Function and Phenomenology: Closing the Explanatory Gap

This paper critiques the view that consciousness is likely something extra which accompanies or is produced by neural states, something beyond the functional cognitive processes realized in the brain. Such a view creates the 'explanatory gap' between function and phenomenology which many suppose cannot be filled by functionalist theories of mind. Given methodological considerations of simplicity, ontological parsimony, and theoretical conservatism, an alternative hypothesis is recommended, that subjective qualitative experience is identical to certain information-bearing, behavior-controlling functions, not something which emerges from them. This hypothesis explains the isomorphism between the structure of experience and neural organization, while providing a naturalistic account of qualia as relational properties of informational states, not a separate ontology of phenomenal essences. On this functionalist view, the hard, empirical problem of consciousness is to discover precisely which neural functions constitute subjective experience. (This paper originally appeared in the Journal of Consciousness Studies in 1995 and was reprinted in Explaining Consciousness, Jonathan Shear, Editor, MIT Press, 1997.)

To truly explain consciousness, we must find a convincing place for it in the natural world and ultimately in the scientific description of that world as expressed in physical, biological, and information theory. The default assumption when undertaking an explanation of consciousness should be that there is nothing ontologically special about it, nothing which sets it apart from the rest of nature as we presently conceive it. In his keynote paper, "Facing Up to the Problem of Consciousness," David Chalmers assumes much the opposite: subjectivity is something more than the naturally evolved neural processes which seem likely, given the available evidence, to instantiate it. Qualitative consciousness - qualia or phenomenal experience - is said to 'arise from' or 'accompany' or 'emerge from' these processes. This dualism of underlying process versus accompanying subjectivity creates the 'explanatory gap' that so worries Chalmers and other philosophers: why should subjectivity arise from some processes and not others? My claim is that the central mystery about qualitative consciousness supposedly in need of explanation is an artifact generated by this presupposition about its nature - that it is an 'effect' of an underlying process - and it is precisely this that we must question if we are to find the true place of consciousness in the world.

My strategy will be first to show that Chalmers' initial assumption about consciousness compromises widely held methodological canons of scientific theory construction that he himself avows. Next, I will suggest some reasons why this assumption has gained such currency, despite its serious shortcomings. In parts three and four, using the principles of Chalmers' theory of consciousness as discussion points, I will sketch an alternative picture, that subjectivity is certain sorts of physical, functional, informationally rich, and behavior controlling (cybernetic) processes. Interestingly, Chalmers does much to support this picture in the latter part of his paper. Only the assumption that qualitative experience somehow arises from such functional processes prevents him from reaching the conclusion that I will defend here, that experience is identical to them. It is the elucidation of this (proposed) identity by the empirical investigation of what precisely these processes do - in contrast to unconscious processes - that will eventually constitute a robust explanation of consciousness. On this picture, there is no question needing an answer about why just these processes give rise to consciousness, since indeed it never 'arises' or 'emerges' at all. Consciousness is what we consist of as physically instantiated subjects, not something extra that our brains create.

I

The hard problem of consciousness, according to Chalmers, is that of experience. 'It is widely agreed that experience arises from a physical basis, but we have no good explanation of why and how it so arises. Why should physical processing give rise to a rich inner life at all? It seems objectively unreasonable that it should, and yet it does' (p. 6). Here, at the very outset, a specific conception of consciousness sets the stage for the rest of Chalmers' investigation. On this widely held view, experience emerges out of and accompanies certain neural functions, but is assumed not to be identical to these functions. The pressing question thus becomes, as Chalmers puts it, 'Why is the performance of these functions accompanied by experience?' (p. 8, original emphasis). If experience is taken to be something over and above neurally instantiated functions, something extra which accompanies them, the 'central mystery' of consciousness becomes the 'explanatory gap...between function and experience.' (Levine 1993 takes more or less the same position, pp. 130-135.) This in turn leads Chalmers to suppose that 'To account for experience, we need an extra ingredient in the explanation' (p. 12, original emphasis). He despairs of finding an explanation of consciousness within existing scientific theory, since physicalist and functionalist accounts will forever omit this extra ingredient:

For any physical process we specify there will be an unanswered question: Why should this process give rise to experience? Given any such process, it is conceptually coherent that it could be instantiated in the absence of experience. It follows that no mere account of the physical process will tell us why experience arises. The emergence of experience goes beyond what can be derived from physical theory. (p. 13)

Of course, it is conceptually coherent that experience might be absent in the presence of certain physical, functional processes, but only on the assumption that they may not be, and probably are not, identical. But this is precisely the fundamental issue that must not be prejudged. As David Cole has pointed out recently,

When a critic [of functionalism] supposes that there could be, 'ex hypothesi,' a system that instantiates the functional architecture yet fails to have the experience, the critic is 'ex hypothesis' supposing that any functionalist theory is false. To simply suppose that the theory is false is question begging, and cannot be the basis of an adequate argument against functionalism. (1994, p. 297, original emphasis)

