies a quite reliable guide to
(at an, rate the development of successful, 'mature' science) has been life? After alL it might be claimed, most of the consequences that I am
'essentially' cumulative, that the deposed theories themselves. and likely to be interested in-for example, that I can't afford a month's
their success/ill empirical consequences, have generally lived on, albeit holiday in Switzerland~will in fact be consequences both of the false
in 'modified form'. after the 'revolution'. If. on the contrary, theory theory, that I hold, and of the truth. None the less, plausible or not. there
change in science has often involved 'radical' shifts- something like the are formidable formal difficulties here." Every false theory, of course, has
complete rejection of the genuinely theoretical assumptions (though infinitely many false consequences (as well as infinitely many true
comhined of course with retention of the successful empirical content)- and there arc things that my 'nearly true' theory gets totally wrong. For
then realism is in dire straits. Before going further, let's be clear on example. the truth is that my total fortune expressed in pounds sterling is
the dependence of realism on the claim that theory change has been a prime number. whereas the 'nearly true' theory I hold says-entirely
'essentially cumulative'. it's composite. Moreover, the argument seems commit
Assume, first, that the realist has convinced us that the development of ted to the claim that if theory T 'approximates' theory 1", which in turn
theoretical science has indeed been 'essentially cumulative '. He could then o
argue for his realism roughly as follows. The development of the 'mature' , Two recent attempts to overcome these difficulties arc Oddie 1<)0\1> and Niiniluot
1987--\hough hoth attc""pts mvolve substantive. nonlog icaL and therefore challengeable

, See Tichy 1974 and Miller J 974. assumptions,

 147
STRUCTURAL REALISM
146 JOHN WORRALL

old theory could be said to 'approximate' the new only

'approximates' T'. then T 'approximates' Tn, (The theories which evenlu
admittedly vague and therefore elastic notion of
aily supersede our presently accepted ones, might lhemselves-presum_
will---eventually be superseded by still further theories. The realist breaking-point.
At first glance, this claim appears to be correct. Consider. for example.
needs to be assured that any presently accepted theory will continue to the history of optics. Even if we restrict this history to the modern era.
look approximately correct, even in the light of the further theories in the there have been fundamental shifts in our theory about the basic constitu
sequence, not just in the light of its immediate successor.) But is this tion of light. The theory that a beam of light consists of a shower of tiny
transitivity assumption correct? After aiL if we took a series of material particles was widely held in the eighteenth century. Some of its
graphs at one-second intervals, say, of a developing tadpole. each photo- empirical consequences -such as those about simple reflection, refraction,
in the sequence would presumably 'approximate' its predecessor, and prismatic dispersion-were correct. The theory was, however. rejected
and yet we start with a tadpole and finish with a frog. Does a frog 'approxi in favour of the idea that light consists, not of mailer. but of certain

mate' a tadpole? I propose, however, that, for present purposes, we put alI vibratory motions set up by luminous bodies and carried by an all-pervad

these difficulties into abeyance. If he can sustain the claim that the devel ing medium, the 'luminiferous aether'. It would clearly be difficult to argue

opment of the 'mature' sciences has heen 'essentially cumulative'. then the that the theory that light is a wave in a mechanical medium is an 'exten

realist has at least some sort of argument for his claim. sion', or even an 'extension with slight moditkations', of the idea that

If, on the contrary. the realist is forced to coneede that there has been consists of material particles: waves in a mechanical medium and particles

radical change at the theoretical level in the history of even the mature travelling through empty space seem more like chalk and cheese than
sciences. then he surely is in deep trouble. Suppose that there are cases of do chalk and cheese themselves, Nor was that all: Fresnel's wave theory
mature theories which were once accepted, were predictively successful, was itself soon replaced by Maxwell's electromagnetic theory. Maxwell.
and whose underlying theoretical assumptions none the less now seem as is well known, strove manfully to an account of the electromagnctic
unequivocally entirely false. The realist would have encouraged the earlier field in terms of some underlying mechanical medium; but his attempts
theorist to regard his theory's empirical success as giving him grounds for and those of others failed, and it came to be accepted that the electromag
regarding the theory itself as approximately true. He now encourages netic field is a primitive. So again, a fundamental change in the accepted
scientists to regard their newer theory's empirical success as giving them account of the hasic structure of light seems to have occurred-instead
grounds for regarding that newer theory as approximately true. The older of vibrations carried through an elastic medium, it becomes a series of
and newer theories are radically at odds with one another at the theoretical wave-like changes in a disembodied electromagnetic field. A mechanical
level. Presumably, if we have good grounds for thinking a theory T ap vibration and an electric (,displacement') current are surely radically
proximately true, we equally have good grounds for thinking that any different sorts of thing, Finally, the acceptance of the photon theory had
theory T' radically at odds with T is false (plain false, not 'approximately consisting again of discrete entities, but ones which obey an entirely
So the realist would be in the unenviable position of telling us that
we now have good grounds to regard as false a theory which he earlier new mechanics.
In the meanwhile, as theories were changing light from chalk to cheese
would have told us we had good grounds to believe approximately true. and then to superchalk, there was a steady, hasically cumulative dcvelop
Why should not his proposed judgement about presently accepted theories ment in the captured and systematized cmpirical content of optics. il' The
turn out to be similarly mistaken? material oarticle theory dealt ~"tid'lctorilv with simple reflection and re
Assuming, then, that the realist is not talking about 'good grounds' in
some defeasible, conjectural sense." realism is not compatible with the
existence of radical theoretical changes in science (or at any rate in mature
science). The chief argument against realism-the argument from scien
tific revolutions---is based precisely on the claim that revolutionary
changes have occurred in accepted seientilic theories, changes in which the

, See bdow, I'll_ ISO 15!.

