Which One of You Was It? The Duhem-Quine Problem and the Underdetermination of Theory

The Hook: 60 Nanoseconds That Blamed Nobody

On 22 September 2011 the OPERA collaboration uploaded a preprint to arXiv that is still available today under the number 1109.4897: Measurement of the neutrino velocity with the OPERA detector in the CNGS beam. Neutrinos had been sent from CERN through the rock to Gran Sasso, across a baseline of – in the paper's own words – "about 730 km". And according to the measurement, they had arrived earlier than light could have. The accessible overview accounts give an advance of roughly 60.7 nanoseconds for the first announcement; a refined analysis in November 2011 produced 57.8 nanoseconds.

If it held up, this was the end of special relativity in its familiar form. It did not hold up. In February 2012 the collaboration named two possible sources of error in the apparatus, and in July 2012 the final result of the same measurement in the same paper read: \((6.5 \pm 7.4_{\text{stat}}\,^{+8.3}_{-8.0\,\text{sys}})\) nanoseconds – that is, zero within the uncertainty.

What is interesting is not that the measurement was wrong. What is interesting is what was logically the case during the months in between. The prediction "neutrinos are not faster than light" was incompatible with the observation. So something was wrong. But what?

The list of candidates is long and heterogeneous: special relativity. The assumption that the two GPS-based clocks at the two ends run in step. The assumption that the baseline has been surveyed correctly. The assumption that signal propagation delays in the cables were subtracted properly. The assumption that the oscillator on an electronics board holds its nominal frequency. The statistical analysis of the proton pulse shape. And a long list of further things nobody writes down, because they go without saying.

The experiment said: at least one of these is false. It did not say which. And that was not down to poor engineering.

That the collaboration saw it exactly this way is in their own text: "We deliberately do not attempt any theoretical or phenomenological interpretation of the results." They had measured a discrepancy and expressly refused to turn it into a claim about theory.

This inability to say which has a name, an originator with a date, and a sequel that ploughed up twentieth-century philosophy. It is called the Duhem-Quine problem.

Part 1: Duhem and the Hypothesis That Never Stands Trial Alone

A Physicist Writes Philosophy

Pierre Maurice Marie Duhem (1861–1916) was a physicist, a historian of science and a philosopher; he described himself, as the Stanford Encyclopedia of Philosophy reports, as a "theoretical physicist". In 1906 his major work in the philosophy of science appeared in Paris with Chevalier et Rivière: La théorie physique, son objet et sa structure. A second edition followed in 1914, and Philip P. Wiener's English translation appeared in 1954 with Princeton University Press under the title The Aim and Structure of Physical Theory.

Duhem's central insight, in the version quoted by the SEP article on underdetermination from Wiener's 1954 translation (p. 185) – in a heavily condensed form whose omissions are marked here:

"A physicist decides to demonstrate the inaccuracy of a proposition; […] if the predicted phenomenon is not produced, the only thing the experiment teaches us is that among the propositions used […] there is at least one error; but where this error lies is just what it does not tell us."

And in summary form, two pages later (1954, p. 187):

"In sum, 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."

The SEP article on Duhem himself quotes the same point with an even sharper image: if the predicted phenomenon is not produced, then not only is the proposition in question thrown into doubt, but "also the whole theoretical scaffolding used by the physicist". And, as a warning to anyone who wants to try anyway:

"To seek to separate each of the hypotheses of theoretical physics from the other assumptions upon which this science rests, in order to subject it in isolation to the control of observation, is to pursue a chimera."

A chimera. Not difficult, not laborious, not expensive – a mythical beast.

The Logic Behind It Is Banal; the Consequences Are Not

The argument is so simple that it is almost a disappointment. No hypothesis \(H\) makes an observational prediction on its own. Predictions arise only from \(H\) together with a bundle of auxiliary assumptions \(A_1, \dots, A_n\): measurement theory, instrument properties, boundary conditions, background theories, mathematics. Formally:

$$ (H \land A_1 \land \dots \land A_n) \rightarrow B $$

If \(B\) is not observed, classical logic yields only \(\lnot(H \land A_1 \land \dots \land A_n)\) – equivalently, the disjunction \(\lnot H \lor \lnot A_1 \lor \dots \lor \lnot A_n\). Which conjunct is false remains open.

