This thesis is concerned with some apparent obstacles to the empirical study of sensory qualities. I argue, in particular, that they are not real obstacles. There are three parts to my argument. Firstly, there is the question of method—specifically, of whether the analysis of terms, a method peculiar to philosophy though currently suspect within it, is relevant to the scientific enterprise. Drawing on the work of Frank Jackson, I defend the view that the “working definitions” that are necessary to empirical research, whether explicit or implicit, rely on the same considerations that are used to arrive at conceptual analyses in philosophy; namely, our ordinary conceptions, or everyday criteria, or ”folk” theories. Sometimes it is not obvious what criteria we are using in our judgements that two tokens fall under the same type. These are the cases in which explicit analysis can be called for. Sensations are one such case; disagreements over the best analysis of sensory concepts are, in effect, disagreements over the most appropriate working definitions of sensations, and undermine empirical claims about their nature. The analysis of sensory terms is therefore a necessary part of the empirical study of sensations.

Secondly, there is the question of the right analysis of sensory terms. I argue against the received view that they are primitive, in the sense of being unanalysable, or purely demonstrative, or purely recognitional, or purely first-person. These views can not, I argue, account for the fact that sensory concepts are passed on from individual to individual regardless of their demonstrative, recognitional, or first-person content. I agree that there is some such content but not that it can be the whole story. I argue, furthermore, that “two concept” views of consciousness, such as are held by Block, Chalmers and others, do not solve this problem. However, I argue that if sensory concepts include the first-person indexical, then the two sorts of concepts proposed by these philosophers can be incorporated into a single indexical concept. The semantic parts of indexical terms can then account for many of the conflicting intuitions that plague this area of philosophy. I offer an indexical analysis of sensory terms that is in fact suggested by J.J.C. Smart’s topic-neutral analysis and show how it renders such an analysis compatible with common intuitions about, for example, inverted spectra and multiple realisability.

Lastly, there is the question of the explanatory gap—the apparent impossibility of a genuine explanation of sensory qualities in physical or functional terms. I propose that the explanatory gap is the result of a failure to see that sensory qualities are a kind of appearance—the way our perceptions of the world appear to us—and are therefore relational rather than intrinsic properties. I draw on “inner sense” views of consciousness, particularly Lycan’s, in my defence of this proposal. However, in contrast with Lycan’s approach, I argue that inner sensings are best seen as an explanation of sensory qualities—of qualia—rather than as a mechanism of consciousness. I offer a dispositional account of those qualities along the lines of dispositional accounts of colour.

On the basis of all of these considerations, then, I conclude that there is no reason to think that sensations are special with respect to their empirical study.

Preface and Acknowledgments

The department of Philosophy at Monash has been a stimulating place to have spent most of the last five years. Graham Oppy has been my principle supervisor throughout my candidature, and I am very grateful to him for our regular meetings through thick and thin, and for his support, encouragement and flexibility. Ian Gold became my associate supervisor half way through my candidature and has been terrific in giving very practical support, advice and encouragement. I am indebted to John Bigelow and Lindy Antippa for keeping tabs on me when I wandered off to America. While at the University of Maryland (College Park), I attended an enormously stimulating graduate seminar taught by Georges Rey, which had a significant effect on my approach to conceptual issues. I learned a great deal in my time at UMCP, from Georges, fellow graduate students Alessandro Giovanelli and Saam Trivedi, and also from a reading group that included Keith Campbell, Jerry Levinson and Jesse Prinz.

More recently, I am grateful to Jack Smart for comments on an early draft of what is now chapter four; to Neil McKinnon for comments on drafts of the first and last chapters; to John Heil for comments on drafts of the third and fourth chapters; and to Phillipe Chuard for comments on an earlier draft of the final chapter. I am indebted to Angus Nicholls, Louise Mills, Anne Taylor and Claire McCausland for their thorough proof-reading.

Practically all of my time spent working on the thesis was in a space given to me by what is now the Monash Research Graduate Centre. Joanne Ligouris managed “the Centre” for most of my time at Monash, and did it in the most friendly and competent manner imaginable. Monash University was generous to me in their provision of, firstly, an Australian Postgraduate Award, and subsequently a Monash Graduate Scholarship.

On a more personal level I would like to thank my father for his steady encouragement, and Louise Mills, Janet Hsuan, and Lloyd Singaretnam for their personal support and friendship. Annik Foreman had the bad luck to share an office with me, and was very tolerant of my untidiness, especially in the final stages of writing up. I am fortunate to have been working on a daily basis alongside a group of talented Ph.D. students from a great variety of disciplines and backgrounds, who have made my time at Monash exceptionally enjoyable and stimulating; for their collective company and engagement I would like to thank Achmed, Amelia, Angus, Anne, Antonella, Atsuko, Brie, Caroline, Chris, Danny, Darriel, David(s), Dini, Emma, Fahrid, Fiona, Grace, James, Jo, Justin, Kathleen, Les, Mark, Matthew, Milsom, Naraset, Paul, Phoung, Rob, Rowena, Simon, Soe Tjen, Sue and Toby. Finally, and especially, I must thank Ryo Otomo for her deep friendship, wise counsel, and wonderful cooking.

Introduction

None of these accounts offers any insight into what consciousness is, above and beyond its hypothesised dependence on a certain brain system…Nor do they tell us why a certain brain system…should be necessary for consciousness. Nevertheless, they are not vacuous or question-begging. Although they do not answer metaphysical questions about consciousness, they are substantive claims about the neural correlates of conscious experience. [1]

Martha Farah

Language is as it were the atmosphere of philosophical investigation, which must be made transparent before anything can be seen through it in the true figure and position. [2]

J.S. Mill

The problem of consciousness would seem more tractable if not for sensations. David Chalmers’ label, “the hard problem”, reflects the mood both in science and in philosophy. Martha Farah’s extreme pessimism (“none of these accounts offers any insight into what consciousness is”!) is offset by a quasi-mystical belief in some quarters that this or that interpretation of quantum mechanics will explain the presence of subjectivity in the universe. There is wide disagreement, not just over which theory is more correct, but over which discipline is most likely to offer a solution. In philosophy, the differences between the various positions seems to have solidified, with Ned Block going so far as to say that the divide between those who do and those who do not believe in qualia represents “the greatest chasm in the philosophy of mind—maybe even all of philosophy”. [3]

In the present day, most people who have studied the matter will agree that, at least in some rough and ready sense, sensations are brain processes. U.T. Place (1956) and J.J.C. Smart (1959) famously argued that the best empirical evidence demonstrated this. It is widely thought that, a few years later, Putnam (1967) refuted Smart and Place in an important sense. In that paper, Putnam defended the view that sensations are functional states rather than specifically neural ones. His argument was ostensibly an empirical one—he observed that the similarities between, for example, an octopus’ pain and a human’s pain is likely to be at a more general level than the ‘physico-chemical’. But this observation does not account for Putnam’s role in the near demise of Place and Smart’s view, since that view is compatible with the brain processes in question turning out to be functionally characterised and species-independent. [4] Of far more importance to Putnam’s attack, ironically given his stress on the empirical, was a conceptual argument. He argued that when we intuitively count animals as instantiating particular mental states, such as states of thirst or pain, we do so by the fact that those animals seem to be in states that fill particular functional roles. Moreover, Putnam pointed out, brains may not even be necessary to fill such roles. [5] Earlier, Putnam (1963) had maintained that the identification of mental states with brain states is a category mistake, tantamount to identifying the “logical” states of a Turing machine with its “structural” states (having argued that psychological states are Turing machine states). Putnam’s claim was not that mental states, including sensations, are not brain processes, but rather that they are not necessarily brain states; that they are not essentially biological.

This claim has passed into orthodoxy—except where sensations are concerned. Block and Fodor (1972) argued that the category mistake claim fails in the case of sensations since we standardly do not, and certainly the language does not, take sensations to be Turing machine states. [6] Block (1980) argued that we do not take sensory states to be any kind of purely functional state. On this question there has been very little movement in the last twenty years. In particular, the “category” question—that is, the question of what our use of sensation words says about sensations—has received little attention. Part of the reason for this is scepticism about the a priori, and the consequent emphasis in philosophical circles on the empirical. A priori analyses of sensation concepts, an important part of Smart’s argument, are now typically considered to be either beside the point [7] or simply unavailable. [8] For reasons that I will try to make clear, I think both points of view are mistaken.

I believe, moreover, that the conceptual argument put forward by Smart was never properly refuted. It is still empirically plausible that sensations are in fact brain processes. On the basis of that and what we mean by sensation words, it is plausible that sensations could not be anything but a type of brain process. [9]

In arguing for this, there are three main issues to be addressed. Firstly, what is the relevance of the ‘analysis’ of sensory terms in the question of whether sensations are brain processes? Secondly, if it is relevant, what is the right analysis of sensory terms? Thirdly, could a type of physical process satisfy that analysis? With respect to the first issue, I will argue that the availability of an analysis, or something very like it, is what gives direction to empirical research and is therefore essential to the scientific enterprise. With respect to the second issue, I will argue that a ‘topic-neutral’ analysis of sensory terms along the lines of Smart (1959) is essentially correct, with a small addendum and with particular emphasis on the inclusion of the first-person indexical. Finally, I will propose a dispositional account of sensory qualities in order to refute the ‘explanatory gap’ arguments that no physical process could, on its own merits, be empirically adequate for the “sensations are brain processes” hypothesis.

This thesis is therefore separated into three parts. Although I hope the effect is cumulative, each part might be defended on its own merits. This is particularly true of Part Three. The first part, which I have called the methodological part, deals with the role of the analysis of terms in claims of the kind that “sensations are brain processes” represents. These are claims in which the referent of some common noun––water, gold, pain, etc.––is alleged to be identical with, or constituted by, some theoretical postulate of science. Aside from general worries about analyticity, which is outside the scope of this thesis, [10] it is commonly thought—and in particular has been argued recently by Block and Stalnaker (1999)—that some of Kripke’s arguments in Naming and Necessity (1980) obviated the need for analysis of the concepts expressed by the common nouns involved in these cases, since their reference may be fixed, Kripke argued, historically rather than by a description. But, even if Kripke is right, this is only ideally the case. Kripke’s account of the apparent contingency of identity statements such as “water is H₂O” requires the existence of some set of properties that are closely associated with, in this case, water—that it is clear, drinkable, and so on. Given that we have very little idea of the historical circumstances surrounding the genesis of the word “water”, that it names the local clear, drinkable liquid is the best we can do. If Lavoisier had been challenged on whether the stuff he had found to be decomposable into hydrogen and oxygen really was water, his only defence would have been to point out that it was taken from the river, and was clear and drinkable, and so on. In other words, whether or not the reference of “water” is fixed by its associated description, that description would need to have been true of whatever stuff was being studied as water. The justification that water is H₂O would thereby have built into it the associated description which consequently could not be proved false. The descriptions associated with natural kind terms, which Kripke identified as being responsible for the apparent contingency of identity statements concerning them, fill the role of a “working definition” of those kinds. The goal of the empirical work is to make that working definition redundant; it makes little sense to say that it might prove it false.

The current group of philosophers who are challenging the role of descriptions, or of analysis, in scientific identity claims do so largely on the basis that there is an alternative route that is inspired by the historical view of reference. Whether or not the historical view is correct, I maintain that it is not enough to motivate that alternative and that in this case all roads lead to analysis.

Part Two deals with sensory terms in particular. In it I ask the question, what are we reporting when we report that we are experiencing a sensation? Once we accept that the empirical study of sensations requires the availability of an answer to this question, the inadequacy of a certain popular view becomes evident. This is the view that the terms we use to talk about our sensations are, in some substantive sense, primitive. There are different ways in which this idea is expressed. Chalmers (1996), anticipating his “fundamental theory”, likens sensory concepts to concepts of space and time in that, as he says, they cannot be described or analysed into “simpler” concepts. One of Block’s (1995a) two concepts of consciousness refers to the way sensations feel and is treated by him as a demonstrative—“really all one can do is point to the phenomenon.” (p. 230). For Loar and Tye, sensory concepts are “recognitional” – they are concepts of inner states that we recognise ourselves to be in from time to time. All of these views have the immediate problem that sensory concepts are shared, despite the fact that we can neither describe our sensations to others, point them out intelligibly to others, nor recognise them in others. For all of the “primitive” views, the content of sensory concepts is arrived at with sole reference to the self, and from that point of view is hard to see how that content is shared across individuals.

This problem is somewhat ameliorated, or perhaps hidden, by the view held by most of the above philosophers that there are in fact two concepts for each sensation—the primitive concept and one that is more easily shareable, usually a functional concept. Chalmers and Block most clearly emphasize this “splitting” of sensory concepts, but it is a view held by many of the major figures who write on the subject—Papineau (1993), Nagel (2000) and, with qualifications, Loar (1990) and Tye (1999), for example. This strategy does help to explain how we might arrive at a shared conception of pain at some level, but it leaves open the possibility that the unanalysable part of sensory concepts might have a different referent to the analysable part. Chalmers (and arguably Nagel) actually endorse this possibility. The others have strategies for avoiding it, depending on the sense in which they take sensory concepts to be primitive, but for a number of reasons it is better—or so I will urge—to avoid both of the ideas that there are primitive sensory concepts in any of the proposed senses and that sensory terms express two distinct concepts.

Having said that, there is a way of respecting the intuitions behind both ideas, and that is to treat sensory terms as indexicals. When I report that I am experiencing a sensation as of blue, in part I am reporting my recognition of a mental state that—as things stand—only I have access to. But in describing it as “blue” I am buying into a common conception of blue sensations as those inner states of the kind that we tend to recognise when we look at the sky on a clear day, and which is not in itself straightforwardly tied to the respect in which I recognise it. The prima facie possibility of an “inverted spectrum” is evidence for this. Consistent with both of these aspects of my report, it may be that our common understanding of “blue sensation” is similar in structure to our common understanding of words like “home”. When someone tells me they are going home I often do not really know where they are going, or what that place is like. But I understand them perfectly well. “Home” is indexical; it depends for its reference, like “I” and “here”, on the specific context in which it is uttered [11] . It has a common meaning, which is how I can understand your use of it without knowing where you live. It also has, in a sense, a private referent; for you, where you live. To know the meaning of an indexical, like “home”, is to know what the referent would be in any given context, while my meaningful use of it is restricted to its referent in my own case. Similarly, and perhaps more obviously, I cannot very easily use “I” to refer to you.

When I report that I am experiencing a blue sensation, then, if that phrase is indexical, you need not know its referent in my case to understand me. Equally, the common meaning of “blue sensation” will be abstracted away from the particular kind of experience that I have when I look at the sky on a clear day—just as particular kinds of dwelling are abstracted away from the common meaning of “home”. That you recognise yourself to be in a particular kind of mental state when you look at the sky is necessary for you to be able to talk meaningfully about blue sensations—similarly, that you live somewhere is necessary for you to be able to talk meaningfully about home. When you think about your blue sensations you call to mind a particular sensory state, to which I have no access. That is the sense in which sensory terms are primitive. But your recognition of blue sensations is not, in itself, a recognition that they are what is meant by “blue sensation”. That particular kind of state, while being the referent of “blue sensation” in your case, is not part of the common understanding of that phrase. The common understanding—that is, abstracted away from particular kinds of states—may simply be that when I am experiencing a blue sensation I am referring to the state I am in usually when I look at the sky on a clear day. The “I” is important and is what makes sensation reports indexical. It is also, I believe, what ties together the notion that sensory concepts are primitive with the notion that they are analysable. A consequence of this is that the working definition for “blue sensation” will of necessity leave out that aspect of blue sensations which we most immediately recognise—that is to say, the way they feel. This does not mean that the study of sensations may ignore that aspect, but does serve to prevent neuroscience from being pre-empted by philosophers who want to insist that what we mean by “blue sensation” prevents the construal that it may turn out to be an ordinary physical/functional kind of state.

It is in any case common sense that no amount of conceptual analysis, no working definition, can reveal which process, physical or otherwise, is responsible for sensory experience. Provided only that scientists have the right working definition of what a sensation is, it is an empirical question. At the same time, any empirical theory will be in doubt if it lacks the resources to explain important aspects of the phenomenon it is explicitly about. An everyday example suffices to make this point: if, according to proven theories, collections of H₂O molecules ought to be opaque, then it would be a real question whether or not water is really nothing but H₂O molecules, since as a matter of fact water is transparent. In the same way, if our best neuroscientific theories give no reason to think that the brain processes correlated with pain should feel painful, then it remains an open question whether brain processes are responsible for the feeling of pain, and consequently whether sensations are brain processes.

It is this problem, the explanatory gap, that is the reason for the current abundance of theories of consciousness, and is the focus of Part Three. As this problem was originally put by Levine (1983) it is a conceptual problem—for Levine, it is the problem that since we know a priori that the properties that we ordinarily attribute to sensations, their feel, are essential properties of them, ‘feels’ cannot be made redundant in picking sensations out. In contrast, the discovery that water is H₂O rendered transparency redundant in picking out water. This conceptual claim, aside from what I regard as the implausibility that any particular “feel” is embedded in the shared concept of a sensation, is mistaken for a number of reasons, first among them that it involves an a priori claim about an a posteriori matter—that is, the matter of which are essential properties of a manifest kind. For any manifest kind, like water, or gold, or pain, it is an empirical question whether the attributes by which we normally pick it out are also the attributes that mark it, and therefore are essential to it, as a scientific kind. To deny this with regard to sensations is to block from the beginning a science of sensations, which I will argue there is no need to do.

The flip side of that coin is that neither can we rule out a priori that their feel is not an essential property of sensations. Moreover, the apparent lack of any theory that looks as though it might provide the resources to make references to “feels” redundant, is perhaps itself some evidence that feels really are an essential property of sensations. This would have been the case for transparency had it turned out, counterfactually, that water lacked any kind of substructure—for in that case, seeming like water in normal conditions would have been the only test (along, perhaps, with “lacking substructure”) for the presence of water, and would therefore have counted as an essential property.

In the case of sensations, there would be a case that their “feel” is an essential property if there is no biological, computational or functional theory according to which certain brain processes ought to feel a certain way. Some philosophers maintain that this is inevitable. Though I am not alone in thinking that this is premature, I do think that there is a sense in which it is justified. If we view the way sensations feel as a species of appearance, as the way our sensory states appear to us, then as a matter of logic no description of sensory states could account for their feel. It is a matter of logic because appearance properties are relational, yet “feels” are often treated as if they were intrinsic. This would explain, I will argue, why it seems impossible to imagine a neuroscientific explanation of sensory qualities.

To make sense of the idea that “feels” are appearances, I draw on higher-order perception, or “inner sense”, theories of consciousness, as proposed by Armstrong (1968) and particularly as defended by Lycan (1996). According to that view, our awareness of sensory qualities is brought about by the literal perception of sensory states. It requires the postulation of a perceptual process that is internally directed. Although I am sceptical of the idea that an inner sense is responsible for consciousness, as part of an explanation of sensory qualities I think it is extremely promising. Recently, the neurophysiologist Wolf Singer (1998) has speculated along these lines. If we perceive our sensory states then it follows that there is some quality that we perceive them to have. It does not, however, follow that the perception of that quality contains meaningful information about its physical basis. In the same way, our perception of colour does not tell us about colour’s physical basis. And like a common view of colour, the physical basis of sensory qualities might be interpreted dispositionally, as that set of properties that are responsible for certain perceptual states. It is impossible to find such a set without reference to their role as the objects of a perceptual process. As a set of properties, they may be of scientific interest only by virtue of that role.

If I am right, sensory states do not have sensory qualities purely because they are the kind of states they are, but also because we perceive them to be a certain way—to have a certain feel. The explanatory gap is a consequence of the failure to see this. Moreover, if the way sensations feel is an inner appearance, it also allows for the possibility that perception itself is entirely representational without transgressing Block’s insistence that our awareness of external perception outruns our awareness of its content.

Whether sensations are brain processes may thus depend on whether brain processes best explain perceptual content. What I offer, then, is a theory compatible with functionalism insofar as the best explanation for perceptual content is a functional explanation, and compatible with the identity theory insofar as the explanation is specifically biological. Needless to say, the two are compatible with each other if the best explanation is given in terms of biological function. On this specific issue I wish merely to claim that the question remains open.

Consistent with Mill’s credo, language is as it were the atmosphere of this thesis. Although I certainly do not believe that I have made transparent the terms we use to discuss sensory qualities, I have tried to clarify the relationship between the debate about those terms and the debate about their referents. Part of that task involves engaging in both debates. Strongly held intuitions tend to structure the flow of argument and, moreover, must be tackled seriously to avoid the sense that the subject has been changed. I believe that in the case of sensory qualities many apparent conflicts of intuition are merely apparent and can be resolved by an adequate theory of sensory terms and the states and qualities they refer to. Building on the work of others, this thesis is intended to be a step in the direction of that theory.

Part One –Methodological

1. Kind Words –Kripke vs. Jackson?

n the face of it, sensory experiences are an appropriate object, in and of themselves, for scientific study. They are the object of a particular sort of curiosity—empirical curiosity. Moreover, it is what they have in common, or seem to have in common, that elicits that curiosity. Our empirical interest in sensory experiences is therefore an interest in them as a natural kind. But empiricism demands that we allow for the possibility that what seems to be a natural kind may turn out to be a collection of divergent phenomena, which the nature of our restricted perception has inadvertently lumped together. It is the world, not our access to it, that makes a group of objects the members of a natural kind. It is therefore an empirical question whether the members of what seems to be a natural kind do in fact belong to one. At the same time, the semantic properties of the words we use to talk about apparently natural kinds constrain the direction of arguments about them, and therefore the direction of empirical work concerning them. The first part of this thesis is concerned with the relevance and nature of those constraints in general, given that both are a matter of contention. Afterwards, I will defend a particular set of constraints in the case of sensory experiences.

I will call the referent of any natural kind word, regardless of whether it does in fact pick out a natural kind, a manifest kind, [12] and a kind explicitly picked out by a set of empirically discovered properties a scientific kind. Water is a good example of the first, and will be used as such many times in this thesis; H₂O will serve as an abbreviation of the second. [13] For any manifest kind we can ask the question, which scientific kind is it, if any?

How, in a general sense, is the answer to be justified? According to some, the general strategy initially involves inferring a priori that the manifest kind in question generally fits some description, referred to variously as the ‘folk theory’, ‘ordinary conception’, or ‘conceptual analysis’ (Lewis 1972; Chalmers 1996; Jackson 1998). Once it is empirically established that anything fitting that description also falls under a scientific kind, it can be taken to have been established that the manifest kind is one and the same as the scientific kind (or, following Johnston, that instances of the manifest kind are constituted by instances of the scientific kind). Moreover, the reason that statements like “water is H₂O” are necessary is that part of their justification involves an analyticity—namely, in this example, an analysis of “water” that uncovers a reference-fixing description. Those adopting this strategy are the a priori physicalists.

In apparent opposition to this strategy sits the a posteriori physicalists, a position inspired by Kripke (1980). Naming and Necessity is best known for containing an argument that some necessary truths are not knowable a priori, and thereby driving a wedge between the classes of necessary truths and those knowable a priori. The statements that he argued can be both necessary and a posteriori are of three kinds: identity statements involving proper names (such as “Bob = Robert”), empirical statements involving natural kind terms (such as “Gold has atomic number 79”), and empirical statements involving indexicals (such as “This table is made of wood”).

Kripke is often taken to have ‘discovered’ a kind of necessity that does not involve entailment, either logical or analytic [14] . Of particular relevance to the current question is that the need for the a priori is apparently bypassed. Instead, statements of the “water is H₂O” (and “sensations are brain processes”) kind might be justified solely through an inference to the best explanation and methodological considerations such as simplicity. This possibility has the welcome consequence that to show a good enough correlation between occurrences of sensation S and occurrences of brain process B might be enough to be justify the hypothesis that the two are identical.

