Вы находитесь на странице: 1из 30

Stud. Hist. Phil. Sci., Vol. 29, No. 1, pp.

33-62, 1998
Pergamon
0 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0039-3681/98 $19.00+0.00

Duhem’s Physicalism

Paul Needham*

Duhem is often described as an anti-realist or instrumentalist. A contrary view has


recently been expressed by Martin (1991) (Pierre Duhem: Philosophy and History
in the Work ofa Believing Physicist (La Salle, IL: Open Court)), who suggests that
this interpretation makes it difficult to understand the vantage point from which
Duhem argues in La science allemande (1915) that deduction, however impec-
cable, cannot establish truths unless it begins with truths. In the same spirit, the
present paper seeks to establish that Duhem is at any rate not the kind of
anti-realist he is often presented as being, and that his views are like those Quine
sees fit to call realist. An interpretation of Duhem’s views on explanation and
precision in science, and their bearing on the epistemological status of theory, is
advanced which leads naturally into his critique of conventionalism. His attitude
towards atomism, which should not be judged from a post-1925 perspective, is
considered part of the unified view he strove after and appropriately called
Duhem’s physicalism, standing in contrast to the kind of reductionist conception
usually associated with atomism.

1. Introduction
Perhaps the most familiar way Duhem’s purported instrumentalism has been
expressed is by labelling him a conventionalist. A primary source of this view
among philosophers seems to be Popper, who established something of a
tradition of bundling Duhem together with PoincarC as a perpetrator of the
conventionalist strategy (Popper, 1972, p. 78, note 1). Thus, Worrall suggests
that what he calls instrumentalism ‘is to be found in the work of the
turn-of-the-century French conventionalists-principally Pierre Duhem and
Henri Poincart’ (Worrall, 1982, p. 203). But it is by no means uncommon to see
historians taking the use of this label as given. Dolby, for example, simply takes
it as read that Duhem is a ‘conventionalist positivist’ (Dolby, 1984, p. 388). And
Charles Gillispie is quoted on the cover of the recent reprint of the English
translation of Duhem (1914) claiming that ‘The central proposition of this
famous book is that physical theories are conventions serving to economize
scientific thought rather than descriptions or explanations of the way the world
is made’.
*Department of Philosophy, University of Stockholm, S-106 91 Stockholm, Sweden.
Received 13 November 1996; in revisedfbrm 13 March 1991.

PII: soo39-3681(98)ooo14-9

33
34 Studies in History and Philosophy of Science

Justification for applying the conventionalist label to Duhem, beyond


Popper’s say-so, is something its proponents are remarkably casual, and
sometimes careless, about. Worrall (1982, p. 203) renders what he sees as the
essence of Duhem’s argument for instrumentalism by construing a sentence
from the middle of Duhem (1914) as a conclusion drawn from a premise in the
form of a sentence occurring 161 pages earlier! Christie (1994) has more recently
gone so far as to quote Duhem’s statement of the conventionalist position as
though he were affirming the view himself. Although this would be a step in the
direction of matching what Poincare’s commentators do, namely attempt to
provide a statement of the conventionalist thesis they attribute to him (cf.
Giedymin, 1982), Duhem is actually describing Poincare’s position in order to
go on to attack it-a point which is discussed in Section 5 below. Pursuing
Duhem’s argument for rejecting conventionalism reveals considerations which
are not easily accommodated in an instrumentalist view and so present a
challenge for the instrumentalist interpretation. The finer points of interpret-
ation of experiment and theory which Duhem emphasises do, however, call for
a certain subtlety, which stands in conflict with a naive realist view, and this
may be enough for the Popperian charge of instrumentalism. Nevertheless,
although many realists have wanted to insist on a literal interpretation of
theories, questions arise (even on a piecemeal view like Popper’s) of where this
puts anyone willing to interpret some greater or lesser part of theory literally
but reluctant to view all theory literally. The holistic considerations Duhem
adduce show this criterion to be even less straightforward; a realistic realism
cannot speak naively of simplistic literal interpretation without qualification.
Duhem’s argument against Poincare is formulated in terms of an example,
the law of multiple proportions, which is related to the issue of how chemical
formulas are to be understood. This raises the issue of the status of atomism,
which was, it should be remembered, largely a matter for chemistry in the 19th
Century. As is well known, Duhem did not hold that there were atoms and was
antagonistic towards some of his contemporaries who held the opposite view.
Others, such as van? Hoff, he chided for not making their position on the issue
sufficiently clear. What he actually says on the matter is seldom discussed,
however, and certainly not by those who subscribe to the popular view which
classifies him among the ‘positivists ... for whom unobservable entities under-
lying the observed phenomena were anathema’ (Achinstein, 1994, p. 417).
Nowhere, to the best of my knowledge, does Duhem offer the mindless
argument here attributed to him that since we can’t see atoms there can’t be
any. It is, in fact, difficult to understand how any such argument would be
consistent with Duhem’s general holistic attitude to the testing of theories. If,
on the other hand, it is granted that the statement that the philosophical stance
of realism requires acceptance of atomism is not a timeless tautology by which
philosophers of all ages are to be judged, then the fact that Duhem held an
Duhem’s Physicialism 35

alternative view of chemical structure, as well as the kind of arguments he


promoted against the atomic account of chemical structure, must be taken into
consideration in the assessment of his position.
This is an appropriate point to mention that the ridiculous situation to which
Perrin gives expression in his Atoms (Perrin, 1913, p. 16), of the elimination of
all mention of atomic theory from the French school curriculum during the
latter part of the 19th Century, was highlighted much earlier by Duhem. He
says, by way of complaint, of the instigator of this policy,

Leaving to his brilliant rival Wurtz the glory of being the apostle of chemical atomism
in France, Mr Berthelot, driven by his evil genius, declared himself the adversary of
the new doctrines and notations. He used his great authority to stop them at the doors
of our Faculties, of our Primary Schools and of our Grammar Schools. He became
isolated in an organic chemistry of his own ... (Duhem, 1897, p. 390).

Duhem had himself come to grips with the representation of chemical structure
by chemical formulas in his Le mixte et la combinaison chimique (Duhem, 1902),
much of which had already appeared in his 1892 article ‘Notation atomique
et hypotheses atomistiques’. This latter title indicates a distinction between
notation and substantive theory which might suggest a conventionalist inter-
pretation. But it would be wrong to assume that Duhem’s rejection of atomism
amounts simply to a rejection of the theory in the spirit of instrumentalism,
while the notation is retained as merely ‘useful’. Atomism is not the only
possible ontological view of matter, so that rejecting it does not amount to
instrumentalism. Duhem explicitly contrasts atomism with the Aristotelian
view of matter in the 1902 book, and the same contrast is there in the 1892
article, although Aristotle is not mentioned by name.’ Now no one dreams of
calling Aristotle an instrumentalist on the grounds that he opposed atomism
and offered an alternative in its place, and Duhem should be judged by the same
standard. If Duhem restricts himself to merely pointing out the Aristotelian
alternative in the 1892 article, he betrays a distinctly more sympathetic attitude
in the 1902 book, emphasising the idea that the components from which
mixtures are formed are present only potentially, and not actually, in the
mixture, but now the Aristotelian view of the transmutability of the elements
gives way to a conception of the elements more in line with Lavoisier’s. This
interpretation raises many questions; the modern reader naturally wishes that
Duhem had himself pressed the details further, and perhaps considered
alternative non-atomic accounts. But one of his general lines of criticism of
atomic theories was their vagueness, and I think it would be fair to say that his
view was at least as clear and detailed as those he criticised.
The fact that his view of chemistry was not an atomistic one means that it was
not a reductionist one as this term has come to be understood, in the spirit of

‘For fuller documentation, see Section 1 of Needham (199613).


36 Studies in History and Philosophy of Science

Dirac’s famous pronouncement that ‘quantum mechanics is now almost


complete’ save for the problems of conciliation with relativity ideas, which
give rise to difficulties only when high-speed particles are involved, and are therefore
of no importance in the consideration of atomic and molecular structure and ordinary
chemical reactions. The underlying physical laws necessary for the mathematical
theory of a large part of physics and the whole of chemistry are thus completely
known (Dirac, 1929, p. 714).