 STRUCTURAl REALISM 149
148 JOHN WORRALL

The picture of the development of science certamly seems, then. to be

fraction and little else; the classical wave theory added interference
one of essential cumulativity at the empirical level. accompanied by
and diffraction and eventually polarization effects too: the electromagnetic
changes of an entirely non-cumulative kind at the top theoretieallevelsY
theory added various results connecting light with electrical and magnetic
Thib picture of theory in the past would seem to supply good
effects: the photon theory added the photoelectric effect and much
inductive grounds for holding that those theories presently accepted
else besides. The process at the empirical level (properly construed)
in science will. within a reasonably brief period, themselves be
was essentially cumulative. There were temporary problems over
theories which retain (and extend) the f'ITInirieal success of
whether or not the classical wave theory could deal with the
which had previously been taken to support the ('essentially') rectilinear theories, but do so on the basis of under.
at odds with those presently accepted. This is. of course, the 50
propagation of light), but these were invariably settled quickly and
positively.11 called pessimistic indllction-usually regarded as a recent
discovery, but in fact already stated clearlv by Poincare. How ean th..:re
Or take the Newton-Einstein case again. At the empirical level it does
b..: good grounds for holding our present theories to be 'approximately'
seem intuitively reasonable to say that Einstein's theory is a sort of 'exten
or 'essentially' true. and at the same time seemingly strong historical
sion with modifications' of Newton's. It is true that. even at this level, if we
inductive grounds for revardinQ those theorics as (probably) ontologically
take the maximally precise consequences about the motion of a given
yielded by the two theories. they will always strictly speaking contradict false?
Unless this picture of theory change is shown to be inaccurate. then
one another. But for a whole range of cases (those cases, of course, in
realism is surely untenable, and basically only two (very different)
which the velocities involved are fairly small compared to the velocity of
bilities open. Thc lirst can be motivated as follows. Science is the Held in
light). the predictions of the two theories will be strictly different but
whi.::h rationality reigns. There can be no rational acceptance of claims of
indistinguishable. It is also true. of course, that Newton's
a kind which history us g,rounds to think are likely later to be
_ cases of corresponding relativistic equations. How
rejected. The successful empirical content of a once accepted theory is in
ever. there is much more to Newton's theory than the laws of motion and
general carried over to the new theorv. but its basic theoretical claims are
the principle of universal gravitation considered simply as mathematical
not. Theories. then, are best construed as making no real claims beyond
equations. These equations were interpretcd within a set of very
their directly empirical consequences: or, if they are so construed, accept
theoretical assumptions which involved amongst other things the assump
ance of these theoretical claims as true or approximately true is no part of
tion that space is infinite. that time is absolute. so that two events simul
the rational procedures of science. We are thus \cd into some sort of either
taneous for one observer arc simultam:ous for all. and that the inertial
mass of a body is constant. Einstein's theory entails. on the contrary. that or 'constructive' anti-realism.
Such a position restores a pleasing, cumulativc (or quasi-cumulative)
space is finite (though unbounded), that time is not absolute in the
development to science (i.e. to the 'real part' of science): but it does so at
Newtonian sense. and that the mass of a body increases with its velocity.
the expense of sacrificing the 'no miracles' argument entirely. After all. the
All these arc surely out-and-out contradictions.
theoretical science which the pragmatist alleges to be insubstantial and to
11 The case of rectilinear propagation of light provides an illustrative example hoth of the a purely codilicatory role has. as a matter of fact. often proved
c:-iscntial empirical continuil), of 'mature' science and of whal it is ahout this process that leads That is. interpreted literally and therefore treated as claims about the
Feycrabcnd and Kuhn to misrepresent it. Certain Iheories become firmly enlIenched at
certain stages of the development of s,icnce. so much parts at knowledge', that structure of the world. theories have yielded testabh: consequences over
they, or at any rate particular experimental ~ilU<itions in!f'rpreted in lij?ht. are n.~aJily talked and above those they were introduced to codify. and those consequences
of as 'facts'. This WaS certainly true' of the 'fact' that lil'h!. if left to itself. is rectilinearly
propagated. Here then is surelv a 'fact" which was 'lost" in the wave revolution, since Fresnel's
have turned out to b.:: correct when checked empirically. Whv'! The prag
entails that light is IIlwa:,..< diffrac'lcd that in most circumstances the difference matist asserts that there is 110 answer.
between the diffraction pattern and the of geometrical optics is well below the
observational level. But this last remark the game away. The idea that light is PUlflcar~ and Duhem. rather lost
That this
rt'-dnnha~iz('d hy Popper and those inllw..: nccd hy him
propagated was never an result (not a 'crude fact' in Poincare's 12

real !.:mpirical 'ray tracings', inahility to sec round corners ~)~ sight of bv the

bent opa<jue tubes, de. were not 'lost' bul simply re-explained as a result the shllt (,,,"eh as J'ohn '

Sec Putnam 197H: 2:': and PoincarL: ! l}O:,: 14uotl'd hdow. p. 157).
to the wave theory.

