Apriorism About Modality: Reply to Scott Sturgeon Carrie Jenkins Draft only - please do not quote or cite without permission These comments fall into two sections. The first section is a mixed bag; in this section I shall pick up on a number of interesting issues and ideas raised in Sturgeon’s paper, making some brief remarks on each. In the second, I turn to what struck me as the main focal point of the paper, the objections (from ‘Magic’ and ‘Extra Magic’) to infallibilist apriorism about modal knowledge. I shall offer a more sustained discussion of these objections. Most importantly, I shall attempt to locate them as part of a long-standing debate over a priori knowledge. I A Few Remarks The first infallibilist apriorist view put up for discussion is the view that the following schema has no counter-instance: (p) pfac(φ) ⊃ φ In words: the prima facie a priori coherence of φ materially implies the possibility of φ. Most of the later, more sophisticated, proposals also take the form of claims of one-way material implication. But all instances of a schema like (p) would come out true provided it were the case that no φ met the condition described in the antecedent. It seems to me that no interesting form of infalliblist apriorism will amount to a view that could be rendered true by the mere fact that nothing is a priori coherent in the relevant sense (or the fact that everything is possible). The infallibilist apriorist believes in an epistemologically significant connection between a priori coherence and possibility. This is not well captured by a material conditional, since for all it says there may be no epistemologically interesting link at all between a priori coherence and possibility. At first blush, it might look as though the material biconditional version of infallibilism mentioned in the paper, which appears much later (see p. **), avoids this sort of worry. This is the claim that φ is genuinely possible iff it is conceptually possible. A scenario where nothing is conceptually possible is not one where this biconditional comes out true (at least, assuming that some things are still genuinely possible in that scenario). However, the problem also affects the use of material biconditionals to try and capture the thesis that a priori reflection can be a good epistemic guide to the modal facts (i.e. a way of acquiring knowledge of those facts). To see this, consider the following analogy. Gullible (an infinitely long-lived individual) goes to study with the Modality Guru (another infinitely longlived individual). Gullible believes that Guru will, over an infinite amount of time, utter all and only things that are possible. Gullible blindly accepts as possible everything Guru says, even though Gullible has no good reason to trust Guru. As it happens, Guru does in fact utter all and only the possible things. Now consider the schema: (g) guru(φ) ≡ φ In words: Guru’s uttering φ materially implies the possibility of φ and vice versa. Clearly, this is true in the envisaged situation. Equally clearly, asserting its truth is not a way of defending the

claim that blindly trusting Guru is a good epistemic guide to the modal facts (i.e. a way of acquiring knowledge of those facts). In the envisaged situation, blindly trusting Guru is not a good epistemic guide to the modal facts, even though that material biconditional is true. The crux of my point is that using a material biconditional to capture an epistemological thesis, as Sturgeon does, is tantamount to assuming a very crude form of reliabilism, susceptible to counterexample in the way described above. Sturgeon’s emphasis on mere alignment between conceptual possibility and genuine possibility, in his closing passages, is another symptom of the same malady. My second observation concerns the purported difference between Sturgeon’s (k) (for which see p. **) and his (o) (see p. **). (k) claims that ‘Kripke coherence’ materially implies possibility. Kripke coherence is limit coherence ‘in the light of’ Kripke truths. A Kripke truth is a true claim which is non-modal, is such that it and its negation are limit coherent, and is such that it is a priori obliged that the matter it raises is non-contingent. Something’s being limit coherent ‘in the light of’ Kripke truths is a matter of its being limit coherent with the Kripke truths. (See footnote **, where a claim ¬K’s being ruled out by a Kripke truth is taken as showing that ¬K is not Kripke coherent.) Sturgeon then argues that there are Kripke coherent claims which are not genuinely possible, and hence counterexamples to (k). His example is L: the claim that Lewisian worlds truthmake claims of genuine modality. He goes on to argue that at least one of L and ¬L is not genuinely possible (an argument which I shall comment on in a moment). He also claims that both L and ¬L are Kripke coherent. Here is his argument (which appears on p. **): ‘Take apriori reflection to the limit, throw in Kripke truth, both L and ¬L stay coherent’. Evidently, then, Sturgeon does not think that one of L and ¬L is itself a Kripke truth. But why? They both seem to be Kripke claims by Sturgeon’s lights, and since Sturgeon assumes the Law of Excluded Middle at various points during the paper, he should think that one of them is true. Why do I say they are Kripke claims by Sturgeon’s lights? Because Sturgeon should think that they meet all the conditions he has given for being a Kripke claim. Firstly, they are nonmodal. (They are about modal metaphysics, of course, but they are not claims to the effect that something is possible or necessary.1) Secondly, they are (by Sturgeon’s lights) both limit coherent (after all, he thinks they are limit coherent with the Kripke truths, which is stronger). And thirdly, Sturgeon gives an argument on pp. **-** which attempts to show, on purely a priori grounds, that ‘whichever of L and ¬L turns out to be true, it is genuinely necessary as well’. Indeed, he explicitly says that it is a priori obliged that the matter they raise is non-contingent (see p. 12, where he says that property (κ2) is had by L-style claims). Sturgeon’s further examples (on p. **) of ‘systematic misalignments between apriori 1