The precise reach of this observation matters, because this is where people tend to overreach. Modus tollens strikes the conjunction as a whole as long as none of the auxiliary assumptions has been independently secured. Secure them one by one, and the same inference does indeed strike a part – which is exactly the way out described in Part 6. The logical form produces no doom; it produces a to-do list.

What Duhem Did Not Claim

Here begins the first of several confusions that shape this debate. Duhem deliberately restricted his discussion to physics. The SEP puts it this way: he restricts "his generalizations to physics", and it quotes him saying that we shall stay "always within the limits of physical science". The SEP article on underdetermination is more direct still: Duhem held that the underdetermination he had described was "a challenge only for theoretical physics" – subsequent philosophy of science then came round to the view that the predicament "applies to theoretical testing in all fields of scientific inquiry".

But this restriction should not be read as a denial of holism outside physics. In Physical Theory and Experiment Duhem expressly compares physics with physiology and regards the difference as logically unimportant. In Wiener's translation, verbatim: "From the standpoint of logic, 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." Anyone working with a thermometer, a manometer, a calorimeter or a galvanometer performs the same act of faith; Duhem applies the same reasoning to chemistry and crystallography.

What does not belong to him is the extension to all knowledge, including logic and mathematics. That is Quine's step.

Duhem's Own Solution – and What to Make of It

If the experiment does not say which assumption falls, who does? Duhem's answer was bon sens, the good sense of the experienced scientist. Not a rule, not an algorithm, but trained judgement.

As an epistemological answer this is thin, and the SEP says so bluntly: Duhem's "rather hopeful appeal to the 'good sense' of scientists themselves […] promises very little if any relief from the general predicament of holist underdetermination".

I am of the opinion that this verdict overlooks a division of labour. At OPERA, bon sens did not solve the problem, but it did determine the search order: auxiliary assumptions first, theory second. What solved the problem afterwards was separate measurement. Judgement supplies the prioritization, not the decision – which is less than Duhem hoped for, but more than nothing.

Part 2: Quine and the Radicalization

Two Dogmas, One Paper

In 1951 The Philosophical Review, volume 60, issue 1, from page 20, carried a paper by Willard Van Orman Quine. The Crossref database lists it under the title Main Trends in Recent Philosophy: Two Dogmas of Empiricism – the first words apparently come from a departmental heading of the journal, not from Quine. It opens:

"Modern empiricism has been conditioned in large part by two dogmas."

The two dogmas: first, the cleavage between analytic truths (true by virtue of meaning) and synthetic ones (true by virtue of fact); second, reductionism, on which every meaningful statement is reducible to some content of experience. Quine's verdict arrives at once: "Both dogmas, I shall argue, are ill founded."

The Place Where Duhem Appears

Section V contains two sentences one has to know; the second is its closing sentence. First the diagnosis:

"[…] our statements about the external world face the tribunal of sense experience not individually but only as a corporate body."

And then the consequence:

"But what I am now urging is that even in taking the statement as unit we have drawn our grid too finely. The unit of empirical significance is the whole of science."

To the first of these passages Quine attaches an endnote, and in the history of philosophy it is the point at which two theses become one. Verbatim:

"This doctrine was well argued by Pierre Duhem, La Theorie physique: son objet et sa structure (Paris, 1906): 303-328. Or see Armand Lowinger, The Methodology of Pierre Duhem (New York: Columbia University Press, 1941): 132-140."

An editorial subtlety matters here, and it is often suppressed: in the widely used online edition of the text, the endnotes are marked by version – "a" for the 1951 first printing, "b" for the 1961 version. The Duhem note carries the label 17b, and for notes 13 through 22 there is no "a" counterpart at all. On that numbering, the Duhem reference therefore appears only in the later version, not in the first printing. Anyone dating it to 1951 is dating it too early.

What is also remarkable is where the reference sits: in the endnotes, not in the text. The name "Duhem" does not occur a single time in the main text of Two Dogmas. The expression "Duhem-Quine thesis" is a later amalgamation by the profession, not a self-description – who coined it, I was unable to document in this run.

And it joins together two things of very unequal size. Duhem speaks of groups of hypotheses in physics. Quine's unit is "the whole of science" – everything.

The Image That Changed Everything

Section VI, Empiricism without the Dogmas, opens with perhaps the most quoted image in analytic philosophy:

"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."

