From Towards a Science of Consciousness 3         Section 1: The Expalanatory Gap       CogNet Proceedings

Conceivability, Identity, and the Explanatory Gap

Joseph Levine

Materialism in the philosophy of mind is the thesis that the ultimate nature of the mind is physical; there is no sharp discontinuity in nature between the mental and the nonmental. Antimaterialists assert that, on the contrary, mental phenomena are different in kind from physical phenomena. Among the weapons in the arsenal of antimaterialists, one of the most potent has been the conceivability argument. When I conceive of the mental, it seems utterly unlike the physical. Antimaterialists insist that from this intuitive difference we can infer a genuine metaphysical difference. Materialists retort that the nature of reality, including the ultimate natures of its constituents, is a matter for discovery; an objective fact that cannot be discerned a priori.

The antimaterialist conceivability argument traces back (at least) to Descartes's famous demonstration of the distinction between mind and body.1 Descartes argued that since he can coherently conceive of a situation in which his mind exists but his body does not, there must in reality be a genuine, metaphysical distinction between the two. Of course one can justifiably take issue with Descartes's claim that he really can coherently conceive of himself as a disembodied mind. But the most common materialist response, as mentioned above, is to challenge his inference from what's conceivable to what's possible. Why think that what's possible, a metaphysical, mind-independent fact, should necessarily coincide with what's conceivable, an epistemic, mind-dependent fact?

While I think this materialist response is right in the end,2 it does not suffice to put the mind-body problem to rest. Even if conceivability considerations do not establish that the mind is in fact distinct from the body, or that mental properties are metaphysically irreducible to physical properties, still they do demonstrate that we lack an explanation of the mental in terms of the physical. The idea is this. It seems conceivable that there could be a creature physically like us (or functionally like us) and yet for which there is nothing it is like to be this creature; or, for whom sensory states are very different from what they are like for us. If we really understood what it is about our physical, or functional structure that is responsible for our sensory states being like what they're like (or being like anything at all), then it would no longer be conceivable that such a creature could exist. Thus what the conceivability argument demonstrates is the existence of an explanatory gap between the mental and the physical.3

In this chapter I want to consider an objection to the explanatory gap argument. According to the objection, there is no explanatory gap because the phenomenon that is allegedly lacking an explanation is not really a proper candidate for explanation. In order to present the objection, I will first need to present the original explanatory gap argument in a little detail. After presenting the objection, I'll argue that it doesn't succeed in removing the problem. Finally, I'll briefly explore the implications of my reply to the objection for the metaphysical question concerning the actual identity of mental properties with physical (or functional) properties.

Water vs. Qualia

One way to appreciate the explanatory gap is to contrast the case of explaining the existence of various sensory qualia (e.g., the way pain feels, or the way red things look) in terms of underlying physical (or computational) processes with the case of explaining other macro phenomena in terms of underlying microphysical processes. So, for instance, let's compare how we explain the boiling point of water at sea level with how we might explain the reddish character of certain visual sensations. Consider then, the following two explanation sketches:

ESI: Boiling Point of Water
(1) H2O molecules exert vapor pressure P at kinetic energy E
(2) At sea level exerting vapor pressure P causes molecules to rapidly escape into air
(3) Rapidly escaping into air is boiling
(4) 212o F. is kinetic energy E
(5) Water is H2O

(6) Water boils at 212o F. at sea level

ESII: Presence of Reddish Qualia
(7) S occupies brain state B
(8) Occupying brain state B is to experience a reddish quale (state R)

(9) S is experiencing a reddish quale

Notice that I have presented both explanation sketches as arguments, where the explanans (i.e., the statements that do the explaining) function as premises from which the explanandum (i.e., the statement describing what is to be explained) is deductively derived. This is no doubt a tendentious characterization of scientific explanation, but I don't believe that the crucial issues at stake here really depend on any of the tendentious features.4 Also, I call these "explanation sketches" because they clearly do not contain all the relevant information.

At first blush, there is nothing significant to distinguish the two explanation sketches. (Of course there's more detail in the first one, but that's an artifact of the example.) In both cases we have deductively valid arguments. Also, in both cases there is crucial use of a "bridge" premise ((5) in ESI and (8) in ESII); that is, a premise that identifies the phenomenon to be explained with some phenomenon describable in the relevant micro-vocabulary. So what is it about ESII that leaves us with the feeling that an explanatory gap remains, but not with ESI? The answer lies in the nature of the bridge premises.

