Consciousness and Cognition##

Consciousness and Cognition[*]

David J. Chalmers

Department of Philosophy
University of Arizona
Tucson, AZ 85721

*[[I wrote this paper in January of 1990, but did not publish it because I was never entirely happy with it. My ideas on consciousness were in a state of flux, ultimately evolving into those represented in my book The Conscious Mind (Oxford University Press, 1996). I now think that some parts of this paper are unsatisfactory, especially the positive theory outlined at the end, although a successor to that theory is laid out in the book. Nevertheless, I think the paper raises issues that need to be addressed.]]


The problem of consciousness is perhaps the largest outstanding obstacle in our quest to scientifically understand reality. The science of physics is not yet complete, but it is well-understood. The science of biology has explained away most of the mysteries surrounding the nature of life. Where there are gaps in our understanding of these fields, the gaps do not seem intractable; we at least have some idea of the direction in which solutions might lie. In the science of mind, things are not quite so rosy. Much progress is being made in the study of cognition, but consciousness itself is as much of a problem as it ever was.

The term "consciousness" often serves as a catch-all for all that is mysterious about mentality. When using the term, one must therefore be careful not to collapse important distinctions. The most important distinction in the study of consciousness bears upon the approach we take to studying it: we may take either the first-person or the third-person approach. We might say that the third-person approach treats consciousness as a scientific problem, while the first-person approach treats it as a metaphysical problem. These two different viewpoints that we can adopt lead to very different treatments of the phenomenon, each of which can supply crucial insights. But the reconciliation of the viewpoints seems to be a difficult matter. Frequently, proponents of the two approaches seem to be talking past each other. In this paper, I will try to explicate carefully the relation between the two approaches, and to argue that they are not so irreconcilable as they might seem.

The third-person approach

The third-person approach has much to recommend it. On this view, consciousness is treated as a problem of science, just like heat, life, or nuclear physics, and amenable to the same methods of inquiry. Researchers in the new discipline of cognitive science have had much success in taking this approach to the study of mental processes in general. It is natural to hope that it might shed some light on the problems of consciousness, too.

The raw materials of cognitive science are much the same as those of any science - data gathered from external observation. These data can take a number of different forms. The most obvious kind of observable data are behavioral, and the study of human behavior is concentrated in the field of psychology. Another source of data, less accessible than behavior but extremely useful for the third-person approach, is observation of brain structure and function. The field of neuroscience is beginning to have significant success in explaining how the brain supports cognition. A third, less direct method in cognitive science is that of cognitive modeling - the construction of models, usually computational, that cohere to some extent with behavioral or neurophysiological data. The field of artificial intelligence is concerned with producing such models, and researchers in psychology and neuroscience often use computational models in their work.

This tripartite investigation of behavior, brain function and cognitive models has led to a significant increase in our understanding of diverse aspects of cognition, such as vision, memory and language comprehension. It is not surprising that these third-person methods might also be used to investigate consciousness. Such an approach to consciousness is urged by Dennett (1978; 1991), who goes so far as to sketch cognitive models for the phenomenon. Another third-person approach has been taken by Jackendoff (1987) (although Jackendoff explicitly recognizes that a third-person approach will not necessarily tell the whole story). Besides these philosophers, some psychologists (e.g., Johnson-Laird 1983) and computer scientists (e.g., Hofstadter 1979) have tried their hand at reining in the mysteries of consciousness with third-person accounts.

At the heart of the third-person approach is the philosophical position of functionalism. Roughly stated, this is the view that the correct way to understand mental processes is to perform causal analyses, revealing the abstract causal structure behind brain function. Such causal structure can be understood in objective terms, and duplicated in quite different materials (including, perhaps, in computers). Central to this approach is the view of mind as system. Insofar as this view addresses consciousness, it takes it to be a particular aspect of the function of a complex system; for example, it might be taken to be that process whereby a system is able to scan its own processing. The view of mind as system is certainly attractive, and it is a view that I share. The question is whether this approach taken alone is enough to explain the deepest mysteries of consciousness.

The first-person approach

Despite the attractiveness of the third-person approach, there is a feeling that it is side-stepping the really hard problems. The truly difficult questions only seem to arise when we take the first-person approach - when we consider, as Nagel famously put it, what it is like to be who we are, and what it might be like to be something quite different: another human, a bat, or a computational system. It is with these subjective questions that the deepest questions arise. Why is being me like anything at all? And why is it the way that it is? These questions are the real content of the problem of consciousness. There may be other senses of the term "consciousness", but in the sense in which it is most commonly used, to refer to the real mysteries of mentality, it is these first-person questions that are being raised.

It is still easy to fall into confusion or to equivocate when talking of "consciousness", so here I will divide the first-person problem into three parts: the problems of sensory qualia, subjective mental content, and the existence of subjective experience.

(1) The problem of sensory qualia

Qualia are the qualitative aspects of our mental states, most obviously of our sensations. The paradigm qualia are those of color sensations; other favorites are the taste of chocolate, the sound of middle C, pleasure and pain. All of these are poorly understood. When we look at a red patch, this sets off a particular pattern of neural firings in our brain. Why should this physical process be accompanied by a rich, subjective sensation? Given that it is accompanied by a sensation, why is it this sort of sensation (the red sort) rather than that sort (the green sort)? There are two issues here: why qualia exist at all, and why particular qualia accompany particular processes. Is the correspondence of qualia to processes arbitrary, or is there some systematicity that we do not understand?

Jackson (1982) has provided the most recent reminder of the qualia mystery, with a sharpening of the argument of Nagel (1974) before him. A future scientist, living in a time when neuroscience is completely understood, might learn everything there is to know about physical brain-processes. But if she has lived all her life in a black-and-white room, she will still not know what it is like to see red; when she sees red for the first time, she will learn something. It seems that the third-person approach, at least as currently understood, cannot tell us about the nature of qualia.

(2) The problem of subjective mental content.

When I think about a lion, something takes place in my subjective experience that has something to do with lions. Again, a straight physical account gives no reason to believe that such an experience should take place. What should a pattern of neural firings have to do with lions? But somehow, my thoughts are about something; they have subjective mental content. It is easy to make attributions of mental content to a system, justified perhaps by causal relations with the external world, but for subjective mental content we need something stronger. We need brain-states to carry intrinsic content, independent of our systems of external attribution; there must be a natural (in the strongest sense, i.e., defined by nature) mapping from physical state to content.