Knowledge is a fabric that touches experience only at the edges. Or: a field of force whose boundary conditions are experience. From which it follows immediately:

"A conflict with experience at the periphery occasions readjustments in the interior of the field."

And, pointedly:

"No particular experiences are linked with any particular statements in the interior of the field, except indirectly through considerations of equilibrium affecting the field as a whole."

No single experience hangs on a single statement. What experience does is put the field as a whole under tension; where in the interior something gives way is not thereby settled.

The Two Sentences People Fight Over

Then come the two formulations that have counted ever since as Quine's real thesis:

"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."

And Quine means it. His example is not some remote borderline case but logic itself:

"Revision even of the logical law of the excluded middle has been proposed as a means of simplifying quantum mechanics; and what difference is there in principle between such a shift and the shift whereby Kepler superseded Ptolemy, or Einstein Newton, or Darwin Aristotle?"

That is the radicalization in a single sentence. Duhem says: in physics, the hypothesis hangs in a thicket of auxiliary assumptions. Quine says: the law of excluded middle hangs in the same thicket, just further in.

It should be recorded, though, that Quine later softened this edge. On the account of the SEP article on Quine, he goes on in later work to propose a definition of analyticity himself – a sentence is analytic for a speaker "if he learned the truth of the sentence by learning to use one or more of its words" (Quine 1991, p. 270) – holds on that basis that all first-order logical truths "would then perhaps qualify as analytic" (ibid.), concedes that certain revisions of belief do involve a change of meaning, and recognizes analyticity "in its obvious and useful but epistemologically insignificant applications" (Quine 1991, p. 271). The SEP stresses at the same time that this does not withdraw the core critique. The 1951 version is thus the sharp one, not the last one.

What This Means for Ontology

Quine consistently draws the consequence for the question of what there is, too. Physical objects do not come out of experience by definition but "simply as irreducible posits comparable, epistemologically, to the gods of Homer". Then follows, with dry humour, a declaration on his own behalf:

"Let me interject that for my part I do, qua lay physicist, believe in physical objects and not in Homer's gods; and I consider it a scientific error to believe otherwise."

And only after that, the epistemological levelling:

"[…] in point of epistemological footing the physical objects and the gods differ only in degree and not in kind."

Anyone suspecting relativism here is disappointed a few sentences later. Quine makes a choice, and he makes it on performance:

"The myth of physical objects is epistemologically superior to most in that it has proved more efficacious than other myths as a device for working a manageable structure into the flux of experience."

The paper ends accordingly:

"In repudiating such a boundary I espouse a more thorough pragmatism. Each man is given a scientific heritage plus a continuing barrage of sensory stimulation; and the considerations which guide him in warping his scientific heritage to fit his continuing sensory promptings are, where rational, pragmatic."

This is not "everything is equally good". It is: the decision is not made logically but pragmatically – and it is made.

Quine returned to the paper once more in 1991, in Two Dogmas in Retrospect (Canadian Journal of Philosophy 21(3), pp. 265–274). The text opens with the sentence: "In retrospecting 'Two Dogmas' I find myself overshooting the mark by twenty years." Exactly what Quine corrects there beyond the points named above could not be checked in this run; the full text was not reachable.

Part 3: Two Problems That Are Constantly Confused

This is where the real gain in understanding of recent decades lies, and it is a distinction. The SEP article – written, incidentally, by P. Kyle Stanford – puts it up front:

"Perhaps the most important division is between what we might call holist and contrastive forms of underdetermination."

Holist underdetermination is Duhem's problem. It arises, verbatim, "whenever our inability to test hypotheses in isolation leaves us underdetermined in our response to a failed prediction or some other piece of disconfirming evidence". It concerns the response to a failure: which assumption do I give up?

Contrastive underdetermination is something entirely different. It concerns, again verbatim, "the quite different possibility that for any body of evidence confirming a theory, there might well be other theories that are also well confirmed by that very same body of evidence". Here there is no failure but a surplus: two theories, one body of data.

To this is added a second axis, which applies to both varieties – durability. Is the choice "transiently underdetermined simply by the evidence we happen to have at present", or "permanently underdetermined by all possible evidence"? That is the difference between a research assignment and a limit in principle.