If we begin our investigation into the nature of consciousness with Chalmers' picture in mind - the picture of experience 'arising from' (hence not identical to) functional neural processes - naturally we will be led to doubt that functional explanations can fully account for experience. But the burden is on Chalmers to justify his starting point and its implicit dualism, the dualism that leads him to suppose that there must be an 'extra ingredient' in any explanation of qualitative consciousness which goes beyond descriptions of functions and physical processes. As Francis Crick observed recently, 'Whether there is something extra remains to be seen' (Crick, 1994 p. 12). Why, if we are conducting a more or less scientifically motivated investigation of a phenomenon, should we begin by assuming that this phenomenon is probably not of a piece with the rest of nature as we currently conceive it? The natural starting point, on the contrary, is to assume parsimoniously that we need not add experience, as Chalmers recommends we do, to our 'fundamental' ontological categories. Generally, we shouldn't posit as fundamental that which we are seeking to explain. Instead, we should start with what seems clearly a simpler and more straightforward hypothesis, namely that experience is identical to certain neurally instantiated cybernetic functions. The way is then open to understand consciousness by empirical investigation of whatever functions are found to correlate with experience, whether waking or dreaming. Such an explanation would best embody the virtues of 'simplicity, elegance, and even beauty' that Chalmers cites as hallmarks of good theory.

Chalmers says that it is 'a conceptual point that the explanation of functions does not suffice for the explanation of consciousness' (pp. 13-14, original emphasis), but again this is a conceptual point only under the particular initial conception of experience that Chalmers adopts, a conception that in effect assumes a conclusion about the nature of consciousness that although widely accepted, can hardly serve as a methodologically sound starting point. After all, the basic explanatory motive in science and philosophy is to incorporate heretofore inexplicable phenomena into an existing theoretical framework, modifying the framework only as minimally necessary to effect the incorporation. This motive is defeated by assuming at the very start that consciousness is a phenomenon that transcends the explanatory reach of existing theory.

The debate about where to begin in explaining consciousness can thus be framed as a competition between powerful and widespread intuitions that there is something 'extra' about consciousness in need of explanation and commitments to standard scientific explanatory practice. I suggest that we are better off allied with the latter than the former, and that the history of successful science is on my side. It is not that the identity of phenomenal consciousness and cognitive function is obviously true, since if it were we'd all be functionalists. Upon further investigation it may turn out (although I doubt it ) that experience does not correlate with any particular set of functions, or that in some essential respect it floats free of any physical or functional property. I am only arguing that it is methodologically more circumspect to start off with a simpler hypothesis, one which does not posit a special nature or essence of subjectivity to be explained.

For some, this may appear to prejudge the issue in the opposite direction, since it seems to deny the existence of the explanandum itself, at least in the form they are used to conceptualizing it. But my suggestion is only that we remain open about the nature of consciousness as the investigation gets underway, and that our starting hypothesis be shaped by methodological constraints, not by concepts or intuitions motivating the supposition, at the start of the investigation, that there likely exist fundamental properties or entities not already included in naturalistic theories.

II

What is so compelling about Chalmers' picture of subjectivity that it tends to override methodological considerations? That is, what leads many philosophers, scientists, and humanists to strongly doubt, at the outset, that qualitative consciousness might simply be identical to certain types of functional organization and hence suppose that it is some sort of contingent effect or accompaniment? This is a different question from asking why they might merely doubt that such an identity holds true. The answer to the latter, of course, is simply that the identity of phenomenal experience (qualia) and cognitive function is not obvious, and should be subject to the same legitimate methodological skepticism accorded any other hypothesis. But the answer to the former lies in appreciating the effect of several distinct trends in our philosophical tradition, some general and some specific, which bias our intuitions about the nature of consciousness against the possibility of identity.

First and perhaps foremost is the residual, and by now mostly subliminal, grip of Cartesianism on basic assumptions about mental life in general. This influence leads us to regard talk of the mental not just as a useful, predictive, 'intentional stance' (Dennett 1987) sort of discourse, but as talk that refers to an independent, or at least parallel realm which interacts with or accompanies physical processes. Pain, pleasure, perceptions, emotions, thoughts, beliefs, and other mental phenomena present themselves as a related set of states and events which for everyday predictive purposes seem virtually autonomous from the physical and functional (or, to use Dennett's typology, the physical and design stances). We need not ordinarily bring in physicalist or functionalist talk in order to get by in the interpersonal world, so it is unsurprising that the commonsense assumption about the autonomy - and hence the ontological separateness - of the mental still generates a philosophical bias despite the best efforts of many philosophers to uproot it.

I am not suggesting that dualism (of either substance or properties) is a priori untenable, although it has well-known difficulties, but only making the uncontroversial point that our philosophical tradition has been heavily influenced by Cartesian intuitions about the ontological divide between the mental and the physical, intuitions reinforced by ordinary mentalistic discourse. Thus when it comes to choosing a picture of consciousness to start our investigations (and we have to choose some picture, after all) we may well be biased in favor of one which puts consciousness in the autonomous realm of the mental, even if only in a rather subtle, sophisticated way. Chalmers' statements to the effect that certain neural processes 'give rise' to qualia constitute exactly such a picture, one which splits the mental from the physical/functional and which takes consciousness to be some sort of unextended, non-spatial property that likely eludes currently available scientific explanations. The resulting 'naturalistic dualism' Chalmers defends is Cartesian at its core, and despite his claim that such a position is 'entirely compatible with the scientific view of the world,' dualisms have fared badly as science proceeds to unify our conception of humankind in nature.

