1 See Bayer, 2007 Bayer, B. 2007. ‘How Not to Refute Quine: Evaluating Kim’s Alternatives to Naturalized Epistemology’. Southern Journal of Philosophy, 45: 1–23.  [Google Scholar]: pp. 5–6 for more.

2 In one line of thinking, if Quine doesn’t intend to establish the foundations of science, he should at least show how to ‘dissolve’ the problem of scepticism, perhaps by showing, in the manner of Wittgenstein (1969 Wittgenstein, L. 1969. On Certainty, Edited by: Anscombe, G. E. M. and von Wright, G. H. New York: Harper Torchbooks.  [Google Scholar]: p. 2e), that it doesn’t make sense to doubt what the sceptics ask us to doubt. Yet Quine’s estimate of the Wittgensteinian strategy is not supportive. He pejoratively characterizes it as offering philosophers a ‘residual philosophical vocation in therapy’, and urges that contrary to the Wittgensteinian idea that philosophical problems are delusions, ‘epistemology still goes on’ after the death of foundationalism (Quine, 1969a Quine, W. V. O. 1969a. “‘Epistemology Naturalized’”. In Ontological Relativity and Other Essays, Edited by: Quine, W. V. O. 69–90. New York: Columbia University Press.  [Google Scholar]: p. 82).

3 See also Quine, 1960 Quine, W. V. O. 1960. Word and Object, Cambridge, MA: Harvard University Press.  [Google Scholar]: p. 2 and Quine 1974 Quine, W. V. O. 1974. The Roots of Reference, LaSalle, IL: Open Court Press.  [Google Scholar]: pp. 1–4.

4 Stroud contends, in particular, that to explain the origin of some subject’s knowledge, two conditions must hold: we, the explainer, must know that the subject’s belief is true; and we must be able to show that it is not an accident, that the subject’s posit turns out to be true because of some connection to the truth. Yet Quine’s naturalistic investigation also reveals and is further motivated by the thesis that our subject’s sensory inputs are ‘meagre’ in comparison to his outputs. If, as Quine suggests, our position is similar to the subject’s, then there is a serious question about whether we, the investigators, are in a position to fulfil the two conditions of explaining the origin of knowledge, not only for our subject, but for ourselves. Our own conclusions about the subject’s circumstances – about whether his connection to facts is accidental or not – would also be mere posits, each with its own rival hypotheses.

5 Or perhaps we make our psychology more rigorous, and note the more general distinction between perception and judgment (as Descartes did as well) – and that only the latter can be true or false, discounting the claim that it is perception that is unreliable. I actually think that this and other attempts to block the reductio by appealing to the wider scientific context would be a fruitful response to the sceptic. The present point is that the option is not available to Quine in the grander scheme of his work.

6 This is, at least, Williams’ take on the underdetermination problem. It is possible, however, that there may be a way of showing that the underdetermination sceptic is also overreacting, and pointing to additional scientific evidence that shows that the problem arises only from an impoverished understanding of the nature of sensory perception, or from an overly narrow hypothetico‐deductive conception of confirmation. In any case, this is evidence that Quine does not consider, nor would it be characteristic of him to do so, since he takes it for granted that hypothetico‐deductivism is central if not exhaustive of the scientific method.

7 In their exchange, both Stroud and Quine discuss the inscrutability of reference thesis explicitly. Stroud is worried that on Quine’s view, there is a ‘possibility that the world is completely different in general from the way our sensory impacts and our internal makeup lead us to think of it’. In response, Quine proposes to understand this point in terms of ‘proxy functions and displaced ontologies’ (1981a Quine, W. V. O. 1981a. ‘Reply to Stroud’. Midwest Studies in Philosophy, 6: 473–5. [Crossref] , [Google Scholar]). ‘Proxy functions’ are logical devices Quine has exploited to bolster the argument for his inscrutability of reference thesis.

8 Suppose, for example, that Jones sees a look‐alike of Smith in the room, and claims that Smith is in the room, and so fails to know this even though Smith is truly in the room (he’s hiding). Here we can diagnose the failure of knowledge as resulting from the fact that the person to whom Jones is referring is not actually Smith. Because Jones does not successfully refer to Smith in this circumstance, he cannot know that Smith is in the room. If term reference is systematically inscrutable, so that there are no facts of the matter determining what any of our terms refer to, then we may be faced with universal Gettierization of our knowledge. If there is no way at all to specify facts that our knowledge refers to, there is no way to show any non‐accidental connection between these facts and our justification for believing in them.