The arguments that Kripke gives in Naming and Necessity have led to a widespread belief, in philosophical circles, that one needn’t demonstrate an a priori connection between the mental and the physical in order to establish that mental states are identical with physical states. The recent attacks on the plausibility of such a connection often rely on these arguments. Block and Stalnaker (1999) rely on them explicitly to separate ordinary conceptions from the questions of analyticity and necessity:

What lies behind Kripke’s cases is the fact that the meaning and reference of our terms depend on empirical facts, facts that we might be ignorant or mistaken about. (Block and Stalnaker, p. 6)

Furthermore, where “water role” is our ordinary conception of water,

Perhaps the semantics of ‘water’ is more like [a proper name] than it is like [a descriptive name], in which case there is no way to fill in the details of ‘the water role’ so that it is a conceptual truth that water occupies the water role. And of course it is even more doubtful that any such analysis of the water role would be…a conceptual truth. (p. 16)

Given the unavailability of analyticities, then, the identity between water and H₂O, and therefore between sensations and brain processes, might then be justified along the following lines:

Suppose one group of historians of the distant future studies Mark Twain and another studies Samuel Clemens. They happen to sit at the same table at a meeting of the American Historical Association. A briefcase falls open, a list of the events in the life of Mark Twain tumbles out and is picked up by a student of the life of Samuel Clemens. “My Lord”, he says, “the events in the life of Mark Twain are exactly the same as the events in the life of Samuel Clemens. What would explain this amazing coincidence?” The answer, someone observes, is that Mark Twain = Samuel Clemens. (Block and Stalnaker, p. 24)

The idea, in the case of sensations, is that once enough facts about brain states line up in the right way with facts about sensations, we become justified in putting forward the hypothesis that sensations are brain states. It would then be wrong to say that there is an entailment from the statement “X is a (particular) brain state” to “X is a (particular) sensation”, though the truth of the former does necessitate the truth of the latter, by metaphysical rather than analytical necessity. The inference would be explanatory rather than logical.

Crucially, in this picture, since our ordinary conceptions play no role in fixing the reference of sensations words, it is illegitimate to conclude anything about sensations from them. Our ordinary conceptions might simply be wrong. Whether this view is consistent with actual scientific practice, and whether it is really supported by the Kripkean view of natural kind words, is what I will discuss in the first chapters of this thesis. I begin, then, by discussing the role of ordinary conceptions in so-called “scientific identities”, such as between gold and the substance of atomic number 79. This is a necessary precursor to the later chapters, though it will not concern sensations specifically.

Those who defend a substantial role for the a priori in identity claims between manifest kinds and scientific kinds often use water and H₂O as an example. But it is conspicuous that the discovery that water is H₂O did not, at least not obviously, involve a priori arguments. There is therefore a prima facie case for the view that a priori arguments are unnecessary, and may be serving to clog up the research rather than facilitate it (witness Farah’s tortured qualification to her own research in the quotation at the beginning of the Introduction)—that scientific discoveries, being empirical, need not deal in conceptual analyses.

That is at any rate something like the view held by Block, Stalnaker, Byrne, Van Gulick, Tye and others. Block and Stalnaker express it most directly and tend to rely on the view of “a posteriori necessity” put forward by Kripke. It is an ill-founded reliance and an unsustainable position, as I will try to demonstrate in the next two chapters.

1.1 Necessity and Manifest Kinds

1.1.1 Rigid designation

Kripke put forward the view in Naming and Necessity (1980) that natural kind words, such as “gold” and “tiger”, refer directly to a kind, independently of the way we ordinarily talk about them. According to Kripke, in most cases the properties we ordinarily think of gold (for example) as having—yellowness and so on—are contingent, while the reference of the word is fixed by its essential (that is to say, necessary) properties, which are knowable only through scientific investigation. “Gold”, then, refers to the very same kind of substance as, for example, “a substance composed of atoms of atomic number 79”. The relationship between the “manifest” property, gold, and the scientific/theoretical property, having an atomic number 79, is identity, on this view; they refer to the same type of substance, named on the one hand, described on the other. What makes this possible is the linguistic mechanism of rigid designation.

A word rigidly designates if it refers to the same thing in every possible world. Paradigmatically, a proper name is a rigid designator: the name of a person refers to that person in every possible world [15] . But names are not the only linguistic types that rigidly designate: indexicals such as “here”, “now”, and “I” also have the same referent in every possible world as at the world of utterance. According to David Kaplan [16] , whose opinion I am inclined to follow, names and indexicals refer to objects independently of their sense or associated descriptions. How rigid designators achieve this is still at least to some extent an open question, though Kaplan suggests the following mechanism: even though “I” can be said to mean something like “the person now speaking or writing”, the content of a proposition expressed using “I” includes the person referred to, not the meaning in that sense; the person is “loaded into” the proposition. The same goes for the other rigid designators. The rigid designation then ensues in the following way (in Kaplan’s own striking metaphor):

If the individual is loaded into the proposition (to serve as the propositional component) before the proposition begins its round-the-worlds journey, it is hardly surprising that the proposition manages to find that same individual at all of its stops, even those in which the individual had no prior, native presence. The proposition conducted no search for a native who meets propositional specification; it simply ‘discovered’ what it had carried in. In this way we achieve rigid designation. (p. 569)

That is not to say that there isn’t something like a Fregean sense attached to a rigid designator, merely that it does not seem to fix the referent [17] . “I” and “the person now writing” are not synonymous, even though there is, in a sense, nothing more to “I” than its reference to the person now writing. Note also that if a word’s sense contained an indexical, then that word could thereby rigidly designate.

1.1.2 Natural kind words as name-like

According to Kripke, natural kind words are somewhat like names in the way they refer. So, if Kripke is right, this lump of gold that I am holding, this particular lump, could not have been what it is without being gold; to imagine it changing to lead is to imagine it becoming a different lump. On the other hand, to imagine it shrinking to half the size is to imagine the same lump being half the size it was previously. So, “gold” as the name of a substance rigidly designates.

What does that mean, and why is it relevant? The short answer is that if natural kind words rigidly designate, designating not particular objects but types of object, then they pick out the necessary, or essential, properties of the objects that have them. What this could mean, to anticipate a little, is that scientific investigation picks out those essential properties, while everyday discourse may pick out the merely contingent ones. If so, we have managed to make the relationship between the two ways of talking clear, and established that the cluster of properties picked out by scientific investigation are the very set of properties that characterise the type named by the natural kind word. So, to be a sample of H₂O just is to be water—“water” refers to H₂O in every possible world. It would be convenient if “pain” functioned in the same way; if we could say, after scientific investigation, that pain just is brain state such and such. Kripke, and many following him, have thought that that is out of the question, but I will argue that it is not. In any case, for Kripke’s argument to work, he still needs to tell a story that includes how it is that natural kind words rigidly designate, what the role of scientific investigation is, and how we are to think about the properties we ordinarily associate with particular kinds, such as yellowness with respect to gold. Needless to say, he has such a story.

1.2 Kripke’s historical view of natural kind words

1.2.1 Naming the kinds

According to Kripke’s (idealised) sketch, natural kind words get their reference in much the same way that proper names get theirs, namely, by a kind of baptism. People, or things, are named by someone pointing at them, or some equivalent of pointing, and saying something like “this person shall be called ‘John’”. The crucial thing in this picture is the demonstration, the pointing out. It is the demonstration that, to use Kaplan’s terminology, brings the actual person into the content of the name; thereafter, to use the word is always to refer to the person originally picked out. This is to be contrasted with descriptions such as “the third person south of the fountain”: whoever happens to actually be the third person south of the fountain is not any part of the meaning of the phrase since, had that person not been there, some other person would have been the third person south of the fountain. Ordinary descriptions pick our whatever fits the description, either in fact or counterfactually. Not so with baptisms. Had my mother pointed at some other baby and said “that baby shall be called ‘John’”, it is not true that someone other than me would have been John, since although it might have been someone else pointed out, in fact it was me, and that is what matters. The proposition “John lives in Melbourne”, again in Kaplan’s terminology, includes the very person actually named; no ordinary description seems to do to job, because any such description might fail to hold of me without my thereby ceasing to exist. That is why ‘John’ picks out the same person in every possible world. And that is roughly what Kripke thinks is the case for natural kind terms. This is what he says in Naming and Necessity:

If we imagine a hypothetical (admittedly somewhat artificial) baptism of the substance, we must imagine it picked out as by some such ‘definition’ as, ‘Gold is the substance instantiated by the items over there, or at any rate, by almost all of them’. (p. 135)

To name a type in this manner, by baptism, is to refer to the same type in every possible world. Assuming that a type is individuated by some unique set of essential properties, to name a type is to pick out the same set of properties in every possible world.

Let us grant that. There is still more work to be done: how do we know what those properties are? Vacuously, there is the property “being gold”. But that is no help, anymore than it is helpful to say, upon being asked who John is, that he is the person called “John”. The key is the fact that we have samples of the thing we’re calling “gold”. If we assume that the samples fall under a single kind, we need to know what it is about the samples that makes them be instantiations of the same substance. Taking it to be a fact about the actual world that different substances are individuated by their microstructural properties, it is the microstructure that makes for the samples being a kind. It is therefore the microstructural properties of gold that are its essential properties, and what those properties are can only be a matter for scientific investigation.

1.2.2 Necessary versus contingent properties

When we talk about gold, then – if Kripke is right – we refer, rigidly, to a substance that has that has a certain microstructure – for one thing, it consists of atoms of atomic number 79. Being made of atoms with atomic number 79, then, is a necessary property of gold. What about the properties, such as yellowness, that we normally think of gold as having, and normally use to differentiate gold from other substances?

According to Kripke, when gold was originally so named, those naming it needed a way of identifying other samples of the same kind. Towards that end, they naturally checked the original samples for identifying characteristics (yellowness, pliability etc.), and thereafter associated those characteristics with gold. Kripke emphasises that they did not, in doing so, affect the meaning of the word “gold” in any sense, since it remains coherent that something could possess all of those characteristics originally identified as belonging to gold and yet not be gold. Those baptising the samples “gold” were able to defer the question of what actually made them gold; that is, the question of what identified gold as a kind, distinct from other kinds of substances. The baptism was successful as long as they had a by-and-large successful way of separating gold from things that weren’t gold. And naturally one only needs to be successful at separating actual samples of gold from actual samples of non-gold; what gold might counterfactually be like is irrelevant to the question of which properties to associate with gold, at least for the purpose Kripke had in mind. Thus, the fact that we tend to think of gold as being yellow has no bearing on whether gold is or is not yellow in other possible worlds.

Of course, it is true that we tend, in a sense, to not merely associate “yellow” with “gold”; we tend to pick out samples of gold on the basis of, among other things, their yellowness (as opposed to being valuable, which is perhaps not a quality used to pick out samples of gold and therefore, in a sense, merely associated with the substance gold). But the way we pick out a person named “X” is irrelevant to the question of whether or not they are X; the only thing relevant is whether or not they are the person originally baptised X. It is of course possible to be mistaken about which person a proper name picks out. Similarly, it possible to be mistaken about what kind of thing a word picks out. The properties we associate with a type, or a person, may not be properties that are essential to them. The fact that I live in Melbourne could fail to hold of me without my failing to be me, but the fact that I was born in the twentieth century, so some thinking goes, could not be false of me, since if someone like me in all other respects was born in some other century, they would have been someone else, and not me, not this very person now writing. Similarly, so Kripke argued, gold has some properties that are essential to it, without which it would not be the very type of thing we in fact call gold, and some that are contingent, without which it would still be the very kind of thing we in fact call gold, but under a different presentation.

1.2.3 The necessity in associated descriptions

So, in Kripke’s view, the properties we originally considered to hold of natural kinds, the properties we ‘associate’ with those kinds, are merely a rough-and-ready way of distinguishing them from other kinds, and were never meant to hold of counterfactual situations. They are, in other words, contingent.

But an interesting question must be raised here: could gold turn out, after scientific investigation, to all along have had very few, or none, of the properties it appears to have? Here is what Kripke says:

Just as something may have all the properties by which we originally identified tigers and yet not be a tiger, so we might also find out tigers had none of the properties by which we originally identified them. Perhaps none are quadrupedal, none tawny yellow, none carnivorous, and so on; all these properties turn out to be based on optical illusions or other errors... (p. 121)

But this is too quick [JOD1] . If we imagine that all of the properties we thought were true of tigers, such as those Kripke mentions, are in fact merely apparent properties, and that the appearance is caused by a reaction between the atmosphere, the light, and specks of moon dust. Would we, in this case, say that tigers turn out to be specks of moon dust? I doubt it. We are more likely, I suggest, simply to say there are no such things as tigers, and that the word “tiger” fails to refer; tigers are mere illusion. That is not to say, however, that Kripke’s general view about the reference of natural kind terms can’t accommodate this intuition; he needs merely to point out that the original baptism was unsuccessful. We can imagine something similar occurring in the case of a proper name baptism. Imagine, for example, that I visit a hospital and, on seeing what appears to be an infant in a crib, point towards it in say “you there, in the crib, are Albert!” Unbeknownst to me, however, the only thing in the crib is a tiny holographic projection device that makes it appear as though an infant is occupying it. In that case, have I named the projection device “Albert”? It seems not; I have named nothing at all. Even though there was something I was pointing at, it was not sufficiently similar to the thing I had in mind to be called “Albert”. [18] Such failures of naming are, I suppose, relatively commonplace, and there is no reason to suppose that the naming of kinds is any different. There are limits on how much a thing, or a kind, can vary from what it is thought to be and still be pointed out and named.

If so, Kripke’s view is relatively unaffected. But since Kripke’s stated view does not seem to allow for the possibility that a natural kind word may fail to name anything, we will have to settle with calling it the Kripke* view [JOD2] . So, according to Kripke*, natural kinds are named by the baptism of a set of ‘original’ samples, which are checked for identifying characteristics. Those characteristics are then associated with the kind (and therefore the name of the kind), but are not intended as criteria covering counterfactual situations, and are therefore merely contingent properties of the kind. The natural kind word itself, via the baptism, rigidly (that is, in every possible world) attaches to whatever property it happens to be that makes the sample picked out a natural kind. In the case of the actual world, that property (or conjunction of properties, as is more likely the case) is likely to be microstructural, and discoverable only through scientific investigation. It is up to science, then, to tell us what the properties are that individuate kinds, and what are thereby the necessary properties of those kinds.

Note that the difference between “water is wet” and “water is a compound” is not just that the former is part of what might in the current literature be called the “folk theory of water” (that is, part of our ordinary conception of water) while the latter is part of the scientific theory of water. The scientific study of water might reveal that water has the very odd property of turning green at 1 million degrees Kelvin. Still, we might think that in other possible worlds, water (the very kind we’ve named “water”) fails to have that particular property. In other words, scientifically obtained true statements about kinds do not automatically come with the tag “necessary” attached to them. Nevertheless, our judgements about what individuates kinds seems to depend crucially on investigation as to what the actual world is like, and that is the most that Kripke needs.

To recap briefly, for Kripke* gold is necessarily the substance which actually has a fair number of properties originally associated with the word “gold” [JOD3] . “Originally” is important because we may today associate a different – and incompatible – set of properties compared with the properties originally thought to hold of what was baptised “gold”. It is the characteristics originally identified that help us, through investigation, find samples of the kind of thing originally baptised. Those characteristics, or associated properties, do not fix the reference of the term, on this view, but they are useful in finding the referent.

There is another, currently popular, view of the route by which natural kind terms refer. It is a view that I largely concur with, though the similarities between it and the Kripke-like view are not, it seems to me, universally appreciated. Seen in the proper context, I will argue, the views are not as dissimilar as is often assumed. In particular, I will show that the recent debate between so-called “a priori physicalism” and “a posteriori physicalism” does not hang on anything substantial in Kripke’s theory of a posteriori necessity.

1.3 Jackson and Two-Dimensionalism

1.3.1 The basic idea

The two-dimensional view, defended by Jackson (1997, 1998a, 2001) and Chalmers (1996, 1999) [19] , is consistent with Kripke’s view insofar as it takes rigid designation to be the key to understanding such necessary a posteriori statements as “water is H₂O”. Where it differs, is in how rigid designation is understood to be achieved.

To get the flavour of two-dimensionalism, compare these two sentences,

(A) The fastest runner in the world is also the best long-jumper.

(B) The current fastest runner in the world is also the best long-jumper.

Notice that there is some ambiguity in (A); it could either mean that whoever is the fastest runner is also the best long-jumper (which is certainly false), or that the person who happens to be the fastest runner in the world at the moment also happens to be the best long-jumper (which is possibly true). Clearly, (B) is the second disambiguation. Take also this sentence:

The same ambiguity exists in (C), though of the two disambiguations, namely,

(C*) Possibly, whoever is the fastest runner is not the fastest; or

(C**) Whoever is the fastest runner, that person might not have been the fastest.

only (C**) is consistent (and true!). For any possible world, the person who is the fastest runner is the fastest runner, though for any particular person who happens to be the fastest runner, there is another possible world in which that person is not the fastest (Maurice Greene happens to be the fastest over 100m, but he might not have been; he might’ve taken up swimming instead).

The point of these examples is that many sentences that seem to involve ordinary descriptions actually incorporate elements which have the effect of rigidifying that description. The description “the fastest runner” played different roles in the above sentences depending on whether it was meant to be a general description ranging across all possible worlds, or whether it was meant to pick out an individual in a world as the subject of the sentence. The two dimensions referred to by two-dimensionalism can be thought of as, on the one hand, the description as such—the fastest runner—and, on the other hand, the individual picked out in a particular world by the description—in the actual one and at the moment, Maurice Greene. Insofar as the role of a description is to pick out a particular individual in a particular world, the modal properties of a sentence in which it is used will be the same as if a proper name had been used. In other words, a description, when used to pick out an individual in a world, can rigidly designate that individual.

1.3.2 Kind words as rigidified descriptions

Davies and Humberstone (1980) called attention to this phenomenon, and suggested that it may be useful in describing the cases of so-called “a posteriori necessity” that Kripke had made famous. They proposed that when a description, such as “the fastest runner” is prefixed by “the actual”, that description takes on many of the properties of a name; it becomes, in their words, a “descriptive name”:

suppose that ‘water’ is a descriptive name with its reference fixed by the description ‘the chemical kind to which that liquid belongs which falls from the clouds, flows in rivers, is drinkable, colourless, odourless, …” so that with w1 as the actual world the reference of ‘water’ with respect to w2 is that chemical kind of stuff which in w1 falls from clouds,… To understand ‘water’ it would not be necessary to know which chemical kind actually has those properties and so it would be an a posteriori discovery that water is H2O… (p. 18)

So, for “water” to have the semantic properties suggested in this passage, it would need to have both a description and a rigidifier. The description as such (namely “the chemical kind to which that liquid belongs which falls from the clouds, etc.”) picks out different chemical kinds in different possible worlds – much like “the fastest runner” refers to different people at different times. But when it is rigidified, as in “the actual chemical kind to which that liquid belongs…”, it names a particular kind, and therefore refers to that kind in every possible world – much like “the current fastest runner” refers to a particular individual, who is that same individual at all times even if not the fastest runner at all times.

If all natural kind terms are descriptive names in this sense, or if you like, rigidified descriptions, then we no longer need samples of a natural kind to have been baptised with the name we currently use in order to explain the necessity of a posteriori statements such as “water is H₂O”, or “gold has atomic number 79”. If “gold” is a rigidified description, whose description is something like “the metal belonging to that kind of substance which is yellow, pliable, found in hills ...”, then to understand what is meant by “gold” is not to know what individuates gold as a natural kind; scientific investigation may be required. Nevertheless, “gold” refers to that same natural kind in every possible world. Therefore, it is necessarily true (if it is true at all) that gold has atomic number 79, while at the same time it is knowable only a posteriori, after empirical investigation.

1.3.3 Jackson on two-dimensionalism and conceptual analysis

Now if two-dimensionalism is correct and natural kind words are rigidified descriptions, still the descriptions to be rigidified cannot be plucked from the air. The recent work of Frank Jackson (e.g., 1997, 1998c and especially 1998a) has in part constituted a sustained defence of conceptual analysis, conceived as reflection on actual and possible cases—a somewhat maligned practice in philosophical circles since Quine and Putnam’s attack on the analytic/synthetic distinction. It is conceptual analysis, Jackson has argued in many places, that yields the right descriptions. He maintains that the descriptions we need are constituted by our “ordinary conceptions” of the world around us:

But how should we identify our ordinary conception? The only possible answer, I think, is by appeal to what seems most obvious and central about [for example] free action, determinism, belief, or whatever, as revealed by our intuitions about possible cases. Intuitions about how various cases, including various merely possible cases, are correctly described in terms of free action, determinism and belief are precisely what reveal our ordinary conceptions of free action, determinism, and belief, or, as it is often put nowadays, our folk theory of them. (1998a, p. 31)

The conditions under which we apply the concept ‘water’, in this way of thinking, will contain a list of properties we associate with water, such as filling the lakes and river, falling from the sky, being clear and tasteless, and so on.

This point is sometimes put by saying that, in some sense, we knew what water was before we knew what water was. The idea is that, before we knew that water is a collection of H₂O molecules, we had an idea in our minds of what it is for something to be water. It is hard to deny – the alternative seems to be that prior to the twentieth century, separating water from other substances was complete guesswork! Whatever that was, whatever we had in our minds about what it was for something to be water, that is part of what it means to understand the word “water” (in English of course), on the Jackson view.

It is worth pointing out here that the idea that we knew, in some sense of “know”, what it was to be water before we knew that water is H₂O, is by no means anathema to Kripke’s view. There is an important place in Kripke’s picture for descriptions associated with natural kind terms. They are, for Kripke, “the properties originally known to hold” of the kind, and have become, generally speaking, the properties we ordinarily associate with it. Those associated properties are what explain, for Kripke, the apparent contingency of necessary a posteriori statements. The properties associated with water, such as falling from the sky, may well be true of substances other than water in other possible worlds; since we think of water as falling from the sky and such, it is easy to be misled into thinking that whatever falls from the sky (and such) is water, and thereby to be misled into thinking that water may not be the same as H₂O in some possible worlds. Kripke denies that it is analytic that water is clear, falls from the sky and so on, but his view requires him to say that reflection on what it is for something to be water will reveal a list of properties that, by and large, hold of water in the actual world. Without that list attached somehow to the concept ‘water’, he has no way of explaining the apparent contingency just mentioned. And all Jackson’s conceptual analysis of ‘water’ requires in this context is that it yields exactly that, viz., a list of properties that, by and large, hold of water in the actual world. There is no reason to suppose that Jackson’s and Kripke’s respective lists will differ other than with respect to how they are understood to be attached to the word “water” (both lists will exclude, for example, properties that we learned held of water long after we learned the meaning of “water”, such as that it exists on a moon of Jupiter). For Jackson, the list is contained in the concept ‘water’; to learn the meaning of “water” is to learn the list. For Kripke, the list is separate from the concept; one must learn the meaning of the word “water” and then learn the list of its associated properties [20] . Whoever is right, the same kind of armchair reflection on water is necessary according to both views, even if one calls it “conceptual analysis” and the other does not. This point will become important later, but for now I will leave it aside.

In any case, the list of properties of water is ordinarily (or “originally” for Kripke) taken to have does a lot more work in Jackson’s picture than it does in Kripke’s, even if it is nonetheless essential to both. In Jackson’s view, the list of properties is one part of the concept ‘water’. The other part is the rigidifying part. The concept of water, according to Jackson, is not merely that it is the clear liquid that falls from the sky and so on; it is, moreover, the actual clear liquid that falls from the sky and so on. The “actual”, or something like it, is the reason “water” rigidly designates, in this view.

Many kinds of words exhibit the two dimensions of two-dimensionalism in some way, according to Jackson; but for most of them, the two dimensions are not distinguishable. The word “square”, to use one of his examples (From Metaphysics to Ethics, p. 49) applies to the same kinds of things from the perspective of any possible world. That is because “square” does not rigidly designate. For words that do rigidly designate – such as natural kind words – the two dimensions disjoin, since they will pick out different kinds of things depending on which possible world is the actual world [21] .

1.3.4 The respects in which Kripke and Jackson agree

So Jackson and the supporter of Kripke* view of reference can agree that the set of properties associated with a natural kind term have some importance to the reference of the term, but only in the sense that they serve to pick out samples of the thing that science must investigate to discover its essential properties. Gold, then, is the substance that does in fact have a fair number of the properties associated with it, though those properties form a set neither of necessary nor sufficient conditions for a thing’s being gold.

According to Jackson, conceptual analysis tells us what properties a thing needs to have in order to be counted as gold in the actual world (not, importantly, in counterfactual worlds). According to Kripke, at the time of the original baptism, the original sample is checked for its identifying characteristics, which become its associated properties. How do we now know, after all this time (since the baptism), what those properties are? Kripke wants to stress that they are not part of the meaning of the terms, but they are closely tied anyway; to learn the word is practically always to learn the associated properties, while being aware that its reference is just whatever is the kind of original object that was baptised. Still, the properties associated with “gold” are useful but not essential for finding the referent of the word; an etymological dictionary ought to be more useful than a standard dictionary. Not so for Jackson. On his view, the properties currently associated with “gold” directly fix its reference. To discover gold’s essential properties, what gold is necessarily, we need to find samples of it to study. We pick out those samples using the set of properties we associate with the word “gold”, no member of which we know a priori to be essential to a thing’s being gold. Conceptual analysis, according to this picture, is the only way we have for connecting the way we currently talk about gold and the way that scientists talk about gold. Jackson (1998a) remarks thus [JOD4] :

conceptual analysis is the very business of addressing when and whether a story told in one vocabulary is made true by one told in some allegedly more fundamental vocabulary. (p. 28)

According to Kripke, conceptual analysis in the sense that Jackson means it (reflection on possible and actual cases) is helpful but not always essential and sometimes downright misleading. The connection between “vocabularies”, in his picture, is made when we know that scientists are talking about the very same kind of thing that was originally baptised. Conceptual analysis is all well and good, but a good historical account of the history of a word’s use will always trump it.