Another factor strengthening the intuition that qualia emerge from physical or functional processes is the supposition that there most likely exists a clear demarcation between those sorts of states likely to 'carry' phenomenal consciousness and those that don't. Qualia could well be absent in systems that are as smart as we are, but that are differently organized or instantiated, or so it is often supposed. (See, for instance, Flanagan 1992, pp. 129-152 for a recent defense of this thesis.) Block's Chinese nation and Searle's Chinese room thought experiments trade on the intuition (although it begs the question against functionalism) that bizarre cognitive systems, however functional in real time, couldn't possibly constitute subjects (Block, 1978; Searle, 1980). Once the wedge is driven between function and qualia (or, as in Searle's paper, between function and intentionality) in the bizarre cases, the stage is then set to suppose that qualia are only a contingent accompaniment to cognition, even in systems functionally very similar to us. Likewise, it seems plausible to many that qualitative states are probably absent in systems that fall short of the functional complexity given to humans and (perhaps) the higher animals. Although the line of demarcation is obscure in both cases, the basic assumption is that if a system fails to be sufficiently like us in some respect - in its physical instantiation or functional design - then the chances are it's not a subject; it doesn't produce or give rise to qualitative states.

If only a certain class of cognitive or functional processes (those more or less like ours) are taken as plausible correlates of subjectivity, this reinforces the notion that something special about those processes produces subjectivity as an 'effect,' whether epiphenomenal or efficacious. Consciousness emerges only at a certain level of complexity, or only out of a certain type of functional design, or only as the product of a certain type of physical instantiation. Drawing a line in advance between systems which are thought to be likely candidates for consciousness and those that are not pushes us towards the picture of subjectivity as a contingent accompaniment to functional processes.

There is, of course, a good deal of anthropocentrism in drawing the line where those convinced by absent qualia thought experiments have placed it. But this may appear a reasonable prejudice, given that it seems we can only be certain of the existence of qualitative states from the single example of human experience. Surely it would be irresponsible as a starting assumption to grant qualia to just any old cognitive system, say a horsefly, a frog, or Hal of 2001. Perhaps. But there is a difference, I suggest, between this sort of methodological conservatism (which in general I support) and the uncritical acceptance of the 'similarity to ourselves' criterion as a benchmark for assigning consciousness to a system. To turn the issue around, why should we assume, from the one example of human subjectivity, that qualia are not present in dissimilar cognitive systems, real or imaginary, that manage well in the world? It is initially at least as plausible, to paraphrase Paul Churchland, that any cognitive system operating at or near our level must contain a 'computational-executive organization' which would be a 'home for qualitative states' (Churchland, 1989, p. 38).1 And more basically, why assume that consciousness 'accompanies' or is 'produced by' functional processes in the first place?

In response to the first question, I propose that we remain agnostic about the possible subjectivity of non-human and less complex systems, otherwise the anthropocentric bias may blind us to the true nature of consciousness. To the second, the focus of this paper, I propose likewise that we do our best to keep the issue open, not to assume at the outset the non-identity of function and qualia. Since the narrowing of candidates for qualitative subjectivity based on similarity to us suggests that (only a certain class of) cognitive functions generate qualia, it can be seen that staying agnostic about the subjectivity of other sorts of systems is a good way to keep this deeper issue open. Qualia may not be generated at all; they may simply be certain types of functional organization, the range of which may not be limited to what is familiarly human.

A third factor motivating the 'emergence' picture of qualitative consciousness is that only by having its own, independent existence could it possibly play the important causal role in our lives that it seems to. Involved here are fairly deep and emotional issues of human autonomy and specialness, especially the fear that if consciousness is nothing over and above physically instantiated function, then we lose our privileged status as rational agents riding above the flux of brute causality. As persons, we tend to identify ourselves with our conscious capacities, and moreover tend to believe that we are in some sense in control of these capacities. Consciousness, conceived of as a product of a very restricted class of functional, cognitive processes (ours), generated by a very restricted class of physical objects (human brains), is what crucially distinguishes us from the rest of insensate, mechanistic nature. If it turns out that subjectivity and the sense of self is 'merely' function, then it becomes terrifyingly (for some) clear that no principled distinction may exist between us and a very clever robot, at least on the question of who has 'inner states.' To the extent that we want, unconsciously or consciously, to preserve our special status vis-a-vis the robot we may opt for a picture of subjectivity, including qualia, that preserves for it a central causal role and restricts it to creatures very much like us. This picture of consciousness is just what the doctor ordered to keep 'creeping mechanism' at bay. (Dennett 1990 makes this point, pp. 523-4.)

But of course we must not let such fears prejudice our initial conception of consciousness or restrict our investigations. If it turns out (as I propose) that consciousness has no independent causal role over and above the functions which instantiate it, and if such functions could be realized in creatures (or artifacts) quite unlike us in some respects, then the implications of that for our personal and moral status must be dealt with as a separate issue.

I suspect that this same worry also works to make Chalmers' anti-reductionist approach to explaining consciousness an attractive alternative to standard scientific practice. After all, a reductive explanation of subjectivity, one which identifies it with a class of functional processes that are in principle realizable in a wide range of instantiations, obliterates having an inner life as the basis for a special human status. (Others remain, however.) There are good arguments against certain types of facile reductionism, of course, and I don't mean to imply that language referring to belief, intention, feeling, and thought is in any sense eliminable. But as a scientific strategy for unifying knowledge, the reductionist impulse is hardly to be eschewed but rather to be encouraged, one would suppose, and reductionism does not diminish us when we become the objects of knowledge. To 'reduce' mental phenomena to functional processes via some plausibly evidenced identification is, after all, not to eliminate them, but simply to redescribe them from a third person perspective. Why such a redescription might seem threatening is an interesting question for another time.

