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

How must consciousness supervene on the physical to validate the materialist view that all mental properties, including conscious phenomenal ones, derive from underlying physical structure? It would seem at the least that no two beings with the same underlying physical causal structure and organization could differ in any mental property (leaving aside all the qualifications and nuances that would have to be built in to accommodate the "wide" or contextual dimension of mental states and their contents.)¹

However neo-dualists, like David Chalmers,² argue that more is needed since causal supervenience could meet that condition yet not suffice for materialism. It's compatible with property dualism as long as there are nomic connections linking physical and mental properties. According to the property dualist, mental properties are distinct from physical properties in the same way that electromagnetic and gravitational forces are distinct. Just as we can have lawlike links between fundamental physical properties, so too the neo-dualist claims we can have lawlike relations among fundamental physical and mental properties. Given the existence of such natural laws, it may be nomically impossible to have a mental difference without a physical difference. If so, systems globally alike in all physical respects must as a matter of nomic necessity be alike in all mental respects.

The neo-dualist contends that materialism is nonetheless false since mental properties do not logically supervene on physical properties but only nomically supervene on them. There are logically possible worlds, he alleges, that are just like ours in all physical respects but differ mentally (especially with regard to conscious mental states). It is these logical possibilities that establish the distinctness of mental properties; if mental properties were merely a special type of physical property then it would be logically impossible for them to vary while holding all physical properties constant, but since they can vary (or so the neo-dualist claims) they can not be physical properties of any kind. At most they are a distinct class of nonphysical properties that are tied to physical properties by nomically necessary but not logically necessary links.

I agree that mere nomic supervenience would not by itself secure materialism. Thus the key issue becomes the sort(s) of supervenience that can be shown to hold or not to hold between the mental and the physical. Is there a form of supervenience that the dualist can show not to hold that will establish his/her position? Or alternatively is there a form of supervenience that the materialist can show does hold that demonstrates his/her view? And how can one determine which dependencies and possibilities there are?

The burden of proof will vary with the specific claim is at issue. To prove the truth of materialism, one would need to show that any imagined world that purports to separate the mental from the physical in fact embodies a contradiction and violates the law of identity. One and the same thing can not at the same time both be and not be; if phenomenal consciousness is just a complex physical property, then its existence can not be distinct or independent from the totality of the physical facts. The materialist's aim may be, however, more modest: simply to challenge various alleged possibilities with which the dualist claims to refute materialism. His burden in that case is far less; he need show only that the dualist lacks an adequate basis for regarding the alleged possibilities as genuinely free from contradiction. The mere fact that none is obvious will not suffice, and so the focus shifts to how well the dualist can defend his purported possibility.

Dualists since Descartes ³ have used tried to use conceivability arguments to show the nonmateriality of mind. Descartes claimed that he could clearly and distinctly imagine his conscious mind continuing to exist in the total absence of his physical body. He thus concluded that the two could not be one and same. In recent years imagined zombie cases have been called upon to play a similar argumentative role. Chalmers ⁴ for example asks us to imagine a world W2 satisfying the following conditions:

1. There are beings who are molecule-for-molecule duplicates of conscious humans in the actual world,

2. Our molecule-for-molecule doppelgangers in W2 are not conscious, and

3. The physical laws that hold in W2 are the same as those that hold in the actual world. On the basis of this claim about what he can conceive Chalmers concludes that such molecule-for-molecule zombies are logically possible.

The standard reply distinguishes between two senses of "conceivability," a weak or subjective sense and a strong or objective sense. Arguments such as the zombie one rely on two premises or claims: first a claim that the existence of one sort of property without the other is indeed conceivable, and second the linking proposition that what is conceivable is logically possible. For example,

P1. I can conceive of a without b (e.g., zombie world)
P2 Whatever is conceivable is logically possible or more specifically
P2* If I can conceive of a without b then it is logically possible for a to exist without b.
_________________________
C3 Therefore it is logically possible for a to exist without b.
P4 If it is logically possible for a to exist without b, then a is distinct from b (not identical with b)
_________________________
C5 a is distinct from b (not identical with b)

Such an argument must either equivocate on its use of the word conceivable or fail to support at least one of its first two premises. To make (P1) plausible, one needs to read it in the weak subjective sense, as meaning merely that imagining a without b, (or any other mental difference without a physical difference), does not seem to involve a contradiction, that is, merely that we can tell ourselves what seems to be a coherent story. But while this makes P1 plausible, it undercuts the second step (P2), which links conceivability to logical possibility. That an imagined state of affairs does not seem to involve a contradiction does not entail that it in fact involves none. If we opt for the stronger or objective notion of conceivability, which requires that there not be any contradiction (as opposed to merely seeming such,) the link to logical possibility goes through directly. But then the earlier claim that one can conceive of a without b (or the mental without the physical) collapses into question-begging. Thus if read "conceivable" strongly we lose P1, if we read weakly we lose P2, and if we equivocate the inference to C3 is invalid. Thus the argument is unsound.