And a third – strength. Larry Laudan has argued, on the SEP's account, that defenders of the underdetermination thesis defend only its weaker versions while drawing their dramatic consequences from far stronger ones. His point in one sentence, as the SEP renders it:

"Laudan suggests that we might reasonably hold the resources of deductive logic to be insufficient to single out just one acceptable response to disconfirming evidence, but not that deductive logic plus the sorts of ampliative principles of good reasoning typically deployed in scientific contexts are insufficient to do so."

That is the crux. Purely deductive underdetermination is nearly trivial: it follows from the form of the test premise alone and carries no radical consequences by itself. The interesting question is whether deduction plus the usual non-deductive principles – simplicity, explanatory power, conservatism, fruitfulness – leaves the choice open. For that, the logical observation alone is no argument.

The SEP grants Laudan that these versions need distinguishing – "Laudan is certainly right to distinguish these various versions of holist underdetermination" – but holds the situation to be more complicated: "the underlying situation is somewhat more complex than he suggests". The decisive reservation is a Quinean riposte: the appeal to ampliative principles is problematic as a response to Quine himself, because on Quine's view those principles hang in the web of belief too and are therefore equally revisable. Whoever invokes simplicity in order to close the underdetermination is using a principle that, on Quine's account, does not stand outside the dispute.

Part 4: The Counterattack I – There Is No Road from Revisable to Analytic

Five years after Two Dogmas, the same journal carried the best-known reply: H. P. Grice and P. F. Strawson, In Defense of a Dogma, The Philosophical Review 65(2), April 1956, from p. 141.

Their own argument, as the SEP cites it, is methodological: "As Grice and Strawson (1956) and Putnam (1965 [1975]) pointed out, it is unlikely that so intuitively plausible a distinction should turn out to have no basis at all in fact." A distinction that intuitively plausible is unlikely to be wholly groundless.

The genuinely sharp objection to Quine's inference pattern comes from the reconstruction in the SEP article on the analytic/synthetic distinction (Georges Rey). Quine's argument there has this form:

"But no sentence is absolutely immune from revision; all sentences are thereby empirical, and none is actually analytic."

And the objection to it is of almost embarrassing simplicity:

"Indeed, a claim might be in fact analytic and justifiable independently of experience, but nevertheless perfectly well revised in the light of it."

The inference from revisability to non-analyticity therefore does not go through: on this reconstruction, analyticity and unrevisability are two different properties. The SEP puts its finger on the equation:

"Where Kant and others had traditionally assumed that the a priori concerned beliefs justifiable independently of experience, Quine and many other philosophers of the time came to regard it as consisting of beliefs unrevisable in the light of experience."

For otherwise something absurd would follow:

"[…] this would imply that people taking something to be analytic or a priori would have to regard themselves as being infallible about it, forever unwilling to revise it in light of further evidence or argument. But this is a further claim that many defenders of the traditional notions need not embrace."

That the effect was nonetheless enormous is recorded by the same article: Quine's position led "several generations of philosophers" to despair not only of the analytic/synthetic distinction but "of the category of a priori knowledge entirely". And the SEP does not close with a rehabilitation either: it has "yet to be shown that appeals to the analytic will ever be able to ground 'analysis' and the a priori in quite the way that philosophers had hoped".

That is the honest interim balance: Quine's attack has not been refuted, but the radical conclusion does not follow – and the hole he tore has not been closed to this day.

Part 5: The Counterattack II – Empirical Equivalence Is a Finding of Today

The second major objection targets the contrastive variety. Larry Laudan and Jarrett Leplin published Empirical Equivalence and Underdetermination in 1991 in The Journal of Philosophy 88(9), from p. 449 (DOI 10.2307/2026601).

Their approach has two levers.

First: what counts as an "empirical consequence" of a theory is not a fixed property of the theory. It depends on what happens to be observable and which auxiliary assumptions happen to be available – and both change. What is a merely logical consequence today may become an empirical one tomorrow because an instrument has been invented. Auxiliary assumptions are both augmentable and defeasible. Consequently, any finding that two theories are empirically equivalent is context-dependent and provisional. It is a statement about the state of science, not about the theories.

Second: even genuine empirical equivalence does not compel equality of confirmation. Evidence can support a theory indirectly – via a more general theory with which it is connected, without the evidence being an observational consequence of the theory itself. Anyone who reduces evidential relations to logico-semantic ones loses precisely this route.