III

Having adduced some possible explanations (there are undoubtedly others) for why Chalmers' picture of consciousness has such currency, I wish now to defend an alternative, one which I believe is a better candidate for our attention this early in the game. I will argue for the identity of function and consciousness - what I will call the functional identity hypothesis - and as the argument develops, further reasons for doubting that qualia arise from or are produced by functional processes will come to light. (See Lloyd 1992 for a recommendation to adopt such an identity hypothesis.)

As mentioned earlier, Chalmers has set out some of the preliminary argument for such a position in his paper, and he describes what in my view is the key characteristic of functions likely to instantiate consciousness: their role in representing informational content essential for the control of complex, adaptive behavior.2 The basic functional identity hypothesis is that qualitative experiences are what it is to be a set of multi-modal, discriminative, representational processes which deploy information for the control of behavior. I use this extended form of identity expression ('what it is to be') quite deliberately in order to emphasize that we, as experiencing subjects, are instantiated by such processes.

To make this point clear, note the similarity, and dissimilarity, of this expression to Nagel's (1974) popular characterization of subjectivity as 'what it is like to be' a such-and-such. The difference between the two is not trivial. Nagel's formulation suggests the notion of an inner life which the subject somehow witnesses, or has direct epistemic 'access' to, so that for instance it (and only it) could say what it was like to have a particular quale. Nagel's central thesis about subjectivity is that this 'first person perspective' cannot be captured by science. By contrast, the formulation I wish to defend hinges crucially on the proposal that as subjects we are constituted by and identical to cognitive processes which themselves instantiate qualia, hence qualia are what it is for us to be these processes. Under this proposal we can't, finally, say what it's like to have qualia since we don't have a first person perspective on them: we don't 'have' them at all, neither do they 'appear' to us, nor are we 'directly acquainted' with them. We, as subjects, exist as them. The ineffability of qualia, among their other properties, is thus a consequence of and explained by the functional identity hypothesis, and qualia generally are not beyond the reach of science. (Part IV will return to these points.)

I believe Chalmers is very much on the right track in what he calls his two 'non-basic' psychophysical principles, although his commitment to the emergence picture of consciousness prevents him from realizing their full explanatory potential. He calls these the principles of 'structural coherence' and 'organizational invariance;' both concern the functional deployment of information in a cognitive system. Structural coherence amounts to the well-established, if not yet completely fleshed out, correspondence between variable features within a sense modality and the empirically associated covariant neural structure and processing. For instance, the structure of our phenomenal experience of color corresponds to the structure of the neural state space of our color processing system (as described by Paul Churchland 1989, pp. 102-110). There is at least a rough isomorphism between the sort of variability we report in our experience and the organization of neural events associated with this variability.

Chalmers theorizes (correctly, I think) that the raison d'etre of cognitive functional processes is to embody informational content which 'is brought to bear in a widespread way in the control of behavior.' Consciousness, it seems, is usually involved with most, if not all, higher order cognition and behavior, including memory, anticipation, speech, learning, planning, and complex motor activity. As Flanagan (1992) puts it 'What is consciously accessible is primarily just what we have the most need to know about: conditions in the sensory environment, and past facts, and events' (p. 134). Such knowledge, whether recognitional, propositional, or performative, requires some sort of representational system involving information about the world and the body. All in all, the connection between consciousness, information and behavior, although not watertight and admitting of exceptions, seems a fruitful line of investigation.3 The functional identity hypothesis makes a strong claim about this connection: that subjectivity is constituted by those central representational processes which transform and enhance sensory information to the point where it normally dominates in the control of behavior. (See Van Gulick 1993 pp. 147-50 for some interesting ideas on what sorts of processes these might be.)

If it is plausible that functional processes correlated with consciousness are primarily informational and behavior controlling, and if the principle of structural coherence holds, (and so far it seems empirically the case that there exists isomorphism between experience and neural organization), it follows that the structure of qualitative experience mirrors the informational content essential to the control of complex behavior. The burning question, however, is whether indeed qualia mirror the informational content of functional processes as some sort of separate, parallel entities in consciousness, or whether qualia are this informational content. Chalmers says, in support of the first position, that 'There are properties of experience, such as the intrinsic nature of a sensation...that no structural distinction can exhaust' (p. 18). So although the structure of experience might well be isomorphic to functionally derived informational content, experience itself cannot simply be this content, since (he believes) it has additional, non-functional mental properties, such as an intrinsic phenomenal nature. Such mental properties are irreducible to physical or functional properties, which is to say we can't explain them or redescribe them using only physical or functional predicates without leaving out something crucial. As Nagel (1986, p. 15) has expressed it, 'The subjective features of conscious mental processes - as opposed to their physical causes and effects - cannot be captured by the purified form of thought suitable for dealing with the physical world that underlies the appearances'. The truth of this claim is, of course, the central bone of contention between physicalism and functionalism on one side, and substance and property dualism and dual aspect theories on the other.

