SENSATIONS AND GRAIN PROCESSES

SENSATIONS AND GRAIN PROCESSES^*

George Graham, University of Alabama at Birmingham

Terry Horgan, University of Memphis

Introduction.

This paper celebrates an anniversary, or near anniversary. As we write it is just more than 40 years since U. T. Place's “Is consciousness a brain process?” appeared in the British Journal of Psychology, and just less than 40 since J. J. C. Smart's “Sensations and brain processes” appeared, in its first version, in The Philosophical Review (Place 1962/1956, Smart 1962/1959). These two papers arguably founded contemporary philosophy of mind. They defined its central preoccupation (the ontology of consciousness), introduced its regnant ontology (materialism/physicalism), offered its initial logical techniques (e.g. appeals to the concepts of identity and event) as well as empirical reliances (on neuroscience), and, finally, offered its most seminal sectarian doctrine (central state materialism). No history of philosophy of mind can afford to neglect them. (See for discussion Macdonald 1989.) It is to be near both the letter of Smart's paper and the spirit of his and Place's concern for how best to develop a philosophical understanding of consciousness that we entitle our own paper 'Sensations and grain processes'. (That is not a typo, as will be seen momentarily.)

In the years since Smart's and Place's contributions, the landscape of philosophy of mind has changed in many ways. The empirical reliances of very recent philosophy of mind have expanded to include the cognitive sciences (not just neuroscience); central state materialism alternately has been displaced by causal role and functional specification theories of mind; fresh logical techniques have been introduced (e.g. the concept of supervenience); and it is less clear than it was in the 1950's whether materialism is to be preferred to some more ecumenical ontology (such as naturalism).

This paper is about the current status of the philosophy of consciousness (which we take to be phenomenal consciousness, for purposes of the paper; of which more momentarily); not so much the case for materialism (which we take to be, for the most part, complicated by considerations which we shall adduce below) as what the philosophical program for doing the philosophy of the conscious mind is and where it can, and most importantly can't, rely on cognitive science. There is quite a lot of ground to cover in a short space. So let's begin by demarcating the subject matter and outlining the paper to follow.

Subject and Outline.

Phenomenal consciousness is the "what it's like" aspect of our mental lives. It is especially salient in bodily sensations and in certain perceptual experiences. Jackson (1982) gives as examples of phenomenal states (so-called qualia) "the hurtfulness of pains, the itchiness of itches, pangs of jealousy,... the characteristic experience of tasting a lemon, smelling a rose, hearing a loud noise, or seeing the sky" (p. 127). But phenomenal consciousness is more extensive in scope than this. As Bieri (1995) remarks, using the term "sensing" for mental states that have a phenomenal aspect, "Sensing comprises a variety of things: sensory experiences like seeing colours and hearing sounds; bodily sensations like lust and pain; emotions like fear and hatred; moods like melancholy and serenity; and finally, desires, drives and needs, i.e., our experienced will. All these states are not only there; rather, it is like something to be in them" (p. 47). The specific thisness of such states--"It is like this," as said or thought by someone while undergoing the state and attending to it from the subjective, first-person, point of view--is its phenomenal character.

Phenomenal consciousness strikes many philosophers as puzzling or mysterious, more so than other aspects of mentality (including other aspects that are often characterized using the language of consciousness, such as attention or second-order awareness of one's first-order mental states). Specific philosophical puzzles arise about its ontological or metaphysical status, given our overall scientific worldview; about its causal efficacy, if any; about how it is to be explained in naturalistic terms; and about why it should have emerged at all, in the course of evolution.

In Section 1 of this paper we describe an interdisciplinary research program in cognitive science that bears directly on the goal of achieving a scientific understanding of phenomenal consciousness: identifying what we call the 'causal grain' of phenomenal states, at both the neurophysical and the functional-representational levels of description. Scientific progress on this research program is foreseeable, and is directly relevant to philosophical issues about phenomenal consciousness; moreover, at present it is hard even to imagine other relevant kinds of scientific progress. In Section 2 we briefly elaborate the three main philosophical puzzles about phenomenal consciousness that have made it seem especially problematic to many philosophers--puzzles involving ontological status, causal role, and explainability. In Section 3, the heart of the paper, we argue that as far as one can now tell, from our current epistemic existential situation, even if the causal grain of phenomenal consciousness were to become fully understood within cognitive science, various theoretical options concerning qualia that are presently live theoretical options in philosophical discussion would all still remain live theoretical options. We conclude, in Section 4, with some methodological remarks about the status of competing philosophical approaches to phenomenal consciousness, in light of our argument in Section 3. With respect to evolutionary approaches specifically, there turn out to be certain important, tractable-looking, questions about the evolution of states with the distinctive causal grain associated with qualia--questions that can be posed and addressed independently of the competing philosophical positions about phenomenal consciousness.

1. The Causal Grain of Phenomenal Consciousness.

Here is a large-scale, long-term, but potentially empirically tractable project for cognitive science: to identify the specific causal role or roles associated with phenomenal states--to identify what, in some sense, those states do. We use 'associated with' and 'in some sense' in a deliberately neutral way, in order not to prejudge the issue whether these states themselves, as opposed to other states that co-occur with them (e.g., representational states, or neurophysical states), have causal efficacy; that is one of the philosophical issues to be discussed below. Our usage also is not meant to prejudge certain identity questions, e.g. whether states of phenomenal consciousness are literally identical with "associated" representational or neurophysical states.

Let us call this scientific program the grain project, since it involves investigating the causal roles associated with phenomenal consciousness at several levels of detail or resolution. One level is common-sense psychology (so-called folk psychology). This is the surface or coarse grain, so to speak; it includes various familiar platitudes about the phenomenal aspects of experience--for instance, that the hurtfulness of pain is unpleasant, that pink looks more like red than it does like green, that the smell of rotten eggs is disgusting, and so forth. Folk psychology need not be completely correct in what it says about qualia, but its homely platitudes are innocent until proven guilty.

