Confirmation Holism and Its Epistemological Impact

When a scientist runs an experiment to test a hypothesis, the assumption is straightforward: gather evidence, check whether it confirms or refutes the hypothesis, and move on. But what if this picture is fundamentally incomplete? What if no single hypothesis can ever be tested on its own, and what if the evidence from any experiment can never point unambiguously to one theory alone? This is precisely what confirmation holism – one of the most consequential ideas in the philosophy of science – proposes. Far from being an abstract puzzle, it reshapes how we understand the relationship between evidence, observation, and knowledge itself.

What is confirmation holism? 🔗

Confirmation holism, also called epistemological holism, is the view that no individual scientific hypothesis or statement can be confirmed or disconfirmed by empirical testing in isolation. Only a whole system or network of interconnected beliefs and theories can be tested against experience. The idea did not emerge from a single moment of insight but from the convergent arguments of two very different thinkers – Pierre Duhem and W. V. O. Quine – whose combined thesis became one of the foundational challenges to classical empiricism.

Duhem’s original insight 🔗

Pierre Duhem, a French physicist and philosopher working at the turn of the 20th century, was the first to articulate the problem with precision. In his landmark work The Aim and Structure of Physical Theory (1906), Duhem argued that a physical experiment does not isolate and test a single hypothesis but examines it alongside a complex “theoretical scaffolding” of auxiliary assumptions, mathematical frameworks, and background theories. If an experiment yields unexpected results, Duhem held, it is impossible to say with certainty which part of this scaffolding has failed. The core hypothesis, the instruments used, the background laws assumed – any one of them could be responsible for the anomaly.

Critically, Duhem confined his claim to the physical sciences. He did not think the problem extended to all fields of inquiry, and he explicitly excluded formal disciplines like logic and mathematics from his account.

Quine’s radical extension 🔗

W. V. O. Quine took Duhem’s insight and extended it far beyond physics. In his pivotal 1951 essay “Two Dogmas of Empiricism,” Quine argued that the holist challenge applies not just to the confirmation of physical theories but to all human knowledge whatsoever – including mathematics and logic. For Quine, the entire body of scientific knowledge constitutes a single “web of belief” where statements face the tribunal of experience not individually but collectively.

Quine described his vision of knowledge as a field of force whose interior adjusts in response to peripheral conflicts with observation, emphasizing that any statement can be preserved as true if sufficiently drastic adjustments are made elsewhere in the system – and conversely, no statement is absolutely immune to revision. This means even our most basic logical and mathematical commitments are, in principle, revisable in light of experience. It was a radical claim, and it remains controversial.

The Quine-Duhem thesis: what it actually says 🔗

The Duhem-Quine thesis holds that it is impossible to isolate a single hypothesis in a “bundle” of hypotheses for testing. Any empirical test of a hypothesis H requires assuming the truth of background assumptions B – that the experiment functions as intended, that related scientific theory is accurate, and so on. What actually goes to trial, when we run an experiment, is the conjunction of H and B together. If the experiment fails, logic tells us that at least one member of the conjunction is wrong, but it cannot tell us which one.

Consider a historical illustration. In the early 17th century, opponents of Galileo’s telescopic observations argued that the instrument might be distorting what was seen rather than revealing real celestial objects. This was not mere obstinacy – it was a coherent epistemological move. Their bundle of background assumptions included doubts about the telescope’s reliability. Galileo’s observations could not, by themselves, settle the dispute, because what the observations meant depended on which background assumptions were accepted. In terms of the Duhem-Quine thesis, studying the defensibility of auxiliary assumptions alongside the primary hypothesis is necessary in arriving at any viable working hypothesis.

The asymmetry between Duhem and Quine 🔗

Though often grouped together, Duhem and Quine held genuinely different positions. Duhem believed the holist problem was real but confined to physics, and he thought scientists with good scientific judgment – what he called “good sense” – could navigate theory choice rationally even without a strict algorithm. Duhem resolved underdetermination by appealing to a cluster of virtues associated with good sense, which scholars have since interpreted as a form of virtue epistemology. Quine, by contrast, offered no such epistemic anchor. His holism was more radical and left theory choice more unsettled, driving concerns about rationality in science that persist to this day.

