Wise incomprehension

It is very hard to devise a theory of consciousness that is not open to decisive objection. This is not because consciousness is so amorphously ill-defined that anything goes and we find it impossible to choose among a plethora of options. Rather, no matter what theory we come up with, it always seems to run into some shattering difficulty. The problem of consciousness is like a chess game in which a series of forced moves always ends in checkmate, more or less humiliating. Sometimes it seems that the best we can hope for is some teetering ad hoc contrivance that just manages to evade outright refutation - for the moment at least. Philosophy is like that, we know; but with consciousness the constraints are especially tight.

David Chalmers's book is an attempt to develop a theory that escapes knockdown refutation, while tolerating some counterintuitive and uncomfortable features. The book is very well argued, thorough, sophisticated, honest, stimulating - and almost plausible. It is certainly one of the best discussions of consciousness in existence, both as an advanced text and as an introduction to the issues. One feels that Chalmers has done about as good a job as could be done on this most intractable of problems. That said, I do not think the position he defends ultimately works, and for reasons that are not surprising. Still, there is much to be gained by following his argument: checkmate, yet again, but an impressive game nonetheless.

The book has two central theses, one negative, the other positive. The negative thesis is that materialism is false, because the mental is not logically supervenient on the physical. The mental is not explained and necessitated by the physical in the way that the observable macroproperties of water are explained and necessitated by the molecular structure of water. Since facts about consciousness are not entailed by physical facts, the former are something over and above the latter. This is argued to follow from the conceivability of zombies - entities physically just like us but without any consciousness: since these are logically possible, the physical facts alone cannot conceptually guarantee the presence of a conscious life. We cannot then come to know anything about conscious experience itself just from knowing all the physical truths of the universe; nor, a fortiori, is it possible to analyse experience in physical or functional terms. Experience is irreducible. It follows that dualism of some form must be true.

The positive thesis is that this dualism consists in fundamental laws that connect physical and mental properties by mere natural (not logical) necessity. We cannot reductively explain experience in physical or functional terms, but we can suppose there to be a contingent empirical law-like connection between them. This is nomological dualism instead of the rejected reductive monism. The physical does indeed "give rise" to the phenomenal, but it does so only with the force of natural necessity. Experience is thus a basic feature of the universe, like space and time, tacked on (as it were) to the swarms of particles that constitute matter.

In addition to these two main theses Chalmers speculates that the notion of information might provide some sort of link between the mental and the physical. Since the concept of information he employs is correlative with the notion of causation (the Shannon-Weaver concept of selection among possibilities), it turns out that experience is ubiquitous in the world - which leads Chalmers to endorse a version of panpsychism. Thermometers can now boast consciousness of some primitive form, a result Chalmers declares himself willing to live with. He also ingeniously defends a version of functionalism that makes experiences lawfully correlated with (but not reducible to) computational-functional properties. The argument here turns on the implausibility of dissociating qualia from the subject's first-person access to them, as would have to be so if experience could float free of a subject's cognitive processing.

There are two large problems with the theory as presented. The first, which Chalmers fully acknowledges, is that epiphenomenalism about experiences is entailed. Since my zombie and me share our physical and functional constitution, nothing in our behaviour differs, so that the doings of both of us can be explained without ascribing conscious states to either of us - yet I have them and he does not. In particular, we make the same judgements - including, for example, "I am conscious and currently having a red experience" - despite the vast difference in respect of conscious experience. But now it follows that my utterance of this is not explained by what makes the judgement true, since my zombie's utterance cannot be so explained - it being false in his case. My experience thus turns out to be epiphenomenal with respect to my self-ascriptions of experience. Chalmers himself spells out this consequence and tries his best to draw its sting; but he is clear that it would be better if it could be avoided, and he does not succeed in removing the attendant air of paradox. What needs to be noted is that it is the denial of logical supervenience that leads directly to epiphenomenalism; so we need to be very sure that this denial is compulsory.

The second problem, which he nowhere confronts, is that just as the alleged conceptual contingency of the link between the physical and the mental leads to the logical possibility of zombies, so also does it lead to the logical possiblity of disembodied consciousness. For if the link is merely that of natural necessity, then there are possible worlds in which the laws are abrogated - which means that the correlated properties could be instantiated independently of each other. There are pure spirit worlds as well as zombie worlds! I do not know whether this consequence would alarm Chalmers, but I suspect it would - and rightly so. How would such disembodied experiences be connected to the rest of nature? What might their causal powers depend on? How could they have any dynamic role in anyone's psychology? Where would they come from? The trouble is that once the psychophysical link is loosened to mere natural necessity the ontology of mind comes out looking pretty radically Cartesian.