Below the surface level of folk psychology are two finer-grained levels of causal detail. One involves the specific functional-representational roles, in the human cognitive system, that are associated with phenomenal states. (The platitudes of folk psychology do say something about these roles, of course, but presumably there is more to say.) Identifying these roles is a task for empirical research and theory, and is likely to occur in the context of working out a more general theoretical account of human cognitive processing.

The other, yet more detailed, fine-grained level involves the specific neural states and processes that directly subserve phenomenal consciousness. Here again, identifying these states is a task for empirical research and theory, and doing so is likely to occur in the context of working out a more general account of how various kinds of mental states are physically subserved in the brain.[1]

Relevant work on the neural correlates of consciousness proceeds apace, with traditional empirical tools and newly available technology: neural-anatomical studies of cell types and connectivity patterns, recordings from individual neurons, deficit and lesion studies in which special neural structures are destroyed, and most recently neuroimaging techniques.

At the level of cognitive architecture, there is certainly much theoretical work that purports to address itself to the functional-representational roles associated with states of consciousness. However, it seems to us that David Chalmers (1995, 1996) is right in maintaining that much of this work really addresses various other aspects of mentality that are often described in the language of consciousness--e.g., attentional processes, the ability to access and/or report one’s own mental states, the deliberate control of one’s own behavior, etc.---rather than directly addressing phenomenal consciousness itself. Concerning cognitive roles associated with phenomenal consciousness in particular, it appears to us that perhaps the most pertinent discussions are to be found in the recent writings of philosophers like Dretske (1995), Lycan (1996), and Tye (1995). (More on this in Section 3.1.)

Ideally, a completed account of the causal grain within cognitive science would integrate the three levels of detail. Also, it would itself be integrated into a more general and well articulated general account of mental processing, addressing a variety of mental phenomena and their interconnections with the phenomenal aspects of mentality. In research on color vision, just to take one example, one approach is to assume that functional-representational grain level descriptions can be developed through behavioral studies of color mixing and matching behavior and then subsequently connected with underlying neural processes. One of the more heralded findings of modern color science is that properties of retinal receptors can explain many of the facts of color mixing and matching. (For discussion see Clark 1998.) There also has been considerable discussion within color science of whether both the functional-representational and neural levels in color perception are modular in organization and design: whether color vision is functionally and neuronally independent of perception of shape, motion, and other properties of a visual stimulus. (For discussion see Davidoff 1991 and Livingstone and Hubel 1987). We would expect that on the basis of research on color vision, and on numerous other types of phenomenal experience, information gathered from the brain and functional-representational levels will be mutually informative. Just as information drawn from functional descriptions will inform descriptions of the neural level, information drawn from the brain will guide development of functional-representational models of phenomenal consciousness and perhaps even help to inform models at the folk psychological surface level. (For related discussion see Flanagan 1996 and Zawidzki and Bechtel in press.)

The grain project is a rich program of inquiry within cognitive science. It is empirically and theoretically tractable, and progress on this project will certainly enhance our understanding of phenomenal consciousness. But enhancement of understanding is one thing; resolution of deep philosophical problems is quite another.

2. Three Philosophical Problems.

We begin with two familiar philosophical thought experiments, as groundwork for the discussion below: inverted qualia and absent qualia. It seems to make sense--not to be logically or conceptually contradictory--to imagine a “possible world” that is just like the actual world in all physical respects (including being governed by all the same physical laws that prevail in the actual world) but in which qualia are differently instantiated then they are in the actual world. On “inverted qualia” versions of this thought experiment, one imagines that phenomenal properties are instantiated in a way that is somehow inverted relative to their actual-world instantiations: for instance, the qualitative aspects of color-experiences are systematically inverted. (What it's like for someone to see red is what it's like for us to see green; and so forth.) On “absent qualia” versions, one imagines that phenomenal properties are not instantiated at all. The creatures in the given world who are duplicates of humans and other sentient actual-world creatures are zombies, in the sense that the states they undergo lack phenomenal content altogether; although these zombies do instantiate those aspects of mentality that are characterizable in functional-representational terms, there isn't anything at all that it's like, for them, to undergo such states.

The assumption that there seem to be possible worlds just like the actual world but containing inverted or absent qualia raises the question of just what sort of possibility we are considering when we say that apparently there could be inverted and absent qualia. In answering this question it will be useful to introduce a concept that is frequently employed in recent metaphysics and philosophy of mind: the notion of supervenience. Supervenience is an ontological determination-relation between facts or properties at different levels of description: the lower-level facts and properties determine the facts and properties supervenient upon them, in the sense that there cannot be a difference at the higher level without some underlying difference at the lower level. Two types of supervenience are important to distinguish: (1) logical (or conceptual) supervenience, which says that it would be logically impossible for the higher-level facts and properties to differ without some underlying lower-level difference; and (2) nomological (or natural) supervenience, which says that it would be contrary to certain laws of nature for there to be one kind of difference without the other. (For further discussion see Kim 1990, Horgan 1993, Chalmers 1995 Chapter 2.)

In saying that there apparently could be inverted and absent qualia in worlds otherwise just like the actual world we are saying that even if inverted or absent qualia were contrary to laws of nature, it seems logically (or conceptually) possible for there to be inverted or absent qualia. Phenomenal properties of experience appear not to logically supervene on those aspects of mentality that are characterizable in functional-representational or neural terms. They seem to constitute--as we shall suggest momentarily--a kind of metaphysical residue that provides the basis for three philosophical puzzles or problems concerning phenomenal consciousness.