The problem of subjective mental content is not entirely different in kind from that of sensory qualia - the experience of content is itself qualitative, in a way. The main difference is that sensory qualia usually arise during external perception, whereas this sort of mental content arises during thought. (There is also a third-person problem of mental content, which has been raging for years, centering on the question of how we can assign propositional attitudes, such as beliefs and desires concerning the world, to systems and persons. In some ways this is an easier problem, as it may rely on human-defined systems of attributions; these contents may have the status of theoretical entities, rather than states that are presented to us directly. In other ways, the first-person problem is easier, as it may not have to deal with the problem of reference. When I think of a lion, my phenomenology bears some relation to a lion, but the relationship seems more like shared pattern than reference.)

(3) The existence of subjective experience

The two items above are concerned with the nature of our subjective states - why they are one way rather than another. But it is just as deep a problem why subjective states should exist in the first place. Why should it be like anything to be me? If I did not know that subjective states existed, it would seem unreasonable to postulate them. This is perhaps the deepest question of all, and no current theory has come close to dealing with it.

Not many people believe in zombies - humans with normal behavior but without any subjective mental states. These may be logically possible, but it seems implausible that there could be such things in the actual world. At least some people believe there could be functional zombies, however: beings which duplicate the functional organization of humans, perhaps computationally, without being conscious at all (e.g. Searle 1980, Block 1980). The question "what sort of entities can be subjects of experience?" is of great popular interest. For example, is every entity exhibiting intelligent behavior conscious? Could an appropriately programmed computer be conscious? I will argue, by combining first-person and third-person considerations, that the possible existence of functional zombies is implausible.

There are some other commonly-raised first-person problems not explicitly listed above. The problem of self-consciousness (or self-awareness) I take to be a subset of the problem of awareness, of which the difficult aspects are covered by (1) and (2). The problem of personal identity is a separate issue and a very deep one; but Parfit's exhaustive analysis (1986), which combines the first-person and third-person approaches to great effect, gives reason to believe that our first-person intuitions here may be mistaken. It is the three problems listed above that seem to be the residual content of the traditional mind-body problem. I will be using the term "consciousness" broadly to cover the phenomena of all of these problems. If you prefer, replace every occurrence of "consciousness" with "the subjective experience of qualia and mental content."

One difficulty with talking about first-person problems is that for every first-person (conscious) mental state, there is a corresponding third-person (functionally definable) mental state. (Perhaps there are not two different mental states, but simply two different ways of viewing one state; this is unclear. In any event, it is uncontroversial that for every subjective mental event there is a corresponding physical event; the physical event may be viewed via functional abstraction as a third-person mental event). For every subjective sensation there corresponds an objectively characterizable perception. This dichotomy in (ways of looking at) mental states makes things a little confusing, but it will be useful later on.

The relationship between the approaches

Of course, the first-person approach and the third-person approach are not entirely independent of each other. Indeed, I will be arguing that much insight can come through considering them not singly, but juxtaposed. Before doing this, it is necessary to enumerate a few facts about the two approaches.

(1) The third-person approach is sufficient, in principle, to yield a complete explanation of human behavior.
This follows from the explanatory completeness of physical laws - a hypothesis that is relatively uncontroversial. Physical phenomena have physical explanations, and behavior is a physical phenomenon. There is no reason to believe that human behavior is any different in this regard from hurricanes, rivers or stars. With a complete understanding of the laws of physics, and a good understanding of the way physical entities combine into systems, behavior will be understood.

This is an extremely strong conclusion, of course, and it goes a long way towards explaining the dominance of the third-person approach to the study of mentality. Cognitive science is doing its best to break down the removable barrier implied by the "in principle", and there is no reason to believe that within a few centuries it should not succeed. If it does not, it will probably be because of problems of complexity rather than any deep metaphysical barrier. We might trivially restate (1) as

(1a) The first-person approach is not needed to explain anything about human behavior.
Of course, the first-person approach may well prove to be a useful short-cut in explaining behavior, and a powerful tool, but the point is that it is not necessary - it is in principle dispensable, as far as behavior is concerned.
(2) The third-person approach (as currently conceived) is not sufficient to explain all first-person phenomena.
Statements (1) and (1a) make a powerful argument for the primacy of the third-person approach. Nevertheless, this approach cannot tell the whole story. Consciousness, the subjective experience of qualia and mental content, is simply not explained by a third-person account. No amount of neuroscience and cognitive modeling explains the qualitative nature of a sensation of red, or even why such a subjective sensation should exist.

This statement is surely the most natural view, but it is not undisputed. Those who have spoken the loudest in dispute of (2) include Dennett (1982, 1988, 1991) and P. M. Churchland (1986). I believe their arguments to be flawed, but will not argue the issue here; it has been eloquently argued by others. Those who disbelieve (2) may choose to stop reading here, then, in the belief that consciousness is not really a special problem. They might however read on, to find an account of first-person problems which is motivated by third-person issues at every point, or at worst to find a reductio ad absurdum of the first-person position. The remainder - those who believe that the mind-body problem is still a real problem - may continue undisturbed.

It should be noted that (2) refers to the third-person approach as currently conceived - that is, the study of the brain as a physical system with the usual physical ontology, abetted by cognitive models and functional analyses. I would not like to rule out the possibility of an eventual "objective phenomenology", to use Nagel's term, but it is difficult to see what it would look like. At the least, new constructs would be needed, and our ontology would need to be expanded. To get from the physical and the functional to the subjective, we would need metaphysical bridging principles. It does not seem impossible, however, that these could be stated from a third-person viewpoint. The beginnings of such an approach will be outlined later in this paper. In the meantime, we can restate (2) as

(2a) The third-person approach is not sufficient, in the absence of extra metaphysical bridging principles, to explain all first-person phenomena.
The need for such a bridge is put nicely by Jackendoff (1987), who says that computational/functional accounts of mind may have solved the Mind-Body Problem, but what we need to do now is solve the "Mind-Mind Problem". This would provide a bridge from the "computational mind" to the "phenomenological mind" (in the terms I have used, from third-person mental events to first-person mental events). That computational and functional accounts may have already got us halfway there is a point that will resurface later.

(3) Our claims about consciousness are a fact of human behavior.
This is trivially true. (If you interpret "claim" as an irreducibly subjective term, then replace it by "utterance" or something more neutral.) When I say "I'm conscious, and I'm totally baffled by it," that is a fact of human behavior. When somebody writes "The ineffable sensation of red is inexplicable by physical premises", that is a fact of human behavior. This is harmless enough, until combined with (1) to yield

(4) The third-person approach is in principle sufficient to explain our claims about consciousness
or worse,

(4a) The first-person approach is not needed to explain our claims about consciousness.