This line of criticism is often illustrated with familiar scientific examples such as those of water and H₂O or heat and molecular kinetic energy. The fact that a chemically ignorant person might conceive (in the weak or subjective sense) of water existing without H₂O shows nothing about their nonidentity, and because water is H₂O-as established a posteriori-it is not logically possible for the one to exist without the other.

Chalmers is fully aware of these objections but believes his own arguments escape such criticism.⁵ First it is worth noting that he uses "logical possibility" with a sense that would rehabilitate the move to C3 while leaving the argument unsound overall. As he uses it, "logical possibility" means little more than conceivability, what he calls "a priori possibility" and which concerns only what we can infer from the concepts involved by a priori means. If "logical possibility" is read in that way, P2 becomes more or less a tautology even on the weak reading of "conceivable," and the move to C3 is automatic. The original problem, however recurs at the next step. If the "logical possibility" of separation means merely its "a priori conceivability" then, as the scientific cases show, it does not entail distinctness, and the general argument fails.

Chalmers acknowledges this and, unlike some of his predecessors, does not rely on any general linking principle such as P4. Moreover he acknowledges that a priori conceivability (what he calls logical possibility) does not in itself entails what he calls "metaphysical possibility," the strong relation needed to license the move from possibility of separation to nonidentity. ⁶

In analyzing the familiar scientific examples, he carefully distinguishes between what he calls primary and secondary intensions associated with terms. The primary intension of the word water involves having various watery characteristics (color, liquidity, taste, smell, etc.). However, through its use in context by our linguistic group to refer to the stuff around here that has those characteristics, it acquires the secondary intension [H₂O]. Thus when someone in our linguistic community claims to be imagining a situation with water but without H₂O, he/she is misdescribing his/her state; when he/she uses "water' it means H₂O (secondary intension). Such a person is in fact imagining a situation in which some substance satisfies the primary intension associated with "water" without being water.

However, according to Chalmers the analogy breaks down in the consciousness/material properties case. The disanology is this: any-thing that satisfied the primary intension associated with "consciousness" would be a case of consciousness. Thus if one can conceive of that concept (primary intension) being satisfied in a world that lacks the materialist's proposed physical properties, then consciousness must be distinct from them. More importantly, if one can conceive of the primary intension associated with "consciousness" not being satisfied in a world that contains all the relevant physical properties, then consciousness cannot be identical with them nor logically supervenient on them.

To be precise, what Chalmers claims is just a bit weaker: namely that as a matter of a priori/logical possibility the phenomenal aspect associated with the primary intension can exist independently of any supposed physical basis and thus must be distinct from it. ⁷ But since that aspect involves the existence of phenomenal awareness, whether or not we identify it outright with consciousness matters little. What counts is the supposed logical possibility that phenomenal what-it's like-to-be-ness can exist or not exist independently of the physical facts.

However, even if successful, this reply would merely weaken the analogy with certain examples (e.g., water/ H₂O) that are standardly used to illustrate the problems with modal conceivability arguments; it would not in itself resolve those issues. Though some analogies are broken, other remain that may suffice to undermine the zombie thought experiment. And more importantly there are residual questions about the adequacy of the concepts that Chalmers and the neodualists use in their thought experiments. Are those concepts well enough developed and anchored to support the argumentative use to which they are put? I think not.

Consider the analogies kept and broken. There are allegedly two sorts of worlds. First those in which the primary intension of "consciousness," the phenomenal aspect, is satisfied despite the absence of the materialist's purported physical referent. These are worlds like those in which there is watery XYZ without any H₂O. This is where the disanalogy intrudes; although we can imagine something watery that is not water, we can't imagine something that involves phenomenal awareness without imagining what amounts to consciousness in the philosophically important sense. The primary intension of "water," refers to relatively superficial properties that demonstratively link us to its referent in context. But the primary intension of "consciousness" incorporates so much that its imagined satisfaction in the absence of its materialist counterpart shows the two to be distinct.