2.1. The Problem of Ontological Status.

The apparent logical possibility of physical-duplicate worlds in which qualia are differently instantiated, or are absent altogether, creates a problem about their ontological status in relation to our overall scientific worldview. For, with the exception of phenomenal properties, it is plausible that the other properties posited in special sciences and in common sense are reductively explainable--and that reductive explanation rests on logical supervenience, typically involving the functional analyzability of the supervenient properties. As Chalmers (1995) observes:

A reductive explanation of a phenomenon need not require a reduction of that phenomenon, at least in some senses of that ambiguous term. In a certain sense, phenomena that can be realized in many different physical bases--learning, for example--might not be reducible in that we cannot identify learning with any lower-level phenomenon, but this multiple realizability does not stand in the way of reductively explaining any instance of learning in terms of lower-level phenomena. . . . In general, a reductive explanation of a phenomenon is accompanied by some rough-and-ready analysis of the phenomenon in question, whether implicit or explicit. The notion of reproduction can be roughly analyzed in terms of the ability of an organism to produce another organism in a certain sort of way. It follows that once we have explained the processes by which an organism produces another organism, we have explained that instance of reproduction. . . . The possibility of this kind of analysis undergirds the possibility of reductive explanation in general. Without such an analysis, there would be no explanatory bridge from the lower-level physical facts to the phenomenon in question. With such an analysis in hand, all we need do is show how certain lower-level physical mechanisms allow the analysis to be satisfied, and an explanation will result. . . . For the most interesting phenomena, including phenomena such as reproduction and learning, the relevant notions can be analyzed functionally. . . . It follows that once we have explained how those functions are performed, then we have explained the phenomena in question. . . . The epistemology of reductive explanation meets the metaphysics of supervenience in a straightforward way. A natural phenomenon is reductively explainable in terms of some low-level properties precisely when it is logically supervenient on those properties. . . . What is most important is that if logical supervenience fails. . . . then any kind of reductive explanation fails, even if we are generous about what counts as explanation. (pp. 43-50)

Although reductive explanation is not by any means a priori (since the lower-level facts and principles invoked are straightforwardly empirical), there is an aspect of the overall explanation that is relatively a priori, viz., the appeal to what is essential about the higher-level supervening property -- typically (as in Chalmers' examples of learning and reproduction) its definitive functional role. Reductive explanation is a matter of logical supervenience.

Now, the problem about qualia is that they seem not to be logically supervenient on physical properties and facts, and thus not to be reductively explainable. This is the moral of the apparent logical intelligibility of inverted-qualia and absent-qualia thought-experiments. Thus, they are a metaphysical anomaly or residue, vis-a-vis our overall scientific account of the world.

By contrast, it does not make conceptual sense to suppose, for example, that there is a physical-duplicate world in which the creatures who are our physical counterparts lack genes. The concept of gene rules this out, given that there are physical constituents of human cells (viz., components of DNA) that play the gene-role. Likewise, it does not make conceptual sense to suppose that there is a physical-duplicate world in which the stuff that is the physical counterpart of this stuff in my glass fails to be liquid. Again, the concept of liquidity rules this out, given that the macro-dispositions that constitute liquidity are the result of the operative inter-molecular physical forces. On reflection, this kind of point looks to be highly generalizable. It is plausible that for most all of our higher-level concepts, there will be logical supervenience at work: in principle, the instantiation of higher-level properties will be reductively explainable on the basis of the facts and laws of natural science (arguably ultimately those of basic physics), together with facts about the nature of higher-level concepts and of the meanings of terms expressing those concepts. (For further articulation and elaboration of this point, see Chalmers 1995 Section 2.5.) But phenomenal consciousness is metaphysically puzzling or mysterious because it does not seem to be logically supervenient on physical facts and properties, and it therefore it does not appear to be reductively explainable.

2.2. The Problem of Causal Efficacy.

If qualia are not logically supervenient on underlying physical facts and properties, then serious doubts arise whether qualia play any causal role in generating behavior. After all, in a physical-duplicate world in which a person's physical counterpart has inverted qualia, or lacks qualia altogether, that counterpart-person still behaves (by hypothesis) exactly as the person in the actual world behaves, despite having different qualia, or none at all. This being so, it appears prima facie that the phenomenal aspects of one's mental life play no genuine causal role at all with respect to one's behavior; rather, the real causal work seems to be done by properties that we share in common with our physical counterparts in these inverted-qualia and absent-qualia physical-duplicate worlds. (For further elaboration of this line of reasoning, see Horgan 1987.)

2.3. The Problem of Explaining Phenomenal Consciousness.

This problem is closely connected to the other two. A thoroughly satisfying explanation of phenomenal consciousness would be a reductive one in which (i) certain lower-level phenomena would be explained, (ii) qualia themselves would then be explained as logically supervenient on these lower-level phenomena, and (iii) qualia would thereby be shown to be causally efficacious. But since the prospects for this kind of explanation are thrown into doubt by the apparent conceivability of inverted-qualia and absent-qualia scenarios, we find ourselves faced with certain recalcitrant looking 'why'-questions about phenomenal consciousness. For instance: even if one could explain, in broadly evolutionary terms, why there would have emerged states with the functional-representational roles associated with phenomenal consciousness, the question would remain why these states have the specific phenomenal character they do, rather than other kinds of phenomenal character (e.g., inverted in certain respects relative to the actual world), or none at all. Moreover, trying to answer such questions by appeal to some distinctive survival/reproductive advantage allegedly accruing to the qualia we actually experience--an advantage in virtue of which these states supposedly have emerged in evolution under pressures of natural selection--seems to be thwarted by the same considerations that throw into doubt the causal efficacy of qualia: in a possible world that is physically just like our own but different with respect to how qualia are instantiated, the same neural states emerge under natural selection as have emerged in the actual world, with the same functional-representational roles; and yet these states have different phenomenal content in those worlds, or none at all. In short, selective advantage seems to attach only to the functional-representational roles associated with qualia, rather than with phenomenal content itself.

So there seem to be residual, intractable, explanatory mysteries about phenomenal consciousness. Why aren't we all zombies? And, given that we are not, why does our experience have the particular phenomenal character it does? And how could phenomenal character make any causal difference to behavior? These explanatory puzzles inevitably arise if inverted-qualia and absent-qualia physical-duplicate worlds are logically possible, as they seem to be.