Nevertheless, as his criticism of Berthelot’s distinction between chemical and


physical processes makes abundantly clear, Duhem maintained a general view
which is appropriately called physicalist, according to which phenomena are
unified by being subject to the same general principles rather than constituting
two different worlds. The point, already suggested by relatively superficial
observations such as that ‘When making use of instruments of physics-a
thermometer, a manometer, a calorimeter, a galvanometer-the chemist and
the physiologist implicitly admit the accuracy of the theories which justify the
employment of these instruments’ (Duhem, 1894, pp. 1855186) is pursued
below. It is difficult to understand the importance Duhem attached to this drive
for unity otherwise than in terms of a search for general explanations and a
dissatisfaction with ad hoc manoeuvres.2

2. Explanation
It might be objected that explanation is something Duhem thought he could
and should do without. However, this is one of the more easily discussed
misconceptions to which his unfortunate choice of terminology gives rise. What
is Duhem concerned to deny with his rhetorically extravagant diatribes against
explanation? Certainly, that the 19th Century fetish for mechanistic models
yields explanations, and a major reason for this was their circularity, if not their
incoherence. Thus, Thomson’s model with spiral springs can hardly elucidate
Navier and Poisson’s theory of elasticity, Duhem argues: ‘One look at it would
be enough to disappoint greatly anyone who might have expected an explana-
tion of the laws of elasticity; how, indeed, would the elasticity of the spiral
springs be explained?’ (Duhem, 1914, p. 75; his emphasis). Duhem goes on to
cite with evident glee what he takes to be Thomson’s own admission that the
model does not, after all, provide an explanation, by quoting the following
passage:
‘Strangely enough, Dolby (1984) seems to recognise implicitly some such problem with his
interpretation, for he argues that as a ‘conventionalist positivist’ who ‘limited his science to the
search for a mathematical description’ (pp. 388-389) Duhem really had no grounds on which to
oppose Berthelot’s applying the distinction between chemical and physical change in just such a
way as to save his law of maximum work. Contrapositively, since Duhem did oppose Berthelot for
thus reducing his law to a tautology (what the logical positivists came to think of as a truth by
convention)-a ‘ridiculous tautology’ (Duhem, 1897, p. 370)-and Dolby offers no independent
support whatsoever for the conventionalist charge. I see this as clear evidence against the
conventionalist interpretation.
Duhem’s Physicialism 31

Although the molecular constitution of solids supposed ... in our model is not to be
accepted as true in nature,
still the construction of a mechanical model of this kind is
undoubtedly very instructive (quoted by Duhem, 1914, p. 75; his emphasis).
Duhem’s criticism is to the effect that mechanical models do not actually
provide explanations, and not that explaining is altogether out of place in
science. Another theme in the diatribe is the conception of explanation typified
by what was offered by the Cartesian physicists, which is where Duhem takes
up his critique of explanation. He rejected the claims of rationalistic insight,
along with the 17th Century dogma that mechanistic speculations about atomic
structure provided the basis of informative explanations.3 At the very best, he
says, atomic tales are consistent with the phenomena. (A contemporary point of
criticism (see Earman, 1989, p. 44) of the modern version of instrumentalism
offered by van Fraassen, echoes Duhem’s view, namely that mere consistency
with the phenomena is not all we normally ask of a theory which is said to deal
with the phenomena.) Consciously emulating Newton’s critique of hypotheses,
Duhem dismisses vague and unclear gestures which, even if they can be clarified
to the point where they can be said to agree with some phenomena, are
repudiated by others. But just as Newton’s commandeering of the term
‘hypothesis’ for this particular purpose is generally regarded as unfortunate, so
Duhem’s expropriation of the word ‘explanation’ to describe what all sides
would agree is unacceptable, is unfortunate in so far as many examples of what
we would call explanations (hypotheses) cannot be rejected for the same
reasons. For, clearly, rejecting Cartesian explanation is not to reject explana-
tion in general, and just as what Newton says about hypotheses is no obstacle
to describing him as employing what we would call the hypothetico-deductive
method,4 so there is no good reason why an exposition of the views Duhem
argued for should be constrained to follow his extravagant usage.5
Duhem’s ‘invincible scepticism regarding the real existence of molecules and
atoms’ (Duhem, 1900, p, 16; from a passage quoted at greater length below) is
often taken as an expression of his rejection of what we would call explanation.
But rejecting a particular explanation is not to be confused with rejecting
explanation altogether; on the contrary, it might well be prompted by a concern
for what actually does explain the phenomenon in question. The instrumentalist
interpretation gives no account of the character of Duhem’s critique, and makes
a mystery of why Duhem should bother with arguments against atomic theories
based on their specific features. As already noted, Duhem nowhere appeals to
‘The purely speculative character of 17th Century atomism is now, somewhat belatedly, being
acknowledged; see Clericuzio (1990) and Chalmers (1993).
“As Duhem (1892c, pp. 147-148) points out himself.
‘It does not require much of an excursion into the realms of speculation to suggest that, just as
the great French theoreticians of the 18th and early 19th Century took over the development of
Newton’s mechanics and established the tradition Duhem so admired, he would have written that
the British physicists of the 19th Century he despised as ‘broad and weak’ took over the Cartesian
tradition, were this not to cast a Frenchman in the same light.
38 Studies in History and Philosophy of Science

the general argument that atomic theories are implausible because atoms
cannot be seen, as might be expected of an instrumentalist, but follows a
strategy reminiscent of Aristotle’s criticisms of atomism, which sought to show
that the doctrine failed to deliver both the explanations it claimed to provide
and those it could reasonably be expected to supply. Now it might be said that
a critic who rejected the whole enterprise of explanation would be interested
in gathering support for this view by showing that particular explanatory
endeavours failed, so that the instrumentalist interpretation being questioned
here is at least consistent with the Duhemian strategy of critique. This would be
a poor explanation of Duhem’s strategy, however. It is surely more natural that
a critic with an interest in explanation should be concerned to eliminate a
proposed explanation on internal grounds to do with the specific details of its
working. The conclusion that Duhem thought explanation in general a
misguided goal is far too strong for an argument appealing to failure due to the
specific details of the putative explanation. At the very least, then, it should be
recognised that the rejection of atomism does not automatically imply a
disinterest in explanation; there remains an onus of proof on the shoulders of
a proponent of the thesis that Duhem rejected the goal of explanation in
general.
There is another general aspect of 19th Century antagonism towards
atomism with which Duhem sympathised, namely that associated with the
question of the explanatory value of specific atomistic theories. This is the line
originally pressed against Dalton by Davy and Wollaston, namely that the laws
of definite and multiple proportions do not call for atomic theory. Like
Thomson’s spiral springs, Dalton’s minimal homogeneous lumps of various
elemental kinds, which he understood atoms to be, raise exactly the same
question they were supposed to answer and therefore do not provide any
explanation. This objection in Aristotelian spirit concerning the explanatory
value of specific atomistic theories leads naturally to the general idea of
parsimony, that one of the values of good theory construction, which has to be
balanced against others, is that its claims should be minimal, confined to what
seems to be relevant and otherwise freed from ad hoc appendages. Of course,
the instrumentalist is also trying to capitalise on the virtues of parsimony. But
this virtue is not the sole prerogative of the instrumentalist, and interest in
economy of thought does not suffice for the title of instrumentalist; it is an
essential feature of good explanation.

3. The Whole Truth

Quine has spoken of what he calls a tight theory formulation, without for a
moment giving up on ontological commitment as a criterion for existence
claims, or what he has come to call his naturalism or immanent standard of
Duhem’s Physicialism 39

truth. It is from a Quinean perspective, I suggest, that Duhem’s many assertions


should be understood about how interpretation of experimental results is to
be conducted in the light of what theories the theorist admits. Proponents of the
instrumentalist interpretation will, of course, see in this word van Fraassen’s
notion of acceptance of whatever saves the phenomena. But Duhem does not
only use ‘admit’ in speaking of the physicist who, when testing a hypothesis,
‘does not confine himself to making use of the proposition in question; he
also makes use of a whole group of theories admitted by him as beyond
dispute’ (Duhem, 1894, p. 189). He also uses it when contrasting such a
physicist’s view with that of ‘a physiologist who admits that the anterior roots
of the spinal nerve contain the motor nerve-fibres and the posterior roots of the
sensory fibres’ (Duhem, 1894, p. 183). But no one suggests the physiologist
is not to be understood as holding these beliefs about the spinal nerve true.
Similarly, he says that ‘theologians .. . did not admit that Galileo’s physical
and mechanical arguments could run counter to their cosmology’ (Duhem,
1893, p. 75). But no one suggests that the theologian here is not denying the
truth of Galileo’s contentions. The onus of proof lies with the proponent of
the instrumentalist interpretation to justify the imposition of his own
preferred reading of ‘accept’ where Duhem uses the word in contexts of the
first kind, given that it differs from his use in contexts of the second and
third kinds.
Too hasty a comparison with van Fraassen’s conception of underdetermina-
tion of theory by observation is a further source of unmotivated interpretation,
overstating sensible notes of caution in much the same way Duhem’s scepticism
towards particular explanations is overstated as a rejection of any notion of
explanation. Thus, even when brought to a reasonable degree of articulation,
Duhem maintains there may still be no grounds for accepting one of two
putatively more fundamental theories. He quotes approvingly Ampere com-
menting on two proposals for the explanation of the equations for the
propagation of heat:

The former shows merely that the equation at issue results from their way of seeing
things, and the latter that it can be deduced from the general formulas for vibratory
movements; neither add anything to the certainty of this equation, but merely show
that their various hypotheses can be upheld (Duhem, 1893, p. 82).