But this wouldn't suffice to refute materialism; it shows that consciousness can be realized by multiple substrates, but that's a materialist platitude. To get a dualist result, one needs to claim that one can imagine consciousness occurring in the absence of any physical substrate at all-a true Cartesian disembodied mind. Materialists typically object to such claims as begging the question. If being in a state of phenomenal awareness is in fact being in special sort of dynamically organized physical state, then when one imagines a world with phenomenal awareness, one imagines some physical substrate whether one recognizes that one is doing so or not. If "being-a-something-that-it's like to be" is a matter of being a certain sort of physical system that interacts with itself in a special way, then when one imagines the former, one imagines the latter whether one conceives of it as such or not. The materialist may be wrong, but to show him so one needs assumptions that don't prejudge the issue.

Imagined worlds of the second sort are even more problematic; these are the zombie worlds in which our exact physical duplicates completely lack consciousness. The materialist will reply that this is just as contradictory as imagining a world in which there is a substance that is physically and behaviorally just like the water in Lake Erie for some extended stretch of time, but that is not liquid. Obviously when one imagines the Lake Erie doppelganger, one in effect imagines its liquidity. Though it is not so obvious, the materialist claims we do the same when we imagine our physical duplicates; whether we realize it or not, in doing so we imagine conscious beings just as we imagine liquid lakes. Is he right? That remains an open question, but to assume he's wrong again seems to beg the question.

The neodualist will deny the parallel. With Lake Erie, we can reductively explain just why and how the underlying physical substrate satisfies all the functional conditions to count as liquid, but we can't do the same for consciousness. Indeed we can not at this time come close to doing so, but that need reflect only the poverty of our current theories and concepts. It shows we have a current explanatory gap, but no metaphysical implications follow. The fault may lie not in the world but in our concepts of it.

We come thus to the second and more general question about the dualist's thought experiments. Are the concepts that he uses adequate to the task to which he puts them? Is our concept of phenomenal consciousness or our concept of its possible physical basis well enough developed to allow us to draw metaphysical conclusions about their relation? As noted above, I think they are not.

Consider first another nonmental example. Imagine a mid-nineteenth-century vitalist who argues as follows:

1. I can conceive of creatures that are just like actual creatures (say actual cats) in all physical respects but that have no ability to reproduce.

2. Therefore the ability to reproduce does not logically supervene on a creature's physical structure.

With the benefit of late twentieth century science we know the vitalist's conclusion is dead wrong; the ability to reproduce does logically supervene on physical structure. More interestingly we can see diagnostically where the vitalist went wrong; he had neither an adequate concept of reproduction nor an adequate concept of the total physical structure of a living organism, and he also lacked an adequate theory of how the two might fit together. He had no idea of the way in which reproduction involves the replication and transfer of genetic information, and he had not the slightest grasp of what we now know to be the biochemical basis of that process, of how gentic information can be coded by sequences of DNA and RNA. What was conceivable from the vitalist's perspective shows little if anything about what is logically (or metaphysically) possible in matters reproductive and physical. The vitalist might have conjoined his concept of the total physical structure of a cat with the negation of his concept of the ability to reproduce without generating any a priori contradictions, but given the radical incompleteness of his concepts vis-à-vis the natures of the two phenomena to which he applied them, nothing really follows regarding what relations of logical (or metaphysical) possibility might hold between those phenomena themselves.

Are the concepts used by Chalmers and the neodualist more adequate than the vitalist's? First we must ask, "what specific concepts do they use?" On the mental side they use a concept of experience derived from our first person awareness of our own conscious mental states and processes. The nature of these concepts derives from their representational and regulatory role in the monitoring and control of our conscious mental life. They surely meet conditions of adequacy in relative to that role. But in what other respects might they be adequate or inadequate? For the neo-dualist's purposes his concept of conscious experience, call it CCE, must be adequate to limn the boundaries of consciousness; whatever satisfies CCE in any world must be an instance of consciousness and whatever fails to fall under CCE must equally fail to be an instance of consciousness. The neodualist treats his concept CCE as the universal (Protagorean) standard of consciousness, but that assumption seems less than plausible or at least problematic. Why should that concept of experience derived from its role in regulating our specific form of consciousness provide a characterization of consciousness sufficiently general to determine the logically possible boundaries of consciousness. Please note I am not saying that the neodualist assumes that all logically possible conscious experience must involve the same specific phenomenal properties associated with human consciousness; he most surely and rightly does not say that. But what he does implicitly assume, though more modest, is still upon reflection implausible, namely that the concepts of consciousness that we command on the basis of their application within our own self-awareness can provide us with a general means of delimiting the logically possible boundaries of consciousness.