In any case, the necessity (if true) of statements such as “gold has atomic number 79” comes about both for Kripke and Jackson because purportedly natural kind words rigidly designate. Where the two differ is in the reason they give for why such words rigidly designate. For Kripke, natural kinds words refer via a causal chain leading back to an original baptism of a sample that could not have been substantially different to what it in fact (actually) was. For Jackson, natural kind words refer via an associated description that includes, or is prefaced by, “the kind of thing that actually, by and large, has the following set of properties…”; the embedding of this passage in quotes, or something like it, within a concept is what makes it a natural kind concept [22] .

The upshot is that, according to both views, it is an a priori truth that if gold has atomic number x in the actual world, then the sentence “gold has atomic number x” expresses a necessary truth. And according to both views, reflection on the properties “originally” associated with gold (whether “originally” reflects either investigation at the time of baptism or embeddedness within the concept) generally yields the referent of “gold” in the actual world, which can be scientifically studied for its properties (for Jackson, of course, it always yields the referent – ergo it generally yields the referent).

Jackson seems to take the two-dimensional view to be the best explanation for the rigid designation of natural kind words. Kripke, on the other hand, takes the historical view to be the best explanation. Of the two, Kripke’s story is less plausible, on at least two grounds. Firstly, it is historically implausible; it is obviously very unlikely that an actual baptism ever took place of gold, water or any other natural kind. Kripke says as much himself in describing as “somewhat artificial” his gold-dubbing example. But mere artificiality is not the problem; it is difficult to imagine anything at all like a baptism of gold taking place. As an historical claim, Wittgenstein’s “light dawns gradually over the whole” [23] seems a more plausible metaphor. In any case, secondly, the historical view seems inconsistent with our actual practice. This is pointed out by Evans (1982; p. 390). According to the historical account, a natural kind term could no more change its reference than history could be changed, but if we were to discover that the word “gold” was actually originally used to baptise grass it is implausible to think that that would cause us to affirm the statement “gold is green”. The reason, of course, is that etymology is simply not sufficient for discovering the reference of natural kind words we currently use [24] ; but if Kripke is right it ought to be.

At any rate a rigidified description view of natural kind terms is a defensible alternative to the causal/historical view. It shares neither the problems just mentioned with that view nor the well-known problems with the Russellian descriptivist view. I will not offer more defence of it here, since my main task is to defend the position with regard to the problem of consciousness in particular. From here on, then, I will assume that a semantic theory of natural kind terms similar to Jackson’s is correct.

Since the idea of a physicalism that is justified a posteriori springs from notions about a posteriori necessity generally traceable back to Kripke, or a view very much like Kripke’s, it will be important to at least have a Kripke-like view in mind. That will particularly be the case when we come, in the next chapter, to the attack on conceptual analysis, which is driven largely by a posteriori physicalists such as Block and Stalnaker in their recent paper. The view I have put forward here suggests that even Kripke has a place in his view for something very like conceptual analysis. Philosophers such as Block and Stalnaker deny that conceptual analysis has a place in a posteriori identifications such as often happen in science but have nothing in the way of an alternative method for doing the work that Kripke and Jackson agree needs to be done, viz. finding the properties ordinarily associated with a natural kind word. But we will come to Block and Stalnaker’s view before long. Now that the groundwork is laid, we are in a position to discuss the current debate between those for whom the relationship between consciousness and the explicitly physical world is an a posteriori matter, and those who take it to be a priori.

2. The Indispensability of Ordinary Conceptions

hysicalism is made true by the way the world is. It is the empirical hypothesis that, were the world to be duplicated physically (and only physically), it would be duplicated in every other way too. [25] It is a thesis that should be distinguished from a different but similar claim that a duplicate of all of the fundamental properties will be a duplicate simpliciter. This latter claim is arguably vacuous; arguably, the only way of defining “fundamental” in the context of the last sentence is something like “that set of properties which underpin all of the others”. It is vacuous because it is hard to see what could make it false. If, for example, dualism is true, then mental properties (or perhaps proto-mental properties [26] —the simples of which mental properties may be complexes) would have to be counted amongst the fundamental properties. Physicalism is the view that whatever makes up the list of fundamental properties, mental properties aren’t on it; neither are social, economic or biological properties. According to physicalism, only physical properties are on that list of fundamental properties, though that there is such a list is true almost by definition [27] . The physicalist does not deny that there are social, economic or biological properties, merely that to the extent that there are such properties they are underpinned by the physical ones. The world is the way it is economically in virtue of the distribution of physical properties. So too with the way the world is psychologically and—in particular for this thesis—the way it is phenomenologically.

According to physicalism, then, phenomenal properties aren’t on the list of fundamental properties, since they are not physical properties. That does not mean that they aren’t fully-fledged, robust properties, merely that their instantiation depends on other properties being instantiated. Once those other properties are instantiated, the phenomenal properties are a fait accompli, since to the extent that they are underpinned by physical properties they are necessitated by them.

Of more importance to this thesis, however, is the epistemology of physicalism rather than its metaphysics. In particular, how do we know that sensations do not represent a counter-example to physicalism, as Chalmers (1996) forcefully argued, and many others before him? There are at least two ways of thinking about a solution to this problem. Both of them involve scientific investigation into the nature of the world. One of them, a posteriori physicalism, also involves scientific investigation into the nature of consciousness qua consciousness – its causes and effects, and so forth. The other, a priori physicalism, involves an investigation into phenomenal concepts rather than phenomenal properties. According to a priori physicalism, when we have a clear understanding of what we mean by words like “pain”, we will discover that the only thing they could plausibly refer to are brain states, given what we know about the brain. An investigation into the properties had by our sensory experiences qua sensory experiences will not help with that process, if this view is correct.

In this chapter I will argue that there are no convincing arguments in favour of a posteriori physicalism. Those who hold it typically rely on Kripke’s view of reference-fixing, ignoring the practical necessity of using ordinary conceptions even for that view. Since, as even Kripkeans must admit, we are ignorant of what is for them the appropriate referential causal chains, ordinary conceptions must inform original research into manifest kinds, including sensations. In the next chapter I will begin to defend a particular view of our ordinary conceptions of sensations.

2.1 Physicalisms A Priori and A Posteriori

2.1.1 Physicalism as an entirely scientific matter

To begin, we should look at the more popular view, a posteriori physicalism. According to a posteriori physicalists, the name of the game is to find the best explanation for the way we discover the world to be, which is in no sense an a priori exercise. In the case of phenomenal properties, then, the aim is to find out what properties sensory experiences have, and then find out if there is anything else in the world that also has (at least most of) those properties. If there is something, then the best explanation is probably that there is only one kind of thing involved, rather than two things with the same characteristics. Here are some examples of how a posteriori physicalists see the process. The first passage, quoted earlier, is taken from Block and Stalnaker (1999), in which they take issue with a priori physicalism; the second passage is from a paper by Robert Van Gulick (1997):

The more one can articulate the structure within the phenomenal realm, the greater the chances for physical explanation; without structure we have no place to attach our explanatory “hooks.” There is indeed a residue that continues to escape explanation, but the more we can explain relationally about the phenomenal realm, the more the leftover residue shrinks toward zero. (p. 565)

In the first passage, Block and Stalnaker are describing what they take to be an analogy with scientific investigation. No need here to worry too much about the right analysis of the concepts we are using to refer to the phenomena under investigation. As long as we’re all clear about what we’re referring to, our energy is better spent looking at what we’re referring to, rather than the way we refer to it. Just as no amount of study into the way “Mark Twain” refers will help us greatly to work out whether Mark Twain = Samuel Clemens, similarly no amount of study into the way “pain” refers will help us greatly to work out whether pain = some brain state.

There is a lot of intuitive plausibility in this position. Plausibly, the discovery that water = H₂O came only after a lengthy investigation into the characteristics of water qua water and also an investigation of the way the world is at a microphysical level. Van Gulick’s point is also credible; surely, the more we knew about water qua water – that it boils at 100°C and so on – the easier it was to point out similarities between it and a substance synthesised from hydrogen and water.

2.1.2 Why scientific criteria are not independent of everyday criteria

So there is certainly something right about the picture described by Block and Stalnaker and Van Gulick. But they may have overstated their case. It helps if we explore a little further the discovery that water is H₂O. We can imagine that the best evidence for the identity of water and H₂O would have been the discovery that water can be synthesised from, and broken down into, a mixture of two parts hydrogen and one part oxygen. Now apart from being able to identify hydrogen and oxygen, the researchers needed to know at least two things: firstly, that the liquid samples that they were breaking down, through electrolysis, into hydrogen and oxygen were indeed water, and secondly that the liquid they were synthesising from hydrogen and oxygen was indeed water.

How did they know that? A posteriori physicalists like Block and Stalnaker, following the Mark Twain/Sam Clemens analogy, will argue that water was identified by those experimentalists in virtue of characteristics scientifically known to hold of water. It seems plausible, however, that they needn’t have; had they simply tasted it, found it to be inflammable and boil at the right temperature, and got it straight from the local river, then they had enough proof that it was really water in their containers. But my knowledge of the history of chemistry isn’t good enough to say for sure whether or not that would have been the case, so let’s grant that there was some ‘scientific’ test performed on the samples to determine if they were water. What epistemological role did that test have in the experiment which, we can imagine, led to the identification of water and H₂O? In order to carry out the test (that is, in order to ‘scientifically’ test that a sample is a sample of water) there needed to have been a set of characteristics ‘scientifically’ known to hold of water. Those characteristics no doubt told us something about the nature of water. But what role would they have played in the experiment? As I have described the process, the characteristics were a way of identifying water. They would have been useful in distinguishing water from “fool’s water”, so to speak (and, of course, non-water generally). But are those criteria independent of the way we ordinarily pick out water; that is to say, the pre-scientific criteria we once (at least) had for distinguishing water from fool’s water?

Let’s imagine so. In that case, our scientific criteria for identifying water are independent of our pre-scientific criteria for identifying water. But this gives us odd results. For what if our scientific criteria for a thing’s being water rejected quite a few samples that our pre-scientific criteria allowed? What if the criteria rejected all of them? It might be objected that that couldn’t happen simply for historical reasons; at some point, at least one sample that fit the pre-scientific criteria must have been given over for scientific investigation as to its nature. So the two may not be strictly independent, this reply goes, but near enough. But what if the sample was itself fool’s water? Or what if there were many different kinds of water, of which the sample was only one kind? The reply here might be that the original sample could not have been a sample of just one; more than one is required, to ensure that, by and large, the samples are not fool’s water and that the different kinds of water, presuming there to have been more than one, were all in the sample. How big should the sample size be? There seems to be no other way to answer the question than this: big enough to ensure that there will be no substantial clash between the tokens picked out by the scientific criteria and the tokens picked out by the pre-scientific criteria. This does not make for independence between the two sets of criteria.

2.1.3 Ordinary conceptions and the historical view of reference

Those inclined to follow Kripke’s historical theory of reference for natural kind terms will have a slightly different worry, and a slightly different solution. Our pre-scientific criteria for picking out water, on that view, is only relevant insofar as it is useful in picking out the kind of thing originally dubbed “water”; that is, the kind of thing instantiated by the objects present at the ‘baptism’. Still, as Kripke himself notes in Naming and Necessity, the sample of objects originally baptised needed to allow for the presence of fool’s water and also for different kinds of water. The baptism must, presumably, take into account the intentions of the baptiser – what he was intending to name. Otherwise, there is no accounting for the difference between the original sample being a mixture of water and fool’s water, and a mixture of two different kinds of water. The baptiser must, in other words, be intending to rule out fool’s water from the kind named, and rule in a different kind of water. And that can only happen if there is something in the baptiser’s thought (I don’t yet say “concept”) that distinguishes water from fool’s water. In other words, even if the original sample of water was kept for future scientists to study for its characteristics there would still be the question of what to call “fool’s water” and what to call “a different kind of water”, and that could only be decided in terms of what criteria the original baptiser had for picking out samples of water. That problem becomes much more pronounced for those cases in which the original sample is lost (which is presumably the case for every actual instance of a natural kind term), for in that case the pre-scientific criteria are the only criteria for picking out which samples are to be studied for their characteristics under the name of “water”, which is precisely the problem discussed in the previous paragraph.

But what of that? One consequence is that the Mark Twain/Sam Clemens analogy cannot be quite right. A better analogy to ‘scientific’ identities would be that Mark Twain’s life was universally believed to contain certain events, and it was empirically shown that very similar events held true of the life of Sam Clemens. That would not quite hold water with historians, of course, since “universal belief” is hardly a scientifically adequate proof that the events in question did in fact hold of Twain’s life. In the case of water, what would constitute proof that what was generally believed to hold of water actually held? As we have just seen, there cannot be very much divergence. If, for example, scientific investigation found that our ordinary way of picking out water was generally unsuccessful, and that in fact the only water in the world was to be found in Loch Ness, then we would simply have to say that the word “water” as we use it is different from the word “water” as scientists use it. So we can in general say that the characteristics we ordinarily (pre-scientifically) take to hold of manifest kinds by and large hold of them, self-evidently [28] . The self-evidence arises out of the way in which natural kind words get their meaning. If they are rigidified descriptions, then conceptual analysis will yield the appropriate characteristics. If, on the other hand, natural kind words attach to kinds via a baptism then the way the baptiser picks out the kind, what he has in mind when he says “that kind of thing” (or our best reconstruction of what he must have had in mind, in the event it was never passed on), will yield the appropriate characteristics.

So a substantial disanalogy between the Mark Twain/Sam Clemens case and water/H₂O is that, in an important sense, our everyday knowledge of water is sufficient to establish its identity with H₂O, whereas in the former case everyday knowledge won’t do. The difference is that there is something about the way natural kind words refer that makes our ordinary knowledge more robust than is usual. It might be objected at this point that some scientific knowledge of water qua water was crucial to establish its identity with H₂O – for example, we needed to know that water behaved in such a way as to form hydrogen and oxygen under electrolysis. Didn’t that at least help a great deal to establish the identity? So surely it was scientific knowledge of the characteristics of water that was necessary, even if it is the case, as it trivially is, that to come to know that the sentence “water = H₂O” is true, one must know what the word “water” means.

2.1.4 Knowing about water, and knowing about “water”

This objection brings us back to semantics. For as Tichý (1983) rightly pointed out, there is a difference between establishing that water is H₂O and establishing that the (English) sentence “Water is H₂O” expresses a truth. The difference is that once it is established that water is H₂O, it may remain to be established that “water” refers to water and “H₂O” refers to H₂O. The point is clearer using another language: it is possible to know that water is H₂O without knowing that the (Japanese) sentence “Mizu wa H₂O dearu.” expresses a truth; plenty of people (including me until a few minutes ago) know the former but not the latter.

Now of course since I am writing in English you might think it hardly needs establishing that “water” refers to water. And you would be right; as long as I am not equivocating, it is self-evident that “water” refers to water. But what does need establishing is that “water” refers to the stuff that has been found to be H₂O, which is not self-evident. What scientists have done, and this no one is disputing, is to establish that there is a liquid that is constituted by molecules of H₂O. But is it those scientists who established that the liquid found to be H₂O is water – is what we call “water”? Here we come back to our earlier point, where that question is answered in the negative. It may seem that I have confused, on the one hand, establishing that a particular liquid is water with, on the other hand, establishing that it is the liquid we call “water”. But, again, since I am writing in English it is self-evident that water is the stuff that is in English called “water”, so given that I am writing in English, whatever it takes to establish that the stuff in my mug is water is exactly what it takes to establish that the stuff in my mug is what is in English called “water”.

The upshot, then, is that while it was scientists who discovered that there is some x such that x=H₂O, that was only half the job, from the perspective of physicalism about water. The other half, to discover moreover that there is some x such that x=H₂O and x=water ought, in principle, to be verifiable by any competent speaker of English. This can be put in another way: in order to prove that water is H₂O, scientists first and foremost needed samples of water, which they could then put under the microscope or electrolyse and so on. It is true that they needed to know that the samples were water, but it is hard to think of better proof than that it came from the local river. For even if there were some ‘empirical’ test that was available, that test could only be based on some other experiments; and those experiments could only have been carried out on samples that were known to be water simply because they were pulled from the local river. In the end, or rather in the beginning, there must have been a choice made as to what was to count as a sample of water for the purposes of scientific investigation as to its nature. That choice could only have been made based on a pre-scientific, ordinary understanding of what characteristics water has, and it underpins the evidence we have today that water is H₂O. [29]

It is again worth pointing out that only very minor adjustments, if any, need to be made to this picture to bring it into line with the Kripkean view of natural kind words. Kripke holds that substances like water must have been named in some kind of baptism ceremony, whereupon it was checked for characteristics that might identify it generally (being odourless, colourless, found in rivers, etc.). Those characteristics became part of the language community’s general knowledge of water. Had we still in our possession those original dubbed samples, we might have given some of it over to scientific investigation. But of course we don’t, and I dare say never did really, so we have only that general knowledge to rely on when we want to pick out samples of water to be investigated as to its nature. If, after investigation, the scientists tell us that only Loch Ness contains water, or even that only the ocean contains it, the Kripkean is in a good position to say that something has gone wrong, because judging from our general knowledge (which is all we have to go on), the original samples dubbed “water” could easily have contained only rainfall and liquid taken from rivers, and most likely were not from Loch Ness. So judgements about whether it was really water that was broken down into hydrogen and oxygen had better not be in conflict with what we would say based purely on our general knowledge of whether something is water. The outcome, for a Kripkean, is very similar to what I have been describing. The difference is that where the “strict” a priori physicalist will talk about those characteristics of water that fall under the concept ‘water’, the Kripkean will talk about those characteristics of water that are a part of the language community’s general knowledge of water. The difference becomes even smaller if, as is surely the case, most of the “general knowledge” will be implicitly passed on, rather than explicitly taught (who was ever taught that water is drinkable?) and if the way it is passed on is closely tied to the learning of the word “water”. For in that case the difference between reflecting on the concept ‘water’ and reflecting on the properties one was taught to associate with water will be virtually nil, if indeed it is not actually nil; the difference may simply lie in how the one cognitive process is best described.

A priori physicalism, then, as we have discussed it, seems to be the right account of how we know that water is H₂O. A posteriori physicalists seem to require the characteristics empirically known to hold of water to be independent of the characteristics thought to hold of water according to competent speakers of English, and we have seen that there is an important sense in which that is false. But how does this issue relate to the justification of physicalism about sensations? That is what we will discuss in the Chapter Three, after we have dealt with some common objections to a priori physicalism that have been raised in the literature.

Ordinary conceptions of manifest kinds are necessary because they are assumed by empirical research. This is the case regardless of whether the conceptions are actually embedded in the concept or merely associated with its referent. In order to study water, a researcher must begin with the premise that water is the clear liquid found locally in rivers and taps. The question as to whether those properties really hold of water does not subsequently arise. Any research that purported to find otherwise would automatically bring into question the idea that, strictly speaking, there is really any water in the world at all. Block and Stalnaker disagree with this idea, as does Byrne, and addressing their concerns will be the subject of the remainder of this chapter.

2.2 Byrne and “Cosmic Hermeneutics”

2.2.1 The objection from the logical gap between distinct vocabularies

Byrne’s (1999) argument against Jackson and Chalmers starts from an odd reading of their claim. According to Byrne, Jackson is committed to what Horgan (1983) calls “cosmic hermeneutics”, the view that statements about water in everyday vocabulary are derivable from statements about H₂O in entirely physical vocabulary. He introduces the issue as follows:

Imagine a Laplacian demon. He knows everything that is expressed by true sentences of some appropriate physical language—every physical fact. He knows everything that is a priori. Let Ψ be a true sentence composed from any vocabulary whatsoever. Using knowledge just stipulated as premises, can the demon deductively infer, and thereby come to know, (the proposition expressed by) Ψ? (p. 347)

Byrne argues against the availability of any such inference, for reasons that are sensible given the way he sets things up. In particular, since the physical language and Ψ are distinct vocabularies there can never be a deduction from one to the other. He is willing to allow for the sake of the argument that “water is the clear liquid around here, etc” may be a priori, but not that “H₂O is the clear liquid around here” is strictly speaking a physical fact. Since neither it is a priori it must be able, as it were, to stand alone as an empirical discovery—a discovery that connects the language of physics with everyday vocabulary.

2.2.2 The solution: overlapping vocabularies

It is doubtful whether Jackson would disagree with Byrne’s rejection of cosmic hermeneutics so defined. But Byrne’s rather austere notion of “physical” vocabulary, from which statements are drawn that are supposed to enable the deduction of statements in Ψ, renders the upholder of cosmic hermeneutics something of a straw man. [30]

When a student studies science at high school he or she certainly learns new words, but clearly not a new language. Insofar as there is a “language of physics” it overlaps the everyday language – it is not possible to entirely quarantine even the purely theoretical statements of the sciences, since they at least require the prepositions, articles and so forth of ordinary speech. In any case the theoretical statements are justified by inferences to the best explanation from a set of statements couched in experimental vocabulary. The existence of H₂O, for example, best explained the fact (among others) that from hydrogen and oxygen one could synthesise a liquid that looked and tasted for all the world like water. Crucially, the fact that H₂O looks and tastes for all the world like water cannot be a further inference. It is part of the inference that there is such a substance as H₂O. The further question of whether any local liquid that looks and tastes like water is in fact water depends to some extent on whether or not a descriptivist theory like Jackson’s is correct, about which Byrne is neutral.

Nonetheless, as we saw in the previous chapter, that water is in fact the clear and tasteless (and so forth) liquid around here will be the starting point even for a Kripkean scientist, since we lack any record of a baptism that might prove it wrong. That water just is the clear, tasteless liquid found locally in taps and rivers will be, as we might say, the working definition of water for any such scientist studying the chemical nature of water. That ‘definition’ will then underpin the rest of the research and be assumed by any theory of water. In this way it is irrelevant to the justification of “water is H₂O” whether or not the ordinary conception of water is truly a priori or not (though it may of course be an interesting issue for other reasons). There are in practice no grounds of appeal other than that conception [31] . Our ordinary conception of manifest kinds is assumed by the empirical research into them and in that sense ordinary conceptions direct empirical research. Without having any idea of which properties gold is ordinarily thought (implicitly or explicitly) to have, empirical research into its nature could not get off the ground simply because the researcher would have nowhere to start.

Facts about H₂O entail facts about water partly, then, because the vocabulary used to express facts about H₂O overlaps the vocabulary used to express facts about water. Although it is not an entailment across vocabularies, it is an entailment across the overlapping parts of those vocabularies. That is cosmic hermeneutics enough.

2.3 The Status of Folk Theories

2.3.1 Block and Stalnaker against a priori knowledge of manifest kinds

Take the following passage from the recent paper, mentioned earlier, by Block and Stalnaker:

Let C be a complete description, in microphysical terms, of a situation in which water (H₂O) is boiling, and let T be a complete theory of physics. Can one deduce from T, supplemented with analytic definitions, that H₂O would boil in circumstances C? To see that one cannot…consider a person on actual Earth, who does not know the story about how water boils – perhaps she doesn't even know that water is made up of molecules. One presents her with the theory T, and a description (in microphysical terms) of a water boiling situation. Can she then deduce that if T is true and a situation met conditions C, then the H₂O would be boiling? No, since for all she knows the actual situation is like the one on Twin Earth. Perhaps, if she were told, or could figure out, that the theory was actually true of the relevant stuff in her environment, she could then conclude (using her knowledge of the observable behaviour of the things in her environment) that H₂O is water, and that the relevant microphysical description is a description of boiling, but the additional information is of course not a priori, and the inference from her experience would be inductive. So we are not persuaded by Levine's [a priori physicalist] argument… (p. 8)

What is brought into question here is very similar to the part of a priori physicalism that worries Byrne; namely, that physical theories are not enough to say when those physical theories are talking about the same thing that our non-physical language talks about. The solution is similar, too. But there is a further worry, which is this: experiments testing some theory about water must, presumably, feature a sample of water. Now if the experiment leads to new knowledge at all, it leads to new knowledge about only the kind of thing instantiated by that particular sample. The passage above draws attention to the inference from the sample’s being a clear liquid, etc., to its being a sample of water.