Clearly, the Kripke truths will include several claims which are about modal metaphysics, so Sturgeon cannot have in mind that no claim which is about modal metaphysics counts as a Kripke claim. For instance, let ‘Universe’ be a name for the actual universe around us, and let ‘World’ be a name whose referent Lewisians attempt to fix as the actual Lewisian truthmaker. Then the claim that Universe is World ought to count as a Kripke claim (at least, Lewisians will think it is one). It is very closely analogous to Sturgeon’s paradigm Kripke claim, the claim that Mark Twain is Sam Clemens.

coherence and genuine possibility’ seem to be offered as further counterexamples to (k), and further motivation for the move to (o). But again, Sturgeon does not establish that these are counterexamples to (k). Take Sturgeon’s example of (C&R) and (C&¬R), where R is such that both it and its negation are limit coherent but it is a priori obliged that the matter it raises is noncontingent, and C is some contingent claim which is independent of R and its negation. Suppose R and ¬R are Kripke claims. Then by Sturgeon’s lights one of them is a Kripke truth. If ¬R is a Kripke truth, then even if, as Sturgeon says, (C&¬R) is limit coherent, it is not Kripke coherent. It is ruled out by a Kripke truth, namely R. Similarly, if ¬R is the Kripke truth, then (C&R) is not Kripke coherent, even if it is limit coherent. To find counterexamples to (k) in this vicinity, Sturgeon would need to show that there are claims meeting the conditions on R which are not Kripke claims. The conditions on R are two of the three necessary and sufficient conditions on being a Kripke truth, so the only option here is to show that some claims meet the conditions on R but fail that final condition on Kripke claims, i.e. are modal. An obvious kind of modal claim which meets the conditions on R is a modalized version of an identity claim, such as: Possibly, Mark Twain is Sam Clemens. Let’s take this as our R, and for ease of expression, let’s use ‘P’ to stand for the unmodalized claim, Mark Twain is Sam Clemens. The trouble is that taking a claim like this as our R will not help us establish that there are counterexamples to (k) along the lines Sturgeon envisages. For then, even if (C&R) and (C&¬R) are both limit coherent, one of them (whichever one is genuinely impossible) is not Kripke coherent. For one of P and ¬P is a Kripke truth. If it is P, then ¬R is genuinely impossible and so is (C&¬R). But in this case ¬R is incoherent with (ruled out, by a priori means, by) a Kripke truth, namely P, and hence ¬R is not Kripke coherent. So (C&¬R) is not Kripke coherent either. If it is ¬P that is the Kripke truth, then R is genuinely impossible and so is (C&R). But now R is incoherent with a Kripke truth, namely ¬P, and hence R is not Kripke coherent. So (C&R) is not Kripke coherent either. So we cannot establish by these means that one of (C&R) and (C&¬R) is Kripke coherent yet genuinely impossible. The third issue I would like to turn to is the discussion of Lewisian modal metaphysics at pp. ****. My point here is much less significant than the preceding two, since it is specific to the particular example Sturgeon uses. Nevertheless, I think it is interesting enough to be worth noting. ‘L’ is the claim that Lewisian worlds truthmake claims of genuine modality. Sturgeon argues that it is impossible that: (12) L & ¬L. The reason he cites is that ‘[i]ts first conjunct entails its second is true only if there is a Lewisian truthmaker at which there is no such truthmaker’ (p. **). But in fact the first conjunct at most entails that there is a Lewisian world at which that world is not a truthmaker for any modal claims. In fact, it doesn’t even seem to entail as much as that. In order L to entail anything of the kind Sturgeon has in mind, L must be the claim that Lewisian worlds truthmake all claims of genuine modality. So one way for ¬L to be true is if Lewisian worlds truthmake some but not all claims of genuine modality. Now, for all Sturgeon has argued, at this world all modal claims are truthmade by Lewisian worlds, so L is true, but there is a world w where some but not all modal