This says nothing about the status of such theories should one of them be
shown to do additional work-to yield new experimental results or encompass
other known phenomena which have hitherto resisted unification with the
theory of the propagation of heat. Duhem recognised fruitfulness as a mark of
success, and merely cautioned against taking apriori prejudice and unmotivated
excess theoretical baggage for knowledge. He was much concerned with the
limitations on the claims reasonably made by the physicist to knowledge of
40 Studies in History and Philosophy of Science

what is true. Overinterpreting Duhem’s caution as an expression of van


Fraassen’s idea that belief in empirical adequacy is all that can be justified
obscures this central line of Duhem’s thought-a line which is pursued in the
next section.
Not only is underdetermination to be distinguished from inductive uncer-
tainty by considering what is compatible with all possible observations; it is
not to be confused with holism-the thesis that statements can only be said
to imply consequences in the context of others, so that incompatible data
can, as far as logic is concerned, only reflect on this whole context. As
Quine clearly recognises, underdetermination is not established by an argu-
ment for holism, but requires a distinct motivation, and anyone seeking to
ascribe to Duhem implicit appeal to this thesis, which he does not state, must
find such a distinct argument in Duhem’s writings. This is not to say that
holistic considerations would not impose requirements on an acceptable
articulation of the underdetermination thesis, so that appeal to empirically
equivalent versions of classical mechanics, for example, would bear on too
narrow a domain to establish a holistically interesting thesis. The difficulties
Quine has encountered in settling on any definite thesis of underdetermina-
tion warn of the problems facing the imposition of any such thesis on
Duhem. Still, by way of contrast with van Fraassen, it might be noted that
Quine, when tempted by his ‘empiricist sensibilities’ to adopt what he calls
the ecumenical position on the issue of underdetermination as he understands
it, was never attracted by van Fraassen’s instrumentalism: two empirically
equivalent systems were each counted true, since their inconsistency can be
rendered merely apparent by Davidson’s device of respelling (Quine, 1990,
p. 13).
Ariew (1984) and Vuillemin (1986) have in recent years suggested that
Duhem should be distanced from Quine, despite the affinity indicated by the
so-called Duhem-Quine thesis, because Quine’s holism is more wide-ranging.
They make a great deal of the contrast Duhem draws between physics and
physiology, although what little he says about physiology as a domain where
holism has no application is contradicted by remarks such as that quoted at the
end of section 1, but more importantly, does not seem to have any systematic
import for his position. The same cannot be said of the contrast they do not
mention, between physics and geometry, to which Duhem appeals more
systematically in making certain points about the nature of scientific knowl-
edge, and which will be taken up in the next section. But criticism of
conservative views of mathematics such as Duhem’s, which is part and parcel
of Quine’s more wide-ranging holism, presents no real obstacle to seeing this
broader view as a development of, rather than as fundamentally antagonistic
towards, a tradition of Duhemian thought, as Quine himself suggests.
Against any such association Vuillemin maintains that
Duhem’s Physicialism 41

To Duhem’s neutral science purged of metaphysical claims Qume opposed a realist


science merging into metaphysics by its ontological commitments (Vuillemin, 1986,
p. 609).

But merely pointing, as Vuillemin does, to Duhem’s disparaging remarks about


‘metaphysics’ does little to show that Duhem denied what Quine affirms.
Whatever else Duhem may have understood by ‘metaphysics’, one clear and
simple point shines through:
As for metaphysical systems, they might suggest a proposition of physics; but physics
alone can decide if this proposition is correct or incorrect (Duhem, 1893, p. 61;
italicised in original).

This is comparable to the rejection of any notion of a higher court of appeal,


distinct from physics itself, which decides truth in physics, and which Quine
insists is not available for mortals, who must claim ‘(p)hysics is factual . . .
simply for want of a higher tribunal’ (Quine, 1986a, b, p. 155). Now for Duhem
metaphysics is, it might be said, a source of truth independent of science, like
religion. But this observation is not inconsistent with the comparison just made.
Simply allowing that whatever is said in metaphysics and religion can be
maintained without fear of conflict with science does not imply that holding
theses in the former domains to be true requires that nothing in the latter
domain can be held true. Being independent does not mean being a higher,
better or unique source of truth. We might think that good sense would have
difficulty in reconciling the striving for logical unity in science to be discussed
shortly with belief in such independent sources of truth; but that is not to say
that the stance is logically incoherent. By the same count, the position is not
strengthened by denying science is a source of truth.
Quine himself went along with Vuillemin’s interpretation, explaining

... there is the difference between Duhem’s fictionalistic attitude towards physics and
my realistic attitude. This is due to my naturalism, which recognises no higher truth
than what we seek in our aggregate scientific system of the world (Quine, 1986a, b,
p. 619).
But he too provided no independent support for this interpretation of Duhem,
and I can only speculate that he was inclined to say so because of Duhem’s
scepticism towards atomism, without considering how he argues. I doubt,
however, that Quine would really want to bind the attitude of naturalism and
realism as he understands them to specific 20th Century theories, and I venture
to suggest that there is much in common between his notion of ontological
commitment and Duhem’s famous vision of the ontological order emerging
with the completion of science. Following the opposition to explanations
without empirical import of the first chapter of Duhem (1914), the second
chapter is more constructive in tone and his concern with the truth could not be
clearer, claiming that as experiment ‘confirms the predictions obtained from
42 Studies in History and Philosophy of Science

our theory, we feel strengthened in our conviction that the relations established
by our reason among abstract notions truly correspond to relations among
things’ (p. 28). Now truths are used to say something of things, things are what
can be said to exist, and ontology is the traditional name given to the branch
of metaphysics concerned with what exists; thus,

the more complete it [physical theory] becomes, the more we apprehend that the
logical order in which theory orders experimental laws is the reflection of an
ontological order, the more we suspect that the relations it establishes among the data
of observation correspond to real relations among things, and the more we feel that
theory tends to be a natural classification (pp. 26-27).

But if Duhem’s attitude towards science is to be regarded as realist in broadly


Quinean naturalistic terms, there are certain qualifications and subtleties built
into his view which, while not fundamentally inconsistent with Quine’s general
position, merit more attention than Quine gives them.
Consistency is a necessary condition of truth. But it is not always easy to see
how theories can accommodate one another. A particularly glaring, if un-
Duhemian, example of this is the concomitant use by working scientists of
macro- and micro-theories, the continuity claims of the former conflicting with
the basic tenets of the latter. This has led many a would-be realist into the
reductionist camp, where what cannot be reduced to micro-theories is rejected
-eliminated. But this is only the brave face put on the official position.
Reductionism is a programme with many outstanding problems, and it is
simply not true that whatever has not been reduced is rejected-not, at any rate,
if the Quinean standards of ontological commitment and holding true in the
spirit of naturalism are applied to theories actually employed. The reduction-
ist’s way of pursuing realism is not particularly realistic. Any interpretation of
Duhem, on the other hand, must be pursued in the light of his non-reductionist
physicalism. The physicalist attitude is borne of a general motivation to
accommodate the whole world within a single theoretical perspective, and
to deal with all phenomena consistently. But this ideal is, as Duhem was only
too well aware, difficult to realise in practice, and the aspirations of realism
must be tempered by the impositions of a realistic attitude if it is to be at all
relevant to living science. Scientists, Duhem says,

have an intuition that logical unity is imposed on physical theory as an ideal to which
it tends constantly; they feel that any lack of logic, any incoherence in this theory, is
a blemish, and that the progress of science should gradually remove this blemish.

Is there a single one among [those who defend the right of theory to logical
incoherence] ... who hesitates for an instant to prefer a rigorously coordinated theory
to a junk heap of irreconcilable theories . ? Therefore, ... like all physicists they
regard the physical theory which would represent all experimental laws by means of
a single, logically coordinated system as the ideal theory; and if they tend to stifle their
Duhem’s Physicialism 43

aspirations towards this ideal, it is solely because they believe it unrealisable and
because they despair of attaining it (Duhem, 1914, pp. 294295).
Despite such problems, signs of realist aspirations are to be found, for example,
in the cavalier attitude scientists often adopt towards the rigors of mathematical
continuity, reflecting an instrumentalist attitude towards certain aspects of
what a literal interpretation of mathematical formalism would seem to imply
for the greater realist good of holding onto as much as possible of what is most
important.
Many remarks might be made on this less than happy position. A famous
Quinean view which might be thought relevant here is that ontological
commitment is revealed by the mathematics actually employed, and regardless
of what the practitioner might himself say of his endeavours, he cannot escape
the existential presuppositions of the mathematics unless he can show how he
can actually do without them. But the applicability of this criterion is
complicated by the fact that what passes for mathematics in actual science is
a far cry from the pristine state in which the pure mathematician delivers
the theorems and techniques of analysis. It is rather a hodgepodge of
considerations which would have Euclid rolling in his grave, where the
mathematician’s strictures on proof are not allowed to prevent the drawing of
wanted conclusions. This raises questions about how to regiment actual
practice and evaluate the results--balancing what can be achieved in simple
cases with the difficulties posed by the complexities of the general case. In this
connection, Duhem brings some very interesting insights to bear in the course
of developing an account of the import of the general use of what physicists call
approximation for what the physicist can reasonably claim to know.
Duhem argues that the commitments and disclaimers involved in knowledge
claims raise subtle issues of interpretation relating to the question of exactly
what is actually held true, leading to a view of scientific knowledge according
to which a naive statement of a law, without mention of limits of error or any
allusion to the delimitation of the range of conditions under which it is
supposed to hold, cannot be justifiably held true. Duhem’s holistic thesis is
intimately connected with this epistemological view, which takes the failure
of the parallel between refutation in science and reductio ad absurdum in
mathematics to have implications about the character of laws. In particular,
the naive view, which would make no distinction of kind between the
mathematically formulated physical law and the ‘definite and precise’ laws
of geometry, cannot be sustained, and account must be taken of what he
calls the relative and provisional character of laws. This epistemo-
logical position, on which Duhem bases his rejection of Poincare’s convention-
alism, is sketched in the next section, and provides a setting in which some of
the issues broached above can be discussed in more detail in the ensuing
sections.
44 Studies in History and Philosophy of Science

4. The Propositions of Physics

The central thesis of Duhem’s holism is that understanding the import of a


statement is a matter of understanding what can reasonably be claimed when it
is held true, that is, of what is actually inferred from it; but sentences do not
imply anything in isolation. These ancillary hypotheses and theories contribute
to what Duhem calls an interpretation of the phenomena ‘according to
admitted theories’ (for example, Duhem, 1914, p. 153). A very important aspect
of this interpretational apparatus is concerned with experimental error, includ-
ing the tracing of sources of systematic error, which leads to an estimate of how
precisely the predicted phenomena are delimited and within what limits the laws
on which the prediction is based can be said to hold. The estimate of precision
restricts the import of a statement to consequences drawn by not exceeding
these limits. This is what Duhem calls the relative character of laws-that in
virtue of which they are defensible by the physicist of a given age only up to a
certain degree of approximation. Laws are also provisional because the bearing
of new conditions, not previously anticipated, must be incorporated into a
better formulation. Little did Boyle appreciate the need to take into account the
intensity of the electric field in which the gas might be placed; he most likely
hadn’t the faintest idea what an electric field was. Accordingly,

The degree of approximation of a law, though sufficient today, will become


insufficient in the future through the progress of experimental methods; sufficient for
the needs of the physicist, it would not satisfy someone else, so that a law of physics
is always provisional and relative . . . the laws of physics cannot be maintained except
by continual retouching and modification (p. 178).