There are actually two accounts I want to offer of the difference between the two bridge premises used in the explanation sketches above. The first is one that I will eventually reject, but it has initial plausibility, adherents among contemporary philosophers,5 and it sets us up for the objection I want to consider. The second one constitutes my reply to the objection.

On the first account, what distinguishes (5) from (8) is that (5) is itself derivable from a combination of statements that are either analytic (i.e., can be known a priori, purely on the basis of competency with the relevant concepts) or descriptions of underlying microphysical phenomena. No such derivation of (8) is possible. Thus the difference between the two cases is this. In the case of water, the crucial bridge premise is itself susceptible of explanation, whereas this is not the case with qualia.

Let me elaborate. On this account, statement (5) can be derived in something like the following manner:

(i) Water is the stuff that manifests the "watery" properties
(ii) H2O manifests the "watery" properties

(5) Water is H2O

By the "watery properties" I mean whatever superficial properties they are by which we normally identify water (e.g., liquidity at room temperature, being located in lakes and oceans, falling from the sky, etc.). There are two crucial features of premise (I) that are responsible for this derivation constituting an adequate explanation of (5): it is analytic, and "watery" is ultimately characterizable in "topic-neutral" terms.

To say that it is analytic, for these purposes, is just to say that one knows it's true purely by knowing what the relevant terms mean; or purely by having the relevant concepts. So it's supposed to be entailed by what we mean by "water" that it manifests the watery properties. A "topic-neutral" expression, for these purposes, is one that does not contain any nonlogical vocabulary that is not already included in the vocabulary of the theory that's doing the explaining. Thus terms like liquid, lake" etc., that might reasonably be expected to appear in any full-fledged characterization of "watery," must themselves be definable in topic-neutral terms (or, more likely, in terms that are themselves definable in topic-neutral terms). The point is that the chain of analytic definitions must bottom out in topic-neutral vocabulary.

Both of these features-being analytic and (ultimately) topic-neutral-are crucial to the explanatory adequacy of the derivation of (5), and, thereby, to the explanatory adequacy of ESI, which depends essentially on (5). Premise (i) serves as a bridge premise for the explanation of (5), and we clearly won't have made any progress in removing an explanatory gap if it stands as much in need of explanation as (5) itself. If, however, it is knowable a priori, because it expresses an analytic truth, then there really isn't any question of explaining it after all. It's true by definition. We couldn't claim analytic status for (5) itself, but so long as it rests on an analytic truth, together with statements describing various micro-phenomena, it is fully explicable. That "watery" is (ultimately) topic-neutral is required for a similar reason. For suppose it weren't; that is, suppose one of the constituent terms in the expansion of "watery" were not definable in topic-neutral terms. If so, then when we turn to premise (ii) we can ask what explains it, and the answer can't be given in terms exclusively of micro-physical processes since (ii) contains at least one unreduced term. So we'll need another bridge premise, and we can then ask what explains it. Thus both requirements-that (i) be analytic and that "watery" be definable in topic-neutral terms-are necessary to the explanatory adequacy of ESI.

Now, suppose we tried to construct an explanatory derivation of (8), along the lines of the one we constructed for (5). It would probably look something like this:

(iii) Qualitative state R is the state that plays causal role C
(iv) Brain state B plays causal role C

(8) Brain state B is qualitative state R

The problem is that (iii) isn't analytic. While it may be true that experiences with a certain reddish qualitative character tend to play a certain causal, or functional role, it doesn't seem to be a conceptual truth that they do. What justifies this claim? Here is where the conceivability argument comes in. It just seems conceivable that one could have a conscious experience of a certain sort without its playing the typical causal role for that state. Perhaps this isn't genuinely, metaphysically possible. Still, the fact that it's coherently conceivable (a premise that the materialist we're interested in grants) shows at least that the claim that this experience plays this causal role isn't analytic; and that's all the explanatory gap argument needs. So (8), unlike (5), still stands in need of explanation. This is why ESI doesn't leave a gap, whereas ESII does.

The Objection

The objection I wish to consider involves a two-pronged attack on the argument. First, the objector takes issue with the claim that statements like (i) above are analytic, or that they can be rendered in topic-neutral terms. Second, the objector responds to the obvious question that arises as a result of the first prong of the attack: namely, if (i) is neither analytic nor topic-neutral, then why is there no explanatory gap in ESI, given its reliance on the bridge premise (5)? The answer is to challenge the assumption on which the question is based, that (5) itself requires an explanation.