In introducing his second principle, that of 'organizational invariance,' Chalmers says, 'This principle states that any two systems with the same fine-grained functional organization will have qualitatively identical experiences' (p. 19, original emphasis). This seems to contradict his earlier claim that the intrinsic nature of sensory experience is beyond the reach of structural distinctions, that is, beyond what the organizational structure of neural processes can account for. Indeed, organizational invariance comes perilously close, it seems, to identifying quality with function (same function, same quality) except that Chalmers continues to insist on the 'emergence of experience' as some sort of entity separate from functional organization.

Chalmers' passing observation that the principle of organizational invariance is 'controversial' is a bit of an understatement, since it is less a principle than a major thesis about the mind. Were it proved, then, as he puts it, 'the only physical properties directly relevant to the emergence of experience [would be] organizational properties' (p. 20, original emphasis). If one were to drop the notion of emergence then this would pretty much amount to the hypothesis I want to defend. Experience is up-and-running functional organization, not something emerging out of it; qualia just are what it is to be a subject instantiated by a set of informationally rich, discriminative, behavior-controlling processes, not a separate ontology of intrinsic natures.

Note that in this purified form, such a hypothesis explains the principle of structural coherence. That is, the reason the structure of conscious experience so nicely mirrors the informational goings on of neural processes is because they are one and the same, under different descriptions, or very loosely speaking, from different perspectives, first and third person. (As the reader will have noted above, I question the literal accuracy of the expression 'first person perspective' as related to one's experience.) Once we discard the idea of emergence the full explanatory power of the functional identity hypothesis starts to become apparent. And working from the other direction (as I will in part IV below), appreciating the extent to which the identity thesis can account for various mental property ascriptions will make it easier to abandon the picture of emergence as explanatorily superfluous.

Chalmers' third, 'basic' and highly speculative principle sketches a deeper explanation for both structural coherence and organizational invariance, but in the end it seems less explanatory than ontologically inflationary. On this principle information 'is truly fundamental, and ...has two basic aspects, corresponding to the physical and phenomenal features of the world' (p. 21). The parallel emergence of phenomenal and physical properties, then, is attributable to a yet deeper substratum of information. But, to return to an earlier point, it seems premature to hypothesize new fundamental entities (in this case universal 'bits') at the start of an investigation when simpler, less inflationary hypotheses might do as well. Positing an underlying basis of information with two aspects may explain the congruence of experience with informationally contentful neural organization, and may support the notion that similar, information-bearing functions might generate similar phenomenal qualities, but we are then left with the equally difficult task of explaining why information should possess two aspects to begin with. It doesn't help to 'solve' one explanatory problem by creating yet another at a more speculative level, at least not until more economical approaches have been thoroughly explored. It is also worth noting that with this proposal Chalmers has shifted from recommending that we accept experience as fundamental (p. 15) to speculating that information is fundamental (p. 21). Either way, all the explanatory work still remains to be done.

IV

I have already suggested how the hypothesis that qualia are identical with higher level functional processes can explain structural coherence: the parallel structure of experience and neural organization is accounted for by supposing, parsimoniously, that experience is to exist as certain types of neural processes. Likewise, the principle of organizational invariance is simply a corollary of the identity hypothesis: two functionally equivalent systems will have the same sorts of qualitative experience because qualia are particular informational values within some sort of functional state space, however it may be realized. (Below I will amend this account of phenomenal quality by questioning the reality of the intrinsic, essential nature of qualia.) But these explications will likely seem beside the point for those whose primary intuition about consciousness is precisely that, as Chalmers insists, experience is something over and above functional processes.

To undercut this intuition, and thus make functional identity more appealing as an initial explanatory hypothesis, it will be helpful to consider some of the properties normally attributed to qualia: their privacy, ineffability, and intrinsic nature. The privacy of qualia, the fact that no one but myself can feel the pain that I feel, is relatively easy to explain as a matter of identity. Flanagan describes it thusly: '... a particular realization [of an experience] will be an experience only for the agent who is causally connected to the realization in the right sort of way... [T]he biological integrity of the human body can account straightforwardly for the happy fact that we each have our own, and only our own experiences' (p. 94, original emphasis).

I would modify Flanagan's account somewhat: the reason no one else feels my pain, even though they might conceivably have a complete description of its neural instantiation, is that only I as a subject consist (partially) of that pain. There is no separate agent or subject apart from the sum total of my experience which could be causally connected to pain in the first place, so naturally no one else could stand in that relation to my experience either. That experience is private is thus explained not as a matter of direct epistemic access but as a matter of instantiation: subjectivity consists of a complex array of coordinated, information-bearing neural states, and as a subject I consist of those states, and no one else can so consist. That I can report being in pain is not, therefore, because I perceive pain (as I might perceive a tree or chair) by virtue of a special causal connection to it, but because I subjectively exist as pain - as stimulated C fibers and associated higher level processing - to some extent. (Under torture I might subjectively exist mostly as pain and the concomitant terror.)