9 It might be alleged that the absence of determinate reference facts need not imply scepticism, if one adopts something akin to the ‘ecumenical’ strategy outlined in n. 10 below. That is, it might be argued that there may simply be no important difference in content between the belief that ⟨p⟩ and the belief that ⟨proxy p⟩. One might take inscrutability arguments as simply expressing scepticism about ‘intensional objects’ such as Fregean or Russellian propositions, but Quine sees his arguments against propositions as distinct from his arguments for the inscrutability of reference. Inscrutability of reference is directed against the reference of terms, not propositions. While it is true that one route to the indeterminacy of translation of whole sentences is through term inscrutability, Quine says that there are multiple routes to the same conclusion, including another drawn from confirmation holism (Quine, 1970 Quine, W. V. O. 1970. ‘On the Reasons for the Indeterminacy of Translation’. Journal of Philosophy, 67: 178–83. [Crossref], [Web of Science ®] , [Google Scholar]). What’s more, Quine sees inscrutability of reference as applying even to the terms of observation sentences, where he does not have a problem with indeterminacy of translation. Anyway, as I have already argued, indeterminacy about the objects of belief (via inscrutability about the reference of terms) spells trouble for identifying a determinate relationship between these objects and our reasons for believing in them, which threatens universal Gettierization and therefore scepticism – this is a problem that stands independently of considerations about propositions. What’s more, Quine himself abandons the ecumenical strategy (1986b Quine, W. V. O. 1986b. “‘Reply to Roger Gibson’”. In The Philosophy of W. V. Quine, Edited by: Hahn, L. and Schilpp, P. 155–6. La Salle, IL: Open Court.  [Google Scholar]; 1992: pp. 99–101).

10 This is particularly clear in a case in which there are two theories, T₀ and T_r, which are equivalent in evidence and theoretical virtues, but known to be incompatible. Bergstrom says that if we know that T_r rivals our home theory T₀, the only rational option is to suspend judgment between the two. We cannot justifiably pick one or the other, and so we do not know which one is true. Bergstrom thinks that the same is true even if we don’t know the nature of T_r, but simply know that it exists that (if, for example, we accept Quine’s underdetermination thesis).

Recently, Bergstrom (2004 Bergstrom, L. 2004. “‘Underdetermination of Physical Theory’”. In The Cambridge Companion to Quine, Edited by: Gibson, R. 91–114. New York: Cambridge University Press.  [Google Scholar]: p. 105) has pointed out that the sceptical implications of underdetermination are clearest when the rival empirically equivalent theories are actually incompatible, i.e., not possibly both true. Earlier (1993 Bergstrom, L. 1993. ‘Quine, Underdetermination and Scepticism’. Journal of Philosophy, 90(7): 331–58.  [Google Scholar]: pp. 343, 345), he does suggest that if T₀ and T_r are not incompatible but merely different, there might still be something irrational about accepting one rather than the other when they account equally well for the evidence. Indeed, if T₀ and T_r are empirically equivalent but not logically incompatible, it is trickier to show the sceptical implication, because then perhaps Quine could take what he calls an ‘ecumenical’ (rather than a ‘sectarian’) line, and say that both of these theories could be said to be true (Quine, 1986b Quine, W. V. O. 1986b. “‘Reply to Roger Gibson’”. In The Philosophy of W. V. Quine, Edited by: Hahn, L. and Schilpp, P. 155–6. La Salle, IL: Open Court.  [Google Scholar]). If both can be said to be true, then there is no question to be agnostic about, no reason to think that one’s present theory is threatened. In fact Quine believes that many cases of apparently incompatible rivals can be reduced to compatible ones, if incompatibilities arising from theoretical terms are eliminated by spelling the relevant theoretical terms differently in each theory (for example, ‘the universe expands’ versus ‘the youniverse does not expand’).

