GORDON BARNES
CONCEIVABILITY, EXPLANATION, AND DEFEAT (Received in revised form 31 October 2001) ABSTRACT. Christopher Hill and Joseph Levine have argued that the conceivabilities involved in anti-materialist arguments are defeated as evidence of possibility. Their strategy assumes the following principle: the conceivability of a state of affairs S constitutes evidence for the possibility of S only if the possibility of S is the best explanation of the conceivability of S. So if there is a better explanation of the conceivability of S than its possibility, then the conceivability of S is thereby defeated as evidence of possibility. Hill and Levine proceed to offer alternative explanations of these conceivabilities, concluding that these conceivabilities are thereby defeated as evidence. However, this strategy fails because their explanations generalize to all conceivability judgments concerning phenomenal states. Consequently, one could defend absolutely any theory of phenomenal states against conceivability arguments in just this way. This result conflicts with too many of our common sense beliefs about the evidential value of conceivability with respect to phenomenal states. The general moral is that the application of such principles of explanatory defeat is neither simple nor straightforward.
Ever since Descartes, judgments of conceivability have formed the basis of much resistance to materialism. In the 1970’s Saul Kripke argued that the conceivability of pain without its neural correlate is good reason to reject the identity theory.1 More recently, David Chalmers has argued that the conceivability of Zombies is good reason to reject physicalism.2 Of course, objections to these arguments abound. Some philosophers deny that the separation of the mental and the physical is conceivable. Others claim that conceivability is worthless as a guide to possibility. However, some philosophers accept both the conceivability of Cartesian scenarios,
This paper is an expanded version of a paper that was read at the 2001 meetings of the Central Division of the American Philosophical Association. I thank Joseph Levine for his helpful comments on that occasion. I also want to thank Marnie Barnes, David Clemenson, Matt Davidson, John Kronen, Sandy Menssen, and Tom Sullivan for many helpful conversations on conceivability and possibility. Philosophical Studies 108: 327–338, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
328
GORDON BARNES
and the general worth of conceivability as a guide to possibility, yet reject the conclusions of these Cartesian arguments. The strategy of these critics is to explain away the Cartesian conceivabilities in such a way as to defeat them as evidence of possibility. In two recent papers, Christopher Hill3 and Joseph Levine4 have taken this approach to conceivability arguments against materialism. Broadly speaking, Hill and Levine share a common strategy. They try to explain away these Cartesian conceivabilities in such a way as to defeat them as evidence of possibility. In this paper I will argue that both versions of this common strategy fail. More generally, however, I want to use these proposals to focus discussion of a general principle in the epistemology of modality. So before I turn to the details of these proposals, I will sketch this principle, and the problem that it poses. Both of the proposals that I will discuss tacitly assume something like the following principle. The conceivability of some state of affairs S is evidence for the possibility of S only if the possibility of S is the best explanation of the conceivability of S. Contrapositively, that is to say that if there is a better explanation of the conceivability of S than its possibility, then the conceivability of S is no evidence for the possibility of S.5 Henceforth, I will refer to this principle as a Principle of Explanatory Defeat (PED). Generally speaking, a (PED) is a principle that says that the putative evidence for a belief is defeated if and when there is a better explanation of the putative evidence than the truth of the belief it allegedly supports. In other words, if there is a better explanation of the putative evidence than the truth of the allegedly supported belief, then the putative evidence is merely putative evidence, not genuine evidence. In this paper I will not challenge any version of the PED. Rather, what I aim to show is that the application of such principles is much less straightforward than one might think. To think otherwise is to overlook a certain necessary condition for being the best explanation of some phenomenon. The condition is simply this: the best explanation of some phenomenon must not undercut too many beliefs that are plausible relative to our overall set of beliefs about the world. If an explanation conflicts with too many beliefs that are plausible relative to our overall view of the world, then no matter how much it simplifies and unifies our understanding of the world, it is not a