Who would maintain that scientific laws are certain under these conditions-
certain in the sense that it can be confidently maintained that they will still be
upheld 50 years from now? Duhem contrasted scientific judgements with the
immediate judgements of ordinary experience, whose certainty can normally be
upheld because they make no particular claims to precision (Duhem, 1914,
Pt. II, Ch. V, section 5).6 Scientific judgements are also contrasted with those of
mathematics.

Physics does not progress as does geometry, which adds new final and indisputable
propositions to the final and indisputable propositions it already possessed; physics
makes progress because experiment constantly causes new disagreements to break out
between laws and facts, and because physicists constantly touch up and modify laws
in order that they may more faithfully represent facts (p. 177).

In assessing the truth value of statements of these various categories Duhem


seems to be concerned with what can be held true ‘for all time and for all men’.
Of a judgement of ordinary experience
‘Duhem’s ‘judgements of ordinary experience’ should be compared with the standing sentences
Quine calls ‘observation categoricals’, such as ‘Where there’s smoke, there’s fire’.
Duhem’s Physicialism 45

we may ask ‘Is it true?’ Often the answer is easy; in any case the answer is a definite
yes or no. The law recognised as true is so for all time and for all men; it is fixed and
absolute (p. 178).

This ‘fixed and absolute’ character which, in Duhem’s view, is also a feature of
mathematical statements, lends itself to the straightforward accumulative
conception of geometric progress. But the provisional and relative character of
laws precludes treating physics naively as making claims which can be defended
as fixed and absolute. Duhem is at pains to emphasise the difficulty of
articulating a scientific judgement, tempered as it must be so as not to overreach
what can reasonably be justified, which is why it is never immediate, never easy
to assess. The familiar mathematical form in which laws are usually presented
is misleading, in so far as they might be taken as mathematical truisms-
truisms whose assessment, if not always easy, is at any rate straightforward. So
presented, laws certainly cannot be upheld as true; they are neither true nor
false. Boyle’s law, taken in this naive way, says that two variables P and V
ranging over real numbers and representing pressure and volume are inversely
proportional to one another, so that a certain real number as value for P
corresponds to a particular real number for V. This cannot be upheld as a literal
truth; but by the same count, nor can it be denied. Any such claims are
indefensible.
This may amount to a denial of what has sometimes been taken as
characteristic of the realist position, namely the affirmation of the truth of laws
literally interpreted. But surely, a position which denies this specl$cally in order
to reflect how laws are reviewed ‘in order that they may more faithfully
represent facts’ is not to be confounded with an instrumentalist critique. For
what is in question is quite clearly a better theoretical representation:

... when we criticize Regnault for not having taken . .. [the compressibility of gases]
into account . .. [w]e are criticizing him for having oversimplified the theoretical
picture of these facts by representing the gas under pressure as a homogeneous fluid,
whereas by regarding it as a fluid whose pressure varies with the height according to
a certain law, he would have obtained a new abstract picture, more complicated than
the first but a more faithful reproduction of the truth (p. 158).

Ordinary testimony fails to give a reproduction of the truth which can make
any claim to precision. Science teaches that the situation can often be improved
and precise claims made by judicious use of mathematical statements. But
truths of pure mathematics, however precise, have a completely different
bearing. If they are to bear on, and perhaps lend some precision to what is said
about, objects in space and time, it must be recognised that a mathematical
statement is endowed with ‘physical meaning only if it retains a meaning when
we introduce the word “nearly” or “approximately” ’ (p. 215).
When used as vehicles for expressing laws, Duhem says that mathematical
statements are subject to a certain degree of indetermination (indgtermination).
46 Studies in History und Philosophy of Science

Part of what this involves is reflected in the way that considerations of what can
be distinguished by the senses lead to saying that a result ‘can be indifferently
represented by it doesn’t matter which of these numbers’, and expressing ‘The
infinity of possible estimates ... by writing for example that one hundred
atmospheres makes an increase in the electromotive force of the battery of
(0.0845 f 0.0005 volts)’ (Duhem, 1894, p. 202). Elaborations might be offered
about the appropriate statistical interpretation of such spreads. But whatever
the exact quantitative significance of such estimates, we must also reckon with
the possibility that the whole result can be shifted due to some cause of
systematic error. It is the task of the theoretical critique to root out and take
account in appropriate fashion of all such sources of error. This may well
require the use of theories apparently dealing with subjects far removed from
the considerations upon which the original conception of the experiment was
based. Since there is no criterion determining when such deliberations have
been completed, beyond the fact that nothing else has yet occurred to anyone,
the experimenter is all too aware of the provisional or incomplete state of the
descriptions he offers. He does not merely not know that he knows; he does not
even know whether he has succeeded in articulating a claim which will prove,
in the long run, to be either true or false. But although past experience suggests
that it is a fair bet that future modifications will be required, presently justified
formulations are presently offered as claims to knowledge, not as expressions
lacking truth value but as claims to know what is true.
Nevertheless, despite the fact that Duhem characterises corrections to
experimental results and modification of laws in physics as improving a
description ‘too simple and too far removed from reality’ by providing another
which ‘better symbolises the concrete instrument’ (Duhem, 1894, pp. 204205)
and which yields, as we saw, ‘a more faithful reproduction of the truth’, he does
say that laws are neither true nor false. The instrumentalist interpretation
thrives on such pronouncements. But it can be met by considering what his
arguments show he was concerned to deny when he says they are neither true
nor false, namely, as has already been suggested, that they are universally and
eternally defensible truths. What also speaks for the appropriateness of this
strategy is that Duhem is not consistent. Sometimes he denies that the
physicist’s statements are true by claiming they are false, rather than lacking a
truth value. This he does, for example, when he closes the 1894 article with a
fable summarising his views of the interpretation of experiment. A botanist
searching for a rare tree comes across two peasants, one of whom tells him
merely ‘There is one of these trees in this wood’, while the other directs him to
‘Follow the third path that you come across, take a hundred paces, and you will
be at the foot of the tree you seek’. Doing as he is told, however, the botanist
‘fails to reach the object of his research; to reach the foot of the tree a further
five paces are necessary’. Of these two pieces of information Duhem says ‘the
Duhem’s Physicialism 41

first was true and the second was false’, although the second was of greater
value (p. 227).
This tale does get the issue right: it concerns truth, and illustrates how easy
it is make a claim which is true by being sufficiently indefinite-in something
like the sense in which an existentially quantified sentence is indefinite relative
to a corresponding instantiation. It is unfortunate, however, partly for not
emphasising that the indefinite statement is true only because some much
more definite one is also true, but mainly for omitting to draw the correct
parallel regarding the second peasant’s statement. Taken naively, as though
it were a statement of geometry, the second statement is indeed false. But
Duhem has gone to great lengths to explain that this is not how the physicist
views his attempts at precision. These are statements ‘whose sense would
remain unintelligible for anyone who does not know physical theory’ (Duhem,
1894, p. 226) and are misconstrued as statements which certain procedures
would easily show to be false. Properly understood, they should be assessed in
the light of an appropriate estimate of the limits of error and the reliability of
the interpretation. The botanist falsifies the second statement if he construes it
naively, as literally saying that the tree stands at a place 9144 cm (=lOO yards)
along the third path. If, under the circumstances, a shortfall of 5 yards falls
within a reasonable ascription of error, it is not proved false. The moral of
Duhem’s paper is surely that this antecedent must be discussed before
any conclusions can be drawn about the truth of the statement in question,
properly understood as a precise claim-that is, one with a specified degree of
indefiniteness.
Although Duhem always speaks of precision in terms of experimental error,
analogous considerations relating to the use of approximations in the applica-
tion and development of theories should also be considered to have a bearing
on the import of theoretical statements.’
The price of precision is that in making claims specific, their sense has to be
articulated in such a way that they are laid open to the possibility of a critique
whose consequences cannot be foreseen. Regnault’s neglect of the compress-
ibility of gases was of relatively minor import, and could be addressed in a
revaluation of the experimental results. Had it come to light that the material
under observation was not at equilibrium, the value of the whole experiment
might have been called into question. Questions might be raised about how
the physicist’s understanding of his considered opinions can be explicitly
formulated-or regimented, as Quine would say. But the point should be clear
enough that the purely mathematical statements dealing with functions on real
numbers and suchlike, which the physicist uses, are not themselves expressions
of the physicist’s propositions. The need to respect physical propositions as
‘Contrary to Vuillemin’s view, this seems to me to argue against the compartmentalised view of
nature he develops in opposition to holism. But it is not possible to expand on this theme here.
48 Studies in History and Philosophy of’ Science

such and not treat them as though they were statements of pure mathematics is
at the root of Duhem’s objection to conventionalism.