Everything that we say about consciousness is, in principle, amenable to the usual kind of physical/functional analysis. There is no need to appeal to any mysterious metaphysical constructs to explain the things we say. On this account, the last 2000 years of debate over the Mind-Body Problem would have gone exactly the same in the absence of subjective experience (if such a thing were possible). In particular, our sense of bafflement about the first-person problems manifests itself in numerous third-person ways - the things we say, the things we write, even the things we think, if regarded in terms of neural processes - and is thus amenable to the methods of cognitive science[*]. The correct cognitive-science explanation of why we claim to be conscious might be a little while coming, but there should be no deep metaphysical problems. This leads us to an important principle:

*[[McGinn (1989) requests an explanation of why it is so difficult to clearly articulate the the problems of consciousness. I believe that the reason is closely tied up with (4) and (4a). Everything we can say about consciousness is physically caused, and somehow, at the back of our minds, we are aware of this. So when the words come out of our mouths, we are aware of how easy it might be to explain our claims without invoking any great mysteries. Further, our words can easily be taken as referring to third-person mental states: perception rather than sensation, and so on. Our claims seem very inadequate to do justice to the real phenomenon.]]

The Mystery Principle: Consciousness is mysterious. Claims about consciousness are not.
This can equivalently be stated as

The Surprise Principle: Although consciousness should surprise us, claims about consciousness should not.

Most people will agree that consciousness is a surprising phenomenon. If it were not for the fact that first-person experience was a brute fact presented to us, there would seem to be no reason to predict its existence. All it does is make things more complicated. By contrast, the things we say about consciousness are common-or-garden cognitive phenomena. Somebody who knew enough about brain structure would be able to immediately predict the likelihood of utterances such as "I feel conscious, in a way that it seems no physical object could be", or even Descartes' "Cogito ergo sum". In short, our bafflement about consciousness can be understood purely as a problem for cognitive science. There are very rich pickings awaiting anybody who takes this path. I believe that it is not as difficult a path as it might seem, and towards the end of this paper I will give a very brief account of why we might expect such bafflement.[*]

*[[There are other reasons why claims about consciousness should not surprise us. It seems plausible that any system which (in a purely functional sense) "perceives" aspects of the world and further has sophisticated linguistic capacity might profess perplexity about the nature of its perception; one might expect puzzlement at both the "qualitative" and "indexical" aspects. And if such a system has the capacity to "think about what it is thinking about" (a functional notion, once again), it might very well claim full-blown self-consciousness. No metaphysics needs to be invoked in our analysis of such a situation. (Computer A: "I know I'm just a collection of digital circuits, but I can't explain this strange feeling I have of having thoughts of my own." Computer B: "Yes, I'm inclined towards dualism myself.") This point deserves further development.]]

The situation with which we are now faced seems to border on the absurd. If everything we say, write and even think about consciousness is explainable by cognitive science, why is there a need to posit any great mystery about consciousness in the first place? Why not concede that the appearance of great metaphysical mystery is simply an illusion, and regard consciousness itself as a cognitive process, on much the same level as memory or learning? It becomes a third-person high-level term, playing a similar role in cognitive science as "heat" does in thermodynamics. As for this weird "first-person experience", then insofar as it is not a purely cognitive phenomenon, it simply does not exist. This is the approach taken by Dennett and others, and there are powerful arguments in its favor, as evidenced above. But this paper is based on the premise that the mind-body problem is still a real problem. We are not yet prepared to concede that (2) is false, and that the third-person approach as currently conceived is sufficient to explain all of the first-person mysteries. As first-person loyalists, we have to bite this bullet and struggle on, trying to produce a coherent account of the matter. I believe that the pressures on a first-person account produced by (4) and (4a), and the third-person constraints that are thus acquired, are of great value in shaping a solution that might be acceptable to partisans of both the first-person and third-person approaches.

This is the low point for the first-person partisan. This is as bad as it gets. But we have to confront the problem posed by (4) and (4a) head on. This has rarely been done in the literature. "Qualia freaks" (Jackson's term) have generally been content to argue for the autonomy of first- person phenomena, from independent considerations, or more frequently to simply assume it. But if we do not take (4) and (4a) seriously, then any account that we produce is open to question. The question which faces us, put slightly differently, is

(4b) How can first-person experience have any bearing on what we say about first-person experience?
or more simply

(4c) How can consciousness have any bearing on what we say about consciousness?

I believe that in the proper understanding of this question lies the seed of a solution to the mind-body problem. All the same, it is surprising how rarely the question has been considered in the literature. (The more general question of how the mental can affect the physical of course a perennial favorite. Question (4c) is much more specific, and raises a different set of problems.) It has been touched on occasionally in various forms, however, with various different reactions. The choices in answering the question spread themselves out as follows.

(A) We might reject premise (1), and deny that the third-person approach can even give a full account of human behavior. This is the approach taken by Elitzur (1989), who argues from the fact that we talk about consciousness to the conclusion that consciousness plays an active role in determining human behavior, and deduces that the laws of physics must therefore play an incomplete role in determining behavior. This is certainly an interesting argument, but it is a implausible and should be regarded as a last resort. The laws of physics are too important to us to be given up so easily. (A similar account of the matter would presumably be given by interactionist dualists from Descartes to Eccles.)

(B) We might answer (4c) by saying, simply, "It doesn't." This is the answer that might be given by Jackson (1982), who argues that qualia are completely epiphenomenal, playing no causal role (although he does not explicitly consider utterances about consciousness). Now, epiphenomenal qualia are already unparsimonious under general considerations, but considered in the context of question (4c) their implausibility increases. Are we to say that our claims about consciousness are completely independent of the fact of our consciousness? That any similarity manifested between our claims and the actual facts are mere coincidence? While this conclusion cannot be completely ruled out, it certainly seems inelegant. It fails what I will call, later, the "Coherence Test" - its explanations of (1) consciousness and (2) claims about consciousness fail completely to cohere with each other.

I think there may nevertheless be a grain of truth in epiphenomenalism, but it needs to be spelt out more carefully. If we can explicate just how consciousness is an epiphenomenon of physical processing, then question (4c) may admit a more parsimonious answer. I will be defending a kind of "double aspect" theory that bears some similarity to epiphenomenalism, while at the same time being more palatable, I hope, to those with a naturalistic view of the world.

(C) We may deny premise (2), and argue that in fact the third-person approach tells us everything there is to know. By denying that consciousness is essentially a first-person phenomenon, we deny that question (4c) poses any problem. If there is anything at all called "consciousness" that plays a causal role, then it does so in exactly the same way that centers of gravity play a causal role in physics, or that temperature plays a causal role in thermodynamics: as a convenient third-person abstraction. This is presumably the approach that the likes of Dennett and Churchland would take.