CONCEIVABILITY, EXPLANATION, AND DEFEAT
329
good explanation, and so a fortiori it is not the best explanation. At any rate, if there is an explanation of the same phenomenon that preserves many more beliefs that are plausible relative to our overall view of the world, then that explanation is a better explanation. How is this criterion of explanatory value relevant in this context? Sometimes an explanation of a particular phenomenon generalizes, which is to say that insofar as it succeeds in explaining that particular phenomenon, the very same sort of explanation will equally well explain all phenomena of the same general type. Second, our beliefs about the world include some epistemic beliefs, beliefs about what we know, and how we know it, beliefs about what we are justified in believing, and how we are justified in believing it. For instance, most of us believe that we know that there is an external world, that we know many things about it, and that we know this on the basis of our sense experiences. Now, conjoin these two observations with the aforementioned principles (PED’s), and one can envisage the following possibility: an explanation of the putative basis for some belief generalizes in such a way as to explain all similar bases for all similar beliefs, and, in conjunction with the corresponding (PED), it thereby undercuts the justification for all similar beliefs. Thus, for example, imagine an explanation of a particular sense experience that generalizes in such a way as to explain all sense experiences equally well. Furthermore, imagine that this explanation of our sense experience does not require the existence of an external world. Then, in conjunction with the corresponding (PED), this explanation will thereby undercut all of our epistemic beliefs about what we know about the external world on the basis of sense experience. Certainly this explanation would thereby lose much of its explanatory value. That is not to say that skepticism is ruled out of court, but only that it is not likely to be justified by its explanatory value.6 The same point applies to our epistemic beliefs concerning modalities. Most of us believe that we know, or are justified in believing, that some things could have been other than the way they are. Moreover, the putative basis of much of our modal knowledge is simply the conceivability of things being other than the way they are. So any explanation of a conceivability judgment that generalizes in such a way that it undercuts many of these epistemic beliefs
330
GORDON BARNES
would thereby lose much of its explanatory value. In short, when we consider an explanation of a conceivability judgment, we should pay attention to its broader epistemic implications. If an explanatory defeat generalizes in such a way that it ultimately undercuts many of our epistemic beliefs about our knowledge of modalities, then it thereby loses much of its explanatory value. So far, this is just a rough sketch of what I will illustrate with the following examples. Before I turn to Hill and Levine, I want to illustrate this point with an historical example. Recall the seventeenth-century debate over Descartes’ conceivability argument. In that debate, Antoine Arnauld famously objected that Descartes’ ability to conceive of himself without any corporeal features shows only that if Descartes’ conception of himself is adequate, then no corporeal features are part of his essence. Thus, Arnauld insisted that Descartes first demonstrate that his idea of himself is adequate – containing all features that are, in fact, part of his essence. Otherwise, Descartes’ ability to conceive of himself without any corporeal features might simply be the result of an inadequate idea of himself. Perhaps Descartes is simply ignorant of the fact that corporeal features are part of his essence. In summary, then, you might say that Arnauld proposed an alternative explanation of Descartes’ ability to conceive of himself without corporeal features. Arnauld’s alternative explanation is that Descartes has an inadequate idea of himself. Arnauld’s challenge to Descartes is to rule out this alternative explanation of his conceivability. Arnauld implies that if Descartes cannot rule this out, then he cannot take the conceivability in question as evidence of a real possibility. As Stephen Yablo has pointed out, “there is a certain irony in Arnauld’s position.”7 To see the irony, we must pay attention to another dispute involving Arnauld. In his correspondence with Leibniz, Arnauld rejected Leibniz’s claim that all of a thing’s properties are essential to it. Here is the crux of Arnauld’s argument against Leibniz. “I find in myself the concept of an individual nature, since I find there the concept of myself. I have only to consult it, therefore, to know what is contained in this individual concept . . . . I can think that I shall or shall not take a particular journey, while remaining very much assured that neither one nor the other will prevent my being myself.” In his rejection of Leibniz’s
