This post presents excerpts from Pierre Duhem’s The Aim and Structure of Physical Theory, first published (in French) in 1906, and translated into English in 1954 (Princeton University Press). See also the following post on Physics and metaphysics.
[I]f the aim of physical theories is to explain experimental laws, theoretical physics is not an autonomous science; it is subordinate to metaphysics. p.10
Now, to make physical theories depend on metaphysics is surely not the way to let them enjoy the privilege of universal consent. p.10
A physical theory reputed to be satisfactory by the sectarians of one metaphysical school will be rejected by the partisans of another school. p.10-11
While we regard a physical theory as a hypothetical explanation of physical reality, we make it dependent on metaphysics. In that way, far from giving it a form to which the greatest number of minds can give their assent, we limit its acceptance to those who acknowledge the philosophy it insists on. p.19
Agreement with experiment is the sole criteria of truth for a physical theory. p.21
The one who contributed most to break down the barrier between physical method and metaphysical method, and to confound their domains, so clearly distinguished in the Aristotelian philosophy, was surely Descartes. p.43
[For whatever is not deduced from the phenomena must be called a hypothesis; and hypotheses, whether metaphysical or physical, or based on occult qualities, or mechanical, have no place in experimental philosophy. Newton’s General Scholium, trans. Andrew Motte, 1729.]
Newton’s disciples, however, did not all adhere to the prudent reserve of their master; several could not remain in the narrow confines assigned to them by his method in physics. Crossing these limits, they asserted, as metaphysicians, that mutual attractions were the real and primary qualities of matter and that a phenomenon reduced to these attractions was truly explained. p.49
Ampére, a more profound philosopher than Laplace, saw with perfect clarity the importance of regarding a physical theory as independent of any metaphysical explanation; in fact, that is the way to keep out of physics the divisive quarrels of the diverse cosmological schools. p.50
… Ernst Mach has defined theoretical physics as an abstract and condensed representation of natural phenomena. G. Kirchhoff offered as the object of mechanics: “to describe as completely and as simply as possible the motions produced in nature.” p. 53
[Thomson/Kelvin:] Although the molecular constitution of solids supposed in these remarks and mechanically illustrated 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. p. 75
Every time people cite a principle of theoretical physics in support of a metaphysical doctrine or physical dogma, they commit a mistake, for they attribute to this principle a meaning not its own, an import not belonging to it. p.287
There you have, then, a theoretical physics which is neither the theory of a believer nor that of a nonbeliever, but merely and simply a theory of a physicist; admirably suited to classify the laws studied by the experimenter, it is incapable of opposing any assertion whatever of metaphysics or of religious dogma, and is equally incapable of lending effective support to any such assertion. p.291
Despite all this, Duhem recognizes one point of metaphysics that makes physics possible:
The study of the method of physics is powerless to disclose to the physicist the reason leading him to construct a physical theory. p. 334
In a word, the physicist is compelled to recognise that it would be unreasonable to work for the progress of physical theory if this theory were not the increasingly better defined and more precise reflection of a metaphysics; the belief in an order transcending physics is the sole justification of physical theory. p.335
From Ernan McMullin’s review of Essays in the History and Philosophy of Science by Pierre Duhem (British Journal for the Philosophy of Science, Vol. 48, No. 4, Dec., 1997, pp. 606-609):
[Duhem claimed] the notion of falsifiability embodied in Bacon’s notion of a crucial experiment, and later presented as the cardinal principle of scientific rationality by Popper, is fundamentally flawed. Testing a physical theory by experiment cannot be decisive in the event of a negative outcome because the theory itself normally consists of a cluster of separable propositions and there are, besides, unstated presuppositions of a variety of kinds that may be implicated. p.607
To attach ontological significance to atoms, electrons, fields, and the like, in [Duhem’s] view was to give metaphysics an improper role in physics. Duhem has found favour with those recent non-realists who are sceptical of explanatory success as a criterion of ontological significance. But the reason he gives for his own scepticism is entirely Duhemian: explanation of material nature is appropriate to metaphysics, not to physics. The credence given by leading physicists to atoms and the like is principally due, he claims, to their prior commitment to neo-atomism, a false metaphysical system. Physical theories themselves should be entirely independent of metaphysics; they are, after all, nothing more than economic ways of systematizing experimental laws. Their measure is convenience, not truth about some underlying causal order. p.608