The denial of analytic status to (i) is part of a general challenge to the analytic-synthetic distinction, which dates back to Quine (1953). For one thing, no one has ever produced a convincing example of a conceptual analysis, aside from marginal cases like "bachelor" or mathematical concepts. After all, what would really go into the expansion of "watery"? Is it really analytic that water falls from the sky, or is liquid at room temperature? Of course, one can always say that we just haven't found the right analysis, or that it is best understood as a cluster concept with no one necessary condition but a weighted sum of necessary conditions. At this point the burden shifts to the advocate of analyticity to show why we should believe there is such a thing as the right analysis to be found.

There are a lot of moves and counter-moves to make at this point, and I can't delve into them here.6 But surely one reason for thinking there has to be an analysis for terms like "water" is that without one we would be at a loss to explain identities like (5). This brings us to the second prong of the attack, removing one of the principle reasons for believing in the necessity of analysis. The argument here is that identities, unlike correlations, do not require explanation. That something is itself is precisely the sort of situation that we accept as a brute fact. What else could it be? What really would count as explaining an object's identity with itself?

Of course it does seem as if we often ask for explanations of identities. Doesn't it make sense to ask why, or how it is, that water is H2O? But whenever such a question makes sense, it is possible to reinterpret it in one of two ways: either as a justificatory question, or as a question about the coinstantiation of distinct properties. So, with respect to asking for an explanation of (5), we might be asking not why water is H2O, but rather why we should think that water is H2O. This is a way of seeking evidence for its truth, not an explanation of its truth.

On the other hand, we might be asking something like this: How is it that this substance made out of H2O molecules appears continuously divisible? Here we are asking for an explanation, but what we want explained is how two distinct properties-in this case, being composed of H2O molecules and appearing continuously divisible-could be instantiated in the same substance. This is a quite proper object of explanation, but notice that it involves a connection between two distinct properties. What, goes the argument, you never have is, strictly, why is this the same as that? To this question, if it were ever asked, the only possible answer is, "because it is, that's why."

If the foregoing is right, and I think it is, then the reason we don't find a gap in ESI has nothing to do with the availability of an analysis of "water," and so nothing to do with our ability to explain the bridge identity (5). There is no gap because bridge identities don't generally require explanations. But if there is no gap in ESI because (5) doesn't require an explanation, why should there be a gap in ESII either? Isn't (8) on a par with (5)? If (5) requires no explanation, then neither should (8). Thus, there is no explanatory gap between the mental (conscious experience) and the physical.

The Reply

The objection presented in the last section was based on the idea that identity claims do not themselves require explanation. Therefore, we can use identity claims as premises in explanatory arguments without thereby introducing new demands for explanation. This view works out nicely for ESI, involving water. The objector argues that we should view ESII, involving qualitative character, the same way. But the problem is that they clearly aren't the same. Something must be wrong with the idea about identity claims on which the objection is based.

Where the objector points to the explanatory adequacy of ESI as a model for ESII, I would emphasize the fact that we don't need to be convinced of the adequacy of ESI, which shows how different it is from ESII. I will introduce the term gappy identity to express this difference. An identity claim is "gappy" if it admits of an intelligible request for explanation, and "nongappy" otherwise. It seems to me that (5) is nongappy, whereas (8) is gappy. I will elaborate.

With respect to (5), imagine that all the micro-physical facts relevant to the behavior of water are known, but someone still asks, why is (5) true? As discussed above, such a request for explanation might really be a request for justification, in which case the explanatory potential of accepting (5)-that we can explain such facts as water's boiling and freezing points-would suffice as an answer. Alternatively, the questioner might be wondering how water could simultaneously instantiate certain (distinct) properties. This too could be answered. But suppose that the questioner refuses both of these attempts at reinterpreting the question. She just insists that she wants to know how water could be H2O. It seems to me at that point that we could only respond with an incredulous look. After all, what could she really be asking? It just is H2O; that's all there is to it.

On the other hand, when it comes to psychophysical identity claims like (8), the situation is quite different. Let's again imagine that we have all the relevant neurophysiological and functional facts. If someone were to press the question, but how is it (or, why should it be) that brain state B is a reddish experience, the question is quite intelligible. Of course some would insist that the identity must be true, since accepting it would explain a lot of phenomena (such as how reddish experiences cause us to call things "red"). But even someone convinced by causal considerations to accept the identity would still understand what someone was asking when requesting an explanation. We don't just stare blankly wondering what they could possibly have in mind. On the contrary, the sense of puzzlement is all too familiar.