CONCEIVABILITY, EXPLANATION, AND DEFEAT
331
superessentialism, Arnauld claims that he can simply consult his own conception of himself and thereby see that taking a particular journey is something that he might or might not do, while remaining one and the same person. What has happened to Arnauld’s requirement that one first certify the adequacy of one’s ideas of things before determining which properties are essential to it? How can Arnauld be so sure that his ability to conceive of himself without taking a particular journey is not the result of an inadequate idea of himself? Perhaps Arnauld is simply ignorant of some of his essential properties. The irony is palpable. We should pay close attention to the path that led Arnauld to this inconsistency. Arnauld’s explanation of Descartes’ ability to conceive of himself as incorporeal can be generalized, such that, together with the (PED), it ultimately undercuts many of Arnauld’s own epistemic beliefs about what he can know to be possible on the basis of conceivability. So what started out as a harmless word of caution threatens to degenerate into a sweeping modal skepticism about the self. Perhaps this should lead Arnauld to rethink his deflationary explanation of Descartes’ conceivability. If that deflationary explanation generalizes in such a way as to undercut many of Arnauld’s beliefs about his own modal knowledge, then maybe it is not such a good explanation after all. In the remainder of this paper I will argue that the same basic problem besets two recent attempts to explain away Cartesian conceivabilities. That is to say, these explanations generalize, and so, in conjunction with the (PED), they undercut many of our epistemic beliefs about our modal knowledge. Thus, I will be arguing that these explanations are not very good explanations at all. More importantly, I hope that these examples will illustrate the difficulties inherent in such naive applications of the (PED). In his discussion of conceivability arguments, Christopher Hill introduces the quasi-technical notion of a “sensory concept.” According to Hill, sensory concepts have two features that distinguish them from all other concepts. First, trivially, if X is a sensory concept, then X denotes a type of sensory state. And second, if X is a sensory concept, and S is the sensory state that is denoted by X, then, in an important range of contexts in which one is justified in ascribing X to an instance of S (specifically, the contexts in which one is led by introspection to ascribe X to one of one’s own sensory states), the sense experience that provides
332
GORDON BARNES
one’s justification for the ascription is identical with the instance of S to which X is applied.8
According to Hill, it is constitutive of the concept of pain that when it is self-ascribed, one is justified in self-ascribing it in virtue of the fact that one is currently in pain. Hill’s main point is that the justification-conditions for the self-ascription of pain are very peculiar in this respect. Unlike any other sort of concept, a sensory concept is such that one is typically justified in self-ascribing it just in virtue of the fact that one is actually in a state to which it applies. No other sort of concept is like this. Certainly no neuroscientific concept is like this. Typically, what justifies one in ascribing a neuroscientific concept is that (1) one has had the sorts of experiences that tend to justify theories that involve the neuroscientific concept, and (2) one has had experiences as of an apparatus which indicates the applicability of the neuroscientific concept.9 These justificationconditions are very different from the justification-conditions for the self-ascription of a sensory concept. The justification-conditions for the self-ascription of a sensory concept involve very different experiences from those involved in the justification-conditions for the ascription of a neuroscientific concept. Armed with these observations, Hill proceeds to explain the conceivability of pain without its neural correlate, in such a way as to defeat it as evidence of a real possibility. Here is the explanation. Given these important differences between the justificationconditions for the self-ascription of a sensory concept, and the justification-conditions