3. Three Persistently Live Theoretical Positions.

We will discuss three approaches to phenomenal consciousness, each of which has some serious currency in the contemporary philosophical literature. In each case we will consider both the attractions or theoretical benefits and the problems or theoretical costs in the given view, and we will explain why, as best one can now tell, both benefits and costs would remain largely intact even if one had available a complete and detailed account of the causal grain of phenomenal consciousness. The upshot will be that each position apparently would remain an epistemically live theoretical option concerning phenomenal consciousness, even if the grain project were to be successfully completed.

By 'epistemically live theoretical option' we mean an option that someone who is well informed of all relevant information, and who carefully weighs the comparative theoretical advantages and disadvantages of the given position in relation to others, could rationally judge to be true, or probably true. This does not mean that all rational persons with full relevant information would make this judgment, however. On the contrary, individual rational persons, all possessing the same relevant information and all judging rationally, could diverge as to which of the theoretically live options (if any) they believe -- because they weigh the comparative pros and cons differently from one another. Thus, the competing positions would be live theoretical options for a community of rational inquirers, even if different members of that community have incompatible -- but respectively rational -- beliefs about which position is true and which positions are false.

Of course, some individual members might find themselves unsure what to believe, because for them no single position sufficiently outweighs all the others in terms of theoretical benefits and costs. This is essentially the state of mind of each co-author of the present paper, with respect to the various competing theoretical positions we will now consider. There are no rules, beyond a certain point, about how theoretical benefits and costs should be weighed. There is room for both individual judgment and suspension of judgment.

3.1. Phenomenal States as Functional-Representational States.

A currently popular approach in philosophy of mind is to maintain that phenomenal states are functional-representational states of a certain distinctive kind -- that phenomenal character is just a specific kind of representational content. Such views construe mental representation itself in broadly functional terms, as a matter of the typical causal role--possibly an environmentally situated role--that the states play in a creature's cognitive economy.

For concreteness we will focus on the specific version of this approach developed by Michael Tye (1995). (The points we will make are generalizable to other versions, as we will explain below.). Tye says:

Phenomenal content, I maintain, is content that is appropriately poised for use by the cognitive system, content that is abstract and nonconceptual. I call this the PANIC theory of phenomenal character: phenomenal character is one and the same as Poised Abstract Nonconceptual Intentional Content. . . . The claim that the contents relevant to phenomenal character must be poised is to be understood as requiring that these contents attach to the (fundamentally) maplike output representations of the relevant sensory modules and stand ready and in position to make a direct impact on the belief/desire system. . . . The claim that the contents relevant to phenomenal character must be abstract is to be understood as demanding that no particular concrete objects enter into these contents. . . . Since different concrete objects can look or feel exactly alike phenomenally, one can be substituted for the other without any phenomenal change. . . . The claim that the contents relevant to phenomenal character must be nonconceptual is to be understood as saying that the general features entering into these contents need not be ones for which their subjects possess matching concepts. . . . Consider. . . color. . . . We have names for only a few of the colors we can discriminate, and we also have no stored representations in memory for most colors either. There is simply not enough room. (pp. 137-39)

Tye, like many others in current philosophy of mind, construes mental intentionality--the 'I' in 'PANIC'--as a matter of causal covariation between representing state and item represented. “The key idea,” he says, “is that representation is a matter of causal covariation or correlation (tracking, as I shall often call it) under optimal conditions” (p. 101). Concerning the intentionality of phenomenal content, he says the following. (Red₂₉ is a specific, fine-grained, shade of red.)

Which features involved in bodily and environmental states are elements of phenomenal consciousness? There is no a priori answer. Empirical research is necessary. . . . They are the features our sensory states track in optimal conditions. . . . I conjecture that for perceptual experience, [these] will include properties like being an edge, being a corner, being square, being red₂₉. (pp. 137-41)

If Tye's account of phenomenal content as a species of intentional content (viz., PANIC) is correct, then the three philosophical puzzles discussed in Section 2 get cleanly resolved. There is no special problem of ontological status for phenomenal states; since these are just a species of functional-representational states, occurrences of such states are reductively explainable. Likewise, there is no special problem of causal efficacy, since it is not metaphysically possible for there to be world that is physically just like ours but in which qualia are inverted or absent. The problem of explaining phenomenal consciousness becomes tractable, being just the problem of explaining why and how there came to be states with this specific kind of functional-representational role--i.e., states with poised, abstract, nonconceptual, intentional content.

It is important to appreciate, however, that the functional-representational roles associated with phenomenal states could conform with Tye's PANIC story even if phenomenal content itself--the 'what it's like' aspect of experience--is something over and above (i.e. is non-identical with) the associated representational content. Tye's position is that phenomenal content is literally identical with the specific kind of intentional content he describes (viz., the kind that is abstract, nonconceptual, maplike, and poised to influence the belief-desire system). Tye could be mistaken about this identity claim even if he is right about the associated functional-representational role.

What bearing does the grain project have on Tye's position? Well, successful completion of this project would tell us specifically, and in detail, just what sort of functional-representational roles are associated with phenomenal states, and just which neural states subserve phenomenal states. The resulting account might not uncover functional-representational roles that meet Tye's generic PANIC characterization; and if not, then his position would thereby get decisively disconfirmed. Suppose, however, that the associated functional-representational roles for qualia did turn out to conform to the PANIC account, with the completed full grain story filling in the empirical details--by telling us, for instance, just which bodily and/or environmental features get represented when qualia are instantiated, and just how qualia are physically realized in the brain. Although such an outcome would be consistent with Tye's position, it would not decisively confirm it; for, he nonetheless might be wrong in claiming that phenomenal content is identical to intentional content that is poised, abstract, and nonconceptual (even though phenomenal content is associated with such intentional content).