5. Duhem’s Critique of Conventionalism

Despite the fact that Duhem is frequently described as a conventionalist, he


never, as far as I am aware, used this term to describe himself, and he explicitly
distanced himself from the major contemporary advocate of this doctrine,
Poincare. Duhem understands this doctrine to be the view that certain very
general propositions of science are immune from trial by experiment and
observation because they are really definitional specifications of the usage of
key terms in the physicist’s vocabulary. Poincare suggested that the principle
of inertia, for example, is applied in such a way that it cannot be contradicted
by experiment. Apparent discrepancies are accounted for by postulating the
influence of forces such as friction and gravitation. But Duhem fails to see in
this a feature distinguishing some laws from others. Factors introduced to
explain the lack of true inertial movement are of exactly the same kind as the
corrections introduced in the application of precise statements of law. A picture
of the experimental situation is provided with the aid of a host of additional
assumptions, and these may well include the phenomena of friction and
gravitation. An experiment’s failing to agree with predicted behaviour does not
tell us which of these many assumptions is to be rejected, although ‘among the
theoretical elements entertained there is always a certain number which the
physicists of a certain epoch agree in accepting without test and which they
regard as beyond dispute’ (Duhem, 1914, p. 211).
Duhem compares Poincare’s claim about mechanics with analogous views
about the law of multiple proportions. If simple elements A, B and C combine
in the ratio of their masses u:b:c to form a compound M, and they also combine
to form another compound M’, then the mass ratio of the three elements in the
new compound will be xa:yb:zc, where x, y and z are whole numbers. It can be
argued that this law is not subject to experimental test because chemical
analysis will only reveal the composition of a compound M’ to a certain degree
of approximation. Integers x, y and z can always be found such that the relation
between the masses of the three elements is represented by the ratio xa:yb:zc to
any specified degree of approximation. Hence, the law of multiple proportions
can never be shown to be wrong by experiment. But this, Duhem retorts, is to
view the law of multiple proportions as ‘deprived of all physical meaning’
(Duhem, 1914, p. 215). It gains physical meaning only by the appropriate
insertion of the word ‘approximately’, without which the statement degenerates
into the mathematical truism that a real number can be approached arbitrarily
closely by a ratio of integers. The word ‘approximately’ signals that a
simple-minded, direct comparison with observation is out of the question. Any
Duhem’s Physicialism 49

attempt to compare one hypothesis with experience must be mediated by


others; ‘it would be absurd to subject the law of multiple proportions .. . to a
direct test’ (p. 215).
Much the same holds for the conventionalist’s arguments for regarding the
laws of mechanics as definitions. They only show that these laws are beyond the
reach of direct empirical refutation, and in this respect not at all unlike any
other law. To make the further claim that they are immune from the test of
experience ‘would be a serious error’ of trying to ascribe sense to separate
domains of theory and observation independently. Ariew and Vuillemin make
a lot of Duhem’s concession that in less advanced sciences such as physiology,
experiment might result in ‘a raw description of the facts’ (Duhem, 1914,
p. 182). But, at any rate in physics, fundamental laws which form the
cornerstone of whole theories or systems of beliefs are still part of a system
which comparison with experiment can only show is, as a whole, ill adjusted to
the facts. Part of what the conventionalist claims would then be correct, namely
that nothing compels us to reject the cornerstones rather than tampering with
other hypotheses instead. Equally, nothing can guarantee that the constant
adjusting of the ancillary assumptions and postulating of extraneous causes will
not come to be regarded as ad hoc, and an adjustment to a long-cherished
principle deemed more satisfactory. The plausible element in the convention-
alist’s thesis can thus be explained in a way which allows the rejection of the
paradoxical element that laws are definitions on the basis of the principle that
no hypothesis has physical meaning in isolation. Logic cannot dictate where the
adjustment should be made; but by the same count, logic cannot be brought to
the aid of any single hypothesis by elevating its status to that of a definition.
Certainly, it would be ‘awkward and ill inspired’ for the physicist to look
among firmly entrenched views for the culprit when faced with recalcitrant
data; but this does not amount to logical impossibility. The physicist’s choice of
hypotheses is never compelled by logical necessity. ‘Perhaps someday . .. by
refusing to invoke causes of error and take recourse to corrections ... and by
resolutely carrying out a reform among the propositions declared untouchable
by common consent, he will accomplish the work of a genius who opens up a
career for a theory.’ (p. 211) The principle that light travels in straight lines in
homogeneous media was maintained for thousands of years. It is the principle
we rely upon when checking for a straight line by sight. But the day came when
diffraction effects ceased to be explained by the intervention of some cause of
error and optics was given a new foundation.
Duhem does not deny that conventions of a kind have a role to play. But he
distinguishes between those involved in the essentially arbitrary choice of, say,
a scale of temperature and the ‘exacting conventions arbitrarily imposed by the
physicist’ (Duhem, 1892c, p. 157) which he took to be the source of mechanical
theories. His discussion of compositional formulas (which, as we will see, he
50 Studies in History and Philosophy of Science

distinguishes from structural formulas) as based on the laws of constant and


multiple proportions is concerned with the extent to which these laws determine
compositional formulas. Water’s gravimetric composition of hydrogen to
oxygen in the ratio 18 is reflected in the usual compositional formula of H,O.
Clearly, the ratio 2:l in the formula is not uniquely determined by the
gravimetric ratio, and Duhem considers to what extent a comparison with
hydrogen’s gravimetric ratio to, for example, sulphur, chlorine, phosphorus
and carbon in compounds whose usual compositional formulas are H,S, HCl,
PH, and CH, restricts possible indices in the compositional formula for water.
Duhem argues that there is a certain room for manoeuvre, and H,O can be
considered the correct compositional formula for water only with respect to
certain conventions (Duhem, 1892a, p. 397; Duhem, 1902, p. 73). He delimits
the range of scales available, emphasises the need for consistency, the leeway
allowed by logic, and the role of good sense in making a sensible choice. But
there is no question of the law of multiple proportions itself being a convention.
Later on in the 1902 book he presents essentially the same formulation of
the import of the law as that reviewed above, but without discussing
Poincare. After correcting Dalton’s original formulation by lifting the restric-
tion to simple (small) whole numbers, which was no longer justified in view
of the homologous series of organic compounds, he introduces his anti-
conventionalist line of thought as an answer to the question ‘What, then, is the
exact sense of the truth that we have just stated?’ (Duhem, 1902, p. 145; my
emphasis). This sense is what is lost when laws are ‘deprived of all physical
meaning’ (my emphasis) and reduced to ‘mathematical truisms’. Truth of laws
must respect physical interpretation.

6. Duhem’s Understanding of Chemical Structure

The definite amounts of specific elements assigned to a given amount of a


compound by a compositional formula might be taken in different ways, either
as indicating the actual existence in a quantity of the compound of parts which
are the corresponding elements, or in Aristotelian fashion as indicating the
potential existence of kinds of stuff that would be created as a result of certain
so-called decomposition processes. These two broad alternatives can themselves
be elaborated in different ways. Duhem clearly preferred this second line of
approach (Duhem, 1902, ch. lo), although he evidently also thought that it
was not necessary to press either line in order to capture the properties
of compounds with a chemical formula.
Now it was clear by Duhem’s time that there is more to the nature of
compounds than the proportions of elements specified in compositional
formulas, and that the notion of a chemical formula which represents a
compound should be developed to reflect this. The fact that the features
Duhem’s Physicialism 51

captured by a compositional formula are only necessary and not sufficient to


capture the nature of a compound led Berzelius in 1832 to propose calling
‘substances of similar composition and dissimilar properties isomeric, from the
Greek iaopepqs (composed of equal parts)’ (Leicester and Klickstein, 1952,
p. 265). Duhem describes how further structure required to accommodate
the various features associated with isomerism is reflected by elaborating the
‘crude’ compositional formula (la formule chimique brute) into what he
variously calls a formule dPveloppke or a formule de constitution-what will be
called a structural formula here. He thinks that this development does not
require an atomistic interpretation, that there are anyway good reasons for not
accepting the atomic theory, and that it can be understood on the basis of his
preferred Aristotelian interpretation, although he is at pains not to give undue
prominence to this interpretation. He develops the notion of a structural
formula in several stages.
First Duhem considers the structure discernible from the use of chemical
substitution. Dumas proposed calling compounds derived from one another by
replacement, either immediately or mediately via some route of successive
substitutions, as being of the same chemical type. Types are conveniently named
after a compound from which other members of the type are derived by
substitution, Members of the ammonia type, for example, are generated by
replacing an equivalent of hydrogen in ammonia by, for example, organic
radicals or groups such as methyl, CH,, ethyl, C,H,, and so on. If the
compositional formula of ammonia is NH,, then the compositional formula for
methylamine, formed by substitution of an equivalent of hydrogen by an
equivalent of the methyl group, is CNH,. But its membership of the ammonia
type is better displayed by indicating that it can be considered to be derived
from ammonia by substitution of one equivalent of hydrogen, thus: N(CH,)H,.
Similarly, dimethylamine is written N(CH,),H, and trimethylamine N(CH&.
Duhem emphasises the important role played by the prediction and actual
synthesis of new substances in the development and consolidation of the
conception of structure based on substitution.
Members of the water type are likewise displayed as results of substitution of
one or more equivalents of hydrogen. Thus, ethyl alcohol can be represented by
the formula (C,H,)HO, and alcohols in general by formulas of the kind
(C,H,,+,)HO. Ethers, on the other hand, can be regarded as resulting from the
replacement of each of two equivalents of hydrogen in water by groups of the
kind CnHZn+,. In particular, dimethyl ether can be represented by the formula
(CH,),O. This compound is isomeric with alcohol, having the compositional
formula C,H,O in common with it, and the fact that two different substances
with the same elemental composition are distinguished by differences in
physical and chemical properties is reflected by construing two different
formulas each compatible with the same compositional formula.
52 Studies in History and Philosophy of’ Science

