Chapter 6: The Quine-Duhem Thesis

Pierre Duhem (1861–1916) was a French physicist and philosopher of science whose The Aim and Structure of Physical Theory (1906) presented a remarkably prescient analysis of the logic of scientific testing. Duhem argued that in physics, no hypothesis is ever tested in isolation. Every experimental test involves not only the hypothesis being tested but also a host of auxiliary assumptions about the experimental apparatus, background conditions, and connecting theories.

Duhem’s argument can be stated formally. Suppose we wish to test a hypothesis H. We derive a prediction P from H, but the derivation requires auxiliary assumptions A₁, A₂, ..., A_n (about the instruments, the experimental setup, relevant background theories, etc.):

If P is observed, the conjunction is confirmed (but we cannot attribute the confirmation to H alone). If ¬P is observed, we know by modus tollens that the conjunction is false:

It could be H that is false, but it could equally well be any of the auxiliary assumptions. Perhaps the instrument was miscalibrated. Perhaps a background theory is incorrect. Perhaps the experimental conditions were not as assumed. Logic alone cannot determine which element of the conjunction to blame.

“The physicist can never subject an isolated hypothesis to experimental test, but only a whole group of hypotheses; when the experiment is in disagreement with his predictions, what he learns 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.”— Pierre Duhem, The Aim and Structure of Physical Theory (1906), p. 187

The Impossibility of Crucial Experiments

Duhem drew an important corollary: there can be no crucial experiment (experimentum crucis) that decisively chooses between two rival theories. Francis Bacon and Newton had imagined that an experiment could be designed whose outcome would confirm one theory while refuting the other. Duhem showed that this is impossible because the two theories are never the only elements in play. Auxiliary assumptions are always involved, and revising them can always save the “refuted” theory.

Duhem illustrated his thesis with the case of optics. The dispute between the particle theory of light (Newton) and the wave theory (Fresnel) was supposedly settled by Foucault’s 1850 experiment, which showed that light travels slower in water than in air — as the wave theory predicted and the particle theory denied. But Duhem pointed out that what was refuted was not the particle theory alone but the particle theory together with its auxiliary assumptions. A sufficiently ingenious physicist could have modified the auxiliary assumptions to save the particle theory.

Importantly, Duhem limited his thesis to physics. He did not claim that all sciences face this problem. In physiology, he suggested, hypotheses can be tested more directly. This restriction distinguishes Duhem’s relatively modest thesis from Quine’s far more radical version.

W.V.O. Quine (1908–2000) published “Two Dogmas of Empiricism” in 1951, and it became one of the most influential philosophy papers of the 20th century. Quine attacked two assumptions that he saw as foundational to logical positivism — the two “dogmas” of the title:

Quine argued that both dogmas are untenable, and that abandoning them leads to a radical rethinking of the relationship between theory and evidence.

Quine’s attack on the analytic-synthetic distinction is one of the most celebrated arguments in 20th-century philosophy. The distinction, going back to Kant, holds that some truths are analytic (true by virtue of the meanings of their terms) while others are synthetic (true by virtue of how the world is). The positivists relied on this distinction: analytic truths (logic, mathematics) are meaningful but uninformative about the world; synthetic truths are meaningful only if empirically verifiable.

Quine argued that the concept of analyticity is circular. Every attempt to define it relies on equally problematic notions:

• •Synonymy: “All bachelors are unmarried” is analytic because “bachelor” is synonymous with “unmarried man.” But what is synonymy? We cannot define it without appealing to analyticity. The definitions are circular.
• •Definition: Perhaps analytic truths are true by definition. But Quine argued that dictionary definitions merely report existing usage; they do not create analytic truths. And stipulative definitions are arbitrary acts, not discoveries about meaning.
• •Interchangeability: Perhaps two terms are synonymous if they are interchangeable in all contexts without changing truth-value (salva veritate). But this only works in extensional contexts. In intensional contexts (“necessarily,” “believes that”), we again need the notion of analyticity to determine interchangeability.
• •Semantic rules: Carnap suggested that analytic truths are true by virtue of the semantic rules of a language. But Quine responded: what are semantic rules? We can say that certain truths hold “by semantic rule,” but this only shifts the problem: why those rules?

“It is obvious that truth in general depends on both language and extralinguistic fact... Thus one is tempted to suppose in general that the truth of a statement is somehow analyzable into a linguistic component and a factual component. Given this supposition, it next seems reasonable that in some statements the factual component should be null; and these are the analytic statements. But, for all its a priori reasonableness, a boundary between analytic and synthetic statements simply has not been drawn.”— W.V.O. Quine, “Two Dogmas of Empiricism” (1951)

Quine’s conclusion was that the analytic-synthetic distinction is a metaphysical article of faith. There is no sharp boundary between statements that are true by meaning and statements that are true by fact. All statements are potentially revisable in the light of experience, and none is immune to revision.

With both dogmas rejected, Quine proposed an alternative picture of the relationship between theory and experience. This is his famous web of beliefmetaphor:

“The totality of our so-called knowledge or beliefs, from the most casual matters of geography and history to the profoundest laws of atomic physics or even of pure mathematics and logic, is a man-made fabric which impinges on experience only along the edges. Or, to change the figure, total science is like a field of force whose boundary conditions are experience. A conflict with experience at the periphery occasions readjustments in the interior of the field.”— W.V.O. Quine, “Two Dogmas of Empiricism” (1951)

On Quine’s picture, our beliefs form an interconnected web. At the periphery are observation statements, directly linked to sensory experience. At the center are the most abstract and general principles: logic, mathematics, and fundamental physical laws. When experience conflicts with the web, something must change — but we have considerable freedom in choosing what to change.

