this distinction between observation and interpretation in physical experiment by offering two descriptions of an experimental apparatus in a laboratory. One description is given in the vocabulary of the physicist who understands the theory of electricity, and the other description is given in the observational language of the observer innocent of such theoretical understanding. The experimenting physicist actually has two distinct representations of the instrument in his mind. One is the phenomenal image of the concrete instrument that he manipulates in reality. The other is a schematic model of the same instrument constructed mentally with the aid of the symbols from the theories that the physicist accepts. Without knowing the theories that the physicist regards as established and that he uses for interpreting the facts he observes, it is impossible for anyone to understand the meaning he gives to his statements. And when a physicist discusses his experiments with another physicist, who accepts an alternative theory, it is necessary for the two physicists to seek to establish a correspondence between their different ideas and then to reinterpret the experiment.
Twenty years before the development of the quantum theory Duhem cited as an example the two alternative theories of light: Newton’s emission theory and Frensel’s wave theory. He maintained that the observations and experiments interpreted in the concepts of one theory could be translated into the concepts of the other theory. In his philosophy this is possible, not because he anticipated quantum theory, but because he was a positivist, who believed that the two theories could be related to a common theory-neutral phenomenalist semantics.
Duhem’s stratification of the semantics of the language of theoretical science is central and strategic to his philosophy of science. It is not surprising that he stated that the approximate fit between measurements and theory creates a semantical difference. Haavelmo did the same thing for his theory of econometrics forty years later. But it might seem more correct were he to have said that the resolution of the indeterminacy in measurement by the calculated value for a variable in a theory actually resolves a semantic vagueness instead of saying, as he does, that it creates two distinct meanings. But it is surprising to find him concluding that the distinct meaning of the symbol in the theory is a “sign” of the phenomenal meaning defined by the experimental measurement method. It is this latter position that stratifies the semantics of science, so that theory cannot be given a realistic or phenomenalistic interpretation.
Nonetheless Duhem has a reason for taking this position. In his “The Physics of a Believer”, an appendix to Aim and Structure of Physical Theory, he reports that earlier in his career he attempted unsuccessfully to conform to Newton’s methods set forth in Newton’s “General Scholium”. He concluded that physical theory is neither a metaphysical explanation nor a set of general laws, whose validity is established, but rather that theory is an artificial construction manufactured with the aid of mathematical magnitudes. Thus the relation of the magnitudes to the abstract notion emergent from experiment is that of sign to thing signified. The key concept seems to be the idea of artificial construction. The artificial nature of theory gives it an artificial semantics, and this artificial semantics is of a different kind than the natural semantics of language that describes the phenomenal world.
Throughout most of the history of philosophy, philosophers believed that while the multiplicity of languages argues for the existence of a conventional aspect in human language, still, as Aristotle said, while men speak different languages, they have the same cognitive experiences. This is the thesis of a naturalistic semantics; all men have the same cognitive experience when in the presence of the same reality, because there is a natural relation between knowledge and reality. Mach’s theory of sensations and of their identification with elements of the phenomenal world is a contorted variation of this thesis. But Duhem could not fit this thesis to the language of physical theory, even while he, like Mach, maintained it for the language of observation. He viewed physical theory as so artifactual that its meanings could not be natural but had to be artificial. Thus physical theory does not describe either the real or the phenomenal world of nature; it only describes symbols. But he was not led to conclude that theory is meaningless. His reconciliation strategy was to make the artificial semantics of theory describe the language of science, in effect a metalanguage.
Duhem’s Philosophy of Science
Aim of Science
Duhem’s statement of the aim of science is similar to Mach`s: the aim of science is economy of thought. Like Mach, Duhem believes that experimental laws contribute an intellectual economy, because they summarize a large number of individual facts including data measurements. But unlike Mach, Duhem furthermore says that theories also contribute to the realization of the aim of science. The economy achieved by the substitution of a law for individual facts is redoubled for the mind, when the mind substitutes theories for the numerous mathematically expressed experimental laws. A theory is a system of mathematical propositions mathematically derived from a small number of principles, which aim to represent as simply, as completely, and as exactly as possible, a set of experimental laws. Its aim in other words is economy of thought by schematically representing and logically organizing experimental laws.
Scientific Criticism
Duhem developed a sophisticated theory of scientific criticism, and it is central to his philosophy of science. He is very emphatic in defending the autonomy of empirical science from any encroachment by metaphysics or natural philosophy. Metaphysics pertains to realities that underlie the phenomenal appearances hidden by the phenomena, while science pertains only to these appearances. Consequently whatever may be the criteria and procedures for criticizing a metaphysical thesis, they are not relevant to empirical science. In empirical sciences that are nonmathematical, the generalizations such as “Every man is mortal” may be accepted or rejected as simply true or false. But in mathematical physics the equations both of the empirical laws and of the hypothetical theories are not simply regarded as true or false, but are approximate. The amount of underdetermination due to the approximate nature of the values of the variables in these equations will be reduced as experimental and measurement techniques improve. And because measurement instruments depend on physical theory, the improvement in instruments occurs due to the improvement in theory. As the range of this indeterminacy becomes smaller, the equations of either the empirical laws or the hypothetical theories that represent the laws may no longer be able to predict values for their variables that fall within the smaller range of measurement error. When this happens, the equations are no longer satisfactory. Duhem maintains that the only criterion that may validly operate in scientific criticism is the ability of the law or theory to make accurate predictions. This exclusion of all prior ontological or metaphysical criteria from scientific criticism has been carried forward into the contemporary pragmatist philosophy of science. It shows up for example as Quine’s rejection of all “first philosophy.”
In his theory of scientific criticism Duhem rejected the use of so-called crucial experiments as a means of establishing the validity of a theory. His thesis is that if the physicist is confronted with several alternative theories, the rejection of all but one cannot imply the establishment of the remaining one. As an example he cites the two alternative theories of light: one theory is the hypothesis that light is a stream of high speed projectiles, and the other is the hypothesis that light consists of vibrations whose waves are propagated in ether. This is not an anticipation of the Copenhagen duality thesis; Duhem is thinking of the wave and particle theories as alternative theories. His position is that the choice is not mutually exclusive, because no one can ever enumerate completely all of the various hypotheses, which may pertain to a group of phenomena. He thus maintains that several alternative theories may fall within the range of indeterminacy of the measurement data and experimental laws, so that more than one theory may be satisfactory. This represents a pluralistic thesis about science, and in the crucial experiment discussion, it means that even if all hypotheses could somehow be enumerated, elimination could not leave but one to be considered as established. This pluralism is another aspect of his philosophy of physical theory that has been carried forward into the contemporary pragmatist philosophy of science.
His theory of scientific criticism also reflects his wholistic view of theories. This wholistic view not only makes the meanings of the mathematical symbols mutually determined by the context consisting of the