The consideration of questions like (4c) has been used as an explicit argument for the third-person approach in a few places. For example, Foss (1989), arguing against the Nagel/Jackson arguments, argues that a "Super Neuroscientist" could in principle know everything that a being would say about its qualia-experience, and everything that it might say. From this, he draws the conclusion that the Super Neuroscientist would know everything that there is to know about qualia.

Clearly questions like (4c) are a goldmine for third-person arguments, but these analyses seem too simplistic. They "solve" the Mind-Body Problem by denying that there is a problem. This runs counter to our premise that there is a substantial problem to be solved. Nevertheless, I believe there is a grain of truth in these arguments. There is perhaps some sense in which qualia are physical. There is perhaps some sense in which consciousness is understandable from the third-person viewpoint. What must be explicated is the nature of this sense. The arguments above do not solve the Mind-Body Problem, but they do indicate that it might be soluble. However, there is more work to be done. A flat statement of physicalism is in no sense a complete solution. We need to know how physicalism could be true. The nature of the work that remains to be done will be outlined shortly.

Before we leave this section, we should note that a well-known argument of Shoemaker (1975) bears some resemblance to the arguments in (C). Shoemaker argues against the possibility of "Absent Qualia", or functional zombies. Some have argued that this possibility refutes functionalism, by demonstrating that we might duplicate any given functional process without any accompanying subjective experience. Shoemaker responds by arguing that we know about qualia; our knowledge is caused by the existence of these qualia; therefore qualia play some causal role. The conclusion is that Absent Qualia are impossible, as removing qualia would change the causal structure. This is generally construed as an argument for the third-person approach (for functionalism, in particular). Nevertheless, I am sympathetic with this argument and with functionalism, and believe that Absent Qualia are impossible. However, what needs to be explained is why Absent Qualia are impossible, and how qualia could play a causal role. This question is not addressed by functionalism. Again, more work remains to be done. Both functionalism and epiphenomenalism have plausible aspects, but neither gives a complete account of the problems of consciousness. I will suggest that a correct theory of consciousness shares aspects of both functionalism and epiphenomenalism, and that the correct answer to question (4c) lies somewhere between (B) and (C).

The Coherence Test

The Coherence Test is a test that any completed theory of mind must pass. It is motivated directly by questions like (4c).

The Coherence Test:
A completed theory of mind must provide
(C1) An account of why we are conscious.
(C2) An account of why we claim to be conscious; why we think we are conscious.
Further (C3), accounts (C1) and (C2) must cohere with each other.

It should be noted that (C1) is a completely first-person matter; it involves answering the metaphysical question of how subjective experience is possible. (C2), on the other hand, is understandable on strictly third-person terms - it is, in principle, a matter for cognitive science. (If, again, in your vocabulary "think" and "claim" are irreducibly subjective terms, then replace them by appropriate third-person terms that deal with behavior and brain function.) Another way in which to phrase (C2) might be "An account of why consciousness seems so strange to us." Viewed appropriately, this is also in principle understandable by a cognitive science approach. Anyway, it is the coherence condition (C3) that is the most important here. This ensures that completely independent accounts under (C1) and (C2) will be ruled out; that the accounts must bear a substantial relationship to each other. This provides a vital link between the first-person and third-person approaches.

Another way of putting the Coherence Test is that we need accounts of
(C1a) The mind that we experience (a first-person question); and
(C2a) The mind that the system perceives (a third-person question).

One would hope that these "minds", on any completed account, would bear a very close relationship to each other. If the relationship is not strict identity, it should at least be a very close correspondence.

It is not possible to "prove" that a correct theory of consciousness must pass the Coherence Test, but it certainly seems extremely plausible. While it is logically possible that an account of why we are conscious could be completely independent of a cognitive account of why we believe we are conscious, it seems inelegant and unlikely. If the Coherence Test in fact fails for a correct completed theory of mind, then we are living in a world where the things we say about consciousness and the Mind-Body Problem bear at best a coincidental relation to the way these things actually are. Faced with the possibility of this complete unreliability of our verbal reports, it would probably be better to join the third-person camp and regard the mysterious nature of "subjective experience" as a mere illusion.

Given that we accept the Coherence Test, a correct third-person account of cognitive processes puts severe constraints on possible first-person accounts, and possibly vice versa. Of course, we have not explicitly defined the notion of coherence. Instead, I think it is better to go ahead and apply it on a case-by-case basis, and let some understanding of the notion emerge.

Applying the Coherence Test

Functionalism is a controversial doctrine when applied to first-person aspects of mentality. But for the third-person study of mental processes, it reigns supreme. Few people these days doubt that the correct way to analyze human behavior and to explain the workings of the brain is in functional terms. All remotely satisfactory accounts of learning or memory, say, explain these via causal analyses of systems with many interacting parts. Indeed, it is difficult to see what other kind of account there could be. So when it comes to explaining the things we say (and even believe) about consciousness, there is little doubt that a functional analysis will be the way to go. We may enshrine this doctrine as

(5) The fact that we claim to be conscious holds in virtue of certain functionally specifiable properties of the brain.

Of course, we are not sure exactly what those functional properties are, yet - this is part of the reason why functionalism alone is not a completed theory of consciousness. But we are almost certain that the properties will be functional ones, considered at the most parsimonious level. And there is no doubt that if we make the correct functional abstraction from the brain, then duplicating this functional specifications will produce the same kind of claims about consciousness.

We may now use this uncontroversial claim, together with the Coherence Test, as an argument against certain theories of mind.

Theory 1: We are conscious in virtue of low-level biochemical processes.

Such a view of mind is held by at least a few writers on the subject (e.g. Searle 1980, Block 1980). These writers have generally been arguing against functionalism as a theory of mind, or at least as a theory of the first-person aspects. Of course, almost everybody agrees that neurophysiological processes give rise to consciousness in the human case; but that view alone is compatible with functionalism, as these processes might give rise to consciousness precisely in virtue of their functional properties. Proponents of the present view take a stronger stand, suggesting that biochemical properties are directly responsible for consciousness. On this view, only a system with the relevant biochemical properties could be conscious; a silicon isomorph might lack consciousness entirely. Block, for example, claims that qualia might be a consequence of physiological facts which need not appear in a functional account. Searle thinks it very plausible that the "causal powers" of the brain (meaning those powers sufficient to cause consciousness) lie at the biochemical level.

This view, however, runs a great risk of failing the Coherence Test. If, as according to (5), our claims of consciousness are consequences of certain functional properties of the brain, then the statement that consciousness is inherently neurophysiological suggests that reason we make the claims is largely independent of the reason we are conscious. This would seem strange. A possible reply by neurophysiological partisans might be to claim that the neurophysiological locus of consciousness is specifiable functionally, albeit at a very low level. The connectionist movement in cognitive modeling has recently made popular the idea that a correct functional account may inhere at a low level of abstraction. But if we accept that neurophysiological accounts can also be functional, then this is no longer an argument against functionalism per se - a consequence that Block and Searle would be loath to accept. Insofar as our theory holds that the sources of consciousness are not specifiable functionally, then it is a theory that fails the Coherence Test. If consciousness does not arise from the functional, then our claims are independent of consciousness.