H H H H

I I I
H-C-C-C-H
I
H-rrT=o I II I
H H H H 0 H

(a) (b)
Scheme 1. Structurul formulus of isomers (a) propionaldehyde and (b) acetone

Duhem goes on to describe how the notion of substitution can be developed


to encompass the idea of several equivalents of an element being replaced by a
single equivalent of an element or a group, which can be expressed in terms
of the notion of valency. Representing the elements or groups a and b as
being united in a given compound by exchanging n valencies by drawing n
lines between a and b in the formula, the fully articulated structural formulas
are obtained by writing all the valence exchange links between the appropriate
elements. The isomers propionaldehyde and acetone, for example, are
represented by the structural formulas (a) and (b), respectively, of Scheme 1.
These two structural formulas highlight what Duhem calls the difference in
chemical function of the two isomers. For example, oxidation of propion-
aldehyde leads to substitution of hydrogen which changes the aldehyde group
OCH into a carboxylic acid group OCOH, whereas oxidation of acetone yields
acetic acid, (CH,)OCOH, and formic acid, HOCOH, all of which is under-
standable on the basis of the structural formulas. Other isomers may be
quite similar chemically but nevertheless display differences in so-called
physical properties such as density, boiling point and freezing point, as
exemplified by ortho-, para- and meta-dichlorobenzene. Again, this chemical
similarity of distinct compounds is understandable on the basis of the structural
formulas.
Duhem’s view is that there is no need to see these structures as representing
arrays of atoms in molecules. They can equally well be understood as
representing how the compounds would behave in various chemical environ-
ments, just as the original compositional formulas could be taken to represent
what would be produced as a result of certain chemical action on the compound
stuff. Moreover, he considers the atomistic interpretation as he understood it
and dismisses it on the grounds that it can only be saved from contradiction
with experimental results by introducing so many ad hoc qualifications as to
make it worthless as an explanatory and fruitful theory. (Whether the atomistic
view he considers is the best that could be mustered at the time is, perhaps, open
to discussion; the critique in the 1902 book is merely an elaboration of a line of
criticism developed in the 1892 article. But no pre-1925 theory was entirely free
from the sort of objections Duhem raised.)
Duhem’s Physicialism 53

Up to this point, all is plain sailing; the first trial arises when Duhem
elaborates the notion of a structural formula to accommodate stereoisomerism.
The suggestions of Le Be1 and van? Hoff, that the four valencies of carbon
should be represented as issuing from the four corners of a regular tetrahedron
and converging on a central point, are characterised by Duhem as involving a
‘new element taken from geometry’ (Duhem, 1902, p. 128; stereochemistry is
not considered in the 1892 article). Surely, one might think, the fully fledged
three dimensional geometric structure must be taken as a representation of
items actually so arranged in space.

7. Duhem’s Challenge
But Duhem did not, it seems, accept that the stereochemical representation
was in fact a fully fledged geometric picture. For anyone who might make such
a claim, he thought, is committed to holding a coherent view about the bodies
whose spatial relations are purportedly pictured by a stereochemical formula-
the atoms as bodies in space. This is clearly the challenge Duhem issued to van?
Hoff in the following passage, taken from the end of a section in which
Pasteur’s discovery of stereoisomers and van? Hoffs proposal have been
presented.
I have just mentioned the words scientz$c symbolism, and in the foregoing, faithful to
the ideas which I developed here in connection with the atomic notation, 1 have in fact
assigned as the proper object of stereochemistry the creation of a notation, the
construction of schemas appropriate to represent truths of the experimental order. I
have attributed to these schemas no relationship with the constitution of matter itself,
no power of revealing the quidproprium of chemical combinations. Does my way of
seeing things conform in this respect to that of Mr J. H. van’t Howl Does the
knowledgeable professor of Amsterdam and Berlin agree with my invincible scepti-
cism regarding the real existence of molecules and atoms? Or is he not, on the
contrary, convinced that at the foundation of chemical combination there is a
structure, an edifice, whose material is the indestructible atoms of compound bodies?
Is he convinced that chemical research can gradually unveil for us the plan of this
edifice, and that the theory of tetrahedral carbon leads us to knowledge of this plan?
This I would not dare to deny, for Mr van? Hoff nowhere in his book gives the
stereochemical representations for the simple symbols. I would no more dare to affirm
this either, for fear of taking literally what might be intended figuratively (Duhem,
1900, pp. 16-17).
The indefiniteness of atomic hypotheses is one line of objection that Duhem
pressed against the atomic theory. Another is the inadequacy of the details,
when such are to be found. Both lines of criticism are found in the 1892 article
and repeated in the 1902 book.
Duhem introduces the discussion of atomism in the 1892 article by claiming
that, traditionally, proponents of the two opposing philosophical views,
according to which matter is either infinitely divisible or atomic, held their
54 Studies in History and Philosophy of Science

respective opinions to be ‘certain’, ‘even evident’ and ‘incontestable’ (Duhem,


1892a, p. 440). On the former view, when copper and sulphur combine, ‘Two
different substances, continuously filling two different volumes, are transformed
into a third substance continuously filling a volume and possessing properties
which are no longer those of sulphur nor those of copper. In other words, there
is no longer any sulphur or copper; there is only copper sulphide’. On the latter
view, ‘when sulphur and copper combine, neither the copper nor the sulphur
cease to exist; their atoms are imperishable. The atoms of sulphur and those
of copper simply become arranged beside one another, grouped in a certain
way ...’ (Duhem, 1892a, p. 441). Doubts might be raised about whether the first
of these purported consequences really follows merely from the former view as
stated. Duhem seems only to have entertained the Aristotelian version,
however,s and does not elaborate the continuous view of matter further.
Perhaps he did not want to lay himself open to charges analogous to those he
brought against the atomic theory.
As was suggested in section 2, much of Duhem’s polemic against explanation
is a polemic against appeal to a priori insight. If, as he goes on to argue, there
is no evidence which requires atomic hypotheses to explain it, then the only
grounds for maintaining atomism would be a priori ones. Why, in Duhem’s
view, are the atomic hypotheses otiose?
Mathematicians had shown how the use of the word ‘infinity’ in analysis can
be understood without taking it to refer to an object. Similarly, Duhem claimed
that the word ‘atom’ need not be taken to denote, as Quine (1960, p. 90 note 1)
uses the term, or be true of, anything; and the derived words ‘atomic weight’
and ‘atomicity’ can just as well be replaced by ‘equivalent weight’ and ‘valency’,
to which he gives sense without appealing to atoms. He acknowledges, in the
spirit of the passage from Ampere we saw him quoting in section 3, that from
the atomic theory, ‘it is easy to deduce the fundamental laws of chemistry, as
Dalton showed’ (Duhem, 1892a, p. 441). If there is to be any point to the
doctrine, however, it must be able to do more than this and actually explain the
formation of compounds with the ‘atomicities’ attributed to atoms. Alas, ‘It is,
in fact, easier to describe how the atomic school interpose atomicity in the
phenomena of substitution than to ascertain how it explains this peculiar
property of the atom’, and the best that can be found by way of ‘making precise
the intimate nature of these atomicities .. has perhaps the defect of over-
resembling the classical hooks of the Lucrecian atoms’ (Duhem, 1892a, p. 445).
He goes on to quote at length passages from Wurtz describing the ‘capacity for
combination’ represented by the notion of atomicities, all bringing to mind
what is usually said about the dormitive quality of opium. But what is more,
there are numerous facts which ‘are difficult to explain on the view that isolated
atoms possess a determinate number of atomicities, whatever the property
‘Another possibility is outlined in Needham (1996a).
Duhem’s Physicialism 55

might be in virtue of which these atomicities might be explained’ (Duhem,


1892a, p. 450).
The fact that the valency of an element varies from one compound to another
was a serious problem for atomism, although not for Duhem’s account,
requiring a distinction between different kinds of atomicity for a single atom.
For otherwise, in the case, say, of nitrogen, which is sometimes trivalent,
sometimes quintivalent, ‘if the five atomicities were absolutely identical, reasons
of symmetry would render absurd the existence of compounds such as ammonia
where three of these atomicities would be satisfied while the other two would be
free’ (Duhem, 1892a, p. 448). Now

Take ethylamine, in which the ethyl group C,H, is fixed by an atomicity of the first
order of the nitrogen atom. If this substance is combined with hydroiodic acid, its
elements will be fixed by atomicities of the second order, and we obtain the
hydroiodate of ethylamine [i.e. (C,H,)NH,I].

Now take ammonia, in which nitrogen’s three atomicities of the first order are
saturated by three hydrogen atoms. If it is combined with ethyl iodide, iodine will be
saturated by one of the atomicities of the second order and the ethyl group will
be fixed by the other and we thus obtain a body whose composition is the same as in
the preceding example.