Furthermore, as far as one can now tell, the kinds of considerations that currently count as reasons to doubt Tye's proposed reduction of phenomenal content to a species of representational content would still be operative, and would still count against this reductive account, even if a successful completion of the grain project were to vindicate the PANIC approach to functional-representational role. The strong intuition would still persist that there is more to phenomenal states than their associated functional-representational role--viz., their specifically phenomenal character, what it's like to be in them. When one imagines an inverted-qualia world or an absent-qualia world that is physically just like the actual world, the world imagined is one in which there are indeed states instantiated that have all the relevant functional-representational features that are uncovered in the grain story (PANIC states, we are now supposing); but the problem is that what they are like is different (inverted qualia), or there's nothing at all that they are like (absent qualia). Adding further details to the PANIC story, via a suitable completion of the grain project, would not change things, as far as the apparent intelligibility of inverted-qualia and absent-qualia scenarios is concerned. So their intelligibility would still tell against Tye's reductivist position, no less than it does now.

Tye does not ignore the idea that there is "something that it's like" to undergo phenomenal states, and the idea that "knowing what it's like" is something essentially subjective. He does attempt to accommodate these aspects of phenomenal experience. He says:

I call the concepts relevant to knowing the phenomenal character of any state 'phenomenal concepts.' Phenomenal concepts are the concepts that are utilized when a person introspects his phenomenal state and forms a conception of what it is like for him at that time. These concepts, in my view, are of two sorts. Some of them are indexical; others are predicative. Suppose, for example, I am having a visual experience of red₂₉. I have no concept red₂₉. So, how do I conceptualize my experience when I introspect it? I bring to bear the phenomenal concepts shade of red [a predicative phenomenal concept] and this [an indexical phenomenal concept]. Intuitively, possessing the phenomenal concept [shade of] red requires that one have experienced red and that one have acquired the ability to tell, in the appropriate circumstances, which things are red directly on the basis of one's experiences. . . . What about the phenomenal concept this? Possessing this concept is a matter of having available a way of singling out, or mentally pointing to, particular features that are represented in sensory experiences while they are present in the experiences, without thereby describing those features (in foro interno). . . . What one has [in having the indexical concept]. . . . is a way of singling out or discriminating the feature for as long as one attends to it in one's experience (and perhaps for a very short time afterward). (pp. 167-68)

In essence, Tye's approach treats 'knowing what it's like' as involving certain cognitive abilities: in the case of knowing what it's like to see red, (i) the ability to classify things as red directly on the basis of one's experiences (and without collateral information), and (ii) the ability to indexically pick out, in thought, a shade of red that is currently being represented PANIC-wise in one's experience (e.g., red₂₉).

For those who find themselves thinking that the PANIC account leaves out the genuinely phenomenal aspects experience, however, the trouble is that these kinds of cognitive abilities seem likewise to leave them out. Residue remains. Like PANIC states themselves, the kinds of cognitive abilities featured in Tye's account of knowing what it's like would be instantiated by duplicates of ourselves in an absent-qualia world. These zombie-duplicates would possess the ability to classify red things as red just by looking, and also the ability to attend to, and indexically pick out, features that are currently mentally represented PANIC-wise (e.g., the current presence of red₂₉). But there would not be anything that it's like when they undergo PANIC states, or when they exercise these cognitive abilities during the occurrence of PANIC states; hence there would be no knowing what it is like, for them.

To underscore the point, consider the famous thought experiment in Jackson (1982): Mary, a brilliant neuroscientist who knows all about how human color-vision works at the functional-representational and the neural levels of description (she knows the full story about the causal grain of color-qualia), but has spent her life in a black-and-white room (better: with lenses on her corneas that filter out color) without ever having had a color experience herself. Plausibly, she is burning with curiosity to know what color experiences are like. But surely what she is dying to acquire is not the mere ability to determine what color things are directly by looking, or the mere ability to indexically pick out color-features when they are visually presented to her. After all, she already understands how these abilities operate, and she is already adept (we may suppose) at using her scientific instruments as an aid to direct perception as a way to identify both the coarse-grained and the fine-grained colors of things. Rather, she is burning with curiosity because she wants to know what it's like to experience red. The abilities in question are a pale monochromatic ghost of genuine knowing what it's like, just as functional-representational PANIC states are a pale ghost of genuine phenomenal content. Or so it seems, anyway, given the apparent conceivability of a possible world that is physically just like ours but in which qualia are not instantiated.[2]

Now, these persistent and recalcitrant intuitions are ones that Tye is committed to explaining away. He makes a valiant effort, in a final chapter entitled “Can you really imagine what you think you can?'” And his answer, as one may guess, is no. Tye charges: We cannot imagine possible worlds otherwise just like the actual world but with absent or inverted qualia. We merely think we can imagine such worlds. Such worlds are not conceptually possible.

We lack the space here to examine his argument in detail, but for us the upshot is this: Although Tye perhaps succeeds in explaining why our absent-qualia counterparts in a zombie-duplicate world cannot really imagine what they think they can (that is, given zombie concepts of the phenomenal they cannot really imagine a world physically just like their own in which certain putative 'what it's like' aspects of mentality, which they think they themselves experience, are either systematically inverted or are altogether absent), he does not succeed in showing that we ourselves (given our non-zombie concepts) cannot imagine a world that is physically just like our own world but in which qualia are either systematically inverted or are altogether absent. His account seems correct for the phenomenally deprived, but incorrect for ourselves.

The points we have made about Tye's position apparently are applicable, mutatis mutandis, to any version of the generic view that phenomenal content is identical to some kind of functional-representational property; variants include Dretske (1995) and Lycan (1996). For each version, successful completion of the grain project might either decisively disconfirm the account, or else yield results that are consistent with the account and also fill in further detail. But even if things should go the latter way, this would not undercut the apparent conceivability of inverted-qualia and absent qualia-scenarios--and thus would leave intact the reasons for thinking that any attempt to explain away this conceivability as illusory would apply at best to zombie-duplicates who lack qualia, rather than to ourselves. So the apparent imaginability of inverted-qualia and absent-qualia scenarios would provide reason to claim that phenomenal-content properties are distinct from these associated functional-representational properties, and hence that the given account is mistaken in asserting that they are identical. Denying that phenomenal properties are identical to functional-representational properties would remain an epistemically live theoretical option.