Moreover, there is a further dimension of adequacy in which the neodualist's concept of consciousness is even less plausibly up to the task, one that is directly relevant to his thought experiment. The question is whether or not the neodualist's concept of consciousness is adequate for assessing whether or not and how it might be instantiated by a material substrate. Recall the vitalist's concept of reproduction, which failed to include the idea of information transfer. Given the incompleteness of the vitalist's concept it is not surprising that he could not see how reproduction might be a fully material process. The situation, I believe is comparable with respect to the neodualist's concept of consciousness. Here too our current understanding of consciousness qua consciousness is far too incomplete to be able to say with any confidence how it might or might not be physically realized.

This conceptual inadequacy on the mental side is compounded by the fact that the concepts that the neodualist invokes on the physical side of his conceivability experiment are terribly nonspecific and even less adequate to the task at hand. Indeed the physical concepts that are used are little more than dummy concepts, such as "beings like actual conscious humans in all physical respects." No real detailed concepts that might characterize the physical substrate of consciousness are given. It's akin to the vitalist's blanket claim about the possibility of physically identical creatures without the ability to reproduce. The vitalist really knew nothing in any specific way about what the physical substrate of inheritance might be; he knew nothing of biochemistry, nothing of nucleic acids, or their structure or operation within the cell.

The vitalist had no remotely adequate or detailed concept of the physical basis of reproduction. He lacked an adequate conception of what reproduction required. And thus he couldn't understand how to fit together the two inadequately conceptualized pieces of his puzzle. Ditto for the neodualist.

Chalmers explicitly denies that there is any valid analogy between his conceivability argument and vitalist arguments for antimaterialism, though he does not consider any parallels drawn in the way I have done just above, that is, he does not consider an analogy based on parallel conceptual inadequacy. Nonetheless we should consider his basis for denying that there is a valid analogy. He argues that all the phenomena that the vitalist saw as beyond physical explanation were functional in nature; according to Chalmers what the vitalist couldn't see was how any physical structures or processes could perform the requisite function. To refute the vitalist we needed only to show how they could in fact do so.

Chalmers, however, denies that the same might happen with respect to consciousness and the physical because on his reckoning-and this is a key claim-consciousness is not a fundamentally functional process. He admits consciousness and conscious mental states perform various functional roles, but he denies that an account of such roles can explain what consciousness is, that is, he denies we can analyze or explicate the nature of consciousness in functional terms. To support his claim, he considers various functional roles that have been associated with or proposed for consciousness-such as information processing roles or behavior regulating roles-and argues that in each case we can conceive or imagine that role being filled in the complete absence of phenomenal consciousness; that is, he offers a variety of absent qualia arguments each directed at one or another functional role that has been or might be proposed for consciousness. But here again we need to ask about the adequacy of the concepts that Chalmers uses to reach his general antifunctionalist conclusion. I agree that consciousness does not logically supervene on the specific functional phenomena that he considers, but that reflects negatively only on the specific inadequate functionalist proposals he considers not on the prospects for functionally explicating consciousness per se.

The functionalist has his own modest embarrassments. In the absence of a detailed functionalist account of consciousness, he is left in a position like that of a mid-nineteenth-century materialist trying to reply to his vitalist critic. He has a wide explanatory gap to bridge and needs to write a hefty IOU. But carrying the burden of a large promissory note is not the same as having been refuted. And that as yet the functionalist has not been, nor has the materialist-zombie thought experiments not with standing. translated by Elizabeth Haldane and

Notes

1.On the contextual nature of the mental see for example, Burge (1979).

2. Chalmers (1996).

3. As in Descartes (1972) originally published in 1642.

4. Chalmers (1996) p. 94.

5. Chalmers (1996), chapter 3.

6. Chalmers (1996), chapter 4.

7. I am here indebted to personal communication from David Chalmers by email, May 1998.

References