2.3.2 Knowledge of manifest kinds as “Moorean” knowledge

Consider the following: a French experimentalist discovers that some particular stuff in a flask, call it A, has some particular nature, call it x, but since he is monolingual fails to know whether A is commonly called anything in English. Now imagine an English speaker visits the lab and is asked if he knows what A is. He notices that it is a colourless, odourless, tasteless, non-viscous liquid, and after asking around discovers that the flask came to hold it after being left outside in the rain (or perhaps that description is e-mailed to him, with the question, “Any idea what this stuff might be called in English?”). That is enough information to be able to work out that A is almost certainly water. So between them, the English speaker and the French scientist can work out that water has nature x, the one bringing a knowledge of English, the other bringing empirical knowledge of the nature of a particular sample.

In the passage above, Block and Stalnaker propose that the inference from the description of A to the statement that A is water is inductive; that is, the English speaker infers that the stuff in the flask is water because it has all (or many of) the properties of instances of water that he has previously come across. This fits the Kripkean picture of natural kind words that they are following. Recall that, for Kripke, it is not a priori that water is clear and tasteless—it is simply that having transparency and being tasteless are among the properties that have come to be associated with water. Accordingly, the English speaker in our above example remembered that water was supposed to have those properties, and suggested that A was likely to be water. In a sense, the difference here is very real; to Block and Stalnaker, the inference is based on his knowledge of the thing that we in English call “water”; to a priori physicalists, the inference is based on his knowledge of the English word “water”. But, as we discussed in the previous section, in this case there is very little separating the two.

To flesh this point out a little, consider the way that Jackson thinks about conceptual analysis. On the one hand, there is this:

[My] account sees conceptual analysis of K-hood as the business of saying when something counts as a K. (1998; p. 45)

On the other hand, there is this:

My intuitions about which possible cases to describe as cases of K-hood, to describe using the term ‘K’, reveal my theory of K-hood… In as much as my intuitions are shared by the folk, they reveal the folk theory. (1998; p. 37)

Consider also that, according to Kripke, we do have something that might be called a “folk theory” of water. It is the set of properties we take to hold of water, which is passed between English users (at least) generation after generation, viz. that it is odourless, tasteless, found in rivers, etc. To the extent that it is just a theory, which might be proven wrong, it is very close to what Block and Stalnaker and other a posteriori physicalists have in mind [32] (see particularly Stalnaker 2001). But – and this is what makes it look like more than just a theory – it is difficult to think of what evidence we could have that could prove it very wrong. To the extent that the folk theory could be proven wrong it makes Jackson’s physicalism look very a posteriori. But recall from our discussion earlier that if scientists declared that what flows in the rivers and falls from the sky is actually something other than water, even the staunch Kripkean would have reason to doubt them, in effect using the folk theory to trump the scientific theory.

[JOD5] What kind of evidence could prove wrong the theory that the stuff that actually flows in the rivers and falls from the sky is by and large water? If, as I suggest, no empirical evidence could have that role (which did not also show that, surprisingly, we all along meant by “water” something other than the stuff that falls from the sky and flows in rivers and so on), then it is hard not to conclude that our theory of what water is is known a priori, without serious appeal to experience. In this respect, that certain properties by and large hold at least contingently of water can be taken to be a kind of Moorean fact, not seriously open to doubt.

So our folk theory of K-hood in effect, even for the Kripkean, simply stipulates when something counts as a K, since in virtue of the way we come to hold the theory it can’t be proven very wrong. So Block and Stalnaker cannot be right that, in our above example, the English speaker’s inference that the sample is of water is inductive, in any ordinary sense of the word.

It is of course true that, given that we think of water as fitting a certain description (whether or not it gets its reference that way), it is an empirical question whether there is anything in the world that fits that description. But the a priori physicalist is by no means denying, or committed to denying, that an English speaker ought to literally look around to establish which things are trees and which are puddles of water, merely that there need be no investigation into the nature of trees and puddles to work out which is which. Any English speaker should be able to say which is which using nothing more than their grasp of the words “tree” and “puddle” (or, if you like, their folk theory of trees and puddles). The same goes, mutatis mutandis, for water.

There is no basis in the Kripkean theory of reference for denying that knowledge of ordinary conceptions of manifest kinds is a prerequisite for their scientific study. On the contrary, given that we lack information about baptisms it is supported by Kripke’s view. The study of water implicitly followed its ordinary conception and there was no significant disagreement about it, so far as I am aware. But in the case of sensations, due at least in part to philosophical intervention, there is debate specifically about the ordinary conceptions involved. In science it surfaces as debate about the appropriate working definition of a sensation. In philosophy it is generally expressed in terms of a debate about the right analysis of sensation concepts, which is how I will generally express it from now on. One of the most popular analyses refuses, in effect, to endorse any working definition, which I will argue in the next chapter is disastrous for empirical research into sensations. Fortunately, I believe, it is mistaken.

Part Two –Conceptual

3. Against Primitive Sensory Concepts

said earlier that there is a sense in which we knew what water was before we knew what water was. For this to be true, there must be two different senses in which we can know about water. We saw that the sense of knowing about water that came first was simply our ordinary way of identifying water—our common knowledge of water; that it falls from the sky, runs in the rivers, is colourless and odourless, and so on. The sense of knowing that came later I identified as our scientific knowledge of the nature of water; that it is H₂O and so on. We saw that both kinds of knowledge are required in order to know that the sentence “Water is H₂O” expresses a necessary truth. Scientific knowledge tells us the nature of things in the world; common knowledge tells us which of those things are to count as samples of water.

If this is correct, then the study of sensations also requires both kinds of knowledge; in the vernacular of conceptual analysis, it requires knowledge both of sensory concepts and the things to which those concepts might refer. Scientific investigation into the things to which the sensory words might refer—let us take the brain as an obvious candidate—is of course proceeding at a cracking pace. But there remains pessimism as to whether sensations will ever be nailed down to a particular physical basis in the way that water has been. At least one reason for that pessimism has nothing to do with the status of our scientific knowledge of candidates for referents of words like “pain”; instead, it lies in disagreements as to what constitutes our common knowledge of what it is for a thing to be in pain.

3.1 “Consciousness” as a Primitive Concept

3.1.1 Prologue: resolving conceptual disputes

Disagreements about what is supposed to be common knowledge of a thing lends doubt to the idea that there is any common knowledge at all. Certainly, the Kripkean account of natural kind words that we discussed in the last chapter lacks an obvious way of dealing with disagreements of this kind. It is a different story for Jackson, who has available the apparatus of conceptual analysis. When we use a word, in Jackson’s view, we may not be able to make explicit the conditions under which we would use it, but put us in a situation and we will be able to say whether the word applies. I may not know, for example, exactly what qualities make a person wise, but I will have much less trouble telling you which of the people I know well are wise, and which are not. This is not just a matter of vagueness, since if it were simply that wisdom is a vague concept I should have as much trouble telling you who are the wise people as I do telling you what makes a person wise. The reason for this difference is that the conditions under which we use some words can be something less than transparent. Because of that lack of transparency, when we try to enunciate the underlying conditions for our use of a word, we may disagree; not because we use the word differently, but because we can make mistakes in spelling out exactly what it is that makes us use that word in one particular situation rather than another. Perhaps because of the way we learn languages, we very often agree in particular situations about whether a certain word is appropriate. This allows us, in discussions of what the conditions are for using a word, to talk about whether it is appropriate in particular imagined situations. By doing this, we refine our notion of what the conditions are for using the word. [33] Disagreement does not automatically mean that there is no shared set of conditions under which we use a word. It may mean there are no such shared conditions, but it may also mean that one party to the disagree-ment, or both, are mistaken in their view as to what those conditions are, and in that case further discussion may well, eventually, yield agreement.

The apparatus available to Jackson, just described, is not straightfor-wardly available to Kripke. But something like it may be. Even for Kripke, our common knowledge of what water is may not be immediately available to us. For example, if you had asked me to describe water, it may not have occurred to me to mention that it is odourless. Nevertheless, being odourless is one of the properties I associate with water. Many of the properties I associate with everyday things I will never have been taught explicitly. No doubt, in the case of water, I was simply told that the stuff in the glass was water, and the stuff in the river was water. But I may never even have been told that much. I may simply have picked it up when I overheard my mother ask my father for a glass of water, whereupon he went to the kitchen tap, and thus I started to discover what water was. Thereafter, perhaps, I would hear the word “water” used in various situations and gather, after a while, that it was not just the stuff from the tap but the stuff in the rivers and puddles, too. In fact, I learned that it was the kind of stuff that resembled the stuff from the tap in certain ways, though I never thought a great deal about which properties exactly differentiated water from non-water. After a while, perhaps, it just seemed obvious.

If this story is at all plausible, then there is room even in the Kripkean account for disagreement over which properties we “commonly” know water to have. If there is disagreement, it may be because we were never sufficiently aware of the properties we associate with water, which is in turn because we learned those properties implicitly from the people around us. Discussion may well lead to agreement after a time, because debate about when we would call something “water”, and when we would not, may help to make explicit the properties we ordinarily, but implicitly, take water to have. So there does seem to be a way of translating talk of concepts into talk of associated properties, and essentially the same method may be available both to Kripke and Jackson in resolving disagreements as to our ordinary conception, or our common knowledge, of things.

Having said that, it is no longer useful – and, indeed, it is cumbersome to the exposition – to continue comparing Kripke’s account and its resulting methodology with Jackson’s. From now on I will be generally talking about concepts and conceptual analysis rather than implicit common knowledge or ordinary conceptions. Those who regard conceptual analysis as a fool’s game should translate my talk of such into Kripkean terminology. The outcome, by and large, should be the same, for the reasons I have given in the above paragraphs and the previous two chapters.

3.1.2 The received view

How are we to analyse the sensory concepts? What factors determine when we use the sensory words of something? We will have to work this out if we are to show that sensations are essentially physical processes. There are surprisingly few analyses of sensory concepts on the market, due largely to the widespread belief that they are a peculiar kind of concept for which no analysis is possible.

This sentence needs qualification, since those who hold the view that “consciousness”—in effect a general term for sensory experience, which I will follow for this chapter—is unanalysable hold it only of a certain sense of the word “consciousness”. The word “consciousness”, after all, has many different senses. As we will see, the sense of the word that is often taken to be unanalysable is exactly the sense that we are most concerned with here. That sense might best be explained as the notion of a pure sensation, or a phenomenological experience, or the feeling of consciousness. [34] Many philosophers argue that “consciousness” in this sense is a primitive concept in that it has no analysis. Chalmers (1996), Block (1992 and 1995a) and Levine (1983 and 1993) most obviously hold this view. More controversially, I think Loar (1997) and Tye (1990 and 1999) can be read as holding the view, as can Kripke insofar as he holds that there are no properties other than “being painful” that we use to pick out pain states. It is a view that has important consequences, as will become apparent.

If no analysis can be given of the sensory concepts then there can be no a priori passage from truths about the physical structure of the world to truths about the distribution of consciousness. The reduction of water to H₂O relied on the a priori proposition that water is the liquid that actually runs in the rivers, falls from the sky, is tasteless, etc.. If there is no analogous a priori proposition of the form consciousness is…, the reduction of consciousness to physical processes appears doomed. For this reason we must first tackle the view that denies precisely that there is any such proposition available. Since Chalmers holds it most explicitly, his arguments deserve our attention first.

3.1.3 Chalmers on phenomenal qualities

In a book that has become influential in both philosophy and the sciences, Chalmers (1996) proposed that our ordinary conception of mental states is actually a mixture of two components: a psychological component and a phenomenal component. The psychological component identifies mental states functionally – through their causal interaction with other mental states and the world. The phenomenal component identifies mental states through the way they feel. As he says in the first chapter,

pain provides a clear example. The term is often used to name a particular sort of unpleasant phenomenal quality, in which case a phenomenal notion is central. But there is also a psychological notion associated with the term: roughly, the concept of the sort of state that tends to produce damage to the organism, tends to lead to aversion reactions, and so on. Both of these aspects are central to the commonsense notion of pain. We might say that pain is ambiguous between the phenomenal and the psychological concept, or we might say that both of these are components of a single rich concept. (p. 17)

In this passage, Chalmers is acknowledging that the functionalists were on to something in pointing our that the casual network that surrounds a mental state is somehow crucial to its being the kind of mental state it is. But he is also acknowledging the worry of anti-functionalists about consciousness, such as Ned Block (1980), who hold that a state like pain is not any old filler of the proper functional role; it is a particular kind of state, though it may in addition fill a certain functional role. The particularity of pain as a kind of feel, a “phenomenal quality”, is what Chalmers’ book is concerned with.

One might think that “phenomenal quality”, or even “feel”, is something of an unhelpful analysis of the word “pain”. For one thing, “phenomenal quality”, or even “a certain phenomenal quality” doesn’t help us to distinguish a pain from a tickle. So when we talk about pain in the sense of its phenomenal quality, what do we take ourselves to be talking about? Here is Chalmers on “phenomenal quality”, or simply consciousness as such:

Trying to define consciousness in terms of more primitive notions is fruitless. One might as well try to define matter or space in terms of something more fundamental. The best we can do is give illustrations and characterisations at the same level. These characterisations cannot qualify as definitions, due to their implicitly circular nature, but they can help to pin down what is being talked about. (p. 4)

An example of such illustrations and characterisations follows soon after:

We can say that a being is conscious if there is something it is like to be that being, to use a phrase made famous by Thomas Nagel. Similarly, a mental state is conscious if there is something it is like to be in that mental state. To put it in another way, we can say that a mental state is conscious if it has a qualitative feel -–an associated quality of experience. These qualitative feels are also known as phenomenal qualities, or qualia for short. The problem of explaining these qualities is just the problem of explaining consciousness. This is the really hard part of the mind–body problem. (p. 4)

I take it that the circularity Chalmers is referring to is plain. The phrases “consciousness”, “something it is like”, “qualitative feel”, “phenomenal quality”, “quality of experience” and “qualia” are offered simply as synonyms. Chalmers does not intend them to be definitions, and certainly not analyses. Vacuous synonyms is the closest that Chalmers thinks we can come to an analysis of the phenomenal concept of consciousness. Accordingly, there in no analysis to be had of what it is to be a painful feeling in the way that we previously saw there is an analysis of what it is to be water.

Does this spell disaster for the scientific study of consciousness? Prima facie, it should. After all, if there were no way of distinguishing the conscious things from the things that aren’t conscious apart from the fact that we refer to some things as “conscious” and not others, then those who are studying consciousness scientifically would have no reason for ever claiming that all states fitting a certain description were conscious. Recall that the analysis of water given in the last chapter was just our implicit everyday criteria for picking out samples of water. In order to demonstrate facts about water, the scientist needs to ensure that the stuff he or she is demonstrating facts about does indeed fill that everyday criteria. For this he or she relies on an English speaker’s conception (or his or her conception as English speakers) of what it is for a thing to be water. Lacking that conception, there is no way to establish that the statement “Water is H₂O” is necessarily true. If, for example, the ordinary conception of water were simply that of a clear liquid it would only be true, at most, that some water is H₂O.

Is “consciousness” a natural kind word at all? If not, then it is not clear that science will ever be able to claim anything more than that some conscious states are brain processes. To the question, why is a particular brain process conscious?, we could then expect no more enlightening an answer than that we call that brain process a conscious state, and that although there is a reason we call some things conscious and others not, it would not be that all of the things we call conscious share some set of properties. There would therefore be no reason to expect there to be anything in an object’s nature that makes it conscious. But without some idea of what we mean by the word “consciousness”, in the sense we are discussing here, it is hard to establish that it is a rigid designator, and thus if there is any point to a scientific investigation of its referent.

3.2 The “No Conceptual Analysis” Argument against Physicalism

3.2.1 The argument, briefly

Chalmers does indeed think that the failure of an instructive analysis for “consciousness” spells disaster for a certain kind of scientific investigation into consciousness – namely, physical investigation. For in order for us to be able to say, based on empirical evidence, that consciousness is some physical state or other, there would need to be, as Chalmers says, “some kind of analysis of the notion of consciousness – the kind of analysis whose satisfaction physical facts could imply – and there is no such analysis to be had.” (p. 104) But Chalmers does not think that scientific investigation into the nature of consciousness is fruitless.

For Chalmers, given that there is no possibility of conscious states turning out to be merely physical states, scientific investigation should proceed by cataloguing the ways in which consciousness and the physical world interact. Its job is to find correlations between kinds of physical state and kinds of conscious state, and on the basis of those to construct laws governing the relationship between the physical and the phenomenal.

An immediate problem faced by an investigation of this kind is that the identification problem remains. Although Chalmers insists that no sentence of the form “consciousness is physical process X” could ever be justified, he does want to allow that a sentence of the form “consciousness is lawfully correlated with physical process X” could be justified. But since we have no criteria outside our own experience for identifying consciousness, how are we to justify either that all conscious states are accompanied by physical process X or that all instantiations of physical process X are accompanied by consciousness? For any kind of conscious state, there may be innumerable kinds of physical state that accompany it. Because we lack criteria for judging the presence of consciousness in the way that we have criteria for water (colourless, tasteless, etc.), the hypothesis that consciousness lawfully accompanies some particular physical process seems impossible either to establish or to falsify.

3.2.2 Criteria and explanation

Of course there is a kind of criterion we obviously use in our own case to tell if consciousness is present, namely that we are directly aware of it. But that criterion does not extend beyond our own selves which creates a generalisation problem for any purported law. Chalmers argues that the solution to this problem lies in scientific method itself, particularly in the use of induction and inference to the best explanation.

There is a plausibility to the idea that we know which things are conscious simply by generalising from our own case. I am conscious, and you seem very much like me so probably you are conscious too; and so are you, and you, and you, etc. So at least we can reasonably say that all humans beings that seem conscious are in fact conscious. That gives us a fairly large, if somewhat biased towards our own species, sample to study. The method in general, then, is this: there is one thing that I know for sure is conscious – myself. Anything with enough similarity to that thing (that is, anything like me) is probably conscious too (especially, Chalmers reasonably adds, if they say they are). Using this as a guide, I can study the correlations between conscious states in beings like me and brain states. Chalmers denies that we are then in a position to hypothesise that conscious states are brain states but we are in a position, he says, to postulate the existence of fundamental laws holding between brain states and conscious states.

Why one and not the other? The reason is that, according to Chalmers, in order to justify the claim that conscious states and brain states are one and the same, we would need to justify the claim that if something is not a brain state, it is not conscious. This is something we will never be in a position to do, since there is nothing in the concept [consciousness] that tells us that it cannot be satisfied by non-brain states even if it is the case that the only things we happen to pick out as conscious states are brain states. There is therefore nothing in the nature of the brain states that could tell us why this particular brain state is conscious rather than another, or why some other kind of state could not also be conscious.

But doesn’t the postulation of laws connecting brain states and consciousness require justification too? Chalmers does not deny it, but points out that laws are merely generalisations of observed regularities. Since the correlations between consciousness and brain states are—at least in a fashion—observed, they can be ‘promoted’ into a law. But in fact Chalmers wants to go further than that. He wants not just a vast network of correlations promoted into laws, but beyond that a simpler set of underlying laws which form an “explanatory framework” (p. 214). After all, according to Chalmers, we ought not to be content with just documenting and generalising surface regularities when we are “questing after the basic structure of the universe” (p. 214). [35]

In any case, the apparent problem Chalmers takes himself to be solving is this: it is impossible, using a non-psychological vocabulary, to capture what it means for a thing to be conscious. In other words, it is impossible to explain using non-psychological vocabulary why we call one thing conscious and another not. And if we can’t explain why we call something “conscious”, we won’t be able to explain what makes it conscious. After all, if we call A conscious but can’t say what it is about A such that we say it is conscious, there seems to be nothing in principle stopping us from calling B conscious even though it seems in every way different to A [36] .

Of course there must be—and Chalmers insists there is—a reason we call some things conscious and others not. It cannot be magical, or arbitrary, that we use the word in some situations and not in others. As we saw above, Chalmers proposes that consciousness means something like “having a phenomenal feel”, which we can recognise in our own case and inductively extend to other people.

Here, then, is the question we can no longer put off: how do we know that consciousness applies in our own case? How do I know I’m conscious? We were not, after all, born with the knowledge that “consciousness” applies to us. We acquired the word somehow. On what basis did we acquire it? It is no help to say, as Chalmers is inclined to, that the word is a primitive. Even primitive words must be acquired; their conditions of use must be learnt. Otherwise, how can we know that we are using the word properly?

3.1.3 Why knowing when we’re “conscious” must involve criteria

To begin to answer this question, let’s imagine that a person learning English wants to know what “consciousness” means. Are we to say, as Louis Armstrong apparently said of Jazz, “If you don’t know, you ain’t never going to get to know”? That would be absurd. Our hypothetical enquirer doesn’t want to know the nature of consciousness, merely when to use the word and when not to. Surely there is an answer. Perhaps we can say, “Consciousness is the state you are in whenever you’re awake.” We are presuming, of course, that the person knows what “awake” means. Let’s further imagine that they don’t know what “awake” means. In that case, perhaps we ought simply to say “Consciousness is the state you are in now and, roughly, until you go to bed tonight.” But perhaps I’m talking to a zombie! That is, perhaps I’m talking to someone who looks and acts very much like me but isn’t, in fact, conscious. So things seem to have become complicated very quickly. [37]

It helps to take a step back at this point. In the course of answering someone’s query as to when to use the word “consciousness”, we find ourselves in a state of uncertainty as to whether it is legitimate for the person to use the word in their own case. But this is an absurd kind of situation. Weren’t we, after all, at one point in the same position as our enquirer, wondering when to use the word “consciousness”? And wasn’t the rest of the world, at that time, in the same position as we now are with respect to our enquirer? That is, wasn’t the rest of the world in a position of uncertainty as to whether I can legitimately use the word “consciousness” to refer to states of myself? Might I therefore have misunderstood what “consciousness” refers to? Perhaps, at least with respect to what other people mean by “consciousness”, I am a zombie. Perhaps I’m just not conscious according to what most people mean by the word.

This situation is of course something which any theory of conscious must overcome, but I suggest that Chalmers does not overcome it. According to Chalmers, we know that other people are conscious by, on the one hand generalising from our own case, and on the other hand treating their protestations that they are conscious charitably. The first route requires that I know myself to be conscious; that I know the word “conscious” applies to me. The second route requires that there is no misunderstanding between myself and others who use the word “conscious”. In other words it requires that I (and they) are using the word correctly, which in turn requires that I know whether or not the word “consciousness” applies to me. For Chalmers, there seems no story to tell about how I come to know how to use the word “consciousness” correctly. I simply know. [38]

We are granting, of course, that we do know whether or not the word “consciousness” applies in our own case. The question we are addressing is how do we know? Perhaps it is better if we start with some requirements as to what our answer must demonstrate. One requirement, I suggest, is that any two people must have, all things being equal, the same chance of getting it right as to whether the word “conscious” applies to a thing, particularly to themselves. For that to be the case, my grasp of whether or not “conscious” applies to me ought to be the same as your grasp of whether or not “conscious” applies to you. There is nothing here that Chalmers would disagree with. And this requirement suggests another requirement (which in any case is independently obvious): it ought to be possible to learn what the word “conscious” means. This seems a trivial requirement, but yields significant results when combined with out first requirement. The result is this: when I learned the word “conscious” from another person, I had as much right to use it in my own case as they. If so, it seems in a sense irrelevant to my learning the meaning of “conscious” that you refer when you use it to a particular internal state you occupy, to which, as things now stand, I have no access. [39]

It may help to take a more specific example: pain. According to our first requirement, having learnt the word “pain” you and I have equal right to use it. According to our second requirement, it is possible for me to learn from you what “pain” refers to without knowing whether or you and I have similar internal states. So the meaning of “pain” cannot be straightforwardly tied to a particular internal state.

This conclusion might seem like a reductio ad absurdum. Pain is, after all, a particular internal state, and an unpleasant one at that. Obviously I am not suggesting that it isn’t, merely that there is nothing in the meaning of the word “pain” that entails it. There is nothing too worrying about this fact. Chalmers does, after all, have to face something analogous: whereas according to the current suggestion it is neither a priori nor obvious that “pain” refers to the same internal state in you and me, according to Chalmers it is an a posteriori matter whether anyone apart from me is ever in pain, in the sense that it is justified by inferences, such as induction, that are peculiar to a broadly scientific methodology. There is a sense in which both situations are unintuitive, but to the suggestion that the conclusion we have come to is implausible, it seems at least a plausible as Chalmers’ alternative. Given that it follows from what seem to be trivial premises, I propose that we tolerate it.