Caution is in order here, because this is a place where people tend to overreach. Pablo Acuña and Dennis Dieks examined exactly this – in 2013 as a preprint in the PhilSci-Archive (item 9883), published in 2014 in the European Journal for Philosophy of Science 4(2), pp. 153–180 (DOI 10.1007/s13194-013-0080-3). Their assessment, verbatim from the abstract: Laudan and Leplin's reasoning is "essentially correct" – but:

"We argue that Laudan and Leplin's analysis does not refute the problem (as they claim). Instead, what they show is merely that science possesses tools that may eventually lead out of an underdetermination impasse."

And they run the test on a real case – "Lorentz's ether theory and Einstein's special relativity" – with a conclusion I consider the most useful in the whole debate:

"We conclude that empirical equivalence and underdetermination are ordinary scientific problems rather than problems that should be solved by epistemology."

Underdetermination, on this view, is not a philosophical doom but a normal research situation: something that can be dissolved by work – sometimes.

Part 6: How Science Actually Settles the Question of Blame

If logic leaves the question of blame open, how does it get answered? Not by a principle but by a move: the auxiliary assumption itself becomes a hypothesis and is measured independently.

The SEP offers a pair of examples for this that is instructive precisely because it shows the same situation twice with two different outcomes: Uranus and Neptune on the one hand, Vulcan and Mercury on the other.

In both cases a planetary orbit departs from the Newtonian prediction. In both cases the obvious rescue runs: the theory is right, an auxiliary assumption about the number of celestial bodies is incomplete – there is another planet. In the first case the rescue was correct; the SEP writes that "the astronomers Adams and Leverrier were eventually led to discover Neptune in 1846". In the second it was wrong; the postulated innermost planet did not exist, and the advance of Mercury's perihelion "would resist satisfactory explanation until the arrival of Einstein's theory of general relativity", as the SEP puts it.

That is the decisive point: the same strategy, once successful and once a failure – and not for logical reasons but for empirical ones. The answer to the question of blame is not available a priori and yet it is not arbitrary. It is a further research question.

OPERA is the same story in months rather than decades. The auxiliary assumptions were unscrewed one at a time and measured one at a time. According to the publicly accessible overview accounts there were two sources of error, and they pulled in opposite directions: a loose connection in the fibre-optic link from a GPS receiver to the master clock, which shortened the reported flight time by about 73 nanoseconds – thereby producing the apparent effect – and an oscillator on an electronics board that ran faster than its expected frequency of 10 MHz and lengthened the reported flight time, thereby shrinking the apparent effect.

Added to this was the strongest instrument in practice: independent repetition with a different apparatus. The ICARUS experiment, operated at the same site, reported in March 2012 a flight time consistent with the speed of light for seven neutrinos from the short-bunch beam. On 8 June 2012 CERN's research director Sergio Bertolucci declared in Kyoto that the neutrino speed was consistent with the speed of light. The corrected version of the OPERA paper appeared in July 2012 and was published as Journal of High Energy Physics 2012(10), article 93 (DOI 10.1007/JHEP10(2012)093).

So the auxiliary assumption was testable after all – just not by the same experiment that used it.

One limit remains. P. Kyle Stanford named it in 2006 in Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives (New York: Oxford University Press): the practical problem is not choosing among the alternatives we know, but that the decisive alternative has not yet been conceived. Anyone trying to explain Mercury's orbit before relativity can choose rationally among all the options available to them – and the right one is not among them. That is a historical claim, not a logical one, and it is contestable accordingly.

Part 7: A Map of the Positions

The Central Takeaway

In its defensible form, the Duhem-Quine problem is not scepticism but an accounting rule: a negative result first charges the entire setup – and only then a single assumption, once the others have been independently cleared. Anyone who tells it differently has left assumptions out.

Three things follow for daily work.

1. A red test does not localize within the setup. A failing test, an incident, a benchmark regression, a result that will not reproduce – each of them says: "At least one assumption in this setup is false." That is a real and usable restriction, since the set of assumptions in play is finite and enumerable. But the code is only one candidate within it. Test data, mock behaviour, environment variables, the clock, concurrency, the network, the compiler version, the measuring tool itself – all of them stand trial too. The debugging effort that most often fails is the one that knows only a single suspect from the start.