In reply to a related argument by Shoemaker (1975), Block (1980) suggests that qualia can be independent of any given functional account but still make a causal difference. But such a difference will by definition not bear on any aspect of our functional account. Given that our account is specifically constructed to include our claims of consciousness as a relevant aspect, this reply fails. If qualia are independent of this functional account, then qualia cannot make a difference to our claims. Horgan (1984) makes a similar move in defending the claim that qualia are neurophysiological. Considering the case of Martians functionally identical to us but physiologically different, he claims that the issue of whether or not they have qualia might be resolved by observing whether they manifest the same kind of puzzlement about qualia that we do. But of course they will manifest similar puzzlement, as this is determined by their functional architecture. Therefore such observation will tell us nothing more than we already knew.

For similar reasons, even if we hold that the underpinnings of consciousness are specifiable functionally, it may be dangerous to go to too low a level. It may well be that when an account of the processes which lead to our consciousness-claims is found, it will be specifiable at a level higher than the biochemical. If we insist that consciousness itself inheres at a lower level than this, then the source of consciousness is independent of the source of our consciousness-claims, and our theory fails the Coherence Test. (It may be possible that certain aspects of qualia are determined at a lower level than the level of our qualia-claims. Such aspects could not make any causal difference relevant to our claims. For instance, the difference between red-sensations and green-sensations is very difficult to articulate, and so the precise nature of these sensations might be dependent on certain low-level facts. But the existence of such qualia must be dependent on the higher-level functional account, if the correct theory of mind is Coherent.)

We may conclude that theories which place the source of consciousness at too low a level are implausible. A similar argument will apply to any theory which does not satisfy Functional Supervenience: the principle that replicating certain relevant functional processes, even in a different substrate, is sufficient to guarantee first-person mental states. (This is a weaker principle than the Functional State Identity Theory, which holds that mental states are functional states.) If consciousness exists in virtue of some fact which is quite independent of any functional account, then our claims of consciousness do not reflect the fact of our consciousness. This argument parallels Shoemaker's argument against Absent Qualia in some ways, although unlike Shoemaker I do not wish to argue for the truth of the Functional State Identity Theory. But the argument establishes that if there could exist beings which duplicate our relevant causal structure (perhaps in a quite different substrate), but which do not possess subjective states - in other words, if there could exist functional zombies - then the correct theory of mind is fundamentally Incoherent.[*]

*[[It should be noted that the implausibility of functional zombies does not necessarily imply the impossibility of computational zombies. It just might be possible to be a functionalist without believing that computational specifications can ever capture the appropriate causal structure. (Harnad (1989), for instance, believes that the requisite function is inherently analog.) Church's Thesis diminishes the plausibility of such a position - it seems likely that a non-computational causal account always supervenes on a lower-level computational account - and in practice functionalism and computationalism tend to stand and fall together.]]

Theory 2: Epiphenomenalism.

On the face of it, this view does not come close to satisfying the Coherence Test. It holds that we are conscious in virtue of certain dualistic principles which yield mental states as a consequence of physical states. Such principles would seem to be quite independent of the relatively straightforward functional sources of our claims that we have "qualia" and "subjectivity". All parsimony is thrown out the window. The facts that we (a) have qualia, and (b) claim to have qualia, are related only by coincidence.

I do not think that all forms of epiphenomenalism are ruled out, however. It might be possible to have a principled epiphenomenalism, where the mysterious "dualistic principles" above are replaced by something more concrete, showing that mental states arise in virtue of specific, functionally specifiable physical states. If this were done, then it is possible that the Coherence Test might after all be satisfied: we would be conscious and claim consciousness in virtue of similar functional accounts. But all that this possibility shows, for now, is that epiphenomenalism would need a great deal of development to be a plausible theory of mind.

Theory 3: Identity Theories, Functional and Physical

I have been rather sympathetic to functionalism in the preceding discussion, but it is not close to a completed theory of mind. It performs well on part (C2) of the Coherence Test. The source of our consciousness-claims is undoubtedly functionally specifiable. True, the flat statement of functionalism gives no insight into just which functional properties of the brain are responsible for our consciousness-claims, so there is more work to be done on that point - but we might set that aside as a mere technical difficulty, a matter for cognitive scientists. A more profound difficulty with functionalism is that it does not come close to dealing with (C1). It has often been noted (recently by Searle 1989) that functionalism is not a theory which is motivated by the existence of first-person mentality; rather, conscious mental states are viewed as an obstacle for functionalism to deal with (witness all the Absent Qualia and Inverted Spectrum objections). It is possible that functionalism is compatible with a correct theory of consciousness, but taken alone it is not that theory.

Functionalism gives very little insight into why consciousness exists. It is one thing to baldly state "subjective experience arises in virtue of our functional organization"; it is another to give an account of why. Functionalism, in this way, is not unlike its predecessor, the Identity Theory, which stated "mental states are brain states". The flat statement alone gives us no insight into how it could be true. There was, of course, some kernel of truth to the Identity Theory: this was that conscious mental states are indeed supervenient on brain states (that is, reproducing a brain state will reproduce the same corresponding mental state). In a similar way, Functionalism contains a kernel of truth: it is very plausible that conscious mental states supervene on certain functional states. By going from the loose notion of "physical state" to the more specific notion of functional state, functionalism has brought us closer to the goal, but it certainly has not achieved it.

So both functionalism and the Identity Theory pass over the key question of just why we are conscious in the first place. And of course, the fact that it fails on (C1) implies automatic failure on (C3) - without giving an account of the existence of subjectivity, it cannot give an account of coherence. Nevertheless, I believe that functionalism is a valuable theory, of which many aspects are worth retaining. Something simply needs to be added. The account that I will give below might be characterized as "functionalism plus something".

To satisfy the Coherence Test, and thus be a candidate for the title of "completed theory of mind", a theory must provide three things, on my estimation:

(1) A metaphysical account of how subjective experience is possible. This will almost certainly include some new metaphysical construct, over and above physical laws.

(2) A functional account of why we think and claim that we are conscious. This will presumably be in the idiom of cognitive science.

(3) An account of functional-metaphysical coherence, which shows how (1) and (2) cohere with each other.