 STRUCTURAL REALISM
lSI
150 JOHN WORRALL

assert that space and time are absolute, that there are action-at-a-distance
The other alternative for someone who accepts the empirically cumulat_
forces of gravity, and that inertial mass is constant; all this was entirely
ive, theoretically non-cumulative picture of scientific change, but who
wrong, and yet the theory based on these assumptions was highly
wishes to avoid pragmatism is pure, Popperian conjectural realism. This is adequate. This just has to be recorded as a fact. And if you happen to
Popper's view stripped of all thc verisimilitudc ideas, which always sat
lind it a rather surprising fact, then that's your own business--perhaps due
rather uncomfortably with the main theses. On this conjectural realist
to failure to internalize the elementary logical fact that all false theories
view, the genuinely theoretical, observation-transcendent parts of scien
have true consequences (in fact. intinitcly many of them).
tific theories are not just codificatory schemes. they are attempted descrip
Both the pragmatist and the conjectural realist can point out that we
tions of the reality hidden hehind the phenomena. And our present best
can't, on pain of infinite regress, account for everything, and one of the
theories are our present best shots at the truth. We certainly have reason
things we can't account for is why this stuff that allegedly does no more
to think that our presently best theories are our present best shots at the
than streamline the machinery of scientific proof or that turns out to be
truth (they stand up to the present evidence better than any known rival),
radically false should have turned out to be fruitful. There obviously can be
but we have no real reason to think that those present theories are true or
no question of any 'knockdown refutation' of either view. None the less, if
even closer to the truth than their rejected predecessors. Indeed, it can be
a position could be developed which accommodated some of the intuitions
accepted that the history of science makes it vcry unlikely that our present
underlying the 'no miracles' argument and yet which, at the same time.
theories are even 'approximately' true. They do, of course, standardly
cohered with the historical facts about theory change in science, then it
capture more empirical results than any of their predecessors, but this is no
would arguably be more plausible than either nra!2matism or conjectural
indication at all that they are any closer to capturing 'God's blueprint of
the universe'. The fully methodologically aware theoretical scientist realism.
Is it possible to have the best of both worlds, to account (no matter how
pursues his unended quest for the truth knowing that he will almost cer
tentatively) for the empirical success of theoretical science without run-
fail and that, even if he succeeds, he will never know, nor even have
foul of the historical facts about theory change') Richard Boyd and
any real indication, that he has succeeded.
occasionally Hilary Putnam have claimed that realism is itself already the
realism is certainly a modest. unassuming position. It can be
best of both worlds. They have claimed, more or less explicitly, that
formulated as a version of realism in the senses we have so far discussed
the picture of scientitlc change that I have painted is inaccurate. and so the
as saying in fact that we do have the best possible grounds for holding our
argument from scientific revolutions is based on a false premiss: the history
present best theories to be true (they are best confirmed or hest 'corrobor
of science is not in fact marked by radical theoretical revolutions
ated' by the present evidence): we should not even ask for better grounds (at any rate, not the l..:".~_" ~f 'n,~jl1rp' '<cie.nce). On the contrary, claims
than these: but since the best corroborated theory tomorrow may funda
mentally contradict the best corrohorated theory of today, the grounds Boyd:
a matter of
that we have for thinking the theories true are inevitably conjectural and The historical progress of the mature sciences is and
,.".hc/'>T\l-:1hlt,
practically, not just in principle) defeasible. I defended this conjectural more accurate approximations to the truth about and theoretical)
realist view myself in an earlier paper: presentations of the view nh"nnr""n~. Later theories typically build unon the
embodied in previous theories.
(almost invariably) met with the response that there is little, if any, differ
ence of substance between it and anti-realism,14 The main problem, r take Elsewhere he asserts that scientists generally adopt the (realist)
it, is again that conjectural realism makes no concessions to the 'no that 'new theories should ... resemble current theories with respect to
miracles' argument. On the conjectural realist view, Newton's theory does their accounts of causal relations among theoretical entities' (Boyd 197~:
" In discussion Richard Boyd acknowledged that he mnde no claim of approximate cont'
For my defence of conjectural reahsm see Worrall 19H2. The response of 'no real dif nf accepted scientific theories. But I had thought tilat
ference' between conjectural and anti-realism was made many times in seminars and the empirical success of theories give us grounds to think
private discussions (by van Fraassen amongst others). See also, Newton-Smith 1981. where metaphysical'. observation-transcendent) deseripti,'" of the reality underlying
realism is defined as including an 'epistemological ingredient' to thIS comecl s
the phenomena is at any rate approximately correct'! Several of Richard Boyd' comments
approach. I should add that I am of course ivine UD the realist suggested to me. 31 least. that he defends not a full-blown realism. but something like the
present paper only
defended.
structural realism that I try to formulate below. ~