But Bergstrom gives a variety of reasons (related to simplicity and economy) for thinking it strange that one could be warranted in believing the conjunction of T₀ and T_rto be true (1993 Bergstrom, L. 1993. ‘Quine, Underdetermination and Scepticism’. Journal of Philosophy, 90(7): 331–58.  [Google Scholar]: p. 347), and argues that the spelling expedient would not eliminate the existence of incompatible theories – if there are any – but only allow us to deal with their compatible counterparts (pp. 350–1). But as late as ‘On Empirically Equivalent Systems of the World’ (1975) Quine still stresses the importance (or threat) of the possibility of incompatible rivals, even if not all empirically equivalent rivals are actually incompatible. And Quine later (1986b Quine, W. V. O. 1986b. “‘Reply to Roger Gibson’”. In The Philosophy of W. V. Quine, Edited by: Hahn, L. and Schilpp, P. 155–6. La Salle, IL: Open Court.  [Google Scholar]; 1992: pp. 99–101) distances himself from the ecumenical position, though he leaves open the possibility that a sectarian might oscillate between compatible rivals, without believing both at the same time. There are even some philosophers of a verificationist bent who doubt that there could ever be such things as empirically equivalent but incompatible rivals (Dummett, 1973 Dummett, M. 1973. Frege: Philosophy of Language, London: Duckworth.  [Google Scholar]: p. 617; Davidson, 1990b Davidson, D. 1990b. ‘The Structure and Content of Truth’. Journal of Philosophy, 88(6): 279–328.  [Google Scholar]: p. 306). The issues involved in deciding whether these doubts are warranted are difficult and beyond the scope of this paper, so I will not dwell on them further here. See Bergstrom, (2004 Bergstrom, L. 2004. “‘Underdetermination of Physical Theory’”. In The Cambridge Companion to Quine, Edited by: Gibson, R. 91–114. New York: Cambridge University Press.  [Google Scholar]: pp. 101–4) for a summary of the debate.

In any case, I will assume that Quine adopts underdetermination with incompatibility here, especially since it is the version of the thesis that has the most significant implications for the rest of his philosophy. A full‐fledged underdetermination thesis that assumes incompatible rivals, which are crucial to establishing the indeterminacy of translation, and thereby undermining the ‘conceptual project’ in epistemology, is another central motivation for naturalizing epistemology (Bayer, 2007 Bayer, B. 2007. ‘How Not to Refute Quine: Evaluating Kim’s Alternatives to Naturalized Epistemology’. Southern Journal of Philosophy, 45: 1–23.  [Google Scholar]: pp. 11–12, 14–18). If underdetermined theories are all equally true – as the ecumenical position suggests – then there is no reason to say that there is no fact of the matter involved in choosing a translation manual. If all translations are equally right, then there are many facts of the matter on which they are based, and translation is not indeterminate – and Quine loses his argument for naturalized epistemology from the indeterminacy of translation (which he invokes against Carnap’s last‐ditch attempt at empiricist epistemology). So, if underdetermination is watered down not to require the incompatibility of empirically equivalent rivals, it is true that its sceptical implications become less clear, but by the same token it also loses its significance as a motivation for naturalizing epistemology.

11 The only reason Descartes considers such scenarios is that they imply systematic alternatives to common sense which are indistinguishable from common sense on the basis of the available evidence – and the underdetermination thesis implies the existence of the same kind of alternatives.

12 Stroud’s second response (1984 Stroud, B. 1984. The Significance of Philosophical Scepticism, Oxford: Oxford University Press. [Crossref] , [Google Scholar]) is a recapitulation of much already noted in his earlier article (1981 Stroud, B. 1981. ‘The Significance of Naturalized Epistemology’. Midwest Studies in Philosophy, 6: 455–71. [Crossref] , [Google Scholar]).

13 Thus it seems that we should concur with Stroud, along with Davidson (1990a Davidson, D. 1990a. “‘Meaning, Truth and Evidence’”. In Perspectives on Quine, Edited by: Barrett, R. and Gibson, R. 68–79. Oxford: Blackwell.  [Google Scholar]: p. 74) and Koppelberg (1998 Koppelberg, D. 1998. ‘Foundationalism and Coherentism Reconsidered’. Erkenntnis, 49: 255–83.  [Google Scholar]: pp. 266–7), who urge that the containment of epistemology in ontology is no panacea for the naturalistic response to scepticism.