Why, under the assumption of functional identity, might qualia be ineffable? Very much the same logic applies. The subject, since it consists of an ongoing stream of neurally instantiated experience, is not in a position to witness or observe the basic elements of that experience. We cannot, as it were, step back from and describe a quale as we might an external object; thus we can do no more than name basic qualitative experiences ('red,' 'hot,' 'sweet,' etc.) and compare and contrast them to one another. We can't describe the redness of red or painfulness of pain precisely because we can't get a perspective on these qualities. (Again, it is a mistake to suppose that we have a first person point of view of our experience.) They are, so to speak, the counters in the game of perception and so cannot be made the object of play. External particulars, on the other hand, and complex internal states, are describable just because they are constituted by ensembles of qualitative elements. Chairs are (sometimes) brown, assume a given shape in my visual field, are hard to the touch, resonant when struck, etc. That the subject is identical to a set of representational processes can thus explain why the primitive components of that set, what we call qualia, are descriptively opaque and non-decomposable, that is, ineffable: we can't stand in an epistemic relation to those representational states which we consist of as knowers.

This brings us to the last, and perhaps most central (and controversial) of properties attributed to qualia, the notion that they have an intrinsic phenomenal nature. It is here, many suppose, that functionalist theories must founder, since the essential redness of red or painfulness of pain seem to float free of any functional role. We can imagine red objects looking blue (or imagine they look blue to someone else) without supposing that our (or the other person's) discriminative and behavioral capacities would be in any sense compromised. The essence of red, just because it is a non-decomposable, ineffable primitive, is thought not to be amenable to functional or structural explanation. As Van Gulick puts it: 'No matter how much structural organization we can find in the phenomenal realm and explain neurophysiologically, [the critic of functionalism] will insist that the distinct redness of phenomenal red will not have been captured or explained by our theory' (1993, p. 145). The purported intrinsic phenomenal nature of qualia creates the explanatory gap functionalism is supposed not to be able to fill.

The functional identity hypothesis can, however, meet this burden by supporting a challenge to the intuition that there are determinate phenomenal facts about the intrinsic nature of qualia in need of explanation. The challenge, simply put, is to specify what precisely is being claimed to be the case when we point to a patch of red and say our experience is like 'that.' Since there is nothing further we can say about the (purported) phenomenal essence of red, there is no way we can link such an essence with a property or state of affairs which specifies or fixes its occurrence, other than to gesture at red objects. Hence, referring to such an essence plays no explanatory or descriptive role in talk about our experiences, and this may begin to cast doubt on whether such an essence exists, even though it may strongly seem that it does. The ineffability of qualia, it turns out, could be a clue to their not having a determinate intrinsic nature after all.

The intuition that experiences do indeed have some sort of qualitative essence supports the possibility of inverted or 'alien' qualia, that the intrinsic character of an experience might differ across subjects. My red might be the same as your red, or perhaps not. A Martian's 'red' might even be a radically different sort of experience. But since experience is private, intersubjective comparisons of qualia are of course impossible. If we can never tell, finally, whether another subject's red is the same or different than mine, this might make us further doubt the validity of the notion of intrinsic phenomenal natures. Why, after all, should we take seriously a 'fact' - the 'fact' that my experience of red is possibly like or unlike yours - which is in principle impossible to ascertain? Both the ineffability and privacy of qualia, therefore, undermine the plausibility of first person phenomenological facts involving determinate qualitative essences which science cannot capture.

That we can reliably distinguish red from other colors isn't explained by there being an essence of red, a particular quality of redness that, for instance, we might suppose other perceivers of the same red object might or might not be experiencing. The ability to pick out red depends simply on the fact that there exists a range of contrasting colors against which red is distinguished as a relationally defined member of that range. This is borne out by the obvious point that every distinguishable bit of red we see is experienced against a background of non-red. If all the world were red, red would drop out as a discriminable property of experience. Imagine Mary, Jackson's color-deprived neuroscientist, growing up in an environment where everything was red instead of (as originally conceived in the thought experiment) black and white (Jackson 1982). Could she have a concept of 'first person' red? Not until, I suggest, other colors were introduced into her environment against which red could be reliably distinguished. A monochromatic world of whatever color is a phenomenally colorless world, not a world in which the single color could 'declare' itself by an intrinsic phenomenal nature.

A functionalist account of our sensory capacities makes plausible the relational, mutually defined nature of qualitative experience, hence supports the attack on the existence of intrinsic phenomenal essences. As the Churchlands have theorized, sensory state spaces, as realized in neural organization, instantiate a range of possible sensory-informational values within a given modality (vision, hearing, taste, etc.), values which correspond more or less to some range of stimulation from the external world and the body. What distinguishes different values of a given modality is variation along one or more of its component dimensions (three in the case of vision, four in the case of taste), a variation that defines each state of the modality in relation to its other states. These states gain their functional significance by virtue of mapping differences in stimulation originating from the external world or the body. Hence it is the difference between the neural instantiation of red and the neural instantiation of blue (among other colors) which defines them as distinct qualitative experiences, not that red is instantiated by a particular state that by itself defines it as red or that blue is instantiated by a particular state that by itself defines it as blue. If, as I hypothesize, we as subjects exist as such sensory modalities (along with other sorts of cognitive processes) the experience of a particular shade of red is just one in an array of possible states of the color system that gets its qualitative value solely by having a particular place in the array. The 'way' red is, what it is 'like,' is simply to exist as a given color state in contradistinction to other states. There is no need to posit an intrinsic nature of red as separate from its functional role as a placeholder in the neural processing that constitutes the experience of color, and that, partially, constitutes us as subjects.4

Conclusions

The foregoing analysis is meant to help undercut the intuition that there are phenomenal facts in need of explanation that cannot be captured by functional facts. The explanatory gap is closed by showing both that the privacy and ineffability of qualia are indeed explained by the functional identity hypothesis, and that their much vaunted intrinsic qualitative nature does not exist, hence needs no explaining. This helps to support the view that qualia are not phenomenal entities that emerge from or arise out of functional processes, but are instead best conceived of as being those processes. This is not to 'quine' qualia out of existence, but simply to identify them with certain sorts of neurally instantiated representational states. Of course, the meaning of the terms 'phenomenal consciousness' and 'qualia' is not (presently) identical to the meaning of the expression 'neurally instantiated representational states' and its functionalist and physicalist cousins. But talk about qualitative experience refers to the same thing as talk about particular neural processes, although precisely which processes we don't yet know.