As we have seen, theories that consciousness is biochemical pass the buck on (2) and fail miserably on (3). Epiphenomenalism makes a start on (1), but as it stands has no chance on (3). And functionalism, as currently conceived, does not deal at all with (1), and therefore with (3). I hope that I have demonstrated just how far short currently conceived theories fall of yielding a solution to the Mind-Body Problem. The mode will now switch from pessimism to optimism. The three requirements above sound daunting, but I believe that they might be satisfied. In the next section, I will outline the beginnings of a speculative theory which has a chance of satisfying (1), (2) and (3) above. This theory alone will not be a solution to the Mind-Body Problem, but I hope that it points in the right direction.

Mind, pattern, information

The theory I am going to suggest is an example of what is known as a "double aspect" theory. That is, it holds that first-person mental states and third-person mental states are two different aspects of the same thing. Double aspect theories have been attractive to others (e.g. Nagel 1986), but they have usually foundered because we cannot say exactly what those aspects might be . The simple idea "double aspect theory" explains very little. But before I go on to elaborate on this particular theory, it might be useful to say a couple of words about why double aspect theories are so attractive. In short, this is because double aspect theories are at the same time almost epiphenomenalism and almost identity theories, combining the virtues of both with the vices of neither. The theory I will propose is almost an identity theory (by this term I include both "Brain State" and "Functional State" Identity Theories, though the proposed theory is nearer to the latter than the former), in that it holds that first-person and third-person states are the same thing - but different aspects of the same thing, a crucial difference. (You might say: an identity theory that takes subjective states seriously.) It is almost a version of epiphenomenalism, as we can imagine the subjective aspects "hanging off" the non-subjective aspects, allowing the complete autonomy of the physical - while at the same time allowing that subjective states can be causally efficacious, as they are but another aspect of the objective states. (You might say: an epiphenomenalism where mind can matter.) The account which follows is brief, speculative, poorly-defined and incomplete. Nevertheless, I hope that it offers a glimpse of what might be the correct direction for the understanding of consciousness.

The two different "aspects" that I propose are pattern and information. Wherever they occur, pattern and information occur together. All information is carried by some pattern in the physical world; all patterns carry some information. I speculate that they should be regarded as two different aspects of the same thing, which for want of a better term we may call "pattern/information."

My proposal is that third-person (objectively analyzable) mental events are patterns in the brain, while the corresponding subjective (first-person) mental events are information. On the above view of pattern/information, this implies that first-person and third-person mental events are indeed two different aspects of the same thing, the underlying pattern/information. (This has the advantage of not expanding our ontology too far.)

The idea that third-person mental events can be regarded as patterns should not be too controversial. A commitment very much like this is already made by functionalism, though it is not always emphasized. When we give a functional account of a system, we are necessarily abstracting away from many superficial details, and highlighting particular patterns in the physical substrate. When the functionalist says that a mental state is a functional state, she is committing herself to the notion that a mental state is an abstraction - and every abstraction is an abstraction of a pattern. (Of course, not every pattern is a functional pattern, but these are a very important subset.)

It is rarely made clear just what ontological claims a functionalist would wish to make, though. Are they committed to patterns as part of their natural ontology? They might not necessarily have to take this step, as if we are considering mental events only from the third-person viewpoint, then it is not clear that we have to admit them into our ontology anymore than we have to admit centers of gravity - both might be "convenient fictions", in Dennett's terminology. But I am prepared to bite the bullet, and reify patterns. Given that we believe that the mind and conscious experience are part of the basic ontology of things, then some concession is necessarily -- we have to posit some metaphysical construction of the kind mentioned in the previous section. Admitting patterns into our ontology is a surprisingly painless way of doing this - and once it is done, we get minds for free.

Of course, we have done a little more than reify pattern: we have reified pattern/information. I claim that this is a reasonable step, given that they always go together. Put it this way: if there is any chance that a double aspect theory might be the right theory - and as we have seen and will see, there is very much that is attractive about them - then it is hard to imagine two more natural candidates for the "aspects" than pattern and information. To say that "mental events arise from a double aspect" is to say very little, unless we want to admit the mental as a primitive member of our ontology. Instead, by positing pattern/information, we have a much more natural primitive in our ontology, from which the double-aspect nature of mentality is inherited.

Once we have made this posit, then conscious mentality falls out. Third-person mental events are patterns in the brain: the corresponding conscious mental events are the information that these patterns carry. Qualia are just information. (A nice way of putting this is "information is what pattern is like from the inside". I am not quite sure what this means, but it certainly sounds good.) Anyway, conscious mentality arises from the one big pattern that I am. That pattern, at any given time, carries a lot of information - that information is my conscious experience. Incidentally, given any physical substrate there are many different ways of abstracting patterns from it. In my brain, there will be some patterns that do not correspond to mental states at all, and some which correspond to unconscious states. This is not a problem for the theory; there will be more on this later.

Why we think we are conscious

The last few paragraphs may strike the reader as wanton ontological extravagance. But apart from any inherent elegance the theory might have, it is also motivated by the fact that it can pass the Coherence Test. In fact, the above is at least in part motivated by an functional account of why we think we are conscious, and why consciousness seems strange. Logically, the functional account ought to come first, but it makes sense to present the metaphysical theory up front. But what follows can be read independently of what went before. This will be a purely functional account, in the tradition of cognitive science. So at least temporarily, all the metaphysical baggage may be thrown away.

Very briefly, here is what I believe to be the correct account of why we think we are conscious, and why it seems like a mystery. The basic notion is that of pattern processing. This is one of the things that the brain does best. It can take raw physical data, usually from the environment but even from the brain itself, and extract patterns from these. In particular, it can discriminate on the basis of patterns. The original patterns are in the environment, but they are transformed on their path through neural circuits, until they are represented as quite different patterns in the cerebral cortex. This process can also be represented as information flow (not surprisingly), from the environment into the brain. The key point is that once the information flow has reached the central processing portions for the brain, further brain function is not sensitive to the original raw data, but only to the pattern (to the information!) which is embodied in the neural structure.

Consider color perception, for instance. Originally, a spectral envelope of light-wavelengths impinges upon our eyes. Immediately, some distinctions are collapsed, and some pattern is processed. Three different kinds of cones abstract out information about how much light is present in various overlapping wavelength-ranges. This information travels down the optic nerve (as a physical pattern, of course), where it gets further transformed by neural processing into an abstraction about how much intensity is present on what we call the red-green, yellow-blue, and achromatic scales. What happens after this is poorly-understood, but there is no doubt that by the time the central processing region is reached, the pattern is very much transformed, and the information that remains is only an abstraction of certain aspects of the original data.