3.3 An Appended Topic-Neutral Reply

3.3.1 The topic-neutral analysis of sensory terms

If pain does not refer straightforwardly to a particular internal state, then, what if anything does it refer straightforwardly to? As a starting point, it may help to take seriously the following fact pointed to by Chalmers in The Conscious Mind:

In talking about phenomenal qualities, we generally have to specify the qualities in question in terms of associated external properties, or in terms of associated causal roles. Our language for phenomenal qualities is derivative on our nonphenomenal language….Even with a term like “green sensation,” reference is effectively pinned down in extrinsic terms. When we learn the term “green sensation,” it is effectively by ostension—we learn to apply it to the sort of experience caused by grass, trees, and so on. (p. 22)

This suggests, if true, a possible route to a “nonphenomenal” analysis of mental terms. Perhaps we can think of “green sensation” as a concept that is analysable as something like “the kind of experience you typically get when you look, in normal conditions, at grass, trees, parts of the Italian flag, and so on.” There are, at least, two theories of mind that include something like this proposal. The first is the causal/functional view, and the other is the topic-neutral view. We can take, as an exemplar of the first approach, Armstrong (1980), with his view that the concept of a mental state is the concept of a state “that is apt to be the cause of certain effects or apt to be the effect of certain causes” (p. 21). For our present example, the concept of a green sensation would be, if Armstrong is right, that of a state apt to be caused by grass, etc. An exemplar of the second approach is J.J.C. Smart (1959) with his view that the concept of a yellowish-orange after-image is something like this: “There is something going on which is like what is going on when I have my eyes open, am awake, and there is an orange illuminated in good light in front of me”(p. 149). The two approaches are, obviously, very similar. The main difference between them is that Smart does not include within the concept the idea that “what is going on” is caused by the orange. That is, Smart’s account allows that Parallelism is a coherent view, whereas Armstrong’s account does not. Since Parallelism – that is, the view that our mental life and physical processes are well correlated but causally independent of each other – does not seem to be incoherent, I will adopt Smart’s account as the more plausible of the two.

3.3.2 Chalmers’ rejection

Chalmers rejects both analyses, and it will be instructive to look at his reason. On the face of it, there is a difference between my talking about “the experience I typically have” when I look at grass, and “the state I am typically in” when I do so. Chalmers argues that any analysis of conscious experience which includes the former is smuggling in the concept to be analysed into the analysis itself, and so will not do as an analysis. On the other hand, he argues, to talk about “the state I am in when…” leaves it entirely open whether that state is a conscious or nonconscious state; that is, the analysis seems to leave out something crucial. As he says:

If “something going on” is construed broadly enough to cover any sort of state, then the analysis is inadequate; if it is construed narrowly as a sort of experience, the analysis is closer to the mark but it does not suffice for the conclusion (that mental states can be analysed in a topic-neutral way). (p. 360, n13)

The first point, that any analysis that fails to mention something like “experience” can’t do justice to what we mean by “conscious experience”, deserves discussion. The first thing to note is that it isn’t obvious that the analysis needs to include “experience”. To say that to have a blue sensation is for there to be something going like what typically goes on when I look up out of doors on a clear day does not obviously fail to capture what I mean when I talk about my blue sensations. It has not been explicitly specified, it is true, that what goes on when I look up out of doors on a clear day is a kind of conscious experience. But it is in fact a particular conscious experience that occurs at those times, and I am only talking about what has actually happened, accidentally for all I know, at the times when I look up out of doors on a clear day.

Chalmers might well object at this point that being in some state when I look at the sky can’t exhaust the meaning of “blue sensation”, since there might be any number of events perfectly correlated to my looking at the sky but which have nothing to do with conscious experience. For example, imagine that whenever I am out of doors looking up the sky the UV light damages my retina. Now retina damage is not what I mean by “blue sensation”, but the analysis just given does not rule it out.

The question is, is there a way of ruling it out? Let’s look at Chalmers’ suggestion, viz., to add ‘experience’—better, ‘sensation’—into the analysis. [40] So, “blue sensation” is to be analysed, roughly, as “the kind of sensation I have when I look up out of doors on a clear day”. What have we added here? What new information is conveyed? Well, plenty if you already understand what a sensation is; none if you don’t. And therein lies the problem, and is the reason Chalmers is right in insisting that it will not do as an analysis.

3.3.3 A Representationist Solution

Nevertheless, there is something we can add to a Smart-style analysis, something that is mentioned by Smart himself, though it drops out of the analysis he puts forward. The suggestion is this:

Suppose that I report that I have at this moment a roundish, blurry-edged after image which is yellowish towards its edge and is orange towards its centre. What is it that I am reporting? One answer to this question might be that I am not reporting anything, that when I say that it looks to me as though there is a roundish yellowy-orange patch of light on the wall I am expressing some sort of temptation, the temptation to say that there is a roundish yellowy-orange patch of light on the wall (though I may know that there is not such a patch on the wall). (p. 115)

Smart rejected this answer in the end because he seemed to him that, at least partly, we are reporting something about a state we are in rather than merely reporting what we are tempted to say the world ‘outside’ is like. But perhaps we can have our cake and eat it too. Perhaps, that is, we can say that when I report that I am having a green sensation, I am reporting two things: firstly, that I am in a state very like the state I am typically in when I look at grass, trees, etc, etc; secondly, that the state I am in inclines me to believe that I am looking at something which is similar in some respect to grass, trees, etc,. [41] The point could be put similarly, though at the risk of appearing more circular, in the following way: to have a sensation of green is to be in the kind of state you are typically in when you look at green things, and which inclines you to believe that there is a green thing in front of you.

Now it can easily be established (at any rate, I presume) that retina damage does not lead me to believe anything in particular about the world. It follows that, if the above analysis is correct, even if perfectly correlated with blue sensations, retina damage cannot conceivably be a part of what “blue sensations” refers to. If this is allowed to stand as an analysis, it appears to defeat Chalmers’ worry that a topic-neutral analysis cannot be specific enough about the kind of state intended without smuggling the analysandum, “consciousness” or a straight out equivalent, into the analysis. For it appears that any state satisfying both of the requirements we have mentioned is a plausible candidate for what I am referring to when I talk about my sensations. [42]

This “inclinations to believe” addendum, it is worth noting, is consistent with “representational” theories of sensory qualities, according to which sensations of blue are, very roughly, nothing more than states which represent a part of the world as being blue. [43] ^, [44] The general view I am presenting here has obvious parallels with such an approach. Where it differs is not so much in the view that sensations are representational as in the view that they are particular kinds of representations, individuated not just by what is represented, also the way in which things are represented. [45] It is also worth noting that if representational states are to be characterised functionally, which seems likely, the present proposal is consistent with many varieties of functionalism, particularly empirical functionalism.

In any case, if I am right there is a way of pinning down what we mean by “blue sensation” without using explicitly phenomenal language. In that case, perhaps we are in a position to justify the statement “sensations are brain processes” using essentially the same procedure that we outlined earlier for justifying the statement “water is H₂O”. That procedure, recall, was essentially to break up the statement “water is H₂O” into two further statements, “water is A” and “A is H₂O”, where “A” is some description of water according to our ordinary conception. The first statement, “water is A”, comes out as a priori (or at least very robust common knowledge), and the second statement, “A is H₂O” comes out as empirical fact. From the transitivity of “is”, we get the statement “water is H₂O” [46] . In the case of water, the “A” we came up with was “the actual clear, odourless, tasteless stuff we find in the rivers and lakes, and which falls from the sky, etc.” In the case of conscious states, the specific “A” will of course depend on the specific kind of experience we are looking for. Take a sharp pain in the finger as an example. In that case, “A” will be something like this: “the kind of state I am typically in when I cut my finger, and which tempts me to believe that there is damage occurring to my finger.” If that is roughly right, then if our statement to be justified is “pains in the finger are brain processes”, it can be broken up into the following two statements:

(1) Pains in the finger are states of the kind that I am typically in when I cut my finger, and which inclines me to believe that there is damage occurring to my finger.

(2) States of the kinds that I am typically in when I cut my finger, and which tempt me to say that there is damage occurring to my finger, are brain processes.

The first statement is the result of our analysis of what we mean when we talk about phenomenal states, and is therefore justified a priori. The second statement is clearly a posteriori, and therefore a matter of empirical investigation, justified by ordinary scientific method. Together, they yield

(3) Pains in the finger are brain processes.

It appears that we are making progress. But before we move on, we will need to deal with an influential argument against the conclusion we have just come to. It is an argument elucidated by Kripke himself, but repeated often by various writers.

3.4 The Kripkean Objection, and a Reply

3.4.1 Kripke’s attack on physicalism

As we saw in the previous chapter, in Naming and Necessity Kripke sketched a theory of the reference of natural kind terms that showed how it could be that so-called “scientific” identities, such as between water and H₂O, could be necessary and yet justified only a posteriori. Having thus showed in principle how an identity between consciousness and brain states could be demonstrated by science and yet seem contingent, Kripke proceeded to pull the rug on that prospect by giving an argument based on his theory as to why it couldn’t be done. This is the argument Kripke gives:

In the case of the apparent possibility that molecular motion might have existed in the absence of heat, what seemed really possible is that molecular motion should have existed without being felt as heat, that is, it might have existed without producing the sensation S, the sensation of heat. In the appropriate sentient beings is it analogously possible that a stimulation of C-fibres should have existed without being felt as pain? If this is possible, then the stimulation of C-fibres can itself exist without pain, since for it to exist without being felt as pain is for it to exist without there being any pain. Such a situation would be in flat out contradiction with the supposed necessary identity of pain and the corresponding physical state, and the analogue holds for any physical state which might be identified with a corresponding mental state. (p. 151)

The reason we are in this situation with regard to pain, according to Kripke, is this:

In the case of molecular motion and heat, there is something, namely, the sensation of heat, which is an intermediary between the external phenomenon and the observer. In the mental-physical case no such intermediary is possible, since here the physical phenomena is supposed to be identical with the internal phenomenon itself… The apparent contingency of the connection between the physical state and the corresponding brain state thus cannot be explained by some sort of qualitative analogue as in the case of heat. (p. 151-2)

If we were to schematise Kripke’s worry, it might look something like this:

HEAT SENSATION ‘S’

|| (is thought to be) || ( is thought to be)

The cause of sensation ‘S’ ??

|| (happens to be) || (happens to be)

MOLECULAR MOTION BRAIN STATE Y

So, the apparent contingency of the relationship heat and molecular motion turns out to be the actual contingency of the relationship between heat and the cause of a particular sensation. By now this is all very familiar. But when Kripke looked at sensations themselves through this picture, he found that there was nothing to stand midway, as it were, between sensations and brain states. We do not after all, think of sensations as being the cause of some sensation, as we may well do in the case of heat; we think of them, obviously, as the sensations themselves.

3.4.2 Sensory qualities and the intermediary role

An assumption that is very explicit in Kripke’s discussion is that there is only one kind of thing that can stand as intermediary between the scientific theory and the common term, namely, sensations. We identify heat by its effect on us, Kripke insists, and therefore naturally tend—erroneously—to identify heat with whatever could have had that effect on us. It is a tendency we can recognise, on reflection, to be misleading, but nevertheless is the reason that there is a sense in which it seems reasonable to think that heat might not have been molecular motion; that is the sense in which it is reasonable to think that the cause of our feeling hot might not have been heat. But, so goes the argument, we don’t identify “feeling hot” itself through its effect on us – it is an effect on us! It follows that there is no analogous mistake to be made in the case of brain states and feeling hot; there is nothing that might seem contingent about sensations and brain states being one and the same thing, if that is what they are, since being one and the same thing is a relation that holds of everything, necessarily. This point is sometimes put by saying that “scientific” (or a posteriori) identifications require that we are able to separate the appearance of the thing we are talking about, and its reality (in the above case, the separation is between the sensation of feeling hot and heat itself). If that is so, we appear forced into the position of rejecting the possibility of identifying a sensation, say pain, with a brain state since there seems to be, as Daniel Stoljar [47] has recently put it, “a certain very natural and ordinary conception of pain according to which… there is no appearance/reality distinction with respect to pain” (p. 47).

Now it is worth questioning whether Kripke was right in thinking that only sensations could play the intermediary role that Kripke suggests (and we are happy to grant) needs to be played. Take, for example, the above example of the route from heat to molecular motion. The intermediary in that case is purported to be the fact that we think of heat as being the cause of our feeling hot. Could we just as easily have said that we think of heat as being the cause of, say, ice melting? If we can, then the path from heat to molecular would look something like this: [48]

HEAT

|| (is thought to be)

The cause of ice melting

|| (happens to be)

MOLECULAR MOTION

The appearance of contingency is preserved, since it seems not all that difficult to confuse the relationship between heat and molecular motion with the relationship between heat and the cause of ice melting. Moreover, since we are not identifying heat with whatever could cause ice to melt, it is reasonable to think that something other than heat might have been the cause of ice melting; but it is not thereby reasonable to think that something other than heat might have been molecular motion.

If the identification of heat with molecular motion could take this route, then it cannot be a general rule that identifications of this kind must allow for a distinction between the appearance and the reality of the phenomena in question. In the above example, we mention only three things: “heat”, “the cause of ice melting” and “molecular motion”. There is no mention of appearances as such, only reality. [49]

3.4.3 Evidence as the real intermediary

What reason is there for thinking that qualitative facts must play the role of intermediary in the justification of statements such as “heat is molecular motion”, “water is H₂O”, or for that matter “sensations are brain processes”? In Naming and Necessity, Kripke himself fluctuates between talking about the intermediary as a reflection of a kind of naïve epistemic state and of a specifically qualitative state. Take the following passage:

Someone can be in the same epistemic situation as he would be if there were heat, even in the absence of heat, simply by feeling the sensation of heat; and even in the presence of heat, he can have the same evidence as he would have in the absence of heat simply by lacking the sensation S. (p. 152)

Here Kripke is simply talking about evidence, and his point is easily generalisable. Take E to be generally accepted evidence for the presence of some phenomena P. It is possible, Kripke is saying in the above passage, for E to be present without P and P to be present without E. Smoke is good evidence that there is a fire, nevertheless smoke can exist without a fire, and fire without smoke. Moreover, mirroring Kripke’s claim we might say, if smoke were indeed our only evidence of fire, that we could be in the same epistemic situation as we would be if there were fire, even in the absence of fire, simply by knowing there to be smoke; and even in the presence of fire, we can have the same evidence as we would have in the absence of fire simply by knowing there not to be smoke.

Now take the following passage of Naming and Necessity, which immediately follows the passage just quoted:

No such possibility exists in the case of pain and other mental phenomena. To be in the same epistemic situation that would obtain if one had a pain is to have a pain; to be in the same epistemic situation that would obtain in the absence of pain is not to have a pain. (p. 152)

So, for Kripke, the only evidence available for one’s being in pain is simply that one is in pain. There is certainly this much that is true about what Kripke is saying: there is no inference that we need to make in order to justifiably believe, and therefore come to know, that we are in pain. Indeed, the difficulty comes when we try to doubt it! So we can agree, at least for the sake of the argument, that there is a certain kind of evidence for the existence of pain that is all but irrefutable, viz., feeling pain. There is a kind of triviality to the proposition that to truly be in pain is ipso facto to know you are in pain.

So if we take E₁ to be Bob’s feeling that he is in pain and P₁ to be the state of affairs such that Bob is in pain, then our general principle above seems to fail; that is, it does not seem possible (we are granting) that E₁ obtain without P₁. The reason that a posteriori identities seem contingent even though they are necessary, according to Kripke, is that we in effect mistake the strength of our usual evidence. In effect, we take feeling hot to entail the presence of heat, which of course it does not; and that it does not is clear upon reflection. [50] In the case of pain, however, to feel yourself to be in pain does seem, in some sense, to entail the presence of pain. We cannot mistake the strength of the evidence for it in the same way that we can mistake the evidence for the presence of heat.

3.4.4 Identifying sensations in other people

That is all very well for our own case. But it hardly needs pointing out that the kind of evidence for the existence of pain in ourselves is different to the kind of evidence for its existence in others. And this is where I think Kripke’s argument fails.

My evidence that you are in pain is nothing to do with your own feeling yourself to be in pain; I can’t feel your pain, so it can’t count as evidence for me that you are in pain. What is the evidence I use to establish that you are in pain? Well if I see you get hit by a car, all things being equal that is good enough evidence for me that you are almost certainly in pain. Not necessarily, of course; you might just have been injected with morphine and not feel a thing. Just as obviously, you might already have been in pain without my knowing it before you were hit. But I have a general idea of the kinds of situations that accompany pain – God knows I’ve been in a few of them myself, and can extrapolate from there (I can also ask you)! So let’s take this general idea of the kinds of external situations that accompany pain as evidence I might use to determine (not infallibly, of course) whether or not pain is present in another person.

We have now recovered our general principle: call E all the evidence I could possibly have that you are in pain, which does not presumably include any of your feelings; and call P your being in pain. Consistent with our general principle, E and P can exist without each other. The appearance of contingency also reappears: if pain in you is brain process A, and I identify pain in you by the ‘external’ evidence, then I can imagine something other than pain – and therefore something other than brain process A – accompanying all of that evidence. This may not yet give us exactly what we were after, but it does give us this:

PAIN IN OTHERS

|| (is thought to be)

The state accompanying car accidents, etc., etc..

|| (happens to be)

BRAIN PROCESS A

So it seems we are in a position to say how pain in others could be shown to be a brain process. But this contradicts Kripke’s claim that pain simpliciter could not be shown to be a brain state. That is because Kripke is assuming that our best evidence that pain is present comes from our feeling it to be present. That is true in some cases. But the vast majority of pains in the world we do not feel, and yet we know they exist. [51]

Now it would be an odd situation if we found we could imagine that the mental states of others were brain states, but could not imagine the same of our own mental states. What has gone wrong here?

3.4.5 Combining the first-person and third-person criteria

The first possibility is that we don’t, in fact, identify our sensations by their feel. Instead, we identify them entirely by the ‘external’ circumstances we are in. But that would be true only if behaviourism were true, which I will take to be a reductio of that possibility. It may be that we originally learned when to use the word “pain” by noting an ‘external’ state of affairs when it was appropriate, but it is hard to imagine that we use it that way still.

The second possibility is that because there are two ways of identifying the presence of pain –in our own case and in another’s case – there are, in effect, two concepts of pain; call them respectively the subjective concept and the objective concept. In that case, the statement “pain is a brain state” is ambiguous, and its justification will be different depending on which disambiguation is intended. But there are a number of reasons to think that this is an undesirable way of viewing the situation. Firstly, if there is indeed a difference between the subjective and objective conceptions of pain, it is obvious that they are strongly connected. As we discussed in the first part of this chapter, we need to have learned the objective concept before we learned the other – otherwise, how could we have learned which internal state to call “pain”? There is, in other words, something parasitic about the subjective conception of pain, if there is such a thing. Michael Tye’s (1999) notion of sensations as “direct recognitional” concepts seems to fit here. According to Tye, our concept of pain is our recognition that we are in the same state as some previous time. But if this is true, there needed to be an original state that we called “pain”, which we later recognised. What kind of state it was that we originally called pain is obviously not simply a matter of our choosing. It is only by using the objective concept (“the state accompanying car accidents, knife accidents, etc.”) that we first called an ‘internal’ state “pain”. Later uses of “pain” must in some sense refer back to the original association. What I recognise, when I recognise pain in myself, is not therefore any kind of direct recognition that this is the appropriate state to call “pain”. It is a recognition that I am in the kind of state that I am usually in under the same ‘external’ conditions by which I identify others as being in pain. To feel pain is not necessarily to feel oneself to be in a state called “pain”, any more than to look at water is to know oneself to be looking at a substance called “water”. More criteria are needed. It cannot be, then, that there is a “pure” subjective concept of pain by which we pick out when we are in pain. Insofar as the concept ‘pain’ is the meaning of the word “pain” our grasp of it is at least partly distinct from our recognition of its referent. Since it is possible to recognise the state, pain, from previous experience without knowing that it is called “pain” or is the referent of the concept ‘pain’ there must be something more to that concept than one’s memory of having been in a certain state. [52]

The third possibility, then, is that Kripke is simply wrong that we identify pains simply by their feeling, even in our own case. Perhaps, in line with the discussion in the first part of this chapter, we identify pain simply by recognising that we are in a state similar to the state we are normally in when our body is damaged. We recognise it in our own case very directly, and perhaps (perhaps!) irrefutably, but our recognition of it as being appropriately called “pain” comes only via an ‘objective’ set of criteria – the same criteria we use to establish that others are in pain. That is not to say that by “pain” we mean nothing more than “the state accompanying bodily damage”, merely that it is, in Kripke’s vernacular, an intermediary between the ourselves and the phenomenon, in the same sense in which “the cause of heat sensations” is the intermediary between ourselves and the phenomenon of heat. If so, then we have arrived at a refutation of Kripke’s objection to the identity theory (and physicalism generally), since we are able to replace “pain in others” in the above schema, with just “pain”. The appearance of contingency in the statement “pain is brain state A” can be shown in my own case in the following way: in the same way that I may mistake water with the way that I recognise it – its taste, etc. – I may mistake pain with the way that I recognise it – as the state accompanying bodily damage. Just as water is not necessarily tasteless, pain is not necessarily the state accompanying bodily damage. So there is no greater difficulty in justifying “pain is a brain state” than there is in justifying “water is H₂O”.

4. An Indexical Theory of Sensory Concepts

ain is a state that feels a certain way. It is the kind of feeling we often get when we do damage to ourselves. And when we’ve got it, we know we’ve got it. Describing pain is notoriously difficult (I remember always being stumped as a child when my mother asked me to describe my headache), but it is very much easier to recognise.

Or at least, it is easy to recognise in myself. I have a particularly stoic friend from whom I cannot recall so much as a grimace escaping. I’m sure he must feel pain as often as I do, but he goes to great lengths to conceal the fact; a habit that comes from a kind of politeness that seems almost to border on the perverse. The epistemological gap, for me, between my own pain and my friend’s pain seems vast. Now perhaps it is a gap that can be closed – after all, if pain turns out to be some brain state, then provided I have the right brain imaging technology I ought to be able to tell as well as my friend whether he is in pain. But as things stand I am left to guess that he is in pain when things that would normally cause me pain happen to him. For now, that is just the best that I can do.

When we infer that some third party is in some mental state, what kind of inference is it? Even if we take on board the analysis of sensations put forward in the last chapter, it still seems to make sense to wonder whether two people could, as it were, fit the definition of the kind offered for some sensation and yet be experiencing different feelings. If we again take the sensation of green as an example, our analysis yielded something like “the kind of state I am usually in when I look at green things, and which tempts me to think that there is a green thing in front of me”. It seems to make sense to wonder whether two people could be in exactly that state and yet be having very different sensations. But since our analysis was supposed to be of the very sensation we call “green sensation”, in what sense could two people both be having a green sensation by our analysis and yet be having different sensations? In what respect could those sensations differ, given that our analysis was supposed to be of one sensation only?

4.1 Privacy and the Problem of the Inverted Spectrum

4.1.1 The problem

The “inverted spectrum”, is the possibility—or at least apparent possibility—that two people (or groups of people) might see the colours of objects in the “negative” with respect to each other, while they nevertheless both see those objects as being the same colour, and are in the same functional state when they do. According to Block (1990, p. 53), it is possible that “things we agree are red look to you the way that things we agree are green look to me” even if we are in the same functional or representational state. In other words, Block insists that it is possible that two people both see an object as green, and yet have different colour sensations with respect to it. A “purely” representational theory of conscious experience cannot tolerate that possibility since, for such a theory, to have a sensation with respect to the colour of an object is nothing more than to see that object as being a certain colour. If you and I have different colour sensations when we look at grass, then you and I see the grass as being different colours, full stop; we genuinely disagree about what colour the grass is, though we may not realise it.

Many philosophers agree that there is a sense in which inverted spectra are conceivable, and another sense in which they are not. The sense in which they are conceivable relies on the non-functional, non-representational concept of sensation. To Loar and Tye that concept is “recognitional”, to Chalmers it is not analysable, and to Block it is “irreducible”. The sense in which inverted spectra are not possible is the sense that relies on the functional/representational concept of sensation. Block argues that those who hold that inverted spectra are flat-out inconceivable (e.g., Harman) are confusing the latter concept with the former, which is to commit, in Block’s phrase, “the fallacy of intentionalising qualia”. [53] I think that it is possible to satisfy the intuitions of these philosophers without resorting to two separate types of concept. It requires that we see phenomenal concept as partly indexical.

To begin to motivate this idea, imagine that you and I look at grass and typically have different sensations. What else must differ between us, and why? Our first attempt at an answer must note that what I have asked you to imagine is not altogether obvious. According to the analysis proposed for the phrase “green sensation”, it just is the state you are typically in when you look at grass. As long as you and I both agree that grass looks green, and that there is a particular feeling we recognise whenever we look at grass, that analysis dictates that both of our feelings deserve the name “green sensation”. And yet, our imagination seems to insist, surely it is possible that, in some sense, grass evokes a different sensation in you compared with me. What are we to say about this apparent possibility, and how are we to describe it?

4.1.2 Could “green sensation” be disjunctive?

Three possibilities suggest themselves. The first is that “green sensation” names a conjunction of kinds of state, in much the same way that “jade” names a conjunction of kinds of rock (it names both jadeite and nephrite). Now it seems possible that you and I can typically be in different kinds of states when we look at green objects. It is tempting to describe this as the possibility that green sensations are like jade; two kinds of state, rather than one. In that case, “green sensation” would name both the kind of state you are typically in when you look at green things, and also the kind of state I am typically in when I do that. A particular state then qualifies as being a green sensation if it is either of those kinds of state.

But the case of jade is importantly dissimilar to the case we are imagining. Imagine that, according to my lights, when I am in the state that you are typically in when you look green objects, I recognise it as a red sensation. In the case of jade, the surface properties of jadeite and nephrite are sufficiently similar that they are commonly thought to be the same thing. The “commonly” is important. If it were the case that half of the population took jadeite to be jade, and the other half took nephrite to be jade, we would be unlikely to say that “jade” refers both to jadeite and nephrite, as we in fact do. Rather, we would say that there is rampant confusion as to what “jade” really means. Yet the case we are imagining for “green sensation” is closer to this than to what is actually the case for jade. For, as stipulated, it is not the case that there are two states that we are both inclined to call “green sensation” – I at least am inclined to call one of them “red sensation”. There is a single state that you are inclined to call “green sensation”, and I am inclined to call something else entirely, with as much authority as you. If only you and I knew which state the other had in mind, we would know that we were disagreeing. And that is very unlike the case of jade.

4.1.3 Could “green sensation” be ambiguous?

This suggests a second possibility – that “green sensation” in the case we are imagining fails to have a single meaning; that it is ambiguous. Perhaps, that is, if you and I are typically not in the same kind of state when we look at green things, then we mean different things by it. It would be as though you have always meant by “bank” the financial institution, and I have always meant the edge of the river. Such a state of affairs would certainly make for confusion in discussions about the nature of banks.

There are, however, important considerations weighing against this description, the most important of which is that whatever kind of phrase “green sensation” is, we all learn it in the same way, or roughly the same way. If you and I both learn “green sensation” from looking at grass, trees, and so on, it is hard to see how the connotation of “green sensation” could differ between you and me. If we take on board the idea that sense determines reference, it is furthermore difficult to imagine the referent of “green sensation” differing between you and me.

4.1.4 “Green sensation” as indexical: the proposal

But there is a way for that to happen. Notably it happens for indexicals, such as here, now, and I. The referent of an indexical – take here as an example – differs according to the specific context of each use. When here is spoken, it refers of course to the place at which that specific token is uttered. Now the meaning of here is something like “my current location”, but there is an important sense in which, although my current location might not have been Melbourne, it is not quite true to say that here might not have been Melbourne —since that is just where I am at the moment. It is an indexical conception of phenomenal concepts that I will be arguing for in this chapter. It is the only way, I think, to account for our shared intuitions about various possible cases such as, and paradigmatically, “inverted spectrum” cases. Various philosophers account for these intuitions by imagining that for each sensation word there are two distinct concepts. A single indexical concept, or so I will argue, is able to carry out the functions thought to be carried out by two. The argument requires an explication of the views of those philosophers who explicitly maintain that sensory concepts have a dual nature, which I will present in the next section. Then I will present an indexical analysis which shows that there is no duality but rather a single concept with two logically separable components; which moreover, is thoroughly compatible with the empirical identification of sensations and brain states along the lines defended in Part One of this thesis.

4.2 The dual-nature accounts of sensory concepts

4.2.1 “Feels like this” and the splitting of sensory concepts

If pain is really nothing more than the state I’m normally in when my body is damaged, and which leads me to believe that my body is damaged, why is it that when I start to think about what pain is, bodily damage isn’t the first thing that comes to mind? When I start to think about what I really mean by “pain”, what comes to mind is a peculiar kind of feeling – one which I take great steps to avoid. There is a kind of immediacy to my feelings of pain that “the state that I’m normally in when my body is damaged” seems to miss. As a description of what I have in mind when I think of being in terrible agony, the analysis put forward in the last chapter seems a bit too—for lack of a better word—abstract. It may be the closest I can get to describing what I mean, but it is hard to feel that it doesn’t fall short, in the end, of what I truly have in mind when I think about pain.

It would be easy to bite the bullet here, to insist that the topic-neutral analysis doesn’t fall short at all. We have, after all, shown that topic-neutral descriptions, of the kind proposed in the previous chapter, are the best descriptions we have of sensations. If they are inadequate, then how do we manage to talk about our sensations at all? And what is missing from the topic-neutral description? An answer to this last question seems impossible for anyone who takes the worry seriously, since it is precisely the worry that words are insufficient for describing sensations fully. To ask what is missing from the topic-neutral description is, on the face of it, to ask for either a supplement to that description, or an alternative to it, both of which are ruled out by the apparently indescribable aspect of our sensations.

To philosophers who are inclined to worry that a description can’t capture what we really mean by “pain”, demonstratives often serve as alternatives to a description. In indicating what pain feels like, Ned Block (1992) is reduced to saying that it “feels like this” (p. 205). Block’s stress on “this” seems to emphasise that it is being used as a demonstrative; the reader is supposed to call to mind his or her idea of what pain feels like and imagine the Block is pointing to exactly that feeling. But Block is not pointing to your mental state when he writes “this” – he is pointing to his own, which none of us has access to. Even then it is hard to imagine how Block could literally be indicating his own sensations. By way of comparison, on being asked who Wittgenstein is I might reply, “Amongst all of the philosophers I’m thinking about at the moment, Wittgenstein is that one, the one I just picked out”. To someone who already knows who Wittgenstein is, they can imagine the kind of thought I would be having and, in a sense, know what “that” referred to. But by any reasonable measure of what it takes to demonstrate something, my “that” failed really to demonstrate anything at all. I didn’t point anything out, not really.

So Block’s use of “this” is unclear almost to the point of emptiness. But his is a far cry from an isolated case. Indeed, it is hard not to think that the “it” in “what it’s like”—used nearly universally in descriptions of conscious experience—is used in a demonstrative sense as well. The pronouns he, she and it can after all be used in the same way that this and that are used. Imagine, for example, that you and I are talking about our mutual friend, Bob. As we are talking, a man appears in the hallway acting strangely. If I were to ask, “What’s he doing?”, my question could in fact be ambiguous between my asking what Bob is doing, and my asking what the man in the hallway is doing. I could be using “he” to refer to the subject of our ongoing conversation (“What’s Bob doing?”), or I could be using it as a demonstrative (“What’s that guy over there doing?”). When philosophers talk about consciousness as the what it’s like of our mental states, the “it” is being used in its demonstrative sense and is thus as strictly meaningless (or almost meaningless) as the “this” in Block’s “feels like this”.

Sensations, if philosophers in the vein of Block are right, are things that we can think about, but not describe. We can call them to mind, remember them, recognise them, perhaps even, if Hume is right, imagine somewhat novel ones. But words are insufficient. We each know what pains and tickles are – and we know that each of us knows – but we simply can’t say what it is that we know. What we can say is what our sensations make us do, and what beliefs and desires they make us have. It is the thing itself, the intrinsic nature of the sensation, that is apparently beyond words.

It is easy to sympathise with this kind of worry. There is a real sense in which there is really no answer to the person who asks that you describe, say, nausea. Feelings just aren’t the kind of thing that can be explained to someone who hasn’t already felt them. Many philosophers working on consciousness feel the force of this as an objection to an analysis—in words—of phenomenal concepts. Those same philosophers usually also feel the force of functionalist or representationist analyses (or at least theories) of conscious experience. The result is more often than not an argument that for every phenomenal word there correspond two distinct concepts – a functionalist (or representationist, which I will lump with the former for now) concept and a non-functionalist one; or, if you like, an analysable concept and a non-analysable one. We saw earlier that Chalmers does it. We will see in this section that Block also does it, along with Tye, Loar, Papineau and—though in a sense that suggests my own view—Nagel. It is instructive to have a closer look at their views.

4.2.2 Loar and Tye on sensory concepts as recognitional

We have already seen that for Chalmers, there is no analysis of the non-functional concept of consciousness. Moreover, it is the nature of the referent of that non-functional concept that is the “hard problem” of consciousness studies. Loar (1997), on the other hand, urges that our non-functional concepts for sensations are demonstrative, recognitional concepts and refer to exactly the same things that our functional concepts for sensations refer to. For Loar, functional analyses don’t completely capture what we mean by sensation words, but they do fully describe the things to which sensation words refer. The fact that we can recognise sensations when they recur implies, or even entails, that we have some kind of concept for them, though it may be a primitive sort of concept that not so much picks out general features of the thing as simply causes us to react with an “Ah-ha – there it is again!” when it recurs. When we begin to pick out general features, we acquire a new concept – probably a functional one – for that same sensation. But acquiring the new “theoretical” concept (Loar’s word) needn’t cause us to lose the old one; we simply end up with two concepts, both of which pick out the same thing independently. So, according to Loar, although there is nothing about sensations that our functional concepts fail to capture, our possession of the recognitional concept can make it seem as though there is something missing. It is true, for him, that our functional concepts for sensations do not completely capture the way we relate to our sensations, but only in the sense that we have an additional concept over and above our functional concept. Both concepts refer to the same thing, in different ways.

Michael Tye (1990, 1999 and 2000) holds a representational view of conscious experience. Sensations for Tye are essentially representations of the world and of ourselves—essentially, but not a priori. Tye argues that the representational nature of sensations is revealed through introspective investigation, and that it is not part of phenomenal concepts. Phenomenal concepts are for Tye recognitional, in much the same sense that Loar argues they are recognitional. He does not hold that we each possess two independent concepts of sensations, but there is a real sense in which the way we come to think of sensations as representational in Tye’s account is very similar to the way Loar argues we come to think of them as functional, namely, by observing our sensations and noticing general features of them. But Tye, unlike Loar, is happy to deny that the result of our introspective observation is significant to our everyday conception of conscious experiences.

The difference between Tye and Loar in this respect is a small one and not worth getting stuck on, but it is perhaps worth a few words. Consider the proposition, “Water is wet”. Is it an a priori truth, or knowable only after experience? It seems possible to respond to both options in the affirmative. If we are challenged, we are likely to head for the nearest tap and demonstrate to our challenger that the stuff coming out is indeed wet. That makes it looks very much a matter for a posteriori investigation. On the other hand, if it turned out that water isn’t wet, I’d have to confess that the meaning of “wet” escapes me; if water isn’t wet, I have no confidence in my ability to tell what is. The same situation applies to, say, the feeling of headaches. Take the proposition, “Headaches are felt to be in the head”. It is a proposition that can look a posteriori if when challenged I wait until the next time a have a headache and “observe” it, noticing of course that I do indeed feel it to be in my head. But on the other hand, there is something absurd in the idea that I might discover that it is not felt to be in my head. How do I recognise when I have a headache apart from the fact that they are pains that are felt to be in my head? The resolution of this kind of apparent paradox was the subject of the first chapter. The fact that I sometimes suffer pains that are felt to be in my head is an entirely a posteriori matter, but given that I am sometimes in such states, that they are headaches is knowable a priori. So when Tye conducts his investigations into his own sensations and finds that they are representational, it is reasonable to assume that he knows which sensations to call headaches only because his grasp of English tells him that headaches are the sensations that feel painful and are located in the head.

So although Tye does not say it explicitly, I am tempted to say that his position commits him to both a recognitional and a representational concept of sensation, in much the same vein as Loar’s recognitional and functional concepts.

4.2.3 Block on P-consciousness

Ned Block does not identify two phenomenal concepts in quite the same way, but the overlaps are many. Block (1995a) distinguishes “phenomenal consciousness” from “access consciousness”. Phenomenal consciousness (or simply “P-consciousness”) aligns closely with Chalmers’ concept of the same name; it cannot be defined, and is the source of the real mystery of the mind. Demonstrations, apparently the poor cousins of definitions, are all we have but suffice for communication, according to Block: “really all one can do is point to the phenomenon” (p. 230). P-consciousness is to him that certain indefinable something which we all know exists and which is the source of the deepest mystery in the study of consciousness. His explication of P-consciousness even involves a discussion of the “explanatory gap”, which he includes “partly by way of pointing at P-consciousness: that’s the entity to which the mentioned explanatory gap applies” (p. 231).

Block denies that our concept of P-consciousness is that of a representational or functional state, and urges that anyone who disagrees is confusing P-consciousness with access-consciousness, or “A-consciousness” – which is, conceptually, a functional state. The concept of “A-consciousness” is what Chalmers means by our “psychological” concept of consciousness, what Loar thinks of as our “theoretical” concept of consciousness and, though more roughly, what is for Tye the representational nature of experience. More precisely, A-consciousness is essentially a functional state whose function is to control our thoughts and behaviour given various other thoughts and external input.

In an earlier paper Block (1990) makes a distinction between two types of content of conscious experience, representational content and qualitative content. He distinguishes the two in the following way:

If blood looks red to both of us, then in the intentional sense of ‘looks the same’, blood looks the same to us… The qualitative sense of ‘looks the same’ can be defined via appeal to such notions as “what it’s like,” or alternatively, by direct appeal to the inverted spectrum hypothesis itself. If your spectrum is inverted with respect to mine, then red things look the same to you – in the qualitative sense – as green things look to me. (p. 52)

Qualitative content, then, is just that aspect of conscious experience that goes beyond merely representing the world as being a certain way. In whatever sense it is possible for two people to perceive the world as being the same way, and yet have different sensations, it is the qualitative aspect of sensations in which they differ. Since it applies to objects in the world, “looking red” is for Block no way to define the qualitative content of our experience; it is merely a way of defining the representational content of our experience.

4.2.4 Papineau's first-person concepts and Nagel’s partial rejection

Papineau (1998, p. 381) explicitly agrees with Chalmers on the distinction between “psychological” and “phenomenal” concepts of sensations. Before Chalmers had published, he had in any case committed to a view of that kind. In Papineau (1993, p. 103-118), he writes that our beliefs about sensations tend to involve the bringing to mind of an exemplification of them. When we think about the sensation of red, for example, we bring to mind a mental state that is somewhat like the state we are in when we are looking at red objects (that is, we imagine looking at a red object). Beliefs that involve this “secondary” kind of state—that is, one that resembles its object—are beliefs that utilise what Papineau terms a “first-person” concept. It is possible, Papineau argues, to have a “third-person” concept of a sensation. To utilise a third-person concept of pain is to think of someone’s pain in a way that does not involve imagining what that pain would feel like. It is to think, for example, about their pain behaviour—or, for that matter, their brain state. [54] The two concepts are independent of each other, in the sense that one could posses both concepts without realising that they referred to the same thing—just as one could possess both the concept ‘morning star’ and the concept ‘evening star’ without realising that they are one and the same object. Papineau argues that it is possible to show that physicalism is true without establishing a conceptual connection between first-person concepts of sensa-tions and any third-person concepts whatsoever.

Papineau’s argument for physicalism about sensations relies on causal considerations, such the causal closure of physics. Nagel (2000) expresses sympathy with that approach. What stops it from succeeding, Nagel argues, are the nature of sensory concepts; the causal evidence does push us to identify the mental and the physical, he says, “if only we could make sense of the idea” (p. 442). What stops us from making sense of the idea is that as we currently understand sensory words, it is a kind of category mistake to use them to refer to physical states. The phrase “pain is a brain process” currently has no clear meaning, according to Nagel. [55] That is because—a point that is now familiar—pain is for him an irreducibly first-person concept. But Nagel differs from Papineau, and all of the others mentioned above, in denying that we have distinct third-person concepts of sensations. There is only—but there is always—a third-person part of sensory concepts, which is also describable along the lines of Chalmers’ “psychological” conception of sensations. [56] For reasons related to Wittgenstein’s private-language argument, Nagel rejects the possibility of a purely first-person/subjective concept of a sensation. He argues (p. 451) that “to have the concept of pain a person must apply it to his own sensation in the circumstances that enable others to apply it to him.” The circumstances that enable others to apply it to him are, of course, just the circumstances that satisfy the third-person conception of pain. There must be, he concludes, a logical connection between the description of those circumstances and the subjective conception of pain. But, he points out (p. 443), “that is not to say that I understand how the first person and the third form two logically inseparable aspects of a single concept—only that they do.” As it turns out, this is the very problem I hope to have solved.

I think Nagel is right that sensory concepts are a logical amalgam of a subjective conception of what sensations feel like and third-person conception, though the third-person conception I have in mind is the conception I defended in the last chapter. What connects that conception to particular sensory qualities is the fact, I propose, that sensory concepts are indexical. It is time to flesh out this proposal.

4.3 Sensory Concepts As Indexical

4.3.1 Demonstrating sensory states

Recall the analysis of sensation concepts that in the last chapter I suggested was the correct one, namely that to have a sensation as of green, for example, is

(1) to be in the kind of state similar to the one I am usually in when I look at grass, trees, etc., etc.; and

(2) to be in the kind of state that inclines me to believe there is a green object in from of me.

Now the first part of the analysis requires, at least on the face of it, that I do recognise that I am usually in the same state when I look at grass, trees, and so on – hence the use of the definite article. But notice that what I recognise is a state of me only. I could not plausibly replace the occurrences of “I” in (1) with “we”. I never recognise myself to be in the kind of state that we are usually in when we look at grass. I have no access to anyone else’s state when they look at grass, so I can’t possibly recognise myself to be in the same state as someone else when they look at grass. It is this apparently indexical nature of sensation concepts that allows there to be a sense in which you and I can have different colour sensations when we look at grass while both of us represents the grass to be green. “Sensation of green” is the state I am usually in when I look at grass, which is very likely – but not analytically – the state we are usually in when we look at grass.

If sensation concepts are indexical in this way, it leaves open the possibility that you and I could be in different states when we look at grass and yet both be referring to that state by “sensation of green”. When philosophers refer to a particular sensation as “this feeling”, they are invoking just this aspect of sensations: we recognise when we are in them. Loar and Tye are inclined to think that this recognition is itself a kind of concept. To recognise something, they argue, is in a sense to classify, and therefore to possess a concept, for it. Whether or not this is true, there is no need to take a stand on it to solve the inverted spectra problem. All we need is that we do recognise when we have particular sensations when they re-occur, and that it is the state we recognise ourselves to be in when we look at grass, etc., that we are referring to by “green sensation”.

So perhaps the concept of a green sensation is partly the concept simply of “that state (which I am normally in when I look at grass, etc.).” Without the qualification in parentheses, the “that” makes no sense. Often, “that”, as in “that lamp post”, is accompanied by a gesture of some kind to indicate which object I am referring to. But a definite description suffices; if I want to indicate something, I can either point to it (“that lamp post”), or append a short description (“that lamp post, the one half way down the street in front of the big old house.”). In the case of sensation, actual pointing is difficult, but the analysis offered in the last chapter seems a way of pinning down which feeling I am pointing to when I say “that feeling”.

But sensations are states that, so it seems, only the person actually having them has access to. So if sensation concepts are partly demonstrative, the meaning of sensation words must in some sense be different for each individual. Each person is after all restricted to demonstrating their own mental states, and effectively barred from demonstrating the particular mental states of others. Certainly, when I “point” to my own green sensations, I have no straightforward way of pointing to that same kind of state in you.

4.3.2 The beetle in the box

Do we know the meaning of “green sensation” from our own case, then? That is what the discussion so far seems to suggest. The answer, I think, is in a sense, yes, and in a sense, no. If sensation concepts do have a demonstrative aspect to them, then our analysis ceases to be a simple rigidified description. It has taken on the logic of an indexical. To understand how this could change the nature of the analysis, it will be helpful to have a look at a problem raised by Wittgenstein (1972, §293)—the so-called “beetle in the box” argument. The argument is this:

Now someone tells me that he knows what pain is only from his own case! – Suppose everyone had a box with something in it: we call it a ‘beetle’. No one can look into anyone else’s box, and everyone says he knows what a beetle is only by looking at his beetle. – Here it would be quite possible for everyone to have something different in his box. One might even imagine such a thing constantly changing. – But suppose the word ‘beetle’ had a use in these people’s language? – If so it would not be used as the name of a thing. The thing in the box has no place in the language game at all; not even as a something: for the box might even be empty. – No, one can ‘divide through’ by the thing in the box; it cancels out, whatever it is.

In this thought experiment [57] , we can see – and therefore have evidence for the existence of – our own, but not others’ beetles. We have only the protestations of others that they have a beetle at all. But perhaps some people have empty boxes (we might call such an idea “the absent beetle hypothesis”!). Are those others justified in claiming that they have a beetle in their box? For all they know, everyone’s box is empty and it is the emptiness in the box itself that “beetle” refers to.

It is hard not to think that the “absent qualia” [58] and inverted spectrum hypotheses are related in some way to equivalent “hypotheses” with respect to beetles in Wittgenstein’s example. If “green sensation” is just that state which I am sometimes in, then it seems that the meaning of “green sensation”, like “beetle”, must cancel out. So “green sensation” must mean something more than just “that state”. And that something more is what I have suggested in the previous chapter. “Green sensation” is not just that state; more accurately, it is that state which I am in when I look at green things, and which inclines me to believe there is a green thing in front of me. The italicised clause can be considered to be a kind of whispered afterthought, but it completes the demonstration; it is the substitute for actual, literal, pointing.

Now since the pointing can be done by way of a description of sorts, that description ought to be able to exist stand-alone, as it were. If green sensation can be “that state (which I am in when I look at green things, and which leads me to believe there is a green thing in front of me)”, then it can be the state which I am in when I look at green things, etc., etc.. But it is important again to notice that the “I” cannot be changed into a “we”. Since we (famously) lack access to other people’s states, I can never demonstrate sensations by saying that they are “that state (which we are in when, etc.)”. So “the state we are in when…etc.” cannot, as it were, be derived from a demonstration in the way that “the state I am in when…etc.” can. What that suggests is that the state we are in when we look at grass cannot be what we mean when we refer to sensations as “this kind of feeling”, as Block and many others do.

4.3.3 Public names for ‘private’ objects

To illustrate this point, take Wittgenstein’s example of the beetle in the box. “Beetle”, in his example, refers to the thing in our box – the box that we each hypothetically possess. It functions, or seeks to function, as a garden variety common noun. But it seems prevented from having a proper referent because each person apparently knows what “beetle” refers to from their own case only; there is no way of fitting the commonality of the noun into our epistemic access to its referent. We can agree with Wittgenstein that it is difficult to imagine how “beetle”, with its stipulated meaning, could successfully refer.

But aside from whether “beetle” refers to it, we can easily imagine that everyone knows – in some sense – what is inside their box; they can see it, even if no one else can. It might be hard, even indirectly, to compare the contents of their box with the contents of someone else’s box, but each person ought to be able to tell if, for example, the thing in their box changes its appearance (they can see it, after all, even if talking about it is hard). Now if each person can see what is in their box, they can name it; each person can have a separate name for the thing inside their own box. The name is in some sense a “private” name, and not part of a public language (whether it deserves to be called a name I leave aside for now). In the same way I might have my own “private” names for the bushes around my house, which I tell no one.

Now imagine that two people meet and discover that each of them has a private name for the thing inside their box. They discover that the name they have both come up with is “Bob”. It is pure co-incidence. They start talking about “my Bob” and “your Bob”, but after a while tire of the “my” and simply use “Bob” to refer to the thing inside their own box. When one says such things as “Bob’s not what it used to be”, the other understands that their own “Bob” is not being referred to. In effect, the referent of “Bob” depends on who says it. When John says “Bob”, he is referring to the contents of his box, and when Alice says “Bob”, she is referring to the thing inside her own box.

Imagine further that, astoundingly, everyone turns out to have named the thing in their box “Bob”. No one quite knows what is in anyone else’s box, but everyone knows that “Bob” is each person’s private name for it, whatever it is. Eventually, “Bob” comes to mean, in a sense, “the contents of my box”. Whenever “Bob” is spoken (or written), it refers to the contents of that person’s box – the person who said “Bob”, for each particular utterance.

For the next generation of language users, “Bob” might cease entirely to be a private name, and become an indexical. Should that happen, the reference of “Bob”, like the reference of “here”, will change according to the context of its use, although within each particular context, there will be a fixed rule for determining the referent. In the case of “here” that rule is something like “the location of the speaker”; in the case of “Bob” the rule is “the contents of the speaker’s box”. Those who know the meaning of “Bob” will likely have a very different attitude to what it refers to in their own case, and in the case of others. It is only in their own case that they learn what it refers to. When someone says “Bob”, others know the rule they are using, but not really what the referent is. We may imagine that each person’s box has the same contents, and have that in mind when others say “Bob”, but leave it open whether that is actually the case; that is, we understand the possibility that the contents may differ.

“Bob”, as stipulated here, is part of the public language. But there is a real sense in which we learn its meaning from our own case only. We learn what it refers to from our own case only, though not the way it refers. The sense in which “Bob” means “the contents of the speaker’s box” is the sense in which its meaning is shared and public. The inhabitants of “Bob” world are similar in this respect to those using “here” in a world in which there is no direct contact, and communication is only ever by phone. In that world, no one knows which place “here” refers to when someone else says it, but everyone knows that it refers to the location of the speaker, wherever that is. Everyone knows what “here” refers to only in their own case, though they know that when another person says it, it refers to that person’s location.

In what respect could “Bob” be similar to “pain”? I am hoping that the analogy between “Bob” and “pain” is fairly clear. Both are words that apply equally well in your case and in mine, and yet the word evokes in me an image of what it applies to in my own case only. I know that the word almost certainly refers to something in you, and I’ve good inductive evidence that what it refers to in you is the same kind of thing as what it refers to in me (we seem to be made of the same kinds of things generally, after all – arms, legs, and so on). But I can’t see whether or not that is the case.

If our sensations are private and we can recognise them, we could presumably have private names for them, à la “Bob”. That would allow us to talk about them in some sense, though perhaps no one would know what we were talking about. It would be better if we could somehow bring those private names into the public sphere. And of course an obvious way of doing that would be to relate those private names to external objects. I might say, for example, “Ted” is my private name for the special state I recognise myself to be in when I’m facing grass, trees, and parts of the Italian flag with my eyes open. Perhaps my neighbour calls that state “Anne”. It makes sense for us to come up with a single name for my “Ted” and her “Anne”. Let’s say we call it “Tan”. Is my “Tan” the same as her “Tan”? Only if the state I recognise myself to be in when I look at grass, etc., is the same as the state she is in when she does the same. We may guess that they are the same, since we are the same in very many ways, but neither of us really knows for sure. Still, “Tan” has a place in our common language, even if its referent is accessible only in our own case.

So it is possible for private sensations to be referred to by words in a public language, and for there to be a sense in which we learn the meaning of those words from our own case only. But of course nothing so fanciful as the story I have just told is actually the case for “sensation of green” or “pain”. We don’t bother – generally anyway – to privately name our sensations.

4.3.4 Naming the sensations

Perhaps when we learn to refer to our sensations, we learn to do it through words like “Bob” and “Tan”—words whose rule for referring is the same for everyone who uses it, but which refers to a private, and possibly different, state in each person. The clearest evidence that this might be the case comes from the way we refer to colour sensations. Colour sensations and the colours of objects are often run together. When I look at a rose, my sensation is often described as “red”, even though it is only the rose that is literally red. [59] It is not a great stretch to think that the reason for this is that we automatically refer to our colour sensations through properties of the external objects they ‘track’. It seems plausible, even, that the way we learn the names of colour sensations is parasitic on the way we learn the names of colours. When I walk with my young nephew, I may point at the sky and say, “That colour is called blue”. My nephew thereby, in the right context, learns the English word for the colour blue. [60] He also, plausibly, learns the English word for the sensation of blue. Certainly, no one is going to point to his sensation and say that in English we call it “blue”! To learn by ostension that the sky is called “blue” is to learn by a kind of indirect ostension that a certain internal state is “the sensation of blue”. After our walk, my nephew will no doubt be able to recognise other objects as being “blue” and do so on the basis of the sensation they elicit in him which he also associates with the word “blue”.

This story, though pedestrian, explains how we could use words to refer to private states that we are aware of only by our recognition of them and how any analysis of what is commonly meant by those words might seem to fail to capture what we each, individually, mean by them.

This point is most clearly put using colour sensations as an example, but essentially the same thing can be shown for other sensations, such as pain. When someone reports that they are in pain, it is prima facie a report of a sensation, not – as for colour sensations – a report of the external world. When I say to my nephew that the sky looks blue to me, it might be argued that I am simply reporting my temptation to say that the sky is blue. But when I report that I am in pain, it is less plausible (though not, I believe, implausible) that I am just reporting my temptation to believe that my body is damaged in some way. It seems that I am directly reporting my “internal” state. Still, when I do report that I am in pain, typically some bodily damage is evident. This much will be obvious to my young nephew. The next time he bumps himself, and finds himself in a particular “internal” state, he may well have learned to call that state “pain”. The word “pain” is likely to mean for him the state people are referring to when they injure themselves. When people act in the same way that he is inclined to act when he injures himself, he will be inclined to think that they are in the state called “pain”. Exactly what state other people are referring to when they say they are in pain my nephew will only ever know through inference from his own experiences of bumps and scratches. Through its typical causes, and perhaps its typical effects and so on, we can name a private sensation, which may be a different sensation in each of us, and simultaneously pain in each of us. Nevertheless it is the private sensation, not the typical causes and effects, that are named.

4.3.5 Character, content, and the inverted spectrum again

Sensations are, in an important sense, private states, but that does not stop us referring to them. We recognise when we are in them, and through that recognition use external criteria to manufacture a common name. As long as that common name is indexical by nature, it can refer to each person’s private state. It can seem demonstrative, since it is natural to think only of our own case as the proper referent of the name (as an indexical, it is the proper—and only—referent when we use it). It can also seem to have a general meaning, such as “the kind of state I am in when…”, since that is the rule that must be followed in finding the referent in any particular context—for any particular person.

If all of this is plausible, then an explanation for our clashes of intuitions in “absent qualia” and “inverted spectrum” thought experiments need not appeal to two distinct kinds of phenomenal concept. A single indexical concept suffices. The two kinds of concept proposed by Chalmers, Block, Loar, etc., are actually mirrored in the two parts of an indexical concept of sensation as discussed here. Those two parts have analogues in the sense/reference distinction often applied to other words, although the “sense” of an indexical is not exactly “sense” as Frege described it. In “Demonstratives”, David Kaplan (1989a) proposed to distinguish the “content” of an indexical from its “character”. The content of an indexical —what a particular use of it picks out—may be different for each occurrence. If we take “now” as an example, its content will change depending on when it is uttered. The content of now is 11:36am…while that of now is 11:37am. The content of “now” cannot, as it were, be abstracted away from a particular utterance; it picks out different things at different times. Content can only be applied to tokens. The character of an indexical, on the other hand, stays the same across all tokens of a single indexical type. Again taking “now” as an example, its character is “the current time”. Indeed, having the character “the current time” is what makes “now” the type of word it is.

Kaplan’s proposal is a useful one in this context, because it allows us to explain many of the conflicting intuitions we have about the various thought experiments involving consciousness. It is crucial that occurrences of two indexicals may have the same content, yet differ in their character, or agree in their character but differ in content. If words for colour sensations are indexical, then the possibility of inverted spectra may simply be the possibility that the content of colour sensation words spoken by one person is different to the content of colour sensations words spoken by another person – different, moreover, in type as well as token.

Soo, for two 'inverted' speakers, “green sensation” is for both speakers “the state I am in when I look at grass”, but which in turn is state X for one speaker and state Y for the other. Both will be correct in describing green sensations as the state they are in when they look at grass, and in explicating the concept “green sensation” as exactly that, while at the same time differing about the kind of state that “green sensation” refers to in their respective cases. Both speakers recognise when they are occupying state X and state Y respectively; moreover, they recognise it – at least partly – as the state they are typically in when they are looking at grass. “Sensation of green” is the only name that either speaker knows for that state – X for one, Y for the other – and so is what they use to talk about it, though what the state is “like”, its nature, is in a sense known only to its possessor.

The character of “green sensation” in both cases is “the state I am normally in when I look at grass” – that is what makes them two occurrences of a single type – while the content in one case is state X and in the other state Y. In fact, of course, it is likely that the content of “green sensation” is the same, or at least similar, for all of us, but the logic of “green sensation” allows for it not to be. Character does not, by the nature of indexicals, entail any particular content. That is why, intuitively, inverted spectra are possible. What I have in mind when I talk about green sensations can be described as “the state I am normally in when I look at green things”, but what I really mean is just this state. Similarly, when I use “here” I can be described as talking about “my current location”, but what I really have in mind is just this place. The two explications differ because my current location might not have been this place, and the state I am normally in when I look at green things might not have been this state.

So, in a sense, Tye and Loar are right in thinking of phenomenal concepts as partly recognitional; partly, we do know them only in our own case, by recognising when they recur. Block and Chalmers are also right in thinking that there is an important sense in which there aren’t any words to describe the nature of our sensations, since sensations are in an important way private. They are all also right that the sensations have aspects to them that we can latch on to with words, namely, at what times they occur, what they tempt us to believe, how they incline us to behave, and so on. It is these aspects of sensations that we use to talk about them, though for all we know the same internal state in you and me inclines you and me to believe and act differently, and may accompany various environments in various individuals. [61]

4.3.6 Indexicality, ‘Martian’ pain, and identity claims

If sensory terms are indexical, in the sense described here, one important consequence is that they are not typical natural kind terms. In particular, indexicals do not admit of type identity claims in the usual sense. If, as I have just argued, the logic of sensory terms allows that they might refer to different kinds of states in different individuals, then there cannot be a bar on, say, non-carbon-based Martians referring to their internal states as “pains” and so on. Therefore, it seems, even if it is true that sensations in us are brain processes, the simple possibility that a Martian could nevertheless refer to its internal states using the very words we use, proves that it cannot be necessary that sensations are brain processes. Moreover, I am not inclined to deny this possibility. [62]

Fortunately, the indexical analysis deals with problem particularly well, in the following way. It is reasonably clear that we tend to think of sensations as natural kinds—as things worthy of scientific investigation as to their nature. They are, after all, actually being investigated scientifically. What seems to force the indexical analysis of sensory terms is that sensations are also, at least contingently and at the moment, private. But there is nothing preventing our conceiving of those private states as natural kinds—as the kinds of states I am in when I do this or that. Furthermore, it is natural to assume, and is empirically very plausible, that others are in the same kinds of states as me, in the same kinds of contexts (though no one has ever seen inside my head, for example, the inductive evidence is good that there is a brain in there). The indexical analysis, therefore, has no implications for the ability of brain science to claim to be the study of sensations. With respect to the possibility of Martian pain, the indexicality of sensory terms preserves the plausibility of the “sensations are brain processes” hypothesis at this point: since my use of sensory terms refers particularly to the kinds of states that I am in, if sensations are in fact brain processes in me then as I use the terms sensations are, and could not be anything other than, brain states. Now it is also true, given the indexicality, that as a Martian might use the terms, sensations might not be brain processes. The reason this poses no problem for the thesis that sensations are brain states is that it is uninformative for me to prefix my words with “as I use the term”. It is understood. Moreover, since no one in the current language community is a Martian it is plausible that, as we all use the terms, sensations are brain processes—and therefore that sensations are brain processes simpliciter. It is true that, in a sense, Martians could discover that sensations are something other than brain processes. But that would be sensations as they use the terms, not as we use them. A disagreement between Earthlings and Martians on this point would be tantamount to a disagreement over whether Earth is here or Mars is here—or whether, to force the analogy, “here” is multiply realisable since both are conceivably true. Both are conceivably true, but a person on Earth is still forced to say that here is, and could not be anything other than, Earth. [63] The indexical analysis presented here is thus consistent with the intuition that, in some sense, there could be such a thing as Martian pain. But it does so in a way that remains consistent with, of all things, an identity theory. [64]

I have argued in this chapter that the our concepts for sensation are not different from other natural kinds concepts in any way that impedes a priori passage from brain state to sensation. But it may still be argued that a gap remains. At best, our entailment will be from some physical state or other to the presence of the state I am in when... What is conspicuously missing is an entailment from the presence of a physical state to that state’s having a nature of the sort that we recognise when we are in it. I have been taking it for granted that our recognition of a sensation is unproblematic, but that recognition ought be a recognition of similarity in some respect or other. What we recognise is that the sensation has some particular nature, a “feel”. Because the analysis of sensation concepts I have argued for does not mention sensations having a certain feel, that is not what any physical state has to satisfy in order to satisfy the concept and therefore the reduction from sensations to physical states. To many writers a strategy of this kind skips over the essential problem of sensations, which is to explain in physical terms why a brain state has a certain “feel”. In the next chapter I will argue that this kind of objection is misconceived.

Part Three –Metaphysical

5. A Dispositional Theory of Sensory Qualities

5.1 The Explanatory Gap: Conceptual Version

n his paper, “Materialism and Qualia: the Explanatory Gap”, Levine (1983) argues that it will quite likely never be possible to explain the nature of sensations by appeal to functional or physical properties; not anyway, in the sense that we can explain the nature of water by appeal to the behaviour of H₂O molecules. There will always be a “gap” between the way things are from a physical point of view, and the way things are from a phenomenological point of view. The way things are in physical terms will never make intelligible the way sensations are. Levine has sympathy for physicalism as a metaphysical thesis and argues that this explanatory gap does not prove it to be mistaken, but he does contend that physicalism will remain speculative and unproven as long as the gap remains. He puts the point thus:

Now, if there were some intrinsic connection discernible between having one’s C-fibres firing (or being in functional state F) and what it’s like to be in pain, by which I mean that experiencing the latter was intelligible in terms of the properties of the former, then we could derive our measure of [the extent of physical similarity between pain states in different creatures] from the nature of the explanation. Whatever properties the firing of C-fibres (or being in state F) that explained the feel of pain would determine the properties a kind of physical (or functional) state had to have in order to count as feeling like our pain. But without this explanatory gap filled in, facts about the kind or the existence of phenomenal experiences of pain in creatures physically (or functionally) different from us become impossible to determine. (p. 360)

David Chalmers’ “hard problem”, much discussed since the publication of The Conscious Mind, takes it’s cue from a worry very like Levine’s. Chalmers insists that the real challenge for those studying consciousness is to explain how any story using purely physical or functional terminology could make intelligible why any particular sensation should feel the way it does. No matter what physical or functional properties are imagined to be responsible for some sensation, they will never “add up” to it, according to Chalmers and Levine. By contrast, once we know what molecules are and how they behave, we can understand why water has the nature it has.

5.1.1 The appeal to concepts and the appearance/reality distinction

Although for Chalmers the existence of an explanatory gap between physical or functional properties and phenomenal ones means that it can’t be merely physical or functional properties that are solely responsible for sensations, Levine draws a different conclusion. For Levine, the problem is simply that phenomenal concepts do not support identity statements of the relevant kind. That is, identity statements involving sensations will always seem genuinely contingent despite the fact that if true they are necessarily true. In a passage that is reminiscent of the discussion of Kripke and Chalmers in chapters two and four, Levine (1993) argues in a later paper that in order to “reduce” the phenomenal to the physical/functional in the same way that water was “reduced” to H₂O,

we would need to look for a property that is being reduced and then a property, or set of properties, by which the to-be-reduced property is normally picked out. Of course, this raises a problem. When it comes to something like the qualitative character of the sensation of red, what other property could we point to to play the role of the reference-fixer? We seem to pick out this property by itself. The distinction between the property to be reduced and the properties by which we normally pick it out, or its superficial manifestation, seems to collapse. (p. 133)

So the difference between water and the sensation of red is that we pick out water by its superficial properties which we can name (colourless, tasteless, etc.), whereas the sensation of red is itself a kind superficiality, as it were, for which there are no further superficial properties that we use to pick it out. We have hit the bedrock of superficiality, at which point intelligible explanation by appeal to more superficial properties must cease. This is a problem for physicalism about sensations because in order to show that sensations are entirely physical it needs to be demonstrated that the reference of sensory terms are fixed in such a way that, when all the facts are in, physical processes are all they could reasonably refer to. In order to demonstrate that, we need ‘reference-fixing properties’ that are not, or need not be, essential to sensory states—that are “superficial”. According to Levine, there are no such properties and hence there can be no such demonstration. Consequently, physicalism cannot be proved.

There are obvious connections here with the discussion in Chapter Three, and the solution to Levine’s explanatory gap is, I think, essentially provided there. [65] According to Levine, we pick out water by its superficial properties, which are to him the way water looks and feels. What is superficial about those properties is of course that they hold merely contingently of water. They are properties that really hold of water but it is conceivable, and possible, that water might have existed and not had those properties. But Levine’s implied description of transparency as one of the “superficial manifestations” of water betrays an inclination to see, as Kripke did, the properties by which we normally pick out manifest kinds as the “mere appearance” of their instances. But the difference between the essential properties of water and its contingent properties is not that the contingent properties merely seem to apply. Water really is transparent, and tasteless, and found in the rivers. Unlike the property of being composed of H₂O molecules, it could cease to be transparent, or tasteless, or found in rivers, without ceasing to be water. But, as things are, water happens actually to be all of those things. You might say that the tastelessness of water is modally superficial, but not superficial in the world as it actually is, not in the sense that a stick lying half in water is superficially bent, or that Stalin was superficially a nice person.

Indeed, the reduction of water to H₂O has very little to say about the look and feel of water. It would be very odd if, for example, scientists studying water discovered why water feels the way it does. In some sense biologists ought to be able to tell us why tomatoes look red, but only in the sense that they ought to be able to tell us why tomatoes are red. Those who study the world tell us how the world is, not how it looks. Of course, those with colour vision know how red things look, and can imagine how a tomato looks if they are told it is red. But to know what makes a tomato red is not to know what makes it look red. To explain why any red thing should look red would after all be to bridge the explanatory gap itself, which presumably Levine (and Kripke) does not consider necessary to explain why tomatoes have the “superficial” properties they have.

The failure of “reductive explanation” in the case of sensations lies, for Levine, in the fact that whereas water has superficial properties by which we identify it, sensations do not. Now if “superficial” is taken to mean “contingent”, then in an important sense Levine is plain wrong. Pain comes with an avoidance reaction, sensations of red tend to occur in the presence of red things; sour tastes tend to accompany eye twitches. All of these are contingent [66] but actual properties of our sensations, and which are evidence for the presence of those sensations. Moreover, as I argued in the last chapter they are in a real sense they are the ‘reference-fixing’ properties that Levine is after. They are part of the character of the sensory terms and as such do in fact fix the reference of those terms for any given individual.

5.1.2 Picking out sensations “directly”—reference and recognition

Levine would agree that sensations have the various properties I have just mentioned, but maintains that we don’t pick out our sensations using them. We pick out a sensation, as he says, “by itself” [67] . While it is the job of “reductive” explanations to tell us why water has the qualities by which we pick it out, and which are contingent, there are no corresponding properties of sensations that need explaining. The contingent properties of sensations are there, but we don’t use them to pick out the sensations; instead, according to Levine, we pick sensations out directly.

As a point about the sensations themselves, there is a sense in which Levine is right. We typically know when we are in pain through an act of recognition. But the explanatory is not a direct consequence of this. It involves the additional idea that those acts of recognition are our only way of picking out sensations. It therefore involves the idea that sensory concepts are purely recognitional concepts. We have seen earlier the sense in which this kind of view is mistaken. It rests on the idea that we always knew which sensation to call “pain”, and which to call “green”. It assumes a kind of magical link between our words for sensations and the sensations themselves. Part of the problem seems to lie in a kind of equivocation on the phrase “picking out”. The equivocation is between picking out an object or substance based on its general properties, and picking it out as the referent of some word or other. Let me bring the point out in this way: if I ask you to pick out a sample of water, I am implicitly asking you to pick out a sample of the stuff that is in English called water. That is because, of course, I am using English to ask you. Now you could have picked out the sample of water entirely independently of its English name. To “pick it out” in this sense, though, is merely to successfully recognise it from some past instance by some of its general properties.

There are two points to note about this. The first is that insofar as I pick out sensations as the referent of some word or other, I am not picking them out directly in the sense that Levine intends. In chapters three and four, we saw that words like “pain” refer using just those contingent properties that we mentioned above. It is the contingent, publicly accessible properties of sensations that we use to tell when particular sensations apply to other people, and, in the first instance, even to ourselves. Without those properties, it seems impossible to tell a story about how we learn to use words for sensations.

The second point is related to the explanatory gap in particular: it is unclear what exactly the distinction is supposed to be between picking out a state directly (“by itself”) and picking it out by its properties. He seems to deny that the latter is possible except in the vacuous sense that sensations are themselves properties. This would be analogous to the view that we pick out gold not by its yellowness but simply by its being gold. It would follow from this view that a scientist studying gold has no answer to a challenge as to whether gold is really what he is studying. Such a scientist could prove that there is a substance with atomic number 79, but not that gold is that substance. Moreover, since being gold is an essential property of gold, if there were such a thing as the direct recognition of that property then there would be no possibility of a scientific description that rendered it redundant; ‘direct’ recognition constitutes the recognition of an essential property. The explanatory gap arises because there is no way of demonstrating in such a case that there are other, physical properties that are, as it were, equally essential. Levine argues, then that when we recognise our phenomenal states we recognise them by their essential properties—as opposed, for example to recognising water by its contingent properties, such transparency, flowing in rivers, and so on. [68] It is worth pointing out that this idea, though mistaken in my opinion, is not absurd. If we can accept that natural kinds have properties that make them a natural kind, in contrast with the properties that they could cease to have and still be the same natural kind, then it is not such a large leap to say that there are some natural kinds that we actually recognise by the properties that make them that natural kind, that is to say, their essential properties. If we were in a world that had no substructure, and substances such as gold were infinitely divisible, then the properties by which we ordinarily pick out gold—its colour, etc.—could in fact be what would make it a natural kind.

But how are we to tell when we’ve picked out the essential properties of a state? The first step, one would think, is to gather together samples of that state and see what they have in common, and what as a group they fail to have in common with other kinds of states. But that needs some qualification, since it will to some extent already be clear what they have in common; they have in common whatever are the properties we used to pick them out as samples of a single natural kind. When we look more closely at those states, we look to see whether or not they have a common “underlying” nature. As in the case of jade, they may not. It is a question that at any rate requires empirical research. We may well pick out the essential properties of a thing on the first go, as it were, but we can’t know that we have until the proper (empirical) research is done.

5.1.3 Why essential properties are discoverable only empirically

Levine’s argument for the explanatory gap is not mistaken because it involves the idea that we pick out sensations by their essential properties; it is mistaken because he assumes, erroneously, that we know the essential properties of our sensations a priori. No research could convince him that a pain could exist without having the property by which he recognises it. The mistake is not the idea that we do pick out sensations by their essential properties, it is in the justification of that idea, which is conceptual. For what gives him confidence that what he recognises about sensations is their essential properties? Why not admit the possibility that the kind of state he is thinking of when he thinks of a sensation might not have had the property he recognises it as having, in the same way that gold might not have been yellow and water might not have been colourless?

The refutation of this version of the explanatory gap, then, seems only to require the possibility of error with respect to picking out sensations. If it possible for me to mistakenly believe that I am in a state that feels a certain way, then it is hard to argue that I am picking out that state directly or by virtue of its essential properties. So the idea that we pick out phenomenal states by their essential properties seems to rely on an infallibility thesis. It requires the impossibility that we could pick out different states of consciousness using the one “feel”, and that seems to be empirically, not to mention conceivably, false. Armstrong (1963), Smart (1962) and Dennett (1978) give what I think are convincing arguments against it which I will not repeat here. Of course it is true (as it is for water) that we normally do pick out sensations by the way they feel, but the coherence of the notion that we could occasionally be mistaken despite using that criterion implies that there might be something other than ‘feel’ that makes a sensation what it is. It cannot be, then, that to pick out a sensation by its feel is to pick it out ‘directly’. It might be that sensory quality is what makes a sensations the kind of state it is, in the same way that it might have been yellow metallic-ness that makes gold what it is. But it is an empirical question. If there is something other than sensory quality that makes a sensation the kind of state it is, then that “something other” should make intelligible why a certain brain state is a certain sensation. There is no good conceptual reason to assert the falsity of the antecedent of this conditional. There is therefore no a priori reason to think that the explanatory gap cannot be crossed.

5.2 The Explanatory Gap: Empirical Version

5.2.1 The explanatory gap as a challenge

If the explanatory gap is not a conceptual truth, the possibility remains that it is an empirical one. In the above context this would be equivalent to the idea that by their sensory qualities we happen to have picked out sensations by their essential properties. Sensory qualities as such would in that case be an irreducible part of the scientific description of sensations, and any purely neurophysiological description would be incomplete.

This version of the explanatory gap is more of a challenge than an argument. Those who argue for it tend to maintain that they cannot imagine how a neurological/ computational/ functional description could ever be a full description of a sensory state. Graham and Horgan’s (2000) recent defence of the knowledge argument is along these lines. They insist that Mary the neuroscientist [69] would be “surprised and delighted” at her first red experience no matter what she knows about how the brain works and that therefore “feelings” seem to be an ineliminable part of the scientific description of sensory states. That Mary would be surprised is not seriously argued for by Graham and Horgan, which suggests that they simply cannot imagine otherwise. Many responses to the Knowledge Argument share this assumpt-ion. [70]

It seems to me that the burden of proof is on those who hold that Mary would in fact be surprised by her experience, since it is a rather extraordinary claim about the reactions of a possible future scientist. [71] With that in mind, I would in any case like to argue that a theory of representation could provide empirical science with the tools to describe sensory states in such a way that there is nothing for Mary to be surprised about. In the remainder of this chapter I will advance the view that the way sensations feel are intentional objects – that they are the way our perceptions are perceptually represented to be. They are the result of an inner sense.

5.2.2 The representationist response

Gilbert Harman (1990), Michael Tye (esp. 1991 and 1995) and William Lycan (esp. 1987 and 1996) have been arguing for almost two decades now that sensory qualities are intentional objects but that, contrary to what I have just proposed, they are the way external objects are represented in perception. Harman and Tye are first-order representationists. To see red, on their view, involves being in a perceptual state that represents part of the world as being red. The sensory quality associated with redness is a component of one’s (perceptual) awareness of the redness of objects ‘out there’ in the world. Because Lycan is a second-order representationist, he agrees with at least this much, but Harman and Tye go further. Being aware of the feeling of redness is nothing more, for them, than being perceptually aware of objects as red. There is no mental quality of which one is aware, only qualities of the objects of ordinary perception. I will flesh out these claims before going on to explain the respects in which Lycan disagrees. My own proposal is most easily explicable in contrast with Lycan’s.

Harman’s view is a consequence of his theory of perception, according to which the way the world appears through perception is a matter of its being represented that way. Appearances, in this view, are intentional objects and qualities, and as such are not proper existences in themselves. It is an attractive view since it places the red that we are aware of in visual perception firmly in the objects we perceive to be red rather than in our perception of the object, which is clearly not itself red. [72] Consequently in hallucinations of red nothing actually red need be involved, just as there are no actual unicorns involved in pictures of unicorns (to borrow an example of Harman 1990).

5.2.3 Anti-representationist worries

To this it has been objected, notably by Block (1995b) that we are surely aware of some specifically mental property when we perceive objects. Perceptual experiences have, according to Block, qualities that we are aware of which go beyond their representational properties, and they are the sensory qualities. To see red, he argues, is not just to be aware that there is a red thing in front of one’s eyes – it is also to be aware of ‘what it’s like’ to see the red thing. Block’s use of the inverted spectrum and ‘inverted Earth’ thought experiments are designed to show that representational content could conceivably vary independently of ‘phenomenal feel’.

For his part Harman points to the apparently diaphanous quality of perceptual experience. To concentrate on one’s experience of blueness is inevitably to concentrate on the blueness of the object that one is perceiving. The perception itself, so this argument goes, is transparent – we tend to see through it to the outside world. Tye (1991) puts it thus:

Introspection tells us that the visual experience that represents blue differs from the visual experience that represents red. This “felt” difference is, I claim, solely a matter of content. Since the colours represented by the two experiences are different, the experiences themselves are introspectively distinguishable. The reason, then, that the visual experience of blue “feels” as it does is that it could not “feel” any other way. The “felt” aspect simply cannot be divorced from the representational aspect. (p. 32)

One strategy “anti-representationists” have available to them is to point out that seeing that something is a smooth surface feels different to feeling (that is, with one’s fingertips) that it is smooth. [73] Harman himself is inclined to bite the bullet and insist that the representational content of seeing and feeling flatness are different, and that this difference accounts for the difference in felt quality. But Tye and Lycan respond in a different way, which I would like to pursue.

5.2.4 Reply: the perception of blue as the perception of ‘visual-blue’

In his discussion of the possibility of what he termed “super blindsight”—a hypothetical kind of visual perception that leads to beliefs about external objects but lacks sensory qualities—Tye modifies the idea that perceptual experience is diaphanous in the following way:

When one introspects [the visual experience of blue], one is aware not only of the real colour blue upon which it is directed but also of the fact that it is a visual experience. That is why in repeating what one is introspecting, one will say that one has a visual experience of blue. (p. 125)

Now being a visual experience is, on the face of it, a property of one’s perception rather than a property of the objects of one’s perception. When we perceive then, if Tye is right, we must have access to some information about what kind of perception it is. [74] On the face of it, that access has at least some claim to be access to an intrinsic quality of the experience of the kind that Block is arguing for. But this is not the conclusion that Tye draws. In Tye (1994) and elsewhere he argues that the fact that a visual experience of blue is a visual experience, is actually part of the representational content of the perception. The colour blue, he argues, is in the case of visual perception represented to be a visual property. It is therefore part of the visual perception of blue that it is visual. He argues moreover that because the (visual) blue appearance of objects is an intentional object, the very same visual quality that it refers to could cause an auditory (say) perception without that auditory perception having the same content, opaquely conceived, as the visual perception [75] . In a transparent sense the visual and auditory experience of blue have the same content –blue– but in an opaque sense they may be, even qua representations, functionally different and that should account, so Tye argues, for our intuition that an object’s looking blue would be different to its sounding blue. No extra-representational aspect is needed, on this account. In that case, the qualities in the world that we perceive are not in fact perceived as ‘bare’ qualities, but rather as a mode of presentation of those qualities – visual blue, auditory blue, somatic blue, etc. Lycan takes this view explicitly in Consciousness and Experience (1996):

[I]n any case of sensing, there is (1) at least one…qualitative intentional object, such as the alleged objective redness of [a] tomato. There is also (2) the mode of presentation under which the represented property is represented (in [this] case the visual-redness mode, as opposed, say, to some scientific description of the colour property, if the property is indeed a real property of objects and has a scientific description). (p. 102)

5.2.5 Tye vs. Lycan on the existence of qualia

For both Lycan and Tye the notion of “visual-redness” is sufficient to account for the felt qualities of experience that Block and others are concerned with. [76] But this is where similarity between the views of Tye and Lycan come to an end, because Lycan but not Tye believe that there are qualia, defined as the apparent qualitative properties that our perceptual experiences seem to us to have. [77] Tye denies that there are qualia in this sense because for him our awareness of sensory qualities is nothing more than our awareness of what our sensations represent. It isn’t, for him, our experience that has the felt aspect but rather the object of our experience. So, for him, there are no qualia.

Tye’s view is, I think, unsatisfactory because it doesn’t explain how we could be aware both that our visual and auditory perceptions (from the above example) are perceptions of the same property and that the perceptions themselves are qualitatively different. One cannot simultaneously be aware that George Orwell is Eric Blair and that Orwell is a better writer than Blair. Tye’s only real answer to this objection is to deny the qualitative difference, which amounts to Harman’s biting of the bullet. Now it may be that, in the end, biting the bullet is all a representationist can do, but here I would like to pursue other options.

Lycan, in contrast to Tye, is happy to admit qualia because for him consciousness consists in being aware of another mental state. There is, he argues, a system of representation within the brain such that perceptual states can themselves be the object of an internally directed perception. Moreover, he argues that that is how we can be aware of—conscious of—the difference in the mode of representation of different sense modalities of the same external property; the difference, that is, between visual blue and auditory blue.

5.3 Representationism, Higher-Order Perception, and Sensory Qualities

5.3.1 The inner sense theory

Lycan’s “inner sense” or higher-order perception theory is a theory of consciousness, held also by Armstrong (e.g., 1984 and 1993) and arguably traceable back to Locke (1690/1898). [78] It explains, at least for Lycan, how we become conscious of the felt qualities of our experience. But the qualities themselves are aspects of the representational nature of our perception – the mode through which qualities are perceptually represented.

Lycan’s broader project is to establish the “hegemony of representation”, the view that “once representation…is (eventually) understood, then not only consciousness…but subjectivity, qualia, ‘what it’s like,’ and every other aspect of the mental will be explicable in terms of representation together with the underlying functionally organised neurophysiology.” (1996, p. 11) Though I am sympathetic with this project, I think Lycan’s specific approach to sensory qualities will not get us there.

For both Lycan and Tye all perceptions have sensory qualities, in the sense that any perceptual representation is a representation in some mode or other (visual, auditory, etc.). The mode, being peculiar to its sensory modality, explains for them how the same ‘external’ quality perceived by different sensory modalities could feel different. Whereas Tye has a difficult time explaining how we become aware of that feel as an aspect of the perception (hence his denial of qualia), Lycan has the inner sense, a higher-order perceptual process that has as its object our perceptions of the external world. But the basic idea, that all perceptions qua perceptions have sensory qualities, is something that I think a representationist account of perception is better off without. I will now try to motivate this and in the next section propose an alternative.

5.3.2 …and its problems

The idea that in seeing red we are representing not red as such but “visual-red” is likely to cause problems for a dispositional account of colour, of which Lycan at least is a follower (Tye opts for a non-relational theory). For such a theory red is, roughly speaking, the surface properties of those objects that cause a perception of redness in normal perceivers under normal conditions. The problem is that if our perception is never of red simpliciter but only visual-red, then the dispositions attached to red simpliciter are either indeterminate (red is the property of those objects that cause which kind of perception?) or circular (it is the property that causes the visual or auditory or somatic, etc., perception of redness [79] ). Dispositional accounts of red are often thought to be in danger of circularity because the perception they cause is standardly said to be simply of redness, but “perception of redness” can usually be cashed out as “the perception that one has when one looks at a tomato”. But that is just Lycan’s characterisation of visual-red, and therefore is not an option for him. [80] There is no way to escape the circularity that threatens a dispositional account of red without ‘locking in’ a specific sense modality, but Lycan’s distinction between red-as-it-really-is and visual-red does not allow it.

This is not a knock-down argument, since Lycan could simply follow Tye into a non-dispositional account of colour, but I do want to cite it as a consideration. Of more importance for Lycan, I think, is an argument brought against him by David Rosenthal, which Lycan himself reports in Consciousness and Experience (pp. 27-9). [81] The objection is this: If all perceptions have sensory qualities, and awareness of first-order perception is itself perceptual, then that awareness ought to have its own sensory quality. Moreover for Lycan there is no bar on third-order awareness (or, at any rate, awareness of second-order awareness), so his theory predicts that we can even be aware of the sensory qualities of the inner sense. This flies in the face of common sense and experience. Introspection appears to reveal that our perception of redness has a sensory quality, but it does not reveal that our awareness of those perceptions itself has a separate sensory quality.

Lycan replies:

The inner sense theorist does not contend…that internal monitoring is like external perception in every single respect. And in particular, we should not expect internal monitoring to share the property of involving some presented sensory quality at its own level of operation. (p. 28)

But he is vague about exactly why we should not expect a sensory quality at the inner sense level, since on the face of it every representation has a mode of representation and the mode is what for Lycan produces the sensory quality. The only explicit reason that Lycan gives for its failing to obtain in the case of second-order perception is that “first-order states themselves do not have ecologically significant features of the sort physical objects do”, from which it is supposed to follow that there is no need for those features to be represented via any particular mode.

This reply is a little too ad hoc. There is no real reason independent of the wish to avoid this objection to think that external perception and inner sense differ in the way they represent their objects. Much less is there any good reason for thinking that they differ in whether or not they produce sensory qualities. As for ecological significance, Lycan is on this point backing himself into a corner, for if first-order perceptual states lack ecological significance (which I suppose means that they lack features it would advantage us to detect), it is a real question why conscious awareness of perceptual states has come to play such a large role in our mental lives.

There is some evidence (for example, on page 29 of Consciousness and Experience) that Lycan is prepared to bite the bullet and admit that in some sense our inner sense has its own inner sensation, as it were. But it is a watered-down sense that is not made clear and which we are clearly better off without it if we want to be as faithful as we can to experience and common sense.

There is an alternative to Lycan’s strategy that I think can solve these problems without either biting the bullet or giving up the hegemony of representation. I think the right strategy is to pursue a dispositional account of sensory qualities, along the same lines of the dispositional account of colour and other so-called “secondary qualities”. Such an account, so I will argue, can be largely faithful to Lycan’s general theory of inner sense, and has the advantage that the problem of higher-order sensory qualities is obviated, since it comes with no disposition to be perceived.

5.4 A Dispositional Account

5.4.1 A diagnosis: four problems for Lycan

When Lycan argues that sensory qualities are given by the mode of perceptual representation, he is thinking of that mode as an intrinsic property of the perception (albeit an intrinsic representational property). It is at any rate entirely independent, logically and ontologically, of our awareness of it. Indeed, Lycan has announced that “I am less firmly attached to the ‘inner sense’ theory of awareness than I am to the representational theory of qualia” (1998, p. 483). But as long as these aspects of sensations are separated so cleanly from one another certain problems are bound to persist, including in particular the following:

1. Sensory qualities seem logically tied to our awareness of them. Kripke and others [82] have made much of the apparent fact that pain and our awareness of how pain feels are either one and the same thing or very closely related.

2. It is commonly thought that only first-order perceptions have sensory qualities, but Lycan’s view is hard-pressed to give an explanation for this, or for that matter a clear denial of it.

3. The identification of sensory quality with first-order mode of presentation causes problems either for representationism (if “mode” is construed as “vehicle” of representation – the vehicle is not part of its intentional object, as Lycan insists qualia are) or a dispositional account of secondary qualities, as noted above.

4. It is still not clear how the mode of presentation is responsible for—makes intelligible—perceptual representations feeling the ways they do, or why visual-redness explains any more than just plain redness why the brain’s representation of it should feel any particular way.

This last problem is of course the explanatory gap, which Lycan admits is there though he claims, as do many philosophers, that it is benign. [83] All of these problems can be solved, I claim, under the supposition that sensory qualities are constituted by the disposition of certain first-order perceptions to cause certain second-order perceptions. In the same way, then, that colour is related to certain perceptions, sensory qualities are also related, under this proposal, to certain (higher-order) perceptions. This solves the first problem. Since the higher-order perceptions are not themselves perceived there are no sensory qualities associated with them under this view. This solves the second problem. The third problem is obviated by this approach because sensory qualities are not part of the intentional object of the first-order perception. The version of the explanatory gap expressed by the fourth problem is solved because under this approach sensory qualities, conceived as how sensations feel, nevertheless remain intentional objects. In what follows I will flesh out these claims.

5.4.2 A dispositional solution

There is a sense in which perceptions create appearances. Moreover, some physical qualities can be defined in relation to the way they appear. Dispositional accounts of colour are an example of this. For such accounts, objects are red just in case their surface properties are such that they present a red appearance under normal conditions to a normal perceiver. The “normal conditions” and “normal perceiver” caveats are integral to the definition because there is a gap between an object’s being red and its being perceived to be red. The reason it smacks of absurdity to say that outside the visible wavelengths of light there are colours we are unable to see is that those wavelengths are not associated with any particular perceptual state. One might talk of possible colours, as one might talk of possible dispositions, but not of actual colours that, as it happens, are not connected to a disposition to be perceived. [84]

It some clear sense, it is possible to say that an object may have a red appearance without actually being perceived to be red. But to say that is to assert that it is such that it would appear red if it were perceived. There is thus a logical, if not ontological, connection between being red and being perceived. We can therefore say that red appearances are intentional objects without being committed to subjectivism about colour [85] because, to a dispositionalist, red is not the same as the appearance of red — it is the property (or properties) of things that are responsible for red appearances understood as intentional objects. If this account of colour is correct, as I believe it is, it can be extended to cover sensory qualities and the way they feel.

Kripke famously insisted that pain and the way pain feels cannot be separated. There are at least two version of this thesis in Naming and Necessity. At certain points Kripke asserts that pain and its feeling are one and the same thing—call this the ontological thesis. At other points he argues that it is impossible to pick pain out other than via its feeling—call this the semantic thesis. That the two theses are distinct is clear in the case of red and red appearances: red is not the same thing as a red appearance, but we understand the one partly in terms of the other. [86] With respect to red and red appearances, the semantic thesis holds but the ontological thesis does not.

Kripke’s argument against physicalism begins with a statement of the ontological thesis:

In the appropriate sentient being is it…possible that a stimulation of C-fibres should have existed without being felt as pain? If this is possible, then the stimulation of C-fibres can itself exist without pain, since for it to exist without being felt as pain is for it to exist without there being any pain. (p. 151)

A little later he writes that “in the case of mental phenomena there is no ‘appearance’ beyond the mental phenomena itself.” (p. 154) But his main reason for asserting this thesis is that we pick out sensations by the way they feel and that, for example, “if any phenomenon is picked out in exactly the same way we pick out pain, then that phenomenon is pain.” (p. 155) And this is precisely the semantic thesis which as we have just noted fails to entail the ontological thesis. All that is required for the semantic thesis to hold is that a state be pain just in case it is the kind of state that normally feels painful. The ‘normally’ caveat allows for the possibility that pain and the feeling of pain might not always co-occur—as for example, when one has a continuous mild headache but does not continuously feel it. At the same time to understand “pain” one must understand that pain is nothing more than a state that, as we might say, typically presents a painful appearance—that typically feels painful. In this way we arrive at a dispositional account of sensory qualities along the same lines as the dispositional account of colour and other “secondary” qualities. It is an account that preserves the logical connection between sensations and the way they feel without going so far as to say that they are identical.

5.4.3 An inner sense without an iteration of sensory qualities

The dispositional account of sensory qualities is difficult to make sense of without some kind of second-order framework. In order to carry beyond analogy the idea that the way sensations feel is the way they appear, a case must be made that they appear to someone. The inner sense theory supplies this, for in this theory perceptual states can literally be perceived. And if they are perceived then ipso facto they must have appearances. Moreover those appearances are in an important sense the direct result of the inner sense, in the same way that colour appearances (as distinct from colours) are the result of colour perception. The inner sense does not have its own sensory quality because it does not have an appearance. It does not have an appearance because its states are never themselves perceived. [87] So for any sensation there is only one sensory quality associated with it. There are two kinds of appearance involved, but one of them is of objects in the external world (their colour, shape, etc.) and the other is of sensations (their feel).

5.4.4 Where to locate feels as intentional objects

A dispositional inner sense account of sensory qualities is consistent with the view that “feels” are intentional objects. At the same time it does not deny the intuitive view that that sensory qualities are qualities of perception rather than qualities of the (external) objects of primary perception. [88] According to Lycan’s theory, sensory qualities are an aspect of the representational content of first-order perception. They are the way particular processes represent qualities in the external world. There is a sense in which the current proposal agrees with Lycan, since for both views what the inner sense perceives is the way in which we perceive the world (whether, for example, we are seeing or hearing that a car is getting closer). But whereas in Lycan’s view that “way” is itself a kind of intentional object (it is the visual part of visual-redness, which is what, for Lycan, is most directly represented in the visual perception of redness), for the current proposal it is the physical/functional process that underlies primary perception. If at any rate we assume that visual perceptions are physical processes, then I claim that those physical processes are what is being perceived when we are aware of our perceptions. For Lycan, on the other hand, since sensory qualities are part of the intentional objects of primary perception, it is those intentional objects that we perceive when we are conscious of our sensations.

Since intentional objects are in an important sense not real existences (the object of one’s hallucination is not really an object), it is a highly unintuitive view that has them as objects of our inner sense—as things that we perceive. But such is Lycan’s view, and is a further reason why the dispositional account offered here is to be preferred.

Moreover, since the object of an inner sensing is, according to me, a perceptual state, there is no need for a distinction between visual-red and red. On seeing an object as red, we are perceiving its redness simpliciter. There is nothing about the intentional object of a visual perception of redness to distinguish it from the intentional object of an auditory perception of redness, on this account, provided only that it is the same external qualities that are responsible for both the visual and the auditory perception. In this respect, at least, this account agrees with Harman’s hard-nosed first-order representationism. On the other hand, our awareness of the way we identify redness allows us to define redness as that quality of objects that causes a certain sensation in us when we look at tomatoes, blood, and so on. Under Lycan’s view what perception identifies is not redness simpliciter but “visual-redness”, which as we noted earlier creates problems for a dispositional account of red simpliciter. The dispositional inner sense view of sensory qualities proposed here does away with the need for any such thing.

5.4.5 The explanatory gap

In Lycan’s view, our awareness of the sensation of redness is our awareness of visual-redness as visual-redness. On the face of it, there remains the question, why does visual redness feel like this rather than, say, auditory-redness? The empirical version of the explanatory gap, burden-of-proof questions aside, still seems to have some force. It is still hard to imagine what might explain the difference between visual-red and auditory-red.

According to Lycan, there is a kind of explanatory gap, but it is benign. It is the result, for him, of the perception of visual-redness (and other sensory qualities) being a kind of primitive concept, “a semantically primitive Mentalese lexeme”. [89] Because it is semantically primitive there is no way for its instantiation to be implied or entailed by any other kind of description, or to be made intelligible in terms of another description. So there is a gap, but it is not a danger to physicalism. Lycan’s argument is similar in many respects to Tye’s (1999) argument that the explanatory gap “derives largely from a failure to recognise the special features of phenomenal concepts.” (p. 707)

Apart from the reasons I gave in chapter three to be sceptical of the notion that sensory concepts are “primitive” [90] , I believe we can do better. In particular, I believe a dispositional inner sense account of sensory qualities can do better.

Lycan argues, then, that perceptions create a kind of explanatory gap because they involves semantic primitives. I think he right, but for the wrong reasons. I think a better reason is that perceptions create appearances, and appearances are (at least arguably) disposition-related and therefore irreducible to the intrinsic properties of the objects that have them.

According the current proposal, in being related essentially to a disposition, sensory qualities are analogous to solubility, for which an explanatory gap exists in the following way. A bare description of the molecular structure of salt is not an explanation for why salt is soluble. What needs to be added is what would happen to that structure if it were surrounded by H₂O molecules. And that could not be explained without describing, at least partially, the structure of those H₂O molecules. The “gap” in this case is not of course a real explanatory gap in the sense that there is some aspect of the solubility of salt that cannot be explained. But it a gap in the sense that no amount of information about salt alone could explain why salt is soluble. Anyone hoping to find an explanation for the solubility of salt entirely within the chemical nature of NaCl molecules would find themselves up against a real gap. Furthermore it would be a gap, I claim, of precisely the kind operating in the case of sensory qualities.

To see this, take firstly the case of the (primary) perception of an object with a red appearance. Now no amount of information about the surface properties of the rose could explain the fact that it has a red appearance without it being taken as given that those properties cause perceptions of red in us and that the light is normal. That is because whatever properties the rose has, a change in the ambient light could change its appearance (which is not of course to say that its colour would thereby change, merely its colour appearance). The rose considered in isolation is not wholly responsible for its having a certain colour appearance, so even after a complete physical description of it is given it remains an open question what colour it will appear to be. To close the question it needs to be added that rose and the conditions are such that it will be perceived as red.

5.4.6 “Feels” as appearances

We can now tell a similar story about perceptions themselves. If we are interested in explaining why our perceptions feel the way they do, then our explanandum is not their nature but rather their appearance. But the way it appears is not something that a perceptual process considered in isolation could be wholly responsible for. The “could” here, as in the case of colour appearance, is the logical could. Since an appearance is a relational property, [91] the fact that it holds of an object or process must be made true partly by something else. Hence, as long as our attention is focussed exclusively on the perceptual process itself we could not possibly explain why it feels the way it does. In order to explain that we also need to take into account the way in which we are perceiving that process. [92]

Recall Levine’s assessment, quoted at the beginning of the chapter, of what it would take to fill the explanatory gap: it would take a story in terms of functional or physical properties of the brain that made intelligible the existence of phenomenal properties, and that furthermore:

Whatever properties the firing of C-fibres (or being in [functional] state F) had that explained the feel of pain would determine the properties a kind of physical (or functional) state had to have in order to count as feeling like our pain.

This requirement may simply be mistaken. The story of how pain feels, like the story of how roses look, may not be explicable in terms of the physical or functional properties of the state alone, not because there is anything unusual about pain states, but rather because the nature of the perceiver must also be taken into account. The same rose may appear different