~
 153
STRllCTURAL REAUSM
152 JOHN WORRALL

try is their chief examplc of an immature science, so tlley would be

Similarly, Putnam once claimed (I Y7R: 20) that many historical cases of happY to concede that phlogiston has been entirely rejected by later
theory change show that 'what scientists try to do' is to preserve 'as often science.!6 Presumably, some of the other items on Laudan's list of once
as oossible' the 'mechanisms of the earlier theory' orto show that they are scientifically accepted but now non-existent entities would receive similar
cases" of new mechanisms', I want first to explain why I think
that these claims are wrong as they stand, I shall then argue that valid treatment.
The cogency of this reply clearly depends to a large extent on whether or
intuitions underlie t he claims, but these intuitions are better captured in not some reasonahly precise account can be given of what it takes for a
a rather different position which might be called slrucfllral or syntactic science to achieve 'maturity', Neither Boyd nor Putnam has anything very
realism, precise to sayan this score, and this has naturally engendered the suspicion
Laudan has objected to Boyd and Putnam's claims by citing a that the realist has supplied himself with a very useful ad hoc device:
whole list of theoretical entities, like phlogiston, caloric, and a range of whenever it seems clear that the basic claims of some previously accepted
ethers, which, he insists, once tigured in successful theories but have now theory have now been totally rejected, the science to which that theory
been totally rejected (19R2: 231), How, Laudan wants to know, can newer belonged is automatically counted as 'immature' at the time that theory
theories resemble older theories 'with respect to their accounts of causal
relations among theoretical entities' if the newer theories entirely reject was accepted.
What is needed is a reasonably precise and independent criterion of
the theoretical entities of the old? How can relativistic physics be said to maturity. And this can, it seems to me, in fact bc 'read off the chief
preserve 'the mechanisms' of. say. Fresnel's account of the transmission of sustaining argument for realism-the 'no miracles' argument. This argu
light. when, according to Fresnel's account, transmission occurs via per ment, as I indicated before. applies only to theories which have enjoyed
iodic disturbances in an all-pervading clastic medium. while. according to genuine predictive success, This must mean more than simply having cor
relativity theory, no such medium exists at all? How can later scientists be rect empirical consequences-for these could have been forced into the
said to have applied to Fresnel's theory the principle that 'new theories framework of the theory concerned after the effects they descrihe had
should, , ,resemble current theories with respect to their accounts of already been observed to occur. The undoubted fact that various chemical
causal relations among theoretical entities' when these later theories en- experimental results could he incorporated into the phlogiston theory does
deny the existence of the core theoretical entity in Fresnel's theory? not on its own found any argument. even of the intuitive kind we are
Boyd alleges that the mechanisms of classical physics reappear as limiting considering, to the likely truth of the phlogiston theory. Similarly. the fact
cases of mechanisms in relativistic physics, Laudan replies that. although it that creationist biology can be made empirically adequate with respect to,
is of course true that some classical laws are limiting cases of relativistic say, the fossil record clearly founds no argument for the likely truth of the
ones. Genesis account of creation. Such empirical adequacy can of course easily
there arc other laws and gcncral assertions made oy the classical theory (e.g.. claims be achieved--for example, by simply making Gosse's assumption that God
ahout the density and fine structure of the etiler. ~cnerallaws aoout the character of created the rocks with the 'fossils' there already, just as they are found to
the interaction between ether and matter. models and mechal1lsms detailing the be. (Perhaps God's purpose in doing this was to test our faith,) But the fact
of the ether) which could not colleeivaoly he limiting cases of
The reason is a simple one: it theory cannot assign values to a
that this elaborated version of creationism is then bound to imply the
variable thaI docs not oceur in that theory's language . Classical ether empirical details of the fossil record is, of course, neither a miracle nor an
contained a number of postulated mechanisms for dealing inter alia with the trans indication that the theory 'is on the right track', The explanation for this
mission of light through the ether. Such mechanisms could not possihly appear in a 'success' is, of course. just that it is often easy to incorporate
successor theory like the special theory of n:latlvity which denies the vcry existence already known results ad hoc into a given framework, Nor is the success of
of an etherial medium and which accomplishes the explanatorv tasks performed by
the ether via very different mechanisms. (Lauclan I a theory in predicting particular events of an already known kind enough
on its own to sustain a 'no miracles' argument in favour of a theory, Even
Does the realist have any legitimate come-back to Laudan's criticisms? the most ad hoc, 'cobbled up' theory will standardly be predictive in the
Certainly some of Laudan's examples can be dealt with fairly straightfor of the benefit of the doubtl so far as to say that
wardly, Boyd and Putnam havc becn careful to restrict their claim of Ih '[Wle do not
'cssential' cumulativity to 'mature' science onlv. Pre-Lavoisierian chemis- "phlogiston" referred'
 154
JOHN WORRALL
STRl)CTURAL REALISM 155
sense that it will entail that the various results it has been made to absorb
will continue to hold in the future. (For example, the heavily epicyclic the realist. Let's concentrate on what seems to me (and to others") the
corpuscular theory of light developed in the early nineteenth century by sharpest such challenge: the ether of classical physics. Indeed, we can make
Biot, having had various parameters fixed on the basis of certain results in the challenge still sharper by concentrating on the elastic solid ether
crystal optics, implied, of course, that the 'natural' generalizations of involved in the classical wave theory of light proposed by Fresnel.
results would continue to hold in the future.) Theories will standardly Fresnel's theory was based on the assumption that light consists in
exhibit this weak predictiveness because, Popper or no, scientists do in. periodic disturbances originating in a source and transmitted by an all
stinctively inductively generalize on the results of wellcontrolled experi. pervading, mechanical medium. There can be no doubt that Fresnel him
ments which have so far always yielded the same results. But the success of self believed in the 'real existence' of this medium-a highly attenuated
such inductive man(uvres, though no doubt miraCUlous enough in itself, and rare medium all right, but essentially an ordinary mechanical medium
does not speak in favour of the likely truth likeness of any particular which generates elastic restoring forces on any of its 'parts' that are
explanatory theory. The sort of predictive success which seems to elicit the disturbed from their positions of equilibrium. J9 There is equally no doubt
intuitions underlying the 'no miracles' argument is a much stronger, more that Fresnel's theory enjoyed genuine predictive successnot least, of
striking form of predictive SUccess. In the stronger case, not just a new course. with the famous prediction of the white spot at the centre of the
instance of an old empirical generalization, but an entirely new empirical shadow of an opaque disc held in light diverging from a single slit. If
generalization follows from some theory, and turns out to be experiment_ Fresnel's theory does not count as 'mature' science, then it is difficult to see
ally confirmed. Instances of this are the prediction of the existence and what does.'"
orbit of a hitherto unknown planet by Newton's theory and the prediction Was Fresnel's elastic solid ether retained or 'approximately retained' in
of the white spot at the centre of the shadow of an opaque disc and of later physical theories? Of course, as I have repeatedly said and as realists
the hitherto entirely unsuspected phenomenon of conical refraction would admit, the notion of one theoretical entity approximating another or
Fresnel's wave theory of light. So my suggestion is that, instead of leaving of one causal mechanism being a limiting case of another is extremely
the notion of maturity as conveniently undefined, a realist should take it vague, and therefore enormously elastic. But if the notion is stretched too
that a science counts as mature once it has theories within it which are far, then the realist position surely becomes empty. If black 'approximates'
predictive in this latter demanding sense-predictive of general types of white. if a particle 'approximates' a wave, if a space-time curvature 'ap
phenomena, without these phenomena having been 'written into' the
theory. iH See e.g Hardin and Roscnherg I <J1l2. which tackle, this challenge on behalf of the realist
(sec helow. Pl'. I:;n7).
With this somewhat more precise characterization of maturity, Laudan's This is not to deny. of course. that Fresnel was also hv what was already known
ahout light. It is also true that at the time of work. much remained to be
list of difficult cases for the modified realist can indeed be pared down
diScovered ahout the dynamical properties of clastic soliJs. As a result. Fresnel's theory
considerably further. Laudan must be operating with some much weaker was dynamically delicienl in certain respects (especially when viewed in hindsight). But the
notion of empirical SUccess than the idea of predictive Success just ex fact that he failed to construct a flllly dynamically aucquatc theory of hght as a disturhanee in
an clastic solid medium (or hetter: the fact that his thellfV ran into certain fundamental
plained When he cites the gravitational ether theories of Hartley and namicai prohlems) docs /lot mean that Fr~snd diu not even aim at such a theory, nor that
LeSage as examples of 'once Successful' theoriesY Presumably he means did not intend the thcory he produccd to be interpreted in this waY. He clearlv thou~ht of
simply that these theories were able successfully to accommodate various as a disturbance in an elastic medium. and dvnamical anu
of an abstract. mathematical sort) certainly gu'iJcd his research. along with the emDirical data
already known observational results. But if we require predictive Success of
the strong kind indicated above, then surely neither Hartley's nor LeSage's JS no doubt that. as Whittaker pomtcd out (1951: I Ill). Selmc aspects of Frcsnel's
speCUlative hypothesis scored any such success. 111particular the discontinuity of the normal component of the displacement acroSS the
betwecn two media .. cohere rather hettcr with Maxwell's notion of a displacement
However there is no doubt that, no matter how hard-headed one is currcnt than they do with the idea an ordinary dynamical displacement. But. contra Hardin
about predictive SUccess. Some of Laudan's examples remain to challenge and Rosenhcrrr (who cite Whittaker). this doesn't mean that Fresnel was talking ahout
along; insteau. he was talking -in a Hawed and problematic w~y. -about