claims are truthmade by Lewisian worlds, and where w itself does truthmake some modal claims. My fourth comment in this section concerns Sturgeon’s interpretation of Lewis at p. **. Sturgeon cites Lewis 1986, p. 4, claiming that in this passage Lewis is aligning conceptual possibility with genuine possibility. Here is the passage, as quoted by Sturgeon: [Conceptual] space is a paradise for philosophers. We have only to believe in the vast realm of possibilia, and there we find what we need to advance our endeavours. We find the wherewithal to reduce the diversity of notions we must accept as primitive, and thereby to improve the unity and economy of the theory that is our professional concern - total theory, the whole of what we take to be true. What price paradise? If we want the theoretical benefits that talk of possibilia brings, the most straightforward way to gain honest title to them is to accept such talk as the literal truth. It is my view that the price is right… The benefits are worth their ontological costs. I do not think this passage supports Sturgeon’s reading. As far as I can tell, here and in the surrounding sections, Lewis is merely arguing that the ontological costs of believing in possible worlds are justified by the theoretical benefits of such a belief. Conceptual modality in Sturgeon’s sense doesn’t seem to come into it, let alone the relationship between conceptual modality and genuine modality. Lewis himself does not use the word ‘conceptual’ at the beginning of the passage. He says ‘logical’. And Sturgeon argues earlier in his footnote 4 that Lewis means by ‘logical’ modality what he (Sturgeon) means by ‘genuine’ modality. Fifthly, and finally for this section, let me make a quick remark about Sturgeon’s closing claim (p. **) that ‘[a]priori reflection is a guide to genuine modality to the extent that genuine possibility is plentiful - to the extent, that is, to which it aligns with conceptual possibility.’ It seems that Sturgeon is assuming here that genuine modality cannot be too plentiful to align well with conceptual possibility. But nothing Sturgeon has said rules that out. And there are some examples which could be taken as showing that genuine possibility outstrips conceptual possibility. For instance, non-Euclidean geometry understood as a theory of physical space, and certain results in quantum mechanics, might with some justification be said to be conceptually incoherent and yet not only genuinely possible, but actually true. II The Objections From Magic and Extra Magic I begin this section with a few words about realist epistemology. I will use these as a framework within which to recast what I see as the principal concerns raised in Sturgeon’s paper, in what I hope is an illuminating fashion. Then I shall suggest that the deep issues in the vicinity of these worries are not, contra Sturgeon, worries about the infallibility of the conceivability method of coming to know modal truths. I shall also claim that these deep issues are old ones to which a variety of responses are already on the market. By a ‘realistically construed subject matter’, I shall mean a subject matter about which we are inclined to be mind-independence realists (in the sense of my 2005). Satisfying accounts of how we know facts about some realistically construed subject matter seem to fit a three-step

pattern. First there is worldly input of some kind into our mental processes. Then, at least typically, some kind of mental processing takes place. And finally, we arrive at a belief. Take, for example, the visual story which we tell to account for much of our knowledge of the physical world around us. First, the physical world impacts causally upon our eyes, and the eyes carry signals to the brain. This constitutes the worldly input into our mental processes. Then various sorts of mental processing of that input take place. And the upshot is that we arrive at a belief about what the physical world around us is like which is responsibly based upon the information received. Of course there is a great deal of detail to be filled in around this barebones outline, but (except in sceptical moods) most of us are confident that some story of this general kind can be told to account for visually based knowledge in the good cases. There is widespread disquiet as to how such a method as attempting to conceive of a proposition’s being true - even when this method is made more sophisticated by sensitivity to ‘red flags’ and ‘yellow flags’, and to what is a priori forced (see the ‘highly seasoned apriori infallibilism’ of p. **) - can help us understand a priori knowledge. In my opinion, the correct diagnosis of this disquiet is that knowledge to be had merely by attempting to conceive of something does not appear to fit the pattern for realist epistemology. The second two steps seem fine (or rather, although they are not fully worked out, they are at least not obviously problematic): the mental processing in this case is some sort of introspective investigation of what one can conceive of (whatever we take that to amount to), and at the end of this process a belief is arrived at. But what of the initial step, the input step? Sturgeon’s ‘Magic’ and ‘Extra Magic’ worries are both cast as worries about infallibility for an idealized conceivability method, i.e. worries for those who accept the following material conditional: IC: (Modal belief B is arrived at through the idealized conceivability method) ⊃ (Modal belief B is true). (This is intended as equivalent to Sturgeon’s (a) on p. **). Sturgeon is not similarly troubled by the corresponding claim for idealized vision, namely: IV: (Belief B is arrived at through idealized vision) ⊃ (Belief B is true). The reason Sturgeon is not troubled by IV is (if understand correctly the gist of his comments at p. **) that the kind of idealization relevant to idealized vision has partly to do with conditions on the input step: idealized vision is veridical. By contrast, the kind of idealization relevant to idealized conceivability has nothing to do with any input step; the idealization is all on the processing side.2 So how could a guarantee of truth sneak in? Realism about the subject matter makes this sort of guarantee bewildering. Epistemology and realist metaphysics, as Sturgeon stresses on p. **, just don’t fit together like that. While I agree with Sturgeon that there is an important issue in this vicinity that is in need of philosophical attention, I think his emphasis on infallibility is a red herring. My point here will be an analogue of the point one often makes to students who think that Hume’s worries about induction may show that we aren’t guaranteed to get true beliefs by inductive methods, but they do not undermine the thought that such beliefs are likely to be true. 2