Both problems have a common source: the denial of logical supervenience. It is therefore extremely important that this denial be shown to be undeniable. Chalmers is aware of this and argues that putative notions of a posteriori supervenience, in which there is no conceptual entailment from one level to the other, will not provide a viable alternative. Only logical supervenience can block the conceivability argument to the possibility of zombies. I find him quite convincing on this, but he underestimates how pressing it is to find some way to defend strong metaphysical supervenience, in view of the problems that arise from denying it. The crucial question here is whether all forms of logical supervenience must be epistemically transparent to us. Must our present concepts allow us to appreciate the nature of the supervenience relations that constitute the psychophysical link? Might we not instead be confronted by a case of opaque logical supervenience? If that were so, then there would exist concepts of both the physical and the experiential, and of whatever relations might connect them, such that there is an a priori explanatory connection between those concepts - even though they are not concepts we do or even could grasp. The conceptual dependencies would go outside of the circle of concepts we bring to bear in thinking about mind and body. Indeed, these concepts cannot be within our grasp or else it would be plainly inconceivable to us that zombies are logically possible. In other words, zombies seem possible to us only and precisely because we do not grasp the concepts that render them impossible. There is logical supervenience after all, but it is hidden to our epistemic faculties.

This is surely a coherent position, and it provides an alternative to the other relations Chalmers mentions. In fact, he does briefly discuss something like this at one point, correctly attributing it to me. But he does not see how serious are the consequences of rejecting it, since it seems to be the only viable way to avoid the twin problems of epiphenomenalism and disembodiment, while accepting that we cannot reduce experience to physical properties.

It is not dogmatic materialism that prompts insistence on strong supervenience but the need to escape the two problems cited. Indeed, the thesis of opaque logical supervenience is not materialist at all, if that means that the terms of current or foreseeable physics are adequate to explain consciousness. The view is actually quite compatible with theories that regard the physical as itself just the appearance of some deeper currently unconceived reality - or with idealism for that matter. Of course, the view assumes that we do not know the concepts that are necessary for a satisfying explanation of conciousness; what it does is use this fact to explain why it is that we can be misled into denying logical supervenience, with all the problems that stem from this.

It helps here not to be too wedded to the old framework of "materialism" versus "dualism". Both notions assume that materialism is a useful well-defined doctrine, but it is not, since the notion of the "material" is entirely theory-relative. We do not want to limit our theoretical concepts to those of current physics, but if we make the notion more inclusive it comes to include anything that might be relevant in explaining what happens in the world. There are really a lot of properties that might be identified and used in explanations of consciousness. Perhaps because he sticks to the old materialism-dualism dichotomy, Chalmers finds it hard to imagine how there could be concepts that transcend those now used in physics or commonsense psychology, and hence finds the idea of opaque logical necessitation difficult to accept. The first order of business here is not to declare materialism false, but to question its very significance.

The speculations on information and panpsychism are admitted to be a bit on the wild side, but the problems go beyond mere incredibility. Not only do we see no evidence in nature of the experiential properties allegedly associated with every causal process; it is also not the case that physics finds any need to postulate such properties in explaining the behaviour of matter. If all matter has experiential properties, should not this be relevant to the correct science of matter? Yet there seems no gap in the physics of the inanimate that calls for the ascription of mental properties to things. These alleged properties make no difference to the way a rock falls or water flows or any other purely physical interaction. The only motivation for invoking them is in order to provide an explanatory account of consciousness; they are idle otherwise. Subtract them from that thermometer and you will not observe any change in its behaviour.

Chalmers's defence of a weak form of functionalism uses some intriguing thought experiments, but the conclusion that there is a lawlike relation between functional properties and consciousness is too weak to be of much interest. We might equally claim that there is also a law-like relation between experiences and underlying neural states: if you keep the latter constant you will always get, as a matter of law, the same experiences. No asymmetry is established between the functional and the neural if law-like dependence is all that is asserted; so it is wrong to suppose that any interesting form of functionalism has been established. All we have is a three-way law-like relation between the mental, the neural and the functional.

The only way to avoid being checkmated by consciousness is to assume you do not understand it. Chalmers has done his level best to understand consciousness, but the result, despite its many merits, shows the wisdom of incomprehension.

Colin McGinn is professor of philosophy, Rutgers University.