I have tried to emphasize the explanatory virtues of the functional identity hypothesis, hoping to win converts to functionalism by showing that it can indeed account for subjectivity. Yet there still might seem a residual, fatal question left unaddressed (the frog grinning up at us from the bottom of Austin's beer mug): why should existing as functional processes of a particular kind be identical to qualitative experience? What is it about them that makes them phenomenally conscious, as opposed to unconscious? (This, it will be seen, is Chalmers' original, central question, but posed about identity instead of emergence.) Well, if by 'phenomenally conscious' we insist on meaning 'possessed of a first person, intrinsic, essential nature' then the question stands unanswered, since nothing in functionalism or science will ever show us how to get from physics, chemistry, biology, and cybernetics to consciousness thus defined. The causal and structural aspects of subjectivity might fall to functional explanation, but intrinsic natures will not, since after all, these are custom made to resist assimilation by science. Why? Because scientific explanation works, in part, by showing how the causal relations among elements of a system at one level can account for features at another level. The macroscopic properties of water, for instance, can be explained by the microstructure of water molecules. But since the property of having an intrinsic qualitative nature is defined (by its adherents) as independent of any causal role and as having no structure, no lower level explanation will ever reduce it. As Levine (1993, p. 134) puts it, '[T]o the extent that there is an element in our concept of qualitative character that is not captured by features of its causal role, to that extent it will escape the explanatory net of physicalist reduction.'

I have tried to cast doubt on the conviction that our concept of qualia need include the property of having a factually determinate intrinsic nature independent of function or physical instantiation. It is logically possible that qualia so exist, but there are no good empirical reasons, nor for that matter any good phenomenological reasons, for supposing they do. There are good reasons, on the other hand, for supposing that qualia are instead a matter of the contrast relations among functionally defined, neurally realized representational processes. If such is the case, then the conscious/unconscious distinction lies in whether or not such processes are instantiated. Discovering just which processes constitute phenomenal consciousness is an empirical matter that only neuroscience can answer, by correlating neural states with various conscious capacities.

There may or may not be a clear functional or physical line to be drawn between the conscious and the unconscious, but when we ask why a process is conscious we will no longer be wondering why qualitative essences accompany particular neural events. Rather, under the identity hypothesis, we will be asking why that process needs to involve the functions typically operational when we are conscious. The beating of my heart or the (mostly) silent operation of my digestive system obviously need not involve conscious processes, since neither requires the representation of a complex environment, or motivational economy, or set of future contingencies. Short-term memory, on the other hand, is a conscious process because it requires those representational capacities only normally available when we are conscious. Some sort of qualitative experience (whether of occurrent thought, feeling, or perception) is necessary as input to the short-term memory system, and such experience is likely to be empirically cashed out as a complex set of discriminative and integrative functions which manage information that serves to control moment-to-moment behavior. The functional identity hypothesis thus explains why we have subjective experience: qualitative representations are what get the job done.

The functional identity hypothesis proposes no new fundamental classes of entities or properties or features to be added to naturalistic theory, and so has the virtues of simplicity, ontological parsimony, and methodological conservatism. If it turns out that consciousness depends fundamentally on a physical process, e.g., a phase locked 40 hertz neural oscillation, or if, as I suspect, it becomes identified with a rather sophisticated array of informational processes which could be realized by many different physical systems, in neither case will anything radically new be discovered to play an explanatory role (although figuring out the actual workings of the massively recursive networks embodied by the brain may reveal vast new realms of control theory). Of course it is remotely possible that a heretofore unknown fundamental property or entity, whether of experience or information or quantum coherence, may eventually be shown to exist. But asserting its probable existence in advance of strong empirical support and theoretical necessity would be to start off on the wrong foot.

If indeed qualia are identical to certain functional informational processes, then the hard problem of consciousness (what Chalmers more or less takes to be the easy problem) is the empirical one of discovering just what these processes are. Equally hard for many, perhaps, will be dropping the picture that consciousness is something ontologically special produced by these processes. Instead, we must get used to the idea that as much as our 'inner' lives seem a categorically different sort of thing than the external world (a world which includes our brains), they are, in fact, just more of that external, physical world which we as subjects happen to be. We must also get used to the idea that consciousness does not have a cognitive or functional role over and above the functions which constitute it, and that any system, artificial or natural, which instantiates such functions will be conscious.