Underdetermination: when evidence cannot decide 🔗

Confirmation holism leads directly to one of the most discussed problems in the epistemology of science: the underdetermination of theory by evidence. In the philosophy of science, underdetermination is the idea that evidence available at a given time may be insufficient to determine what beliefs we should hold in response to it – and that all evidence necessarily underdetermines any scientific theory.

The problem is not that evidence is useless. It is that evidence is never exhaustive in its verdict. At any given time, multiple mutually incompatible theories can be compatible with the same body of observational data. One theory may explain the data by positing one kind of underlying mechanism; another may explain the same data by positing a completely different mechanism. The observations alone cannot adjudicate between them.

The Internet Encyclopedia of Philosophy illustrates this with skeptical scenarios: if a hypothetical “brain in a vat” would have exactly the same sensory experiences as a person living a normal life, then no possible observation can distinguish the two situations – the belief that one is living normally is radically underdetermined by the evidence. While this is an extreme case, it captures the logical structure of the problem that arises even in ordinary science.

Total vs. partial holism 🔗

Not everyone accepts the most sweeping version of confirmation holism. Critics of “total holism” point out that it leads to awkward consequences: if confirming the general theory of relativity also confirms every arbitrary conjunction that entails it, then even the claim that “the moon is made of cheese” gets confirmed alongside relativity – which seems absurd.

This concern has motivated the development of partial holism, associated with thinkers like Adolf Grünbaum, Ken Gemes, and Elliott Sober. Partial holists accept that evidence spreads its support across multiple beliefs, but deny that it spreads uniformly across an entire theoretical system. They argue that arbitrary or ad hoc conjunctions have internal barriers that prevent confirmation from bleeding into unrelated claims. The Duhem-Quine problem holds that there is no principled way of knowing how falsification affects individual conjuncts – partial holists seek to supply exactly such a principled account.

Theory-laden observation: how what we see depends on what we believe 🔗

Confirmation holism connects intimately with the concept of theory-ladenness of observation, a term developed most influentially by philosopher Norwood Russell Hanson. Following the Duhem-Quine thesis, Hanson argued that observation does not constitute neutral, objective data but is instead theory-dependent, interpreted knowledge. In his 1958 work Patterns of Discovery, Hanson proposed that what we perceive is not raw sensory input but already structured, filtered information – shaped by the theoretical frameworks we bring to it.

Hanson’s famous example involves the difference between how a Copernican and a Ptolemaic astronomer experience a sunrise. The Ptolemaic observer sees the sun falling behind the horizon because, in their view, the sun is moving; the Copernican sees a fixed sun and a rising horizon. Both are looking at the same event. But the theoretical frameworks they inhabit lead them to organize the same sensory input into fundamentally different perceptions.

Theory-ladenness is especially relevant for the problem of confirmation: if an observation is theory-laden, it already implicitly presumes various theses and therefore cannot act as a neutral arbitrator between theories that affirm or deny those same presupposed theses. This creates a circularity that undermines the classical empiricist idea that raw observation can serve as a neutral foundation for all knowledge.

Thomas Kuhn and the social dimension 🔗

Thomas Kuhn’s influential work amplified these concerns by adding a historical and sociological dimension. Kuhn cited the Duhem-Quine thesis as one of the reasons why scientific paradigms are incommensurable – meaning that scientists working in different paradigms may not even be able to fully understand or evaluate each other’s evidence, because they are literally perceiving and categorizing the world differently. At least since Kuhn’s influential work, one important line of thinking has held that it is ultimately the social and political interests of scientists which serve to determine their responses to disconfirming evidence.

This was a provocative claim. It suggested that theory choice in science is not purely a rational, evidence-driven process but is also shaped by the social contexts, institutional pressures, and shared assumptions of scientific communities. Confirmation holism, in this reading, opened the door to sociology of knowledge and feminist epistemology as legitimate tools for understanding how science actually works.

Epistemological implications 🔗

Confirmation holism has profound consequences for how we understand knowledge, justification, and objectivity. Several implications stand out.

The challenge to falsificationism 🔗

Karl Popper’s falsificationism held that a scientific theory is only genuinely scientific if it can, in principle, be refuted by evidence. Confirmation holism complicates this criterion: it is always possible to rescue a falsified theory by claiming that only one of its underlying auxiliary hypotheses is false, since there are an indeterminate number of such hypotheses – any theory can in principle be made compatible with any particular observation. This means that falsification is never logically conclusive. A theory faces experimental failure, but the failure can always be redirected onto some background assumption rather than the core hypothesis.