2. Testability comes from isolating things. OPERA did not solve the problem by better interpretation but by measuring the auxiliary assumptions separately – and conclusively only with a second, independent apparatus. Transferred: bisection, a second independent measurement path, an attempt to reproduce in a foreign environment, a test that checks only the auxiliary assumption. What cannot be observed individually in the setup cannot be individually cleared when something fails. That is an argument for observability long before it is one for debugging.

3. "Behaves identically" is a finding of today. Two implementations, two architectures, two configurations that look the same under all available observations are not the same – they are so far indistinguishable. Laudan and Leplin's point transfers as an analogy: the equivalence is a statement about your instruments, not about the systems. New load, new metric, new edge case – and the equivalence can vanish. Decisions resting on "makes no difference" should carry a note about under which observations that holds.

And a fourth consequence, more attitude than technique: when a finding seems to topple a central assumption, the most likely explanation is not the toppling but an auxiliary assumption. That is not blind conservatism but an estimate of base rates – and it is exactly what OPERA did when it refused to turn 60 nanoseconds into a theory.

A Closing Question for Reflection

Think of the last bug that cost you more than a day. Reconstruct the list of assumptions that all had to be true simultaneously at the moment the symptom appeared – and do it completely, including the three or four you would not have written down at the time because they went without saying.

Where in that list was the fault? And: where in the list did you look first?

The real question is the third: what would have had to be different in your setup for the assumption that was actually guilty to have been individually measurable? The answer to that is not a philosophical confession but a list of concrete changes – and it is the only thing that has ever turned the Duhem-Quine problem into a solvable task.

Fact Check

102 claims were checked: 36 hard facts (figures, years, titles, authors, volumes, pages, DOIs), 40 verbatim quotations, 12 source attributions and 14 synthesis claims. 24 were corrected or weakened – 6 discarded as false, 16 weakened, 2 marked as not verifiable.

The three most important corrections:

1. A sentence presented as a verbatim quotation, to the effect that Laudan traces radical conclusions back to a "simple equivocation", does not appear in that form in the source cited; the word "equivocation" does not occur in the SEP article at all. The passage is now formulated as a paraphrase.
2. Quine's opening sentence from 1991 reads "…I find myself overshooting the mark by twenty years" – two words were missing. At the same time it was false to claim that what Quine corrected in 1991 could not be established: the SEP article on Quine documents the softening (a definition of analyticity, change of meaning in certain revisions, first-order logical truths), which now appears in the text and expressly qualifies the strong reading.
3. The endnote in which Quine cites Duhem carries the label 17b in the online edition and, on that edition's numbering, appears only in the 1961 version, not in the 1951 first printing – which made the previous dating ("45 years later") false.

Further weakenings: the logical point (modus tollens strikes the conjunction only as long as no auxiliary assumption has been independently secured – otherwise it does strike a part); Duhem's restriction to physics (he restricted the discussion, not holism, and regarded the difference from physiology as "from the standpoint of logic" slight); the claim that Duhem had not anticipated independent repetition (deleted); the presentation of Laudan & Leplin (they do not refute the problem but, per Acuña & Dieks, merely display tools for finding a way out of such a situation); the order of the Quine quotations in Section VI; the details of the OPERA episode (the CERN press releases were unreachable – primarily documented are the arXiv data, the baseline, the final result and the refusal to interpret; the two sources of error, ICARUS and the declaration of 8 June 2012 come from accessible overview accounts); and a table row that wrongly attributed Rey's objection to Grice & Strawson. Left unverifiable were the authorship of the expression "Duhem-Quine thesis" and the extent of Quine's 1991 corrections beyond the documented points; both are marked as such in the text.

Automated AI check against primary sources; no human expert review.