If this distinction between gappy and nongappy identities holds up, then I think we can reply to the objection of the previous section. Granted, the difference between (5) and (8) is not that (5) is derivable from analytic and micro-physical premises whereas (8) is not. There is no analysis of our concept of water underlying our acceptance of its identity with H2O. We accept it because of its explanatory power. (5) itself doesn't require an explanation. However, (5) is different in this respect from (8). (5) is a nongappy identity, a fact that is manifest by our not finding a request to explain it intelligible (that is, once we remove the possible reinterpretations of the request). (8) is a gappy identity, manifest by the fact that a request to explain it seems to be quite intelligible. So, given the intelligibility of a request to explain it, our inability to explain leaves an explanatory gap. Thus, the difference between the two explanation sketches is just this. The one that leaves an explanatory gap is the one that relies essentially on a gappy identity!

Metaphysical Implications

At the start of the chapter I distinguished between the explanatory gap argument we've been discussing and the traditional antimaterialist conceivability argument. The latter attempts to establish a metaphysical thesis, to the effect that mental properties are irreducible to physical properties (in whatever sense of reduction is required by materialism). The former attempts to establish a more modest, epistemological thesis, to the effect that mental properties cannot be explained in terms of physical properties, though they nonetheless might be metaphysically reducible to physical properties. However, given the defense of the explanatory gap argument presented in the last section, it's not hard to see how the metaphysical antimaterialist can turn it to her own purposes.

In section 3 it was argued that identity claims per se never require explanation. Whenever it seems as if someone is intelligibly asking for an explanation of an identity claim, it turns out that their request can be reinterpreted in one of two ways: either as a request for justification, or as an explanation for the coinstantiation of distinct properties. The latter is not really a question about why an identity is true, but why (or how) distinct properties are manifested by the same thing. Now, given the argument of section 4, we can say that this argument only applies to nongappy identities, not to gappy ones. For in the case of gappy identities it does seem as if one can intelligibly ask for an explanation of the identity claim per se.

But, the metaphysical antimaterialist is likely to press, why think the reinterpretation model just proposed for nongappy identities isn't fully general? In fact, given the independent plausibility of the position that when it comes to a pure identity claim there is nothing really to explain, the idea that there are identities for which this rule doesn't hold seems extremely doubtful. Rather, what makes more sense is to account for gappy identities in terms of the standard form of reinterpretation. If an identity is gappy, it's because we really have in mind two distinct properties that are alleged to be instantiated in the same thing, and that's what our request for an explanation targets. But if that is our account of the gappy psychophysical identity claim (8), then there must be two distinct properties for the coinstantiation of which we are requesting an explanation. But this just is to say that being a quale of type R, or some related mental property by which we are acquainted with a quale of type R, is not reducible to a neurophysiological property after all. Thus we now have derived a metaphysical conclusion from an epistemic premise concerning what is or is not intelligibly an object of explanation.

The original basis for rejecting the metaphysical antimaterialist argument was that the inference from what we can conceive to what's genuinely possible didn't seem warranted. The general idea is that the ultimate metaphysical nature of reality-including the nature of the mind itself-is independent of our cognitive access to it. So, the question now is how this materialist move concerning the limits of our cognitive access can be applied to the specific argument about the proper interpretation of gappy identities.

I don't think there's an easy answer here. The existence of gappy identities like (8) is a puzzle that itself requires explanation, and it's not easy to see how to do that within a materialist framework.7 However, I don't think the puzzlement arising from gappy identities justifies dualism either. Let me end by briefly presenting a way of understanding how there could be gappy identities even though materialism is true.

The apparently troublesome fact is that pure identity claims don't require explanation. But to understand this fact properly, we need to appreciate that explanation is not purely an epistemological matter; it has a metaphysical side as well. When I say that phenomenon A explains phenomenon B, one thing I mean is that A's obtaining is responsible for B's obtaining. This is a metaphysical notion. I also mean that I can understand why B obtains, B's obtaining becomes intelligible to me, once I know that A obtains (together with B's connection to A). This is an epistemological notion.

