The Duhem-Quine thesis is so-called because the turn of the century French physicist Pierre Duhem and the twentieth century philosopher Willard Van Orman Quine made similar statements about the role of experiments in science. In reality, however, their emphases were quite different.
The Duhem thesis, as articulated by Pierre Duhem himself, is the claim that “an experiment in physics can never condemn an isolated hypothesis but only a whole theoretical group.” Since Duhem published his work, the Duhem thesis has been interpreted so diversely that it might be more appropriate to discuss Duhem theses. Here I will reserve the term Duhem thesis for Duhem’s own claims about theoretical holism, which are both more interesting and more plausible than most Duhemistic theses.
In his paper “Physical Theory and Experiment,” Duhem argues that physicists must rely on additional theories besides the one they intend to test in a given experiment. The necessity of auxiliary theories arises from the use of scientific apparatus, both physical and conceptual. “Without theory it is impossible to regulate a single instrument or to interpret a single reading,” wrote Duhem. The use of a scientific tool such as a thermometer or a principle such as inertia itself relies on theories, and any experiment which involves such apparatus is a test of the theories on which they are based. As a result, if a physicist uses a hypothesis to make a prediction about the result of an experiment and the prediction is incorrect, the physicist should not automatically conclude that the hypothesis is incorrect as well; the prediction could also be false if one of the auxiliary theories is false. “What he learns,” wrote Duhem, “is that at least one of the hypotheses constituting this group is unacceptable and ought to be modified; but the experiment does not designate which one should be changed.”
One of the many controversies surrounding the Duhem thesis is that of its domain: should it in fact apply only to physics, as Duhem stated? Or should Duhem’s particular focus on physics be ignored and his thesis taken to state that no experiment in any science can “condemn an isolated hypothesis”? Donald Gillies argues that “Duhem is correct to limit the scope of his thesis, but wrong to identify its scope with that of a particular branch of science—namely, physics.” Gillies is right that physics is not now the proper scope for the Duhem thesis. Duhem did, however, have good reasons to limit the domain of his thesis to physics based upon the state of science in the early twentieth century. The sciences have changed in the last hundred years in ways that have shifted the domain to which the Duhem thesis applies.
The physics of Duhem’s time, like the physics of today, had characteristics which distinguished it from other sciences. Duhem himself draws attention to two of these distinctions: the maturity and resulting mathematical nature of physics, and the use of physical theories as the basis for the construction of scientific instruments. In Duhem’s time these were both characteristics which physics held far more strongly than other sciences; they were both strong distinctions. Since then, physics has become less unique in both of these respects, though it still displays these traits more prominently than does, for example, biology. As other sciences have developed these traits, they have entered the domain of the Duhem thesis.
Duhem’s most important reason for delineating physics from other sciences was that it was a mature science. It underwent its deepest revolutions over two hundred years before Duhem with Isaac Newton’s Principia Mathematica and thus had more time to develop than other sciences. Biology, for instance, had fewer than fifty years to develop since Charles Darwin’s Origin of Species and has undergone major theoretical shifts since Duhem, among them numerous discoveries relating to DNA. Biology has become a more mature science.
During the twentieth century physics too has undergone deep changes, including the discoveries of quantum mechanics and relativity. These developments raise the question of how mature physics really is; despite hundreds of years of apparent progress and theory development, there still remain deep unanswered questions. Physics might seem less mature in the early twenty-first century than it did in the early twentieth, but it nonetheless has seen more development than other natural sciences and has more maturity in some sense.
The nature of this maturity need not be ambiguous, for Duhem himself describes the sense in which maturity is relevant to his thesis at the beginning of his paper.
When many philosophers talk about experimental sciences, they think only of sciences still close to their origins, e.g., physiology or certain branches of chemistry where the experimenter reasons directly on the facts by a method which is only common sense brought to greater attentiveness but where mathematical theory has not yet introduced its symbolic representations.
Immature sciences can benefit from simple experiments and observations which do not rely on other theories. In mature sciences, however, these experiments have already been conducted. We already know that if we drop an apple it will fall, but we might not know what happens when one cuts a nerve root in an animal. The former example implies that at least part of mechanics is mature in a deeper sense than Duhem himself states: we are aware of phenomena like gravity (at least in everyday terrestrial manifestations) because of our informal and unavoidable observation of the world, i.e. simply because our eyes are open.