Anyway, here is why color perception seems strange. In terms of further processing, we are sensitive not to the original data, not even directly to the physical structure of the neural system, but only to the patterns which the system embodies, to the information it contains. It is a matter of access. When our linguistic system (to be homuncular about things) wants to make verbal reports, it cannot get access to the original data; it does not even have direct access to neural structure. It is sensitive only to pattern. Thus, we know that we can make distinctions between certain wavelength distributions, but we do not know how we do it. We've lost access to the original wavelengths - we certainly cannot say "yes, that patch is saturated with 500-600 nm reflections". And we do not have access to our neural structure, so we cannot say "yes, that's a 50 Hz spiking frequency". It is a distinction that we are able to make, but only on the basis of pattern. We can merely say "Yes, that looks different from that." When asked "How are they different?", all we can say is "Well, that one's red, and that one's green". We have access to nothing more - we can simply make raw distinctions based on pattern - and it seems very strange.

So this is why conscious experience seems strange. We are able to make distinctions, but we have direct access neither to the sources of those distinctions, or to how we make the distinctions. The distinctions are based purely on the information that is processed. Incidentally, it seems that the more abstract the information-processing - that is, the more that distinctions are collapsed, and information recoded - the stranger the conscious experience seems. Shape- perception, for instance, strikes us as relatively non-strange; the visual system is extremely good at preserving shape information through its neural pathways. Color and taste are strange indeed, and the processing of both seems to involve a considerable amount of recoding.

The story for "internal perception" is exactly the same. When we reflect on our thoughts, information makes its way from one part of the brain to another, and perhaps eventually to our speech center. It is to only certain abstract features of brain structure that the process is sensitive. (One might imagine that if somehow reflection could be sensitive to every last detail of brain structure, it would seem very different.) Again, we can perceive only via pattern, via information. The brute, seemingly non-concrete distinctions thus entailed are extremely difficult for us to understand, and to articulate. That is why consciousness seems strange, and that is why the debate over the Mind-Body Problem has raged for thousands of years.

The above account is far too brief, and is almost certainly wrong in many important aspects. Nevertheless, I believe there is a kernel of truth in there. Even if the above account needs to be thoroughly revised, I think this fact will remain: the facts that we think we are conscious, we claim to be conscious, and consciousness seems strange all hold true in virtue of the fact that the brain is a sophisticated pattern-processor (is a sophisticated information-processor).

Further, this account should not surprise us. Things have to be this way. Any being which "perceives" the world must do so in virtue of pattern/information processing. Once the processing has reached the heart of the cognitive system, further processing can only be sensitive to the information distinctions embodied there. When the being tries to articulate the nature of its perception, it will be reduced to talk of "qualitative distinctions". Similarly, if a being has any reflective access to its internal processing, it will only be in terms of high-level patterns. We should expect such beings to be baffled by this "consciousness". The Surprise Principle is vindicated: Consciousness should surprise us, claims about consciousness should not.

The above account, of course, did not rely on any ontological commitment to pattern and information - it merely used them as "convenient fictions." This was a completely functional account. If I were a third-person aficionado, who believed that the Mind-Body Problem is only a pseudo-problem, I might stop now, and say "See! Consciousness is just an illusion." But I am not, and this account was intended not to explain away consciousness but to cohere with it.

So now, we may run the Coherence Test. We claim that we are conscious in virtue of the brain's ability as a pattern-processor (as an information-processor). We in fact are conscious in virtue of the patterns (information) embodied by the brain. These explanations seem to cohere rather well. Of course, there is no pattern-processing without patterns. It is precisely the same patterns which are at the crux of the pattern processing (those at the heart of our processing system, from which all distinctions are made), which are also the patterns which are our mental states. (I occasionally use "pattern" as a shorthand for "pattern/ information". No great harm is done.) One could not ask for better coherence than this. The reason we talk about mental states is that our processing is sensitive to precisely those patterns which are our mental states.

Another, slightly over-simplified way to put this, is: (1) The brain perceives itself as pattern. (2) The mind (that we experience) is pattern. Therefore (3) The brain perceives itself as the mind.

Patterns in pattern-processors

On this account, there are two criteria for being a conscious entity: a metaphysical (first-person) and a functional (third-person) criterion. The metaphysical criterion is that one must be a pattern. All patterns exist, presumably. It would seem strange to reify some patterns but not others. But not all patterns are conscious. To be a conscious pattern, one must be part of the right kind of pattern processor, bearing an appropriate relation to it.

An interesting and difficult question concerns the status of patterns which are not part of pattern processors. Such patterns may still carry information; is there anything it is like to be them? My answer is a tentative "yes". But it would not resemble what it is like to be a human being, for such patterns are not conscious, in the appropriate functional sense. They are not parts of pattern-processors, so they cannot be aware. This incidentally seems to demonstrate that we should separate the purely first-person notion of "be-ability" from the partially third-person notion of "consciousness". Consciousness itself is dependent on certain functional criteria, but it seems implausible that such third-person criteria should impose restrictions on metaphysical be-ability. Being the number 5 might be like something; if only like being asleep, without the excitement of dreaming.

On the other hand, there are pattern processors which are not embodied in human brains. Might such processors (or the patterns therein) have subjective experience? A connectionist network, for instance, is a pattern processor par excellence. A typical feed-forward network might have served as a perfect simplified example in the above account of the kind of pattern/information-processing that goes on in the human brain. Networks are sensitive to certain patterns in environmental data; they recode these patterns as they pass down the information-processing chain, until output is sensitive only to a certain kind of pattern in the inputs. Do connectionist networks have qualia? My answer is "yes!" This may seem very counter-intuitive at first, but I believe that upon reflection it becomes plausible. There is no principled distinction between the kind of pattern-processing performed by such a network and that performed by a brain, except one of complexity. And there is no reason to believe that complexity should make such a great difference to the existence of qualia. Certainly human qualia will be more complex and interesting, reflecting the more complex processing; but then, even considering our processing of "redness", there is no evidence that much more complex processing goes into this than goes into a typical large connectionist network. Pattern-processing leads to qualia in humans, and there seems no reason to deny that pattern-processing leads to qualia in networks also. (Of course networks cannot as yet reflect on the fact that they have qualia, let alone talk about it, but this is not the point.)

Then there is that old favorite, the thermostat. Do thermostats have qualia? It is not entirely clear that thermostats do the right kind of pattern-processing, but if they do, their qualia are remarkably simple. For thermostats process all physical inputs down to three states: too hot, too cold, and just right. It is not clear just what three-valued qualia would be like. But if we are ever to understand qualia, this is a good test case.