Sturgeon calls this kind of idealization ‘epistemic’, but that may not be the best name. ‘Epistemic’ means having to do with knowledge, and knowledge, being factive, is world-involving in the way Sturgeon wants to avoid here.

For exactly the same reasons Sturgeon thinks we should be worried about IC, namely because epistemology and realist metaphysics just don’t fit together like that, we should also be worried about the claim that it’s even likely that the application of a conceivability method idealized only along processing dimensions should lead to a true belief. Epistemology and realist metaphysics don’t fit together like that either: absent further explanation, you don’t get reliable ways of finding out about the independent world just by looking inside your head. It’s not just that you don’t get infallible ways; you don’t even get mostly-successful ways. So infallibility is not needed to generate the kind of worry Sturgeon is interested in (as, indeed, Sturgeon himself seems to acknowledge on p. **, when he diagnoses the plausibility of (!) as due to the same ‘ep-and-met tendency’ as he thinks motivates infallibilism). In fact, it’s not sufficient either. If we could include a satisfactory input step into our conceivability method, and then include idealization of that step as part of idealization of the conceivability method, then Sturgeon’s worries will be dissolved. For this would set up an analogy with the vision case, where (by Sturgeon’s lights) there is no problem. I recommend that we leave infallibility to one side and focus on the real issue in this area. If we are mind-independence realists about modality, we expect a realist-type epistemology for modal truths. But the conceivability method as normally construed, even when idealized, involves no input step, and hence it cannot fill this role. Now this problem is a very good one, but it’s an instance of a very general problem, namely that of saying how we can have a priori knowledge – which seemingly requires no input from the mind-independent world – of facts which belong to some mind-independent domain. And this is also a very old problem, to which there are many purported answers available already. For instance, rationalists might claim that rational intuition somehow provides input into the conceivability method (see e.g. Bealer 2000, BonJour 1998). Others would argue that the lack of any plausible input step is a reason to reject mind-independence realism about modal truth; they may adopt instead some form of Ayerian or Carnapian anti-realism (see e.g. Ayer 1936, Carnap 1950). Still others would say that the conceivability method cannot be our method of coming to know the (mind-independent) modal truths, since it involves no input step. Rather, they will argue that we must be using ordinary empirical methods (see e.g. Mill 1843, Quine 1951). I have a horse in this race myself: I believe that the conceivability method involves an empirical input step. I believe that empirical input renders our concepts good guides to the world, so that we can recover information about the world from those concepts, through the practice of attempting to conceive of various scenarios. (See e.g. Jenkins forthcoming.) If Sturgeon wants to press his worry in a new way, he owes us an account of what is wrong with the various extant positions which exist in response to it. Otherwise he is, when it comes down to it, simply reminding us that a priori knowledge of realistically-construed subject matters is a philosophical puzzle.

References Ayer, A. J. 1936. Language Truth and Logic. Second edition, impression of 1960. London: Gollancz. Bealer, G. 2000. ‘A Theory of the A Priori’, in Pacific Philosophical Quarterly 81, pp. 1-30. BonJour, L. 1998. In Defence of Pure Reason: a rationalist account of a priori justification. Cambridge University Press. Carnap, R. 1950. ‘Empiricism, Semantics and Ontology’, in Revue Internationale de Philosophie 4, repr. in H. Feigl, W. Sellars and K. Lehrer (ed.s) New Readings in Philosophical Analysis, 1972, New York: Appleton-Century-Crofts, pp. 585-96. Jenkins, C. 2005. ‘Realism and Independence’, in American Philosophical Quarterly 42, 3. ------------ forthcoming. Grounding Concepts. Oxford University Press. Lewis, D. 1986. On The Plurality of Worlds. Blackwell Press. Mill, J. S. 1843. A System of Logic. London: Parker. Quine, W.V.O. 1951. ‘Two Dogmas of Empiricism’ in The Philosophical Review 60, repr. in his From a Logical Point of View: Nine Logico-Philosophical Essays, 1953, edn. of 1980, Cambridge, MA: Harvard University Press, pp. 20-46.