The sorts of experiments that physicists actually do when studying mechanics, however, are much more complicated than dropping an apple. They involve preexisting knowledge of the relationships between physical forces (e.g. gravity) and properties (e.g. mass and velocity), as well as of physical constants. These concepts, expressed through the “symbolic representations” of mathematics, are necessary in order to design an experiment with any substantial sophistication. The Duhem thesis thus applies to physics, or at least to its mature subfields. In the quotation above, Duhem implies that mathematization is a process which sciences generally undergo eventually, and indeed other sciences such as biology now depend much more on mathematical expressions of generalized theories than they did in Duhem’s time. As a result, new biological hypotheses are much more likely to depend on auxiliary theories than they did one hundred years ago, and thus to fall within the domain in which the Duhem thesis is valid.
In addition to the cognitive apparatus provided by auxiliary theories, physics experiments typically require physical instruments, which themselves can only be read and understood in the light of theories. The thermometer, for example, depends upon a theory of temperature as well as upon the assertion, perhaps too simple to term a theory, of a relationship between a thermometer’s appearance and the temperature of its environment. Duhem acknowledges that experiments in other sciences depend on instruments, but argues that “the [auxiliary] theories used, as well as the instruments employed, belong to the domain of physics.” Other scientists such as chemists and physiologists rely upon the theories of another science, physics, when they use their instruments, while physicists rely upon the theories of their own field.
This seems like a philosophically irrelevant distinction, as Gillies asserts, writing that “physiology and chemistry no doubt contain hypotheses subject to the Duhem thesis.” However, Duhem himself admits as much. “From the standpoint of logic,” he writes, “the difference is of little importance; for the physiologist and chemist as well as for the physicist, the statement of the result of an experiment implies, in general, an act of faith in a whole group of theories.” And yet Duhem does not at the beginning of his paper apply his thesis to physiology and chemistry as well as physics; indeed, he specifically contrasts physics with physiology.
The reason for this may involve a bit of philosophical sloppiness. This is not a complete explanation, however. Duhem expects that in a physiological experiment, the hypothesis undergoing testing will be physiological but all of the auxiliary theories—if there even are any—will be physical. If the experiment fails, the physiologist can blame either a presumably novel physiological hypothesis or a well-tested physical theory. The maturity of physics stands, perhaps axiomatically in Duhem’s paper, as a good reason to believe physical theories. This maturity does not necessarily similarly support a particular physical theory, as the maturity of a science does not imply the maturity of each constituent theory, but presumably the instruments used by physiologists during the early twentieth century did not rely on recent or speculative physical theories. In any case, Duhem did believe that his thesis applied to some experiments outside of physics, but he also thought that some experiments in young sciences—such as cutting a nerve root—were so fundamental that they involved “only common sense brought to greater attentiveness.”
The idea that the Duhem thesis applies across the borders of the sciences—that a physiological experiment can depend upon physical auxiliary theories, for instance—can be radically extended into stronger claims. An example is W.V.O. Quine’s statement that “the unit of empirical significance is the whole of science.” The Duhem thesis does not entail the Quine thesis, however. Duhem claims that experiments depend on the auxiliary theories involved in understanding the apparatus involved; Quine claims that experiments depend on all possible auxiliary theories. The former not only seems more plausible, but also has greater utility: if the Duhem thesis holds, a scientist conscious of the theories on which her apparatus depend can list them, and if an experiment fails she and her colleagues can attempt additional experiments and observations to determine which theory, primary or auxiliary, is false. If the Quine thesis holds, in contrast, the failure of an experiment simply shows that something in science is wrong, providing no information—at least through the logic of the Quine thesis—about how to fix it.
The scientific instruments of the early twenty-first century are not the same as those of the early twentieth. They include “instruments” such as genetically engineered laboratory animals, the use of which depend primarily on biological and biochemical theories rather than physical ones. As a result, it is now possible for a test of a biological hypothesis to depend on biological auxiliary theories, and for the Duhem thesis to operate just as thoroughly in biology as Duhem argued that it did in physics.
Changes in science over the last hundred years have thus broadened the domain of the Duhem thesis. Duhem’s arguments for it, however, remain compelling. Although its radical competitors such as the Quine thesis challenge our ability to determine the truth or falsehood of theories at all, the Duhem thesis itself remains a source of insight into what scientists can and cannot learn from experiment and observation.