These two bodies of the same composition are formed in different ways. In the one,
the ethyl group is fixed by an atomicity of the first order and in the other by an
atomicity of the second order. Since these two atomicities of different kinds cannot be
identical, the two compounds cannot be identical either, they should be two isomeric
substances.

But experiment shows that these two compounds are not two different isomers, but
one and the same substance (Duhem, 1892a, p. 449).

Many similar such cases, Duhem says, are to be found, and goes on to expound
the virtues of his own account of valency, according to which nitrogen is only
assigned a valency relative to a compound in which it is bound, three or five
depending on the type to which the compound belongs.
The atomic interpretation also leads to the prediction of differences in crystal
structure which would reflect what, on the atomic account, are differences in
molecular structure, but fail to materialise. Potassium nitrate and potassium
chlorate, for example, have different structural formulas but are isomorphous
salts.
The adequacy of this evaluation of the evidence available at the end of the
last century might be discussed. The absence, for example, of any discussion of
the theory emerging in the mid-1890s of steric hindrance, accounting for
varying reaction kinetics of esterification of di-derivatives of benzene, might
be thought a weakness of the 1902 book. But Duhem cannot reasonably be
accused of arguing against atomism on general positivist grounds.
56 Studies in History and Philosophy of Science

8. The Goal of Comprehensive Theory

It is interesting to note that in the same year that he first published his views
on atomic notation, Duhem was lamenting the fact that

Suddenly, the path they (d’Alembert, Clairaut, Lagrange, Laplace, . ) had laid open
was forgotten; ... the problems that had occupied their minds were regarded as futile
and childish; and while the higher minds took refuge in the realm of mathematical
combinations devoid of all reality, the great mass of students turned to the
ascertainment of facts, to experimentation without theory, without idea (Duhem,
1892b, p. 161).

The negative attitude towards theory of which Duhem complained still


dominated French science at the beginning of the 19th Century. As Nye (1979)
observes, in reacting to a relativist variant of conventionalism,

the French scientific community, including Poincare himself, may well have sought
defensively to emphasize more than ever experimentalism and the dependence of
science on observed data, rather than to take risks in speculative metaphysical
approaches. If so, the influence of the Boutroux Circle in the long run may have
increased the already experimental character of French science at the turn of the
century, a character which still could be seen after the First World War in French
skepticism about quantum mechanics and general relativity theory, including
misgivings about the work of their own Louis de Broglie (Nye, 1979, p. 120).

And we read that, despite a promising 1914 doctoral thesis by RCnC Marcellin
on reaction rates,

no French scientist took any part in the very swift creation of the quantum
theory of
the chemical bond and of the quantum description of chemical reaction paths, which
fitted so well in the general framework of Marcellin’s thesis (Gueron and Magat,
1971, p. 7).

It would therefore be a considerable distortion of the facts to suggest, as is


sometimes done, that the younger generation of French scientists led by Jean
Perrin and Paul Langevin, who criticised Duhem’s stand against atomism, did
so by advocating a theoretical perspective against one overly concerned with
observation. In particular, they did not rise to the challenge that I am
suggesting Duhem posed.
As it turned out, developments since 1925 showed that many of the sorts of
problem facing atomism in the classical perspective could no longer be
formulated, and saved modern atomism from incoherence such as that which
all agreed afflicted the Bohr atom. But the early atomists did not meet Duhem
eye to eye, adducing rather experimental evidence, such as Perrin’s observation
on Brownian motion, in favour of their cause. The suggestion that Duhem
would have dismissed any such experimental evidence, on the grounds that it
did not actually amount to seeing atoms, conflicts with all he says, at great
length, about an experiment being an interpretation of observed phenomena.
Duhem’s Physicialism 51

Mach might have thought that what you can’t hit with a hammer has no claims
to existence, but the criterion will not wear on Duhem. In any case, why, if he
really was an instrumentalist, should such a man not immediately accept the
new corpuscular theory of matter as ‘empirically adequate’ in the manner of
van Fraassen? Some of Duhem’s modes of expression, such as the emphasis on
notation in the passage on van’t Hoff, the polemic against explanation, and so
on, do, admittedly, suggest instrumentalist inclinations. But they make a
mystery of this central question. His response is less mysterious if, in accord-
ance with his holistic attitude to the interpretation of experimental results and
the understanding of physical laws, Duhem’s main concern is seen as the
development of a coherent overall picture. We have seen (section 3) that he was
aware of the difficulties confronting scientists which may well ‘tend to stifle their
aspirations towards this ideal’. But claims about what is true are nevertheless
constricted by this necessary condition of consistency and coherence, and the
question posed by atomism is whether it is acceptable as a balanced scientific
judgement which aspires to truth.
These attitudes are very much apparent in Duhem’s critique of Berthelot-
also an opponent of atomism. Duhem questions the value of presenting data in
the absence of any sort of theoretical framework. Berthelot’s doing so was,
Duhem suggests, a desperate ploy to separate something of lasting value from
his ill fated law of maximum work. But Duhem mercilessly criticises the
inadequacy of isolated items of data which are insufficient to support the
detailed functional connections between variables suggested by thermodynamic
considerations.9 And Berthelot’s rearguard action in defence of the principle of
maximum work, based on a distinction between physical and chemical pro-
cesses that drew on a spurious notion of extraneous energy, was anathema to
a physicalist who sees no explanatory value in a distinction of worlds. Duhem
develops the point by showing how Berthelot’s employment of the distinction
reduces the principle to a tautology. His alternative conception is to develop
proper statements of general laws, which, he insisted, requires recognising
exceptions for what they are and striving to accommodate them by reformu-
lation, rather than treating laws as dealing with their own specific domains.
Obvious as this may seem to be, it does not seem to have struck home for many
a commentator on the history of science, who maintain that Newton’s law of
universal gravitational attraction was first seriously threatened by anomalies in
Mercury’s orbit. As Duhem points out, the mundane phenomena of capillary
action and electrostatic forces provide counterexamples. More specifically, they
provide counterexamples to the statement that any two bodies attract one
“Two experiments on the dissolution of sodium acetate suffice, strictly, to supply the results
which we found in the work of Mr Berthelot. Some thousands of determinations, two or three years
of strenuous work, are necessary to bring the monograph of a saline solution to a successful
conclusion according to the method of Mr E. Monnet’ (Duhem, 1897, p. 388; Monnet was one of
Duhem’s doctoral students).
58 Studies in History and Philosophy of Science

another with a force directly proportional to the product of their masses and
inversely proportional to the square of their distance apart. ‘In order to avoid
capillary phenomena refuting the law of gravitation, it must be modified. It is
necessary to regard the formula of the inverse square of the distance as an
approximate formula; .. . it becomes quite incorrect when the expression of the
action of two elements which are very close is at issue’ (Duhem, 1894, p. 224).
It is illuminating to contrast Duhem’s position with that of Nancy Cartwright
(1983), who has more recently drawn attention to such matters and maintains
an instrumentalist view to the effect that laws are retained (for their explanatory
virtues when they are ‘fundamental’), but are false. Duhem, on the other hand,
rejects them because they are false, saying that the law is modified in order to
‘complete this schematism and to make it more representative of reality’
(Duhem, 1914, p. 174) that is, in order to replace it with a better shot at the
truth.iO
Quoting Duhem to this effect is, as always, problematic. The avid Duhem
reader may wish to point out that the immediately preceding sentence (where
the example under consideration is the gas laws rather than gravitation) begins

He [the physicist] realizes that the faulty relation is merely a symbolic one, that it did
not bear on the real, concrete gas he manipulates but on a certain logical creature, on
a certain schematic gas characterised by its density, temperature, and pressure
(Duhem, 1914, p. 174).

This might be taken to contradict the suggestion just made, that Duhem claims
the ‘schematism’, the law, is modified in order that it can be upheld as true.
Now, such a contradiction would itself be as much a problem for the
instrumentalist interpretation as for the realist one. But it is not necessary to go
as far as the next sentence to find a contradiction of the first part of this
sentence. The same sentence continues

and that this schematism is undoubtedly too simple and too incomplete to represent
the properties of the real gas placed in the conditions given now.

This is more clearly in agreement with the suggestion at the end of the previous
paragraph. Rather than saying Duhem manages to contradict himself in the
course of the same sentence, I suggest the instrumentalist reading of the first
part of the sentence be abandoned. He should not be read as saying that what
he calls the symbolic character of laws, arising because they are provisional,
means that laws cannot ever ‘bear on the real’, but rather that the discovery of
exceptions shows that the earlier formulation does not in fact bear on the real
object, the gas actually manipulated in the experiment. A modification is
required, characterising the gas in more complex fashion, giving a better
account of what is actually being handled in the physicist’s experiments.
‘“This quotation from Duhem (1914) is taken from a section which largely comprises the text of
the section of Duhem (1894, p. 224) from which the sentence just quoted was taken.
Duhem’s Physicialism 59

Preconceptions about the simplicity of laws-perhaps a distinguishing


feature of fundamental laws in Cartwright’s view-ill fit such a conception,
and, appropriately enough, Duhem would not be bound by such ideals.
[T]he motives guiding his [the physicist’s] choice will not have the same nature nor the
same imperative necessity as that which obliges the preference of truth to error. He
will choose a certain formula because it is simpler than the others. The weakness of
our minds constrains us to attach great importance to considerations of this order,
but the times when it was supposed that the intelligence of the Creator is tainted by
the same debility have passed, when any law which expressed too complicated an
algebraic equation was dismissed in the name of simplicity of the laws of nature
(Duhem, 1894, p. 219).