Popper himself acknowledged this difficulty and argued that scientists should avoid what he called “ad hoc” modifications – adjustments made purely to save a theory without independent justification. Imre Lakatos later refined this into his methodology of scientific research programs, which tried to specify when protective modifications were progressive versus degenerative.

The interconnectedness of knowledge 🔗

One of the deepest insights of confirmation holism is that knowledge is not a collection of isolated, independently testable propositions. For Quine, knowledge constitutes a single web of empirical significance, making even the most abstract components epistemologically continuous with empirical inquiry. This means that when evidence challenges one part of our knowledge system, its effects ripple outward – adjustments in one area may require, or permit, adjustments elsewhere. There is no point at which we can stand outside the web and test it from a truly neutral vantage point.

W. V. O. Quine’s incorporation of underdetermination problems into a general account of human knowledge was one of the most significant developments of 20th century epistemology. By dissolving the sharp boundary between theoretical statements and observational ones, Quine also undermined one of the core commitments of logical positivism – the idea that meaningful statements are either analytically true (true by definition) or empirically verifiable. If every statement is embedded in a web that faces experience collectively, neither category survives intact.

Does this lead to relativism? 🔗

A natural worry is that if evidence cannot uniquely determine theory, and if observation is always theory-laden, then perhaps all theories are equally valid – or equally arbitrary. This concern is real but, most philosophers argue, overstated. Underdetermination does not imply that all theories are equally well-supported; rather, it suggests that empirical evidence alone is insufficient to fully adjudicate between rivals. Other factors – explanatory power, simplicity, coherence, predictive success – can still guide theory choice without collapsing into pure subjectivism.

Cognitive psychology research also supports the Hanson-Kuhn position while tempering its stronger claims: top-down theoretical influences on perception are strongest when incoming sensory input is weak or ambiguous, but in cases where sensory evidence is strong and unambiguous, there is little evidence that theory can override observation entirely. Science retains its grip on reality, even if that grip is never unmediated.

Why confirmation holism still matters today 🔗

Confirmation holism is not merely a historical curiosity in the philosophy of science. Its consequences continue to shape debates about scientific methodology, the objectivity of research, and the proper role of values in science. Climate science, for instance, involves precisely the kind of complex, interlocking web of models, instruments, and background theories that holism describes. When results are contested, the dispute often turns not on a single measurement but on competing assessments of auxiliary assumptions – instrument calibration, modeling choices, historical data interpretation.

Similarly, in medicine and psychology, replication crises have exposed how deeply results depend on methodological assumptions that go unstated and unexamined. Confirmation holism provides the philosophical vocabulary for understanding why a single experiment never settles a scientific question and why the context of background theory always matters.

The Stanford Encyclopedia of Philosophy notes that thinkers ranging from Lakatos and Feyerabend to feminist epistemologists have drawn on underdetermination and holism to argue that non-empirical values inevitably enter scientific practice – not as a contamination of pure inquiry but as an ineliminable part of how science actually works. Recognizing this is not a concession to relativism; it is an honest accounting of how knowledge is produced in a world where evidence never speaks entirely for itself.

What do you think? If no hypothesis can be tested in complete isolation from background assumptions, is there any meaningful sense in which a scientific theory can be proven false – or does falsification always remain, at some level, a matter of interpretive judgment? And if our observations are always shaped by the theories we already hold, what would it take for genuinely new evidence to change deeply entrenched scientific beliefs?

References

1. https://www.newworldencyclopedia.org/entry/Confirmation_holism
2. https://grokipedia.com/page/Confirmation_holism
3. https://philpapers.org/browse/quine-duhem-thesis
4. https://en.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis
5. https://scispace.com/pdf/the-duhem-quine-thesis-reconsidered-2b84tfze0z.pdf
6. https://plato.stanford.edu/entries/scientific-underdetermination/
7. https://iep.utm.edu/evidence/
8. https://en.wikipedia.org/wiki/Confirmation_holism
9. https://www.sciencedirect.com/science/article/abs/pii/S0039368116301248
10. https://www.samuelschindler.org/wp-content/uploads/2015/10/Observation-and-Theory-final.pdf
11. https://en.wikipedia.org/wiki/Theory-ladenness
12. https://compass.onlinelibrary.wiley.com/doi/10.1111/phc3.12475
13. https://escholarship.org/uc/item/140110h4