Cross-References in the Vault

• Why the Sun Will Rise Tomorrow: Hume's Problem of Induction, Grue, and the Limits of Learning – the closest relative. Hume and Goodman ask why finite evidence singles out one continuation; Duhem and Quine ask why a disconfirming result singles out one assumption. The same gap, viewed from two sides.
• The Rule That Cannot Explain Itself: Wittgenstein's Paradox and Kripke's Sceptic – the sister argument: there, the rule does not determine its application; here, the finding does not determine its consequence. In both cases what closes the gap is a practice, not a derivation.
• The Cosmic Tension: Why the Universe Seems to Have Two Rates of Expansion – an ongoing case of holist underdetermination in pure form: two measurement routes contradict each other, and the open question is exactly which auxiliary assumption falls.
• The Field That Wasn't There: The Aharonov-Bohm Effect and the Reality of Potentials – empirical equivalence you can touch: gauge-related descriptions predict the same thing, and then an experiment shows that the distinction has consequences after all.
• The Solar Storm That Dated the Vikings: 1021 and the Proof Before Columbus – the resolution in practice: the most useful anchor comes from outside the dispute. That is exactly what the independent experiment delivered at OPERA.
• Three Pages Against 2,000 Years: The Gettier Problem and What Knowledge Really Is – the same pattern of impact: a short text, a sharply framed argument, and an established distinction is permanently damaged.
• The Interest Rate of Morality: The Social Discount Rate and What the Future Is Worth to Us – the sequel in modelling practice: there too it is not the empirical evidence that decides, but a stipulation entered into the calculation as a number.
• Will I Still Be Me Tomorrow? Personal Identity from the Ship of Theseus to the Teletransporter – a related case in which competing descriptions admit the same observations and the decision is made elsewhere.

Sources and Further Reading

• Quine, W. V. (1951): Two Dogmas of Empiricism. The Philosophical Review 60(1), from p. 20 (Crossref lists the contribution as "Main Trends in Recent Philosophy: Two Dogmas of Empiricism"). DOI: https://doi.org/10.2307/2181906 – Online edition of the full text: https://www.ditext.com/quine/quine.html (endnotes there are marked by 1951/1961 version)
• Quine, W. V. (1975): On Empirically Equivalent Systems of the World. Erkenntnis 9(3), 313–328. DOI: https://doi.org/10.1007/BF00178004
• Quine, W. V. (1991): Two Dogmas in Retrospect. Canadian Journal of Philosophy 21(3), 265–274. DOI: https://doi.org/10.1080/00455091.1991.10717246
• Grice, H. P., Strawson, P. F. (1956): In Defense of a Dogma. The Philosophical Review 65(2), from p. 141. DOI: https://doi.org/10.2307/2182828
• Laudan, L., Leplin, J. (1991): Empirical Equivalence and Underdetermination. The Journal of Philosophy 88(9), from p. 449. DOI: https://doi.org/10.2307/2026601
• Acuña, P., Dieks, D. (2014): Another Look at Empirical Equivalence and Underdetermination of Theory Choice. European Journal for Philosophy of Science 4(2), 153–180. DOI: https://doi.org/10.1007/s13194-013-0080-3 – Preprint (2013): https://philsci-archive.pitt.edu/9883/
• Stanford, P. K. (2009, rev. 4 Apr 2023): Underdetermination of Scientific Theory. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/scientific-underdetermination/
• Ariew, R. (2007, rev. 13 Feb 2026): Pierre Duhem. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/duhem/
• Rey, G. (2003, rev. 1 Jun 2026): The Analytic/Synthetic Distinction. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/analytic-synthetic/
• Kemp, G. (2010, rev. 6 Jul 2023): Willard Van Orman Quine. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/quine/
• OPERA Collaboration (2011/2012): Measurement of the neutrino velocity with the OPERA detector in the CNGS beam. arXiv:1109.4897 (v1 22 Sep 2011, v4 12 Jul 2012); published as JHEP 2012(10), art. 93. https://arxiv.org/abs/1109.4897 – DOI: https://doi.org/10.1007/JHEP10(2012)093
• Stanford, P. K. (2006): Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives. New York: Oxford University Press.
• Duhem, P. (1906): La théorie physique, son objet et sa structure. Paris: Chevalier et Rivière; 2nd ed. 1914; in English as The Aim and Structure of Physical Theory, translated by Philip P. Wiener, Princeton University Press 1954.

Note: the underdetermination debate is open. What is solid are the bibliographic details and the distinction between holist and contrastive underdetermination; the verbatim quotations were checked against the online sources listed in this run, and two faulty quotations were corrected. Contested are the reach of the strong Quinean reading and the question of whether Laudan & Leplin dissolve the problem. The transfers to debugging, observability and architectural decisions in "The Central Takeaway" are the author's analogies and not results of the literature cited; where this article goes beyond the secured state of knowledge in other respects, it is marked with "I am of the opinion that ...".

← All articles