I have criticized Laudan on thiS point in WorraIl 1988b. Cf. Laudan 1982: 225 (also p. 115. this 'If that Irrcsnel's prediction of the "white
spol"l docs not count as empirical success. nothing

.........
 156 JOHN WORRALL
STRUCTURAL REALISM 157

proximates' an action-at-a-distance force, then no doubt the realist is right

None the less, there is something right about what they, and Boyd,
that we can be confident that future theories will be approximately like the
say. There was an important element of continuity in the shift from Fresnel
ones we presently hold. This won't, however, be telling us very much.
to Maxwell-and this was much more than a simple question of
rt does seem to me that the only clear-sighted judgement is that Fresnel's
over the successful empirical content into the new theory. At the same
elastic solid ether was entirely overthrown in the course of later
time, it was rather less than a carrying over of the full theoretical content
Indeed, this occurred, long before the advent of relativity theory,
or full theoretical mechanisms (even in 'approximate' form). And what
when Maxwell's theory was accepted in its stead. It is true that Maxwell
was carried over can be captured without making the very far-fetched
himself continued to hold out the hope that his electromagnetic field would
one day be 'reduced' to an assumption of Hardin and Rosenberg that Fresnel's theory was 'really'
about the electromagnetic field all along. There was continuity or accumu
the ether as Fresnel had conceived it. But in view of the failure of a whole
lation in the shift, but the continuity is one of form or structure, not of
series of attempts at such a 'reduction', the field was eventually accepted
content. In fact, this claim was already made and defended by Poincare.
as a primitive entity. Light becamc viewed as a periodic disturbance. not
And Poincare used the example of the switch from Fresnel to Maxwell to
in an elastic medium, but in the 'disembodied' electromagnetic field.
argue for a general sort of syntactic or structural realism quite different
One would be hard pressed to cite two things more different than a
from the anti-realist instrumentalism which is often attributed to him.
displacement current, which is what this electromagnetic view makes light,
This largely forgotten thesis of Poincare's seems to me to offer the
and an elastic vibration through a medium, which is what Fresnel's theory
had made it. hopeful way of hoth underwriting the 'no miracles' argument and ac
cepting an accurate account of the extent of theory change in science.
Hardin and Rosenberg (1982). replying to Laudan, suggest that. rather
speaking, it seems right to say that Fresnel completely mis
than trying to claim that Fresnel's elastic solid ether was 'approximately
identified the nature of light; but none the less, it is no miracle that his
preserved' in Maxwell's theory, the realist can 'reasonably' regard Fresnel
theory enjoyed the empirical predictive success that it did; it is no miracle
as having been talking about the electromagnetic field all along. This is
because Fresnel's theory. as science later saw it. attributed to light the right
certainly a striking suggestion! As someone influenced by Lakatos, I cer
structure.
tainly would not want entirely to deny a role to rational reconstruction of
Poincare's view is summarized in the following passage from Science and
history. Indeed. it does seem reasonable for a historian to reserve
Hypothesis. which begins by clearly anticipating the currently fashionable
the option of holding that a scientist did not jidly understand his own
'pessimistic induction':
theory; but to allow that he may have totally misunderstood it and. indeed.
that it could not really be understood until some 50 years after his death, The ephemeral nature of scientifie theories takes by surprise the man of the world.
to hold that Fresnel was 'really' talking about something of which we know TIleir brief period of pro~pcrity ended. he sees them abandoned one after the other;
he sees ruins piled upon ruins: he predkts that the theories in fashion today will in
he had not the slightest inkling, all this is surely taking 'rational reconstruc a short time succumb in their turn, and he concludes that they are absolutely in vain.
tion' too far. Even 'charity' can be overdonc. 21 Fresnel was ohviously This is what he calls the bankruptcy u( sciI'm!'. (1905: 160)
claiming that the light-carrying 'luminiferous aether' is an elastic solid,
obeying, in essence, the ordinary laws of the mechanics of such bodies: the But this passage continues:
ether has 'parts'; restoring clastic forces are brought into play when a part His scepticism is superficial: he does not take into account the object of scientific
is disturbed out of its equilibrium position. He was obviously claiming this, theories and the part they play. or he would understand that the ruins may still be
and it turned out that, if later science is right. Fresnel was wrong. Hardin good for something. No theory seemed established on firmer ground than Fresnel's,
which attributed light to the movements of the ether. Then if Maxwell's theory is
and Rosenberg's claim has a dellnite air of desoeratinn about it. nrpfF'rred today, docs it mean that Fresnel's work was in vain? No; for Fresnel's
was not to know whether there really is an ether. if it is or is not formed of
PUlnam has a well-known (and notoriously 'principle of charity' (or 'benefit of the
doubt') which says that 'when speakers specify a for a term they lise by a riesuimiml One critic who explicitly does not classify Poinnuc as an instrumentali,t is Zahar (se" his
)1

and, beeausc of mistaken factual beliefs that those speakers have. that description fails to The term 'structural realism' was also used by Grover Maxwell for a position which he
we should assume that they would accept reasonable reformulations of their descriptIOns' (197H: from Russell's later philosophy (see Maxwelll'J70a,b). Maxwell's position grows out of
23-4), different (more 'philosophicar) concerns. though it is dearly related to that of Poincare (one of
the points for further research is to clarify this relationship),