for the ascription of neuroscientific concepts, there are no substantive a priori ties between sensory concepts and neuroscientific concepts. Since there are no substantive a priori ties between sensory concepts and neuroscientific concepts, “it is possible to conjoin either of the concepts with the negation of the other without producing an inconsistency.” Now, if someone can conjoin one concept with the negation of the other, without thereby producing an inconsistency, then unless that person has some a posteriori reason for thinking that the conjunction is impossible, he will believe that the conjunction is possible. Of course, we do not normally have an a posteriori reason to think that pain without its neural correlate is impossible. Therefore, we normally infer, in just this way, that it is possible.10 Finally, Hill contends that this explana-
CONCEIVABILITY, EXPLANATION, AND DEFEAT
333
tion of the conceivability of pain without its neural correlate defeats it as evidence of a real possibility. This is because the mechanism described in this process of reasoning would, under normal circumstances, yield the conceivability of numerous impossibilities. This process of reasoning would yield the conceivability of water without H2 O, gold without the atomic number 79, etc. Consequently, we can see that this process is generally unreliable.11 The first thing to notice about this line of reasoning is that it turns on nothing more than a unique feature of sensory concepts, namely, that the justification-conditions for their self-ascription typically coincide with the truth-conditions for their self-ascription. What typically justifies one in self-ascribing pain is the very state that makes the ascription true. This feature of sensory concepts distinguishes them from neuroscientific concepts. However, the second thing to notice is that this feature is unique to sensory concepts. So this feature distinguishes sensory concepts not only from neuroscientific concepts, but also from any non-sensory concept whatsoever. Thus, Hill’s explanation of Cartesian conceivabilities can be generalized into an explanation of the conceivability of pain without , where the blank is filled in with any non-sensory concept whatsoever. Finally, given the Principle of Explanatory Defeat, Hill’s alleged defeater of the Cartesian conceiovabilities generalizes into a defeater for the conceivability of pain without , no matter how we fill in the blank. In sum, if Hill’s explanation of the Cartesian conceivabilities is correct, then all conceivabilities concerning sensory states are thereby defeated as evidence of real possibilities. In case this is not yet clear, I will use some examples to illustrate the point. Many philosophers believe that inverted qualia are possible. That is to say, it is possible that the distal stimuli that cause you to have sensations as of red objects, cause me to have sensations as of green objects, and vice versa. Why believe that this is possible? The standard argument for this possibility, at least since Locke, is implicitly a conceivability argument. We can conceive of this scenario, and on that basis we judge that it is possible. Finally, many of us share the epistemic belief that we are thereby justified in believing that this is possible. So the standard justification for belief in this possibility is the conceivability of sensations being caused by
334
GORDON BARNES
different distal stimuli. However, if Hill’s explanation of Cartesian conceivabilities is correct, and if the success of that explanation undercuts the evidential value of any conceivability judgment to which it applies, then this standard justification of belief in the possibility of inverted qualia is also defeated. That is because a Hill-type explanation applies in this case as well. Here is how it goes. Sensory concepts are unique insofar as the justification-conditions for their selfascription typically coincide with the truth-conditions for their self-ascription. No other concept is like this. Certainly the concepts of various distal stimuli are not like this. Consequently, there are no substantive a priori ties between sensory concepts and the concepts of various distal stimuli. When there are no substantive a priori ties between two concepts, then unless we have a posteriori reason to think that their separation is impossible, we infer that it is possible. So we infer, in just this way, that sensory states could be caused by different distal stimuli. However, this mechanism is generally unreliable, for it would produce belief in the possibility of water without H2 O, and heat without mean kinetic molecular energy, etc. So the conceivability of inverted qualia is actually no evidence of the possibility of inverted qualia.