On the other hand, there also would be significant theoretical advantages in the claim that phenomenal properties are identical to whatever functional-representational properties turn out to be associated with them under a successful completion of the grain project. The truth of this identity hypothesis would render nonproblematic the ontological status of phenomenal states, since they would be logically supervenient on facts and properties at the physical level of description.[3] There would be no special problem of causal relevance for phenomenal states, since conceptual supervenience would preclude the genuine metaphysical possibility of a world that is physically just like ours but different with respect to how phenomenal properties are instantiated. And the problem of explaining phenomenal consciousness would reduce to the tractable problem of explaining why and how there came to be states with the relevant functional-representational role, in creatures like ourselves. These theoretical advantages of the hypothesis that phenomenal properties are identical to functional-representational properties would be sufficiently powerful that this hypothesis would be an epistemically live option--as would be Tye's contention that when it comes to putative inverted-qualia and absent-qualia scenarios, you cannot really imagine what you believe you can.[4]

3.2 Phenomenal States as Ontologically Sui Generis.

Another influential current approach to phenomenal consciousness treats qualia as ontologically fundamental features of the world, over and above those that figure in the basic laws of physics. This view, probably implicit in Jackson (1982), is explicitly articulated and defended by Chalmers (1996). Briefly, the position goes as follows. It really is logically possible (just as it seems to be) for there to be a world that is physically just like our world but in which qualia-instantiations are systematically inverted in some respect relative to the actual world. Likewise, it really is logically possible (just as it seems to be) for there to be a world that is physically just like ours but in which qualia are not instantiated at all. These are genuine conceptual possibilities because phenomenal properties simply are not logically supervenient on physical properties and facts. On the other hand, these logical possibilities are not nomological possibilities; they are contrary to the fundamental laws of nature. This is because a complete inventory of nature's fundamental properties and laws would include not only the theoretically basic physics-level properties (mass, charm, spin--whatever they turn out to be) and the theoretically basic physics-level laws governing these properties and their interrelations--but also phenomenal properties themselves, together with certain theoretically basic laws of inter-level supervenience for qualia. (Thus, inverted-qualia and absent-qualia worlds are nomologically impossible, even though they are consistent with the fundamental laws of physics, because they are not consistent with the full set of fundamental laws of nature.) Our scientific worldview needs to be expanded beyond materialism: phenomenal properties are no less metaphysically basic than the fundamental properties of physics, and the inter-level supervenience laws governing phenomenal properties are no less metaphysically basic that the fundamental laws of physics.[5] This expanded worldview is a version of scientific naturalism, but it is a denial of scientific materialism.

What should be said about the three philosophical problems mentioned in section 2, given this position? The problem of the ontological status of phenomenal states receives a very dramatic answer: phenomenal properties have the distinctive, and highly elite, status of being among the metaphysically rock-bottom, basic and ultimate, properties instantiable in the natural world. (Materialists, who think that only the most basic properties posited in theoretical physics have this status, are simply wrong.)

The problem of causal efficacy is vexed and difficult, given this metaphysical account of qualia. For, even though qualia are nomologically linked to associated physical and functional-representational properties, there are counter-nomological possible worlds (i) that are just like our actual world in all physical respects (including being governed by all the same physical laws), but (ii) in which qualia are differently distributed or are absent altogether. This being so, there is prima facie reason to think that qualia as such do no real causal work at all--that the real work is done by the functional-representational properties, as realized by neurophysical properties. Chalmers discusses this issue at some length, suggesting that there may be an appropriate way of construing causal efficacy that will apply to qualia and will contravene the initial impression that they can do no real causal work if they are merely nomically--and not logically--supervenient on physical facts and properties. But in the end there may not be any viable such account of causal efficacy, in which case this ontological treatment of qualia would render them epiphenomenal.

Concerning the problem of explanation, the approach yields a mixed verdict. On the one hand, it allows certain explanatory questions concerning phenomenal consciousness to receive scientific answers--provided that the explanatory resources used to provide the answers include the relevant inter-level supervenience laws. For example, in answer to a question like 'Why did phenomenal consciousness arise, in the course of evolution?', an overall scientific answer would have two components: first: an explanation of the emergence, in evolution, of the relevant functional-representational states, as neurophysically realized; and second, an appeal to the applicable inter-level laws telling us that phenomenal properties supervene on those functional-representational states, so realized neurophysically. Also, particular instantiations of phenomenal properties can be explained by reference to relevant lower-level properties together with inter-level supervenience laws. On the other hand, certain explanatory questions about phenomenal consciousness simply have no answer, on this view. For example, if one focuses on the general supervenience laws governing qualia, and one asks why they obtain, the response has to be that these laws simply have no explanation. They are explanatory bedrock in science, in just the same way that the fundamental laws of physics are explanatory bedrock. The inter-level supervenience laws themselves must be accepted (as the British emergentists used to say) “with natural piety” (see McLaughlin 1992).

The principal--and significant--theoretical advantage of this position is that it thoroughly respects the data served up by first-person phenomenal experience. For many of us, the full richness of our phenomenal experience produces a persistent and deep-seated conviction that there is more to phenomenal content than its associated functional-representational role and/or its specific mode of physical realization; hence the apparent intelligibility of absent-qualia and inverted-qualia scenarios. On Chalmers' view, this conviction is correct, rather than being treated as a troublesome mistake that must be somehow explained away.