 158 JOHN WORRALL

STRUCTURAL REALISM 159

atoms. if these atoms really mov.; in this way or that: his object was to predict OPttcal
phenomena.'"' i and r be the made by the incident and refracted beams with the
This Fresnel's theory enables us to do today as wdl as it did before normal to a reflecting surfaee. Fresnel's equations then state
time. The differential "equations are always tiue. they may b.; alwflYs integrated
by the same methods. and the results of this integration still preserve their
value. RII =
+ r)
So far, of course, this might seem a perfect statement of positivistic instru_
mentalism: Fresnel's theory is really just its empirical content, and this is +r)cos(i - r))
preserved in later theories. However, Poincare goes on to make it X'II' = 2sin r .cosijsin(i + r)
explicit that this is not his position.
It cannot be said that this is reducing physical theories to simpio:
Fresnel developed these equations on the basis of the following picture
thesc equations expr.;ss relations. and if the equations remain true.
of light. Light consists of vibrations transmitled through a mechanical
relations preserve their reality. They tcach us now. as til
medium. These vibrations occur at right angles to the direction of the
such and such a relation between this thing and that; only tne something which we transmission of light through the medium. In an unpolarized beam, vibra
then called mOlion. we now call electric current. But these arc merely names of the tions occur in all planes at right angles to the direction of transmission
images we substituted for the real objects which Nature will hide for ever from our
The true relations hctw.;cn these rcal objects arc tile only reality we can attain. but the overall beam can be described by regarding it as the composition of
1(2) two vibrations: one occurring in the plane of incidence and one occurring
in the plane at right angles to it. The bigger the vibrations, that is, the larger
Poincare is claiming that. although from the point of view of Maxwell's
the maximum distance the particles are forced from their equilibrium
theory, Fresnel entirely misidentified the nature of light, his theory accu.
positions by the vibration. the more intense the light. I, R, X, etc. in fact
rately described not just light's observable effects but its structure. There is
measure the amplitudes of these vibrations, and the intensities of the light
no elastic solid ether. There is, however, from the later point of view, a
are given by the squares of these amplitudes.
(disembodied) electromagnetic field. The lIeld in no clear sense approxi.
From the vantage-point of Maxwell's theory as eventually accepted, this
mates the ether. but disturbances in it do obey formally similar laws to
account, to repeat, is entirely wrong. How could it be anything else when
those obeyed by elastic disturbances in a mechanical medium. Although
there is no elastic ether to do any vibrating? None the less, from this
Fresnel Was quite wrong about what oscillates, he was, from this later point
vantage-point, Fresnel's theory has exactly the right structure-it's 'just'
of view, right, not just about the optical phenomena, but right also that

that what vibrates according to Maxwell's theory arc the electric and
these phenomena depend on the oscillations of something or other at right

to the light.
field strengths. And in fact, if we interpret I, R, X, etc. as the
amplitudes of the 'vibration' of the relevant electric vectors, then Fresnel's
Thus, if we restrict ourselves to the Icvel of mathematical equations
equations are directly and fully entailed by Maxwell's theory. It wasn't,
not, notice, the phenomenal level-there is in fact complete continuity
then, just that Fresnel's theory happened to make certain eorrect predic
between Fresnel's and Maxwell's theories. Fresnel developed a famous set
of equations for the relative intensities of the reflected and refracted
tions: it made them because it had accurately identitled certain relations I
between optical phenomena. From the standpoint of this superseding
beams in various circumstances. Ordinary un polarized light can be
theory, Fresnel WaS quite wrong about the nature of light: the theoretical
analysed into two components: one polarized in the plane of incidence, the
mechanisms he postulated are not approximations to, or limiting Cases of,
other polarized at right angles to it. Let 12 , R2, and X2 be the intensities of
the theoretical mechanisms of the newer theory. None the less, Fresnel was
the components polarized in the plane of incidence of the incident, re
quite right not just about a whole range of optical phenomena, but right
flected, and refracted beams respectively: while 1'2 , R'!, and X'2 are the
that these phenomena depend on something or other that undergoes peri
components polarized at right angles to the plane of incidence. Finallv. let
odic change at right angles to the light.
about Fresnel's 'object' (see above. n. 19). However, the normative But then, Poincare argued, his contemporaries had no more
of how a theory ough! to be interpreted is. of course. logically mdepeo fOr regarding Maxwell as having definitively discovered the nature of light,
psychological question of what its creator in fact
as having discovered that it really consists in vibrations of the electromag-