14 I myself am sympathetic to the idea that the sceptic’s doubts are incoherent, and that this by itself is sufficient to diagnose the sceptical illness. But I also agree with Stroud (1984 Stroud, B. 1984. The Significance of Philosophical Scepticism, Oxford: Oxford University Press. [Crossref] , [Google Scholar]: p. 227) that Quine’s ‘view of language and his rejection of the philosophical use of synonymy or analyticity leave him in no position to appeal to what is or is not included in the meaning of a particular term’, and that arguments from coherence do tend to presuppose specific theories of meaning, whether analytic or otherwise.

15 It is odd that both Stroud and Gibson focus their respective interpretations on only the final three paragraphs of ‘Reply to Stroud’, neglecting to discuss the main body of the paper.

16 Here Quine was pulling back from a brink reached in ‘Ontological Relativity’ (1969b Quine, W. V. O. 1969b. “‘Ontological Relativity’”. In Ontological Relativity and Other Essays, Edited by: Quine, W. V. O. 26–68. New York: Columbia University Press.  [Google Scholar]), in which the notion of reference – a concept he originally thought to be respectable and objective in ‘Notes on a Theory of Reference’ (1953a Quine, W. V. O. 1953a. “‘Notes on a Theory of Reference’”. In From a Logical Point of View, Edited by: Quine, W. V. O. 130–8. Cambridge, MA: Harvard University Press.  [Google Scholar]) – appeared to lose all such respectability and drop to the status enjoyed by murkier notions such as meaning.

17 Except, of course, that Quine thinks inscrutability applies to terms even in observation sentences, while indeterminacy does not apply to observation sentences taken as wholes.

18 This is especially clear when, after showing how reference can only be specified by bringing in (equally inscrutable) questions about identity and diversity (‘Is this the same gavagai as that?’, etc.), Quine states:

 Two pointings may be pointings to a numerically identical rabbit, to numerically distinct rabbit parts, and to numerically distinct rabbit stages; the inscrutability lies not in resemblance, but in the anatomy of sentences. We could equate a native expression with any of the disparate English terms ‘rabbit’, ‘rabbit stage’, ‘undetached rabbit part’, etc. and still, by compensatorily juggling the translation of numerical identity and associated particles, preserve conformity to stimulus meanings and occasion sentences.

  (1960 Quine, W. V. O. 1960. Word and Object, Cambridge, MA: Harvard University Press.  [Google Scholar]: pp. 53–4, my italics)

19 Later I will say more about what I take the difference between these to be. For now, what’s clear is that Quine does not think that there is a logical method of deciding between rival theories on the basis of the available evidence (where he assumes the primary evidence to be sensory), and whatever it is that does eventually help us decide must be something other than this logical method. ‘Pragmatic’ is for the moment a placeholder term, denoting whatever it is that allows us to make choices among underdetermined theories. ‘Logical’ justification is also something of a placeholder, referring to whatever it is with respect to which theories are underdetermined: our existing stock of deductive and inductive rules of inference, which cannot decide between empirically equivalent theories. I think that it would be misleading to call this ‘epistemic’ justification, because according to Quine, epistemology is concerned with forms of justification beyond ‘logical’ forms. So pragmatic justification would count as a form of ‘epistemic’ justification on his philosophy as well.

20 My understanding of this point has been informed by numerous conversations with Gary Ebbs.

21 It has been brought to my attention by Patrick Maher that if Quine accepts this as a serious statement about the content of scientific theories, it may have the effect of truly trivializing his statement of the underdetermination thesis. It would imply that empirically equivalent theories are also logically equivalent, and therefore certainly logically compatible. Of course, this is only if the statement is interpreted in a purely phenomenalist manner, not in the structural realist manner that I will shortly mention. As I’ve said, there’s a fine line between the two. In any case, I think that Quine did have a tendency to entertain more and more trivial versions of underdetermination as the years went by. All I can say is that the more trivial they become, the less motivated naturalized epistemology becomes. The trouble is that while Quine lost confidence in underdetermination, he kept confidence in the project of naturalizing epistemology. This is trouble because without the first, there may have been little motivation for the second.

22 Note that this does not imply that Quine has necessarily abandoned epistemology as a normative project. As I have already suggested, he still has the option of naturalizing normativity, of showing how these various theoretical developments have served some adaptive function for us. So they would count as ‘good’ beliefs from the perspective of natural selection, perhaps. Even so, they would only be normative in this new sense, not in the traditional epistemological sense of justifying our beliefs.