Now, when it's said that a pure identity requires no explanation, this too has both a metaphysical and an epistemological side. On the metaphysical side, what this means is that there is no sense in which there is a responsible source for the identity, other than the identity itself. Identities are brute facts; what else could they be? On the epistemological side, what this means is that once we have removed any questions about the coinstantiation of distinct properties (or about justification), we recognize that we are dealing with a brute fact and therefore requests for further explanation come to seem otiose.

The materialist who recognizes the phenomenon of gappy psychophysical identities must say this. Metaphysically speaking, there is nothing to explain. That is, we are dealing with a brute fact and there is no further source (beyond the fact itself) responsible for its obtaining. The fact that we still find a request for an explanation intelligible in this case shows that we still conceive of the relata in the identity claim as distinct properties, or, perhaps, the one thing as manifesting distinct properties. We can't seem to see the mental property as the same thing as its physical correlate. But though our inability to see this is indeed puzzling, it doesn't show, it can't show, that in fact they aren't the same thing. For what is the case cannot be guaranteed by how we conceive of it.

In the end, we are right back where we started. The explanatory gap argument doesn't demonstrate a gap in nature, but a gap in our understanding of nature. Of course a plausible explanation for there being a gap in our understanding of nature is that there is a genuine gap in nature. But so long as we have countervailing reasons for doubting the latter, we have to look elsewhere for an explanation of the former.


1. The most widely known source for this argument is Descartes's Meditations on First Philosophy, Sixth Meditation. See Rosenthal (1991), p. 26.

2. However, I don't think it's nearly as straightforward as many materialists believe. See Jackson (1993) and Chalmers (1996) for counterarguments, and Levine (forthcoming) for extended discussion of the debate surrounding the metaphysical implications of the conceivability argument.

3. Nagel (1974), in a seminal article, emphasized how we don't have a clue how physical processes can give rise to conscious experience, "what it's like" to be us. The term explanatory gap was introduced in Levine (1983), and the argument further elaborated in Levine (1993b).

4. For the locus classicus on treating explanations as deductions, see Hempel (1965). For a wide range of views on the nature of explanation, see Kitcher and Salmon (1989).

5. Among them, Jackson (1993) and Chalmers (1996).

6. I discuss this issue at length in Levine (forthcoming). Also see Fodor and Lepore (1992), Levine (1993a), and Rey (1993). 7. Again, for more in-depth discussion see Levine (forthcoming). Also, for a different and quite illuminating approach to the problem, see Loar (1997).

7. Again, for more in-depth discussion see Levine (forthcoming). Also, for a different and quite illuminating approach to the problem, see Loar (1997).


Chalmers, D. 1996. The Conscious Mind. Oxford: Oxford University Press.

Fodor, J. and Lepore, E. 1992. Holism: A Shopper's Guide. Blackwell: Oxford.

Hempel, C. G. 1965. Aspects of Scientific Explanation. New York: The Free Press.

Jackson, F. 1993. Armchair Metaphysics, in Philosophy in Mind, O'Leary-Hawthorne and Michael, eds., Dordrecht: Kluwer.

Kitcher, P., and Salmon, W. C., eds. 1989. Scientific Explanation, Minneapolis, MN: University of Minnesota Press.

Levine, J. 1983. Materialism and Qualia: The Explanatory Gap, Pacific Philosophical Quarterly 64.

Levine, J. 1993a. Intentional Chemistry, in Fodor, J. and Lepore, E., eds., Holism: A Consumer Update, special issue of Grazer Philosophische Studien, vol. 46.

Levine, J. 1993b. On Leaving Out What It's Like, in Davies, M. and Humphreys, G., eds., Consciousness: Psychological and Philosophical Essays, Oxford: Blackwell.

Levine, J. forthcoming. Conceivability and the Metaphysics of Mind, NoŻs.

Loar, B. 1997. Phenomenal States, in Block, N., Flanagan, O., and GŁzeldere, G. eds., The Nature of Consciousness. Cambridge: MIT Press.

Nagel, T. 1974. What Is It Like to Be a Bat? The Philosophical Review, vol. 82.

Quine, W. V. 1953. Two Dogmas of Empiricism, in From a Logical Point of View, Harvard University Press: Cambridge, Mass.

Rey, G. 1993. The Unavailability of What We Mean I: A Reply to Quine, in Fodor, J. and Lepore, E., eds., Holism: A Consumer Update, special issue of Grazer Philosophische Studien, vol. 46, 61-10.

Rosenthal, D., ed. 1991 The Nature of Mind. Oxford: Oxford University Press.