This example illustrates how Hill’s defeater generalizes into a defeater for any conceivability judgment concerning sensory states. If Hill’s explanation of the Cartesian conceivabilities is correct, then every conceivability judgment concerning sensory states is thereby defeated as evidence of a real possibility. But surely that defeats too much. Surely at least some of our beliefs about what is possible concerning sensory states are justified by what is conceivable concerning them. If so, then Hill’s defeat-cum-explanation fails because it undercuts too many of our beliefs about what we know to be possible on the basis of what is conceivable. In order to embrace Hill’s alleged defeater of Cartesian conceivabilities, one must relinquish all conceivability arguments concerning sensory states. I believe that this is more modal skepticism than many materialists can consistently tolerate. In my discussion of Joseph Levine’s proposal, I will press this point further. I now turn to Joseph Levine’s attempt to defeat the Cartesian conceivabilities.12 First, I must say a word about how Levine understands conceivability and possibility. Levine contends that “A situation S is metaphysically possible just in case it has no accurate representation that is logically inconsistent.” He goes on to say that “S is conceivable for X just in case there is at least one formally
CONCEIVABILITY, EXPLANATION, AND DEFEAT
335
consistent representation of S by which X represents S, and X is unaware of there being any formally inconsistent representation of S.” In summary, then, “our cognitive access to modal facts is primarily a matter of our sensitivity to the logical forms of the representations by which we conceiva them.” As long as we have at least one formally consistent representation of a situation S, and we know of no formally inconsistent representation of that same situation S, S is conceivable for us. Moreover, this is supposed to be some reason to think that there is no formally inconsistent representation of S. After all, “What better evidence could I have that there is no formally inconsistent description of a situation than the fact that I don’t know of one?”13 Levine’s view is that conceivability consists in our sensitivity to the logical forms of our representations. This suggests some interesting ways in which conceivability could fail to indicate real possibility. For example, perhaps the logical form of a certain representation of a situation somehow prevents us from thinking of a formally inconsistent representation of the same situation, even though there is one. This is exactly how Levine diagnoses the conceivability of the mental without the physical, and vice versa. At the heart of Levine’s strategy is a distinction between two modes of presentation: ascriptive and non-ascriptive. A mode of presentation is simply “the means by which a representation connects to its referent.” An ascriptive mode of presentation “involves the ascription of properties to the referent, and it’s (at least partly) by virtue of its instantiation of these properties that the object (or property) is the referent. A non-ascriptive mode is one that reaches its target, establishes a referential relation, by some other method.”14 Ascriptive modes of presentation necessarily involve the satisfaction (by the referent) of a set of conditions that are explicitly represented in the mode of presentation, whereas non-ascriptive modes of presentation achieve representation via other sorts of relations between the representation and the referent, without any explicit grasp of a set of conditions that must be satisfied by the referent. With this distinction, Levine explains how we might conceive of the mental without the physical, and vice versa, despite its impossibility. Suppose that our representations of some mental
336
GORDON BARNES
properties, such as qualia, are in the non-ascriptive mode. That is to say, suppose that we represent our own qualia to ourselves without grasping the necessary and sufficient conditions for being the referent of our qualia-representations. Then it is no wonder that we will be unable to think of a formally inconsistent representation of a situation involving qualia, even if there is one. Our inability to think of such an inconsistent representation is not due to any real possibility, but only to the fact that our representations of qualia do not wear their satisfaction-conditions on their sleeves. Levine concludes that the possibility of non-ascriptive representations of qualia provides an explanation of Cartesian conceivabilities that defeats it as evidence of a real possibility. As with Hill, the underlying assumption here is that if the best explanation of a conceivability judgment does not require a corresponding real possibility, then that conceivability judgment is defeated as evidence of a real possibility. But the trouble with Levine’s alleged defeater is essentially the same as the trouble with Hill’s. It undercuts too many of our beliefs about what we can know on the basis of conceivability judgments. First, notice