And when he later elaborates this passage, he says ‘We are no longer dupes of
the charm which simple formulas exert on us; we no longer take that charm as
the evidence of a greater certainty’ (Duhem, 1914, p. 171). Rejecting a priori
standards of simplicity is not inconsistent with maintaining that physicists
should strive to make their theories as simple as possible, simplicity being, as
ever, just one inductive consideration among others, and Duhem sometimes
appeals to undue complexity when arguing against certain theories. Thus, he
argues (Duhem, 1892~) that restricting theories to those built upon mechanical
notions makes them unduly complex; but that is not all:
The method which puts aside all non-mechanical theories leads to great complica-
tions. And it might well happen that it runs up against impossibilities. What assures
us that all physical notions, that all experimental laws, can be symbolised by a
combination, even a very complicated one, of mechanical concepts alone? . Is it not
for an analogous reason that the most complex mechanical theories have not been
able, as yet, to give a satisfactory account of Carnot’s principle? (Duhem, 1892c,
pp. 156-l 57).

According to Cartwright, however,

Like van Fraassen, Duhem rejects theoretical laws because he does not countenance
inference to the best explanation. . They [both] make a specific and concrete attack
on . inference to the best explanation-and thereby on the scientific realism to which
it gives rise (Cartwright, 1983, p. 88).

But Duhem certainly did not, as we have seen, reject theoretical laws, and I see
no good grounds for claiming that Duhem rejects inference to the best
explanation either. He expresses caution, requiring that explanatory work
actually be done, and rejects, along with Ampere, the idea that a theory can be
said to explain laws merely because they can be deduced from it. Applying these
principles leads him to reject the atomic or reductionist thesis that ‘phenomenal’
laws can be reduced to ‘fundamental’ laws. The acceptability of such a thesis
would have to be motivated by good scientific argument from the available
evidence, and Duhem argued that it was not so motivated. He was, as I say, a
non-reductionist physicalist, and interpreting him as a realist makes sense of
60 Studies in History and Philosophy of Science

this: like Cartwright, he couldn’t see how any serious candidates for simple
fundamental laws to which others are reduced could be maintained as true; but
his course was to reject reductionism, in the form of the belief that explanations
are to be had from simple laws.

9. Conclusion

The many superficial signs of instrumentalism in Duhem’s writings do not


suffice to clinch the interpretation of him as an instrumentalist. Superficially,
there are many signs of realism too. So if Duhem is to be taken seriously at all,
an interpretation of him can only be based on a deeper, systematic reading. In
particular, his rejection of atomism is often superficially understood as an
expression of instrumentalism, just as it would be on the part of a contemporary
scientist. But once it is acknowledged that Duhem cannot reasonably be
criticised for lack of prescience, and clearly recognised that the grounds of his
anti-atomism are not more radically positivist than his atomist opponents,
many of whom were antagonistic towards theory, then Duhem’s credentials as
a realist need to be considered seriously.
He considered his own views to imply a rejection of conventionalism-what
must have been for him the most prominent contemporary expression of
instrumentalism. He found the atomists at best unwilling to present any clear
understanding of what atoms are, and versions of atomism were often ad hoc or
simply inconsistent. He held an alternative conception of the nature of matter to
that of atomists. He was wary not to let this acquire in his thinking the status of
an a priori doctrine and wanted to introduce empirical considerations where ‘so
long as a Cartesian found something in a quality other than “bare extension and
its modifications”, he was certain its true nature had not been found’ (Duhem,
1914, p. 124). Perhaps he overdid this latter aspect, especially in his early
writings, and failed to see the role of what we might call metaphysical con-
siderations-both critical and positive-in his own thought. (As Martin (1991)
remarks, Duhem’s habit of incorporating earlier material almost verbatim
alongside his new ideas in later publications inevitably led to conflict.) Given the
excesses of his critical rhetoric, he may well have been afraid that his own views
might be dismissed as a ‘crowd of images aroused ... by the words “constitution
of matter” ’ (p. 84). The formulations he offers by way of summaries of the
considerations he raises are not always happy, sometimes even contradictory.
His style of writing is hardly conducive to the precise statement of philosophical
theses. But the arguments and lines of thought in Duhem’s writings can be
assessed without undue fixation on these stylistic blemishes.rr
“Parts of this paper were presented at the Swedish national philosophy conference in Ume&
June 1995, and published as Needham (1995). Research on which the paper is based was supported
by the Swedish Humanistisk-samhLllsvetenskapliga ForskningsrBdet.
Duhem’s Physicialism 61

References

Achinstein, P. (1994) ‘Jean Perrin and Molecular Reality’, Perspectives on Science 2,


396427.
Ariew, R. (1984) ‘The Duhem thesis’, British Journal for the Philosophy of Science 35,
313-325.
Cartwright, N. (1983) How the Laws of Physics Lie (Oxford: Clarendon Press).
Chalmers, A. (1993) ‘The Lack of Excellency of Boyle’s Mechanical Philosophy’, Studies
in History and Philosophy of Science 24, 541-564.
Christie, M. (1994) ‘Philosophers versus Chemists Concerning Laws of Nature’, Studies
in History and Philosophy of Science 25, 613-629.
Clericuzio, A. (1990) ‘A Redefinition of Boyle’s Chemistry and Corpuscular Philos-
ophy’, Annals of Science 47, 561-589.
Dirac, P. A. M. (1929) ‘Quantum Mechanics of Many-Electron Systems’, Proceedings of
the Royal Society of London 123, 714733.
Dolby, R. G. A. (1984) ‘Thermochemistry versus Thermodynamics: The Nineteenth
Century Controversy’, History of Science 22, 375400.
Duhem, P. (1892a) ‘Notation atomique et hypotheses atomistiques’, Revue des questions
scienttjiques 31, 391,457.
Duhem, P. (1892b) ‘Emile Mathieu, His Life and Works’, Bulletin of the New York
Mathematical Society 1, 156168.
Duhem, P. (1892c) ‘Quelques reflexions au sujet des theories physiques’, Revue des
questions scienti@ques 31, 139-177.
Duhem, P. (1893) ‘Physique et metaphysique’, Revue des questions scientt$ques 34,
55-83.
Duhem, P. (1894) ‘Quelques reflexions au sujet de la physique experimentale’, Revue des
questions scienttjiques 36, 179-229.
Duhem, P. (1897) ‘Thermochimie, a propos d’un livre recent de M Marcelin Berthelot’,
Revue des questions scient$ques 42, 361-392.
Duhem, P. (1900) ‘L’Oeuvre de M. J. H. van? Hoff, a propos d’un livre recent’, Revue
des questions scient$ques 47, 5-21.
Duhem, P. (1902) Le mixte et la combinaison chimique: Essai sur I’evolution dune idee
(Paris: C. Naud; reprinted Paris: Fayard, 1985).
Duhem, P. (1914) La theorie physique: son objet-sa structure, 2nd ed. trans. by Philip
Wiener, The Aim and Structure of Physical Theories (Princeton: Princeton, University
Press, 1954).
Duhem, P. (1915) La science allemande (Paris: Hermann).
Earman, J. (1989) World Enough and Space-Time (Cambridge, MA.: MIT Press).
Giedymin, J. (1982) Science and Convention (Oxford: Pergamon).
Gueron, J. and Magat, M. (1971) ‘A History of Physical Chemistry in France’, Annual
Review of Physical Chemistry 22, l-23.
Leicester, H. M. and Klickstein, H. S. (eds) (1952) A Source Book in Chemistry,
1400-1900 (New York: McGraw-Hill).
Martin, R. N. D. (1991) Pierre Duhem: Philosophy and History in the Work of a Believing
Physicist (La Salle, IL: Open Court).
Needham, P. (1995) ‘Duhems quineska realism’, Filosofisk Tidskrift 16, 2640.
Needham, P. (1996a) ‘Macroscopic Objects: An Exercise in Duhemian Ontology’,
Philosophy of Science 63, 2011223.
Needham, P. (1996b) ‘Substitution: Duhem’s Explication of a Chemical Paradigm’,
Perspectives on Science 4, 4088433.
Nye, M. J. (1979) ‘The Boutroux Circle and Poincare’s Conventionalism’, Journal of the
History of Ideas 40, 1077120.
Perrin. J. (1913) Les Atomes (Paris: Felix Alcan), trans. by D. L. Hammick as Atoms
(Woodbridge, CT: Ox Bow Press, 1990).
62 Studies in History and Philosophy of Science

Popper, K. R. (1972) The Logic of ScientiJic Discovery (London: Hutchinson).


Quine, W. V. (1960) Word and Object (Cambridge, MA.: MIT Press).
Quine, W. V. (1986a) Reply to Roger F. Gibson, in L. E. Hahn and P. A. Schilpp (eds),
The Philosophy of W. V. Quine (La Salle, IL: Open Court).
Quine, W. V. (1986b) Reply to Jules Vuillemin, in L. E. Hahn and P. A. Schilpp (eds),
The Philosophy of W. V. Quine (La Salle, IL: Open Court).
Quine, W. V. (1990) ‘Three Indeterminacies’, in R. B. Barrett and R. F. Gibson (eds),
Perspectives on Quine (Oxford: Blackwell).
Vuillemin, J. (1986) ‘On Duhem’s and Quine’s Theses’, in L. E. Hahn and P. A. Schilpp
(eds), The Philosophy of W. V. Quine (La Salle, IL: Open Court).
Worrall, J. (1982) ‘Scientific Realism and Scientific Change’, Philosophical Quarterly 32,
201-231.

Вам также может понравиться