Connectionist networks help illustrate another point: that patterns may supervene on functional descriptions. The relevant pattern in networks are patterns of activation over a number of units, where units and their activations are of course functionally specifiable. It may not be the case that all patterns supervene on functional descriptions (that is, are specifiable relative to functional descriptions, and duplicating the functional system in another substrate allows duplicating the pattern), but it seems plausible that all patterns that represent first-person mental states should. The plausibility lies in the fact that patterns need to be processed for us to be aware of them, and processing is a functional notion. The current account is thus compatible with functionalism of a certain variety. You might say that it takes functionalism as a starting point, and adds what is necessary to deal with the problems of consciousness.[*]

*[[No theory of mind worth its salt would pass up the chance to attack Searle's "Chinese Room" problem. For our theory, the explanation is straightforward. When we have a homunculus manipulating symbols on paper, there are two quite distinct sets of patterns: patterns carried by the head of the homunculus, and patterns carried by the complex system of symbols on pattern. The patterns in the paper may well support their own mind. Even if we assume the homunculus could internalize the rules (which would require vastly more memory than any human has), then there would still be two sets of quite distinct patterns, both of which are present in the homunculus's head. It is a fundamental fact that there can be many different patterns present in a given substrate. Thus, in a single head, there might be two quite distinct phenomenologies, without any overlapping first-person mental states.]]

Loose ends

Moving away from the more cognitive aspects to the more metaphysical aspects, we are yet to explicitly answer our original question (4c): how can consciousness have any bearing on what we say about consciousness? On the present account, this is easy. Conscious experience is identified with the "information" aspect of certain pattern/information states. This information can certainly make a difference. It is just another aspect of the pattern, and there is no question that the pattern plays a causal role. Changing the information changes the pattern, and changing the pattern changes many consequences. In fact, as we have seen, our claims about consciousness reflect precisely that pattern/information in which our mental states consist. Consciousness is thus causally efficacious. At the same time, it is quite possible to analyze third-person mental states and thus behavior without ever needing to invoke consciousness, if we so desire. It is possible to cast everything in terms of pattern, without invoking information.

This answer corresponds to none of the answers (A), (B) or (C) given originally. Our options were in fact not quite so limited as they seemed. This again shows the advantages of a double aspect theory - an identity theory which takes consciousness seriously.

It might be noted that this is not quite a double aspect theory in the traditional sense, where the two aspects are the physical and the mental. This theory holds instead that the two aspects are first-person and third-person mental states - both of them mental, though in different senses. We are able to use the resources of theories like functionalism to get us past the first stage, from the physical to third-person mental states, which may be functional states, computational states, or some other patterns. All that is left is Jackendoff's "Mind-Mind" Problem - the bridge from third-person patterns (the "Computational Mind") to first-person mental states (the "Phenomenological Mind"). This is precisely what our posited pattern/information duality achieves.

Finally, it must be glaringly obvious that I have said very little about a very important topic: the relationship between pattern and information. Is this merely a primitive posited relation, or can it somehow be explicated? Some might claim that a given pattern can be interpreted as carrying any information you like. I believe that this is using a different sense of the word "information" to mine. The kind of information I have dealt with here is not "information that" - it does not need to refer. The kind of information I believe in is intrinsic. In this sense, there is a natural mapping from pattern to information.

Exactly what the nature of this mapping is is very difficult to say. This is why the current theory is not yet a solution to the Mind-Body Problem. It may be the case that the pattern/information relation is a brute fact, in much the same way that the laws of physics are brute facts. We have a certain amount of knowledge about their nomological connection, via our subjective experience. This may or may not be enough to understand all the facts perfectly. If it is not, then it may be the case that we can never know precisely what it is like to be an alien being (or a connectionist network), as we will not understand precisely the nature of the information that their patterns carry. The outlook is not necessarily so pessimistic, though, and perhaps a better analysis of the nature of pattern and information will tell us everything there is to know. (If pressed for an informal opinion, I would guess that at least nine tenths of the pattern/information relationship will be understandable analytically, with the last tenth up for grabs. If we are not able to understand precisely what it is like to be a bat, we will be able to come very close.)

This account has not removed the Mind-Body Problem. Nevertheless, my minimal hope is that it shows what the beginnings of a solution might look like. My maximal hope is that it has removed much of the confusion surrounding the problem, and localized the mystery at one key, primitive locus: the relationship between pattern and information.


Block, N. 1980. Are absent qualia impossible? Philosophical Review 89:257-74.

Churchland, P.M. 1985. Reduction, qualia and the direct introspection of brain states. Journal of Philosophy 82:8-28.

Dennett, D.C. 1982. How to study human consciousness empirically, or, nothing comes to mind. Synthese 53:159-80.

Dennett, D.C. 1978. Toward a cognitive theory of consciousness. Minnesota Studies in the Philosophy of Science, Vol. 9. Reprinted in Brainstorms (MIT Press, 1978).

Dennett, D.C. 1988. Quining qualia. In (A. Marcel and E. Bisiach, eds.) Consciousness in Contemporary Science. Oxford: Oxford University Press.

Elitzur, A. 1989. Consciousness and the incompleteness of the physical explanation of behavior. Journal Of Mind and Behavior 10:1-20.

Foss, J. 1989. On the logic of what it is like to be a conscious subject. Australasian Journal of Philosophy 67: 205-220.

Harnad, S.R. 1989. Minds, machines and Searle. Journal of Theoretical and Experimental Artificial Intelligence 1:5-26.

Hofstadter, D.R. 1979. Gödel, Escher, Bach. New York: Basic Books.

Horgan, T. 1984. Functionalism, qualia, and the inverted spectrum. Philosophy and Phenomenological Research 44: 453-469.

Jackendoff, R.S. 1987. Consciousness and the Computational Mind. Cambridge, MA: MIT Press.

Jackson, F.C. 1982. Epiphenomenal qualia. Philosophical Quarterly 32: 127-136.

Johnson-Laird, P.N. 1983. A computational analysis of consciousness. Cognition and Brain Theory 6: 499-508.

McGinn, C. 1989. Can We Solve The Mind-Body Problem? Mind 98:349-366.

Nagel, T. 1974. What is it like to be a bat? Philosophical Review 83:435-450.

Nagel, T. 1986. The View From Nowhere. Oxford: Oxford University Press.

Parfit, D.A. 1984. Reasons and Persons. Oxford:Oxford University Press.

Rumelhart, D.E., McClelland, J.R., and the PDP Research Group 1986. Parallel Distributed Processing. Cambridge, MA: MIT Press.

Searle, J.R. 1980. Minds, brains and programs. Behavioral and Brain Sciences 3:417-424.

Searle, J.R. 1989. Consciousness, unconsciousness, and intentionality. Philosophical Topics 17:193-209.

Shoemaker, S. 1975. Functionalism and qualia. Philosophical Studies 27:271-315.