But this attractive feature brings significant theoretical costs in its wake. Positing fundamental properties outside the domain of physics, and additional fundamental laws of nature beyond those of physics, seriously complicates our overall scientific conception of the world and our conception of ourselves as denizens of the natural order; it is far simpler to suppose that physics alone is the realm of all fundamental properties and laws, and that all higher-order properties and facts are logically supervenient on physics-level properties and facts. Two additional considerations seriously exacerbate this loss of theoretical simplicity. First, as far as one can tell, phenomenal consciousness is unique--or close to unique--as a phenomenon that allegedly requires a departure from materialistic metaphysics. (Chalmers himself defends this claim in section 2.5 of his book, entitled “Almost everything is logically supervenient on the physical.”) Second, with the exception of putative inter-level supervenience laws involving qualia, the other fundamental laws of nature evidently apply directly to the fundamental physical constituents of things--whereas phenomenal consciousness is evidently instantiated only in complex, highly evolved, sentient organisms.[6] The epistemic force of such simplicity-driven considerations was well expressed 40 years ago in Smart (1962/1959), the paper whose title ours echoes. Smart wrote:

The suggestion I wish to resist is...that to say 'I have a yellowish-orange after-image' is to report something irreducibly psychical. Why do I wish to resist this suggestion? Mainly because of Occam's razor. It seems to me that science is increasingly giving us a viewpoint whereby organisms are able to be seen as physicochemical mechanisms: it seems that even the behavior of man himself will one day be explicable in mechanistic terms. There does seem to be, so far as science is concerned, nothing in the world but increasingly complex arrangements of physical constituents. All except for one place: in consciousness. . . . I just cannot believe that this can be so. That everything should be explicable in terms of physics (together of course with descriptions of the ways in which the parts are put together--roughly, biology is to physics as radio-engineering is to electromagnetism) except the occurrence of sensations seems to me to be frankly unbelievable. Such sensations would be 'nomological danglers'. . . . It is not often realized how odd would be the laws whereby these nomological danglers would dangle. . . . Certainly we are pretty sure in the future to come across new ultimate laws of a novel type, but I expect them to relate simple constituents: for example, whatever ultimate particles are then in vogue. I cannot believe that ultimate laws of nature could relate simple constituents to configurations consisting of perhaps billions of neurons (and goodness knows how many billion billions of ultimate particles). . . . Such ultimate laws would be like nothing so far known in science. They would have a queer 'smell' to them. I am just unable to believe in the nomological danglers themselves, and in the laws whereby they would dangle. (Smart 1962 pp. 161-62)

In addition, there is also the very real possibility that under an adequate account of causal efficacy, a view like Chalmers' would end up having to embrace epiphenomenalism concerning qualia. Since the hypothesis that the what-it's-like aspects of our experience have no genuine causal influence on our behavior is grossly offensive to common sense, the threat of epiphenomenalism is yet another consideration on the debit side of the cost-benefit ledger.

Suppose that the grain project were to be successfully completed, fully specifying what functional-representational roles are associated with phenomenal states, and how phenomenal states are physically subserved. As far as one can now tell, a view like Chalmers' would still retain both the theoretical advantages and the theoretical disadvantages just canvassed; further details about causal grain would not change things significantly. This approach would remain a live theoretical option.

3.3. Phenomenal States as Explainable in a Cognitively Inaccessible Way.

A third kind of position, perhaps implicit in Nagel (1974) and explicitly defended in McGinn (1991), is one whose attractions come to light when one asks whether there might be a way to combine the theoretical advantages of each of the views just considered, while avoiding the specific theoretical disadvantages of each. There will be some kind of theoretical cost, to be sure, but a different kind.

The leading ideas are as follows. On one hand, the view asserts that qualia really do supervene logically on lower-order properties--and thus that these supervenience relations are explainable in a materialistically acceptable way, rather than being fundamental laws of nature. Thus, the three problems about consciousness mentioned in Section 2 receive solutions like those they receive under functional-representationalist approaches like Tye's.

On the other hand, the view asserts that although materialistic explanations of inter-level supervenience relations involving qualia exist in principle, nevertheless human beings, because of our innate cognitive limitations, are “cognitively closed” to these explanations; we are constitutionally incapable of formulating and articulating them. There is something about the nature of phenomenal properties that we cannot fully grasp--something that makes them logically supervenient on the physical, even though we humans cannot see why or how this relation of logical supervenience obtains. Thus, the view also exhibits the principal advantage of Chalmers' view: it takes seriously, and accommodates, the hard-to-shake intuitive thought that no amount of theoretical information that we humans could acquire about physics, and/or neural organization, and/or the functional-representational roles of mental states, could possibly provide an explanation of why it should be that when certain neural states are instantiated in humans, they are correlated with the specific phenomenal qualities we know about from our own experience--rather than with different phenomenal qualities (inverted qualia), or with none at all (absent qualia).

On this view, it is indeed the case that there are no possible worlds--not even counter-nomological worlds--that are physically just like our actual world but in which phenomenal properties are instantiated differently (or are not instantiated at all). But we humans cannot grasp why this is not possible. Because of our innate cognitive limitations, we cannot understand what it is about phenomenal properties, and/or what it is about the underlying physical facts and properties, in virtue of which there actually lurks some logical-conceptual contradiction in the idea that the world could have been physically just the same as it is but different with respect to how qualia are instantiated. Physical-duplicate worlds with inverted or absent qualia do indeed seem logically possible. But on the view in question, this appearance is deceiving, being a byproduct of our cognitive closure to the reasons for their impossibility.[7]

The attractions of the cognitive closure approach are significant: in essence, it combines the principal theoretical benefits of each of the earlier two approaches we have discussed in this section, while avoiding the principal theoretical disadvantages of each. But needless to say, these virtues come at an enormously high theoretical price: the hypothesis of human cognitive closure, vis-a-vis the putative materialistic explanation of physical-phenomenal supervenience relations. Given the human race's spectacular history of success in scientifically explaining other aspects of the natural world, many people are bound to find it hard--or impossible--to believe that phenomenal consciousness is inherently beyond the conceptual bounds of human scientific understanding.

Suppose, once again, that the grain project were to be successfully completed, fully specifying what functional-representational roles are associated with phenomenal states, and how phenomenal states are physically subserved. As far as one can now tell, a view like McGinn's would still retain both the theoretical advantages and the theoretical disadvantages just canvassed; further details about causal grain would not change things significantly. This approach would remain a live theoretical option.