that Levine’s explanation of Cartesian conceivabilities trades on nothing more than the alleged non-ascriptiveness of our representations of qualia. Second, this explanation can be applied equally well to any conceivability judgment involving a representation of qualia and some ascriptive mode of presentation. So a Levine-type explanation can be constructed for the conceivability , no matter how we fill in the blank. of qualia without And if this explanation really undercuts the evidential value of this conceivability judgment, then we will have a defeater for the , no matter how we conceivability of qualia without fill in the blank. So once again, we can accept this defeat-cumexplanation of Cartesian conceivabilities only if we are willing to relinquish all conceivability arguments concerning phenomenal states. Surely that would be to relinquish too much. It would undercut too many of our beliefs about what we can know on the basis of conceivability judgments. For example, many of the internecine arguments between various sorts of materialists are actually conceivability arguments. Against teleofunctionalist accounts of qualia, it is argued that a Swampman,
CONCEIVABILITY, EXPLANATION, AND DEFEAT
337
with no evolutionary history, could have qualia. Against common sense functionalists, it is argued that the functional role of qualia could be performed without any qualia at all. Against identity theorists, it is argued that creatures made of silicon could have qualia. All of these arguments tacitly rest on conceivability judgments concerning qualia. However, if we accept either Hill’s or Levine’s explanatory defeater of Cartesian conceivabilities, then we must reject all of these arguments as well. Again, that is because the HillLevine explanations of Cartesian conceivabilities apply equally well to any conceivability judgment concerning sensory states. So if their explanations undercut the evidential value of Cartesian conceivabilities, then they undercut the evidential value all conceivabilities concerning qualia in just the same way. But surely that is to undercut too much. Thus, I conclude that neither Hill nor Levine has actually defeated Cartesian conceivability judgments as evidence against materialism. Many philosophers have suggested or embraced some version of the PED. They have suggested that conceivability is defeated as evidence of possibility if there is a better explanation of the conceivability judgment that does not entail any real possibility. I hope to have shown that this principle, even if true, is not easily applied. The reason is that explanations often generalize, and so explanatory defeat generalizes as well. But if a defeat-cumexplanation ultimately undercuts too many of our beliefs, then it thereby loses much of its explanatory value. NOTES 1
Saul A. Kripke, Naming and Necessity (Cambridge, MA: Harvard University Press, 1980). 2 David Chalmers, The Conscious Mind (Oxford: Oxford University Press, 1996). 3 Christopher S. Hill, “Imaginability, Conceivability, Possibility, and the MindBody Problem” Philosophical Studies 87 (1997), 61–85. 4 Joseph Levine, “Conceivability and the Metaphysics of Mind” Nous 32 (1998), 449–480. 5 This principle could be made more precise in a variety of ways, yielding several distinct, but related principles. However, what I have to say here will apply to any of these variations on the principle. 6 Even if it is possible that a skeptical hypothesis be justified by its explanatory
338
GORDON BARNES
value, it seems that this could happen only if that hypothesis simplified and unified our total set of beliefs enough to offset the degree to which it undercuts, rather than explains, many of our epistemic beliefs. And the more radical the skeptical hypothesis, the less likely this becomes. For instance, consider the hypothesis that all of our experiences are directly caused by a Cartesian evil demon. This hypothesis would certainly unify and simplify our understanding of the world to a great degree, but no one would be tempted to accept it on that basis. Why not? Arguably, the reason is that the hypothesis undercuts, rather than explains, too many of our epistemic beliefs. 7 Stephen Yablo, “The Real Distinction Between Mind and Body” Canadian Journal of Philosophy Supplementary Volume 16 (1990), 149–201. 8 Hill, op. cit., p. 73. 9 Hill, op. cit., p. 74. 10 Hill, op. cit., pp. 75–76. 11 Hill, op. cit., pp. 76–78. 12 Levine, op. cit. 13 Levine, op. cit., pp. 450–451. 14 Levine, op. cit., p. 457.
REFERENCES Chalmers, D. (1996): The Conscious Mind, Oxford: Oxford University Press. Hill, C. (1997): ‘Imaginability, Conceivability, Possibility, and the Mind-body Problem’, Philosophical Studies 87, 61–85. Kripke, S. (1980): Naming and Necessity, Cambridge: Harvard University Press. Levine, J. (1998): ‘Conceivability and the Metaphysics of Mind’, Nous 32, 449– 480. Yablo, S. (1990): ‘The Real Distinction Between Mind and Body’, Canadian Journal of Philosophy Supplementary 16, 149–201.
Department of Philosophy University of St. Thomas St. Paul, MN 55108, USA E-mail:
[email protected]