 160 JOHN WORRALL STRUCTURAL REALISM 161

netic lield, than Fresnel's contemporaries had had for regarding Fresnel as of that predecessor as special cases of its own equations. 25
having discovered the nature of light. At any rate, this altitude towards But the general applicability of the correspondence principle certainly
Maxwell would be mistaken if it meant any more than that Maxwell built is not evidence for full-blown realism-but. instead, only for structural
on the relations revealed by Fresnel and showed that further relations realism.
existed between phenomena hitherto regarded as purely optical on the one Much clarificatory work needs to be done on this position, especially
hand and electric and magnetic phenomena on the other. concerning the notion of one theory's structure approximating that of
This example of an important theory change in science cenamly ap another. But I hope that what I have said is enough to show that Poincare's
pears, then, to exhibit cumulative growth at the structural level combined is the only available account of the status of scientific theories which holds
with radical replacement of the previous ontological ideas. It speaks, then, out realistic promise of delivering the best of both worlds: of underwriting
in favour of a structural realism. Is this simply a feature of this particular the 'no miracles' argument. while accepting the full impact of the historical
example, or is preservation of structure a general feature of theory facts about theory change in science. It captures what is right about Boyd's
in mature (i.e. successfully predictive) science') realism (there is 'essential accumulation' in 'mature' science at levels
This particular example is in fact unrepresentative in at least one than the purely empirical) and at the same time what is right about
ant respect: Fresnel's equations are taken over completely intact into the Laudan's criticism of realism (the accumulation does not extend to the
superseding theory-reappearing there newly interpreted but, as math fully interpreted top theoretical levels).
ematical equations, entirely unchanged. The much 1110re common pattern As one step towards clarifying the position further, let me end by sug
is that the old equations reappear as limiting cases of the new-that is, the that one criticism which, rightly or wrongly, has been levelled at
old and new equations are striclly inconsistent, but the new tend to the old scientific realism does not affcct the structural version. Arthur Fine has
as some quantity tends to some limit. strikingly claimed that
The rule in the history of physics seems to be that, whenever a theory
Realism is dead .. , lts death was hastened by the debates oVt.!r the of
replaces a predecessor, which has however itself enjoyed genuine pre quantllm theory where Bohr's non-realist
dictive success, the 'correspondence principle' applies. This requires the Einstein's passionate realism. (p. 21, this
mathematical equations of the old theory to re-emerge as limiting cases of
But realism has been pronounced dead before. Some eighteenth-cen
the mathematical equations of the new. As is increasingly realized,24
tury scientists believed (implicitly, of course: they would not have ex
the orinciple operates, not just as an after-the-event requirement on a new
pressed it in this way) that realism's death had been hastened by debates
if it is to count as better than the current theory, but often also as a
over the foundations of the theory of universal gravitation. But it is now
heuristic tool in the actual development of the new theory. Boyd (1984) in
surely clear that in this case realism was 'killed' by first saddling it with an
fact cites the general applicability of the correspondence principle as evi
extra claim which then proved a convenient target for the assassin's bullet.
dence for his realism. But the principle applies purely at the mathematical
This extra claim was that a scientific theory could not invoke 'unintelli
level, and hence is quite compatible with the new theory's basic theoretical
notions, such as that of action-at-a-distance, as primitives. A
assumptions (which interpret the terms in the equations) being entirely
realist interpretation required intelligibility, and intelligibility
at odds with those of the old. I can see no clear sense in which an action
at-a-distance force of gravity is a 'limiting case' of. or 'approximates', a )\ Putnam this account of Bovd's position in his 1978. adding that
space-time curvature. Or in which the 'theoretical mechanisms' of action spondcnce 'is often the Illmles! way 10 gel a theory that keeps the
this last remark very difficult to understand. IInw exactly could it be done
at-a-distance gravitational theory are 'carried over' into general otherwise? (I am assuming that what comes out is required to be a theory in some recognizable
theory. Yet Einstein's equations undeniably go over to Newton's in certain sense rather than simply any old collection of empirical statc:menls.) Zahar has shown (see n. 35)
special cases. In this sense, there is 'approximate continuity' of how the corresnondcnce principle can be used as a definitc heuristic principle
uidancc. But a scientist s(;:t out to obtain a theory
structure in this case. As Boyd points out, a new theory could capture its sU(.:cessfui empjncai consequences its pn...decessOI in some other way Ttlan
predecessor's successful empirical content in ways other than yielding the equations as limiting cases-surely he would be
dear idea of how to go about the task. (l am a,SSI
are excluded on the grounds thal they would fail to produce anything that anyone \UI!,,;JUUIll~
See e,g, Zahar 198.\1, and Worrall 1985. as well as floyd 1984. anti're"li,t) would regard as a theory,)

~
 162 JOHN WORRALL
STRUCTURAL REALISM 163

of the basic theoretical notions in terms of some anteced_ of state seem to cry out for
mgrained nature of certain classical metaphYSIcal
the Newtonian case of course this was the framework of as the idea that action-at-a-distance 'cried out' for
l-by-contact Without claiming to be an expert nation was a reflection of a deeDlv in!!.rained oreiudice for
in the foundations of quantum mechanics (and with all due respect for the
mechani cs).
peculiarities of that theory), it does seem to me that, by identifying the
The structural realist simply asserts. in other words, that, in view of the
realist position on quantum mechanics with Einstein's position, Fine is
theory's enormous empirical success, the structure of the universe is (prob
similarly saddling realism with a claim it in fact has no need to make. The
ably) something like quantum-mechanicaL It is a mistake to think that we
realist is forced to claim that quantum-mechanical states cannot be taken
need to understand the nature of the quantum state at aIL and, a fortiori,
as primitive. but must somehow be understood or reduced to or defined in
classical tcrms. a mistake to think that we need to understand it in classical terms. (Of
course, this is not to assert that hidden variables programmes were obvious
But the structural realist at least is committed to no such
non-starters. that working on them was somehow obviously mistaken--no
he explicitly disowns it. He insists that it is a mistake to think that we can
more than the structural realist needed to assert that the attempts at a
ever 'understand' the nature of the basic furniture of the universe. He
Cartesian reduction of gravity were doomed from the start. The only claim
what eventw,lIv in the Newtonian case. There the is that ultimately evidence leads the way: if. all efforts. no scientific
. successful empirically and so per';'tf>n can be constructed which incorporates our favourite
resistant to 'mechanistic reduction' that gravity (understood as a
assumptions. then no matter how firmly entrenched those principles
action-at-a-distance force) became accepted as a primitive irreducible no
be, and no matter how fruitful they may have proved in the past, they must
tion. (And action-at-a-distance forces becamc perfectly acceptable, and
ultimately be up.)
realistically interpreted. components of other scientific theories, such as
It seems to me. then, that. so long as we are talking about structural
electrostatics.) On the structural realist view. what Newton really dis
realism. the rcports of realism's death at the hands of quantum mechanics
covered are the relationships between phenomena eApressed in the math
are greatly
ematical equations of his theory. the theoretical terms of which should be
understood as genuine primitives.e,.
Is there any reason why a similar structural realist attitude cannot be
RE'ITRINCES
towards quantum mechanics? This view would he
orced from the 'classical' metaphysical prejudices of Einstein: that R (I lI73). Realism. Underdetermination and a Causal Theory of Evidence.'
cal variables must always have sharp values and that all physical events are 7: 1-12.
--(1984). 'The Current Status of Scientific Realism: In Lcplin (cd.) 19114: 4\-112.
determined by antecedent conditions. Instead. the view would simply
DUhem. P. (I lI06). The Aim lind Sintclure of Physical Theory. (pag.: references to
be that quantum mechanics does seem to have latched on to the real the translation hy Philip Wiener (New York: Atheneum. 1(2).)
structure of the universe, that all sorts of phenomena exhibited by micro
. 27 It is not in fact clear to mc that Fine', NOA Ithe natural ontological attitude) is ,ubstan
systems really do depend on the system's quantum state, which really
tlally different frolll Slluelulal realism. Structural lcalism perhaps sUppli('s a banner under
does evolve and change in thc way quantum mechanics describes. which hoth tho~e v.ho regard themselves as n.:ali,~ls and those who re~ard themselv('s as nnli
It is, of course. true that this state changes discontinuously in a way which realists of various sorts can untIl!.
Similar remark" about lhe 'anti~rca]jsr consequences
the theory does not further explain when the system interacts with a WIthout rderence to hne-by McMullin ([984:
'n"M~~".~._;~ system'-but then Newton's theory does not in a/so to
but simolv on~llIl~ji'~ that it Occurs. (Tnoee(J, no volume (Worrall [(880)
These last remarks on mechanics were modified and elaborated
of infinite regress.) If such tneet the raJ~C'd in discussion at Ncuch;tle) bv Professor
I wish to Ihank for some suggested improvements t() earlier drall: Elie Zahar
for numerous Ihis paper: and Howard Stein for his
Sec. ill particular. Poincare's discussion of the notion of force (l9{),';: H'I.. 1.l9j.
commcnls on the ver!:\)on
 STRUCTURAL REALISM 165
164 JOHN WORRALL