Key features of Quine’s holism:

• •No statement is immune to revision: Even the laws of logic and mathematics could, in principle, be revised if such revision simplified our overall theory. Quine cited the proposal to adopt a non-classical logic for quantum mechanics as an example.
• •No statement is immune to retention: Any statement can be held true in the face of any experience, provided we make sufficient adjustments elsewhere in the web. We can always blame an auxiliary assumption, reinterpret the evidence, or even revise logic.
• •Pragmatic considerations govern revision:When experience forces a revision, we prefer changes that minimize disruption to the overall web. We revise peripheral beliefs (observation reports) more readily than central ones (laws of logic) because the latter are more deeply entrenched and more broadly connected.
• •The unit of empirical significance is the whole of science: Individual statements do not have their own empirical content. Only the totality of our beliefs confronts experience. This is Quine’s radical extension of Duhem’s thesis.

A major consequence of the Quine-Duhem thesis is the underdetermination of theory by evidence. If individual hypotheses cannot be tested in isolation, then the same body of evidence is always compatible with multiple, mutually inconsistent theories. The data alone do not determine which theory to accept.

Quine distinguished two forms of underdetermination:

The underdetermination thesis has been used to support various philosophical positions:

• •Anti-realism: Bas van Fraassen argues that underdetermination shows we should not believe our best theories are true; we should only accept them as “empirically adequate” (adequate to the observable phenomena).
• •Social constructivism: Sociologists of knowledge (e.g., the Edinburgh “Strong Programme”) argue that since the data do not determine theory choice, social factors (interests, power, culture) fill the gap.
• •Pragmatism: Quine himself drew a pragmatist conclusion: we choose between empirically equivalent theories on the basis of pragmatic virtues (simplicity, conservatism, generality, fecundity), not on the basis of correspondence to reality.

“Any statement can be held true come what may, if we make drastic enough adjustments elsewhere in the system... Conversely, by the same token, no statement is immune to revision.”— W.V.O. Quine, “Two Dogmas of Empiricism” (1951)

The Quine-Duhem thesis poses a direct challenge to Popper’s falsificationism. If no hypothesis can be tested in isolation, then no hypothesis can be falsified in isolation. When a prediction fails, we can always save the hypothesis by modifying an auxiliary assumption. Strict falsification of a single hypothesis is logically impossible.

Popper was aware of this problem and attempted to address it through his concept of basic statements — observation statements that are accepted by convention within the scientific community. But this move concedes that falsification is not a purely logical relationship between a theory and experience; it involves a social element (the community’s agreement to accept certain basic statements).

Lakatos used the Quine-Duhem thesis to motivate his own methodology of scientific research programmes. He argued that scientists rightly protect the “hard core” of their theories from falsification by modifying the “protective belt” of auxiliary hypotheses. The question is not whether a theory has been falsified (it can always be saved) but whether the programme of saving it is progressive or degenerating.

The Quine-Duhem thesis has generated a vast literature. Several important responses deserve mention:

• •
  Grünbaum’s critique:

  Adolf Grünbaum (1960) argued that Duhem’s thesis is correct but that Quine’s radical extension is not. In practice, auxiliary assumptions are often independently testable. The “freedom” to save any theory by modifying auxiliaries is severely constrained by the requirement that the modifications be independently motivated rather than ad hoc.

• •
  Laudan and Leplin on underdetermination:

  Larry Laudan and Jarrett Leplin (1991) argued that the underdetermination thesis is less threatening than it appears. Empirically equivalent theories at one time may become empirically distinguishable as new observational techniques are developed. Moreover, non-empirical virtues (simplicity, explanatory power) can rationally break ties between empirically equivalent theories.

• •
  Moderate holism:

  Many contemporary philosophers accept a moderate form of holism: hypotheses are not testable in strict isolation, but neither is the entire web of belief at stake in every test. In practice, scientists can often identify the most likely source of error and revise accordingly. The freedom to revise is real but constrained.

• •
  Bayesian responses:

  Bayesians argue that holism is compatible with rational theory choice. Even if hypotheses are not testable in isolation, Bayesian updating can assign differential degrees of blame to different parts of the web based on their prior probabilities and likelihoods.

“Duhem’s thesis is correct and important. Quine’s thesis is incorrect and unimportant.”— Adolf Grünbaum (paraphrased), on the difference between moderate and radical holism

The Quine-Duhem thesis, in both its Duhemian and Quinean forms, has had far-reaching consequences for philosophy of science:

• •The collapse of positivism: By undermining both the verification principle (which required individual statements to have empirical content) and the analytic-synthetic distinction (which structured the positivist account of meaning), the Quine-Duhem thesis effectively ended logical positivism as a viable philosophical programme.
• •Challenges to falsificationism: By showing that theories can always be saved from refutation by modifying auxiliaries, the thesis undermines Popper’s claim that falsifiability is the hallmark of science.
• •Support for Kuhn: The thesis supports Kuhn’s observation that scientists do not abandon paradigms at the first sign of anomaly. Normal science involves precisely the kind of auxiliary modification that Duhem described.
• •The turn to holism: The thesis initiated a fundamental shift from atomistic to holistic conceptions of meaning, evidence, and justification. This shift reverberates throughout contemporary epistemology and philosophy of language.
• •Naturalized epistemology: Quine’s rejection of the analytic-synthetic distinction paved the way for his proposal to “naturalize” epistemology — to study how humans actually form beliefs rather than prescribing how they ought to form them.