In a sense there is nothing special about consciousness since, if I am right, nothing extra is produced by or emerges out of this set of functions. They just constitute a marvelously complex and adaptive representational system which keeps us out of trouble, more or less. Yet on the other hand, consciousness is indeed special in that these functions, as carried out in our day to day lives, have, as a historical fact, led most of us to suppose that an ontologically separate world of subjectivity exists. It may be that the sorts of higher level cognitive processes which are found to correlate with consciousness inevitably generate (among language users) a self/world model containing the strong intuition that the self and its experience cannot simply be the body, cannot simply be a bit of the world suitably organized. Explaining consciousness satisfactorily will consist in overcoming that intuition, and in placing experience fully within the natural, physical realm.

Acknowledgements

I am grateful to Mary Ellen Myhr for a close reading of an earlier draft of this paper, and to the anonymous referees for helpful criticisms from which the present version has benefited.

Notes

1. Lycan makes somewhat the same point when considering systems that are functionally very much like us: 'Is it really possible to imagine something's sharing my entire many-leveled functional organization and still not being conscious in the way that I am?' (1987, p. 24 his emphasis). Of course one can imagine this, but the possibility of absent qualia in the face of functional near-equivalence need hardly be the default assumption. Despite intuitions to the contrary, there is no a priori conceptual barrier that separates experience from function.

2. Several other philosophers have taken a similar tack, although they still by and large assume some version of the emergence picture. See for instance Flanagan 1992, pp. 129-152, Van Gulick 1993 p. 152-3, and Van Gulick 1980 for discussions of the role of information in conscious processes.

3. Blindsight experiments seem the obvious counterexamples, since the subject clearly has what Flanagan terms 'informational sensitivity' to objects in the blindfield without 'experiential sensitivity' (a reportable experience) of them (Flanagan 1992, pp. 147-52). But the very limited sorts of behavior possible with respect to blindfield objects actually highlights the centrality of consciousness in mediating most of our interaction with the world. These limitations, when compared with normal behavior, suggest what the special informational functions that instantiate consciousness might be. See Van Gulick 1993, pp. 147-50 for some speculations about candidate functions.

4. If one wishes to retain the notion of intrinsic qualitative nature but deny that this defeats functionalism, Paul Churchland offers a way out: admit that such natures exist but insist that 'such intrinsic natures are nevertheless not essential to the type-identity of a given mental state, and may indeed vary from instance to instance of the same type of mental state' (1988, p. 39) Of course some would object, for example, that Martian pain couldn't be radically different in its intrinsic nature from our pain and still count as pain. But, the reply comes back, if it played the same functional role why wouldn't we call it pain? My point to both sides, however, is that the impossibility of intersubjective qualia comparison renders this debate undecidable, hence irrelevant. See Shoemaker 1991 for a defense of intrinsic quality compatible with functionalism and Dennett 1990 pp. 538-44 for arguments against the existence of intrinsic properties of qualia.

References

Block, Ned (1978), 'Troubles With Functionalism', Minnesota Studies in the Philosophy of Science, IX, C. Wade Savage, ed. (Minneapolis: University of Minnesota Press). Reprinted in Rosenthal 1991, pp. 211-228.

Churchland, Paul M. (1988), Matter and Consciousness (Cambridge, MA and London: MIT Press).

Churchland, Paul M. (1989), A Neurocomputational Perspective (Cambridge, MA and London: MIT Press).

Cole, David (1994), 'Thought and Qualia', Minds and Machines, 4, pp. 283-302.

Crick, Francis (1994), interview, Journal of Consciousness Studies, I, pp. 10-17.

Davies, M. and Humphreys, G.W. eds. (1993), Consciousness (Oxford and Cambridge, MA: Blackwell).

Dennett, Daniel (1987), The Intentional Stance (Cambridge, MA and London: MIT Press).

Dennett, Daniel (1990), 'Quining Qualia', in Consciousness and Contemporary Science, A. Marcel and E. Bisiach, eds. (Oxford: Oxford University Press). Reprinted in Lycan 1990 pp. 519-47.

Flanagan, Owen (1992), Consciousness Reconsidered (Cambridge, MA and London: MIT Press).

Jackson, Frank (1982), 'Epiphenomenal Qualia', Philosophical Quarterly, 32, pp. 127-36.

Levine, Joseph (1993), 'On Leaving Out What It's Like', in Davies and Humphreys 1993, pp. 121-36.

Lloyd, Daniel (1992), 'Toward an identity theory of consciousness', Behavioral and Brain Sciences, 15, pp. 215-6.

Lycan, William G. (1987), Consciousness (Cambridge, MA and London: MIT Press).

Lycan, William G. ed. (1990), Mind and Cognition (Cambridge, MA: Blackwell).

Nagel, Thomas (1974), 'What Is It Like to Be a Bat?', The Philosophical Review, LXXXIII, 4, pp. 435-50.

Nagel, Thomas (1986), The View From Nowhere (New York and Oxford: Oxford University Press).

Rosenthal, David M. ed. (1991), The Nature of Mind (New York: Oxford).

Searle, John (1980), 'Minds, Brains, and Programs', Behavioral and Brain Sciences, 3, pp. 417- 457.

Shoemaker, Sydney (1991), 'Qualia and Consciousness', Mind, C, pp. 507-24.

Van Gulick, Robert (1980), 'Functionalism, Information and Content', Nature and System 2. Reprinted in Lycan 1990, pp.107-129.

Van Gulick, Robert (1993), 'Understanding the Phenomenal Mind: Are We All Just Armadillos?', in Davies and Humphreys 1993, pp. 137-154.