..--(1991). 'Feyerabend and the Facts.' In G. Munevar (cd.), Beyond Reason, 329
Fine, A. (1984). 'The Natural Ontological Attitude.' In Leplin (ed.) 1984: 83-107
repro as Ch. I. ' 53. Dordrecht: Kluwer Academic Press.
Zahar, E. G. (1983a). 'Logic of Discovery or Psychology of Invention?' British
Hardin, C. and Rosenberg, A. (1982). 'In Defence of Convergent Realism.' Philoso_
Journal for the Philosophy of Science. 34: 243-61.
phy of Science, 49: 604-15, _(1983b). 'poincare's Independent Discovery of the Relativity Principle:
Laudan, L. (1981). 'A Confutation of Convergent Realism.' Philosophy of Science,
48; repr. in Leplin (cd.) 1984: 218-49. and as Ch. VI. Fundamenta Scientwe, 4: 147-75.
Leplin. J. (cd.) (1984). Scientific Realism. Berkeley: University of California Press.
Maxwell, G. (1970a). 'Structural Realism and the Meaning of Theoretical Terms.' In
S. Winokur and M. Radner (eds.), Minnesota Studies in the Philosophy ofScience,
vol. 181-92. Minnesota: University of Minnesota Press.
--(1970b). 'Theories, Perception and Structural Realism.' In R. G. Colodny (ed.),
The Nature and Function of Scientific Theories. 3-34. Pittsburgh: University of
Pittsburgh Press.
McMullin. E. (1984). 'A Case for Scientific Realism.' In Leplin (ed.) 1984: 8-40.
Miller. D. (1974). 'Popper's Qualitative Theory of Verisimilitude'. British Journal
for the Philosophy of Science. 25: 166-77.
Musgrave. A. (1988). 'The Ultimate Argument for Scientific Realism.' In R. Nola
(ed.), Relativism and Realism in Sciences. 229-52. Dordrccht: Kluwer.
Newton-Smith, W. (1981). The Rationality of Science. London: Routledge and
Kegan Paul.
Niiniluoto. I. (1987). Truthlikeness. Dordrecht: Reidel.
Oddie, G. (1986). Likeness to Tmth. Dordrecht: Reidel.
Poincare. H. (1905). Science and Hypothesis. Repr. New York: Dover, 1952. (Page
references are to the Dover edn.)
Putnam. H. (1978). Meaning and the Moral Sciences. London: Routledge and Kegan
Paul.
Quine, W. V. O. (1953). 'Two Dogmas of Empiricism.' Tn From a Logical Point of
View, 20-46. Cambridge, Mass.: Harvard University Press.
Tichy. P. (1974). 'On Popper's Definition of Verisimilitude.' British Journal for the
Philosophy of Science. 25: 155-60.
Van Fraassen, B. (1980). The Scientific Image. Oxford: Clarendon Press.
Whittaker. E. 1'. (1951). History of Theories of Aether and Electricity. The Classical
Theories. London: Nelson.
Worrall. 1. (1982). 'Scientific Realism and Scientific Change.' Philosophical Quar
terly, 32: 201-31.
--(1983). 'An Unreal Image.' British Journal for the Philosophy of Science, 35:
65--80.
--(1985). 'Scientific Discovery and Theory Confirmation.' In J. Pitt (cd.), Change
and Progress in Modem Science. 311-14. Dordrccht: Reidel.
--(1988a). Review of Leplin (cd.) 1984. Philosophical Quarterly. 38: 370~6.
---(1988b). 'The Value of a Fixed Methodology.' British Journal for the Philo
sophy of Science. 39: 263-75.
- - ( I 989a). 'Fresnel, Poisson and the White Spot: The Role of Successful Predic
tion in Theory-Acceptance.' In D, Gooding, T. Pinch and S. Schaffer (cds.),
The Uses of Experiment-Studies of Experimentation in Natural Science, 135-57.
Cambridge: Cambridge University Press.
-(l989b). 'Scientific Revolutions and Scientific Rationality: The Case of the
"Elderly Hold-Out'''. In C. Wade Savage (cd.). The Justification. Discovery and
Evoillfion of Scientific Theories. 319-54. Minneapolis: lJniversity of Minnesota
Press, 1990.