Duhem-Quine Thesis

The Duhem-Quine thesis is the claim that scientific hypotheses are never tested in isolation — they face experience only in conjunction with a network of auxiliary assumptions, background theories, and methodological commitments. When a prediction fails, the failure falls on the conjunction, not necessarily on the central hypothesis. The scientist is therefore free to reject any element of the network in response to a failed prediction: the central hypothesis, an auxiliary assumption, a measurement protocol, or a background theory. This holism of confirmation and refutation — first identified by Pierre Duhem for physics and generalized by W.V.O. Quine to all empirical claims in Two Dogmas of Empiricism (1951) — is the most important technical objection to Popper's falsificationism. It shows that Popperian falsification is not logically clean: a single failed prediction does not unambiguously falsify a hypothesis. It does not show that evidence is irrelevant — it shows that the relationship between evidence and theory is mediated by holistic judgment about which element of the network to revise. Lakatos's research programme methodology is the most systematic attempt to specify rational principles for deciding which parts of the network are in the hard core (protected from revision) and which are in the protective belt (subject to revision in response to anomalies).

The Social and Narrative Dimensions of the Web of Belief

The Duhem-Quine thesis is typically treated as a logical point about the structure of scientific testing. But the holism it describes has a sociological dimension that Quine himself underexplored and that becomes visible when the thesis is applied to how scientific communities actually function.

The network of assumptions that surrounds any empirical hypothesis is not merely a set of background beliefs held by an individual. It is a collectively maintained structure of commitments, maintained by practices — textbooks that standardize auxiliary assumptions, instruments that embody background theories, peer review that enforces which elements of the network are candidates for revision and which are not. A scientist who responds to a failed prediction by revising a foundational background assumption (say, the calibration methods of instruments taken as standard across the field) must do so against the resistance of a community whose practices are organized around treating that assumption as settled. The rational revision licensed by the Duhem-Quine thesis is not the same as the socially permitted revision.

This is the sociological sharpening of the thesis developed by sociology of scientific knowledge (SSK) researchers — Harry Collins, David Bloor, and their colleagues in the Edinburgh school. The SSK thesis is more radical: not only does the Duhem-Quine thesis show that single observations cannot uniquely determine theory change, but social factors — professional interest, institutional authority, credibility networks — determine which auxiliary assumptions are treated as available for revision and which are held fixed. The hard core in Lakatos's sense is not fixed by logical structure; it is fixed by credibility economies and by the investment of scientific communities in the theories that form the core.

The narrative extension goes further: the web of belief is not only maintained socially but sustained by the stories a scientific community tells about its own history. A paradigm in the Kuhnian sense is partly a set of theoretical commitments and partly a set of exemplars — canonical problem-solutions that teach practitioners what counts as a correct application of the paradigm. The exemplars are narratives about how the paradigm solved problems. When an anomaly arises, the community's response is shaped by which exemplary stories it considers most authoritative — which successful applications of the hard core most constrain what revisions are permissible.

The consequence: the Duhem-Quine thesis, read fully, is not just an epistemological observation about underdetermination. It is a description of the structure of scientific communities as interpretive communities — groups whose collective narrative histories determine what they can treat as revisable. The underdetermination is not merely logical; it is social. And social underdetermination has a cultural dimension that logic alone cannot close.