4. Concluding Methodological Postscript.

To say that approaches to phenomenal consciousness like those we have considered would all remain live epistemic possibilities even if the grain project were successfully completed is not necessarily to say that there could never be any way to determine which approach--one of the three, or some other view entirely--is correct. In science, epistemic appraisal of competing hypotheses and theories is a matter of overall cost-benefit evaluation--what philosophers call "wide reflective equilibrium." The same is true, we would maintain, for questions of the sort that arise in philosophy--although often these questions prove more recalcitrant, more open to the possibility that reasonable people will be able to disagree because they make different overall cost-benefit assessments. In principle, some particular philosophical account of consciousness--perhaps some variant of those we have considered, or perhaps some quite different view--might eventually turn out to have theoretical benefits so powerful, and theoretical costs so minimal, in comparison to the benefits and costs of competing accounts, that a strong consensus would emerge among rational, well-informed, people that this account is very probably true.

On the other hand, there is certainly no guarantee that rational convergence on these issues would eventually occur, even given sufficient theoretical ingenuity and empirical information. Indeed, the preceding discussion suggests that strongly divergent views about the nature of qualia would remain live theoretical options even in the ideal limit of theoretical inquiry. This seems to us a plausible conjecture.

Although in some ways it is disappointing to realize that carrying out the grain project would not resolve the philosophical problems about phenomenal consciousness, in other ways it is liberating. This very realization could cause those working in cognitive science to be happy with less: just seek to understand the causal grain of phenomenal consciousness, and settle for that. Doing so would advance our knowledge significantly, even though it would leave the philosophical puzzles about phenomenal consciousness still open. It seems like an attractive, empirically tractable, and non-tendentious perspective from which to pursue the grain project, neither dismissive of cognitive science, nor of conceptual intuitions, nor of metaphysics; but, on the other side, not too metaphysically wistful. Evolutionary approaches fit naturally into this scaled-down program, since there are at least two important kinds of evolution-related question to ask about states with the distinctive causal grain associated with qualia. First, why did such states emerge on earth, in the evolution of humans and other terrestrial creatures? Second, with respect to evolutionary landscapes in general (including non-local ones for planets in distant solar systems, for A-life environments, etc.), do states with the relevant causal grain occur upon many of the fitness-peaks that reach an altitude where mentality resides?

The aspirations of Place and Smart were at least partly right. Our science of the conscious mind can get us at least somewhere -- if not all the way to ontology. Anniversaries are if anything humbling.[8]

References

Bieri, Peter (1995). Why is consciousness puzzling? In T. Metzinger (Ed.), Conscious Experience (45-60). Paderborn: Shoningh.

Chalmers, David (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2, 200-219.

Chalmers, David (1996). The Conscious Mind: In Search of a Fundamental Theory. New York and Oxford: Oxford.

Clark, Austen (1998). Perception: Color. In W. Bechtel & G. Graham (Eds.), Companion to Cognitive Science (282-88). Oxford: Basil Blackwell.

Davidoff, Jules (1991). Cognition Through Color. Cambridge, MA: MIT.

Dretske, Fred (1995). Naturalizing the Mind. Cambridge MA: MIT.

Flanagan, Owen (1996). Is a science of the conscious mind possible? In his Self Expressions: Mind, Morals, and the Meaning of Life (12-31). New York: Oxford.

Graham, George & Horgan, Terence (2000). Mary Mary, quite contrary.” Philosophical Studies, in press.

Horgan, Terence (1987). Supervenient qualia. Philosophical Review, 94, 491-520.

Horgan, Terence (1993). From supervenience to superdupervenience: meeting the demands of a material world. Mind, 102, 555-86.

Horgan, Terence (in press). Multiple reference, multiple realization, and the reduction of mind. In G. Preyer, F. Siebelt, and A. Ulfig (Eds.), Reality and Humean Supervenience: Essays on the Philosophy of David Lewis. Lanham, MD: Rowman and Littlefield.

Jackson, Frank (1982). Epiphenomenal qualia. Philosophical Quarterly, 32, 127-36.

Kim, Jaegwon (1990). Supervenience as a philosophical concept. Metaphilosophy, 21, 1-27.

Lewis, David (1966). An argument for the identity theory. Journal of Philosophy 63, 17-25.

Lewis, David (1980). Mad pain and Martian pain. In N. Block (Ed.) Readings in the Philosophy of Psychology. Volume 1 (216-22). Cambridge MA: Harvard.

Lewis, David (1994). Lewis, David: reduction of mind. In S. Guttenplan (ed) A Companion to the Philosophy of Mind (412-31). Oxford & Cambridge MA: Blackwell.

Livingstone, Margaret & Hubel, David (1987). Psychophysical evidence for separate channels for the perception of form, color, movement, and depth. Journal of Neuroscience, 7, 3416-3468.

Lycan, William (1996). Consciousness and Experience. Cambridge MA: MIT.

Macdonald, Cynthia (1989). Mind-Body Identity Theories. London: Routledge.

McGinn, Colin (1991). The Problem of Consciousness: Essays Towards a Resolution. Oxford and Cambridge MA: Blackwell..

McLaughlin, Brian (1992). The rise and fall of British emergentism. In A. Beckermann, H. Flohr, & J. Kim (Eds.), Emergence or Reduction? Essays on the Prospects of Nonreductive Physicalism. Berlin: Walter de Gruyter (49-93).

Nagel, Thomas (1974). What is it like to be a bat? Philosophical Review, 83, 435-50.

Place, Ullin (1956). Is consciousness a brain process? In V. C. Chapell (Ed.), The Philosophy of Mind. Englewood Cliffs: Prentice-Hall (101-109). Reprinted from British Journal of Psychology (1956).

Smart, J. J. C. (1962). Sensations and brain processes. In V. C. Chappell (Ed.), The Philosophy of Mind. Englewood Cliffs: Prentice-Hall (160-72). Reprinted from The Philosophical Review (1959).

Tye, Michael (1995). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. Cambridge MA: MIT.

Zawidzki, Tadeuz & Bechtel, William (in press). Gall's legacy revisited: decomposition and localization in cognitive neuroscience. In C. Erneling & D. Johnson (Eds.), Mind as a Scientific Object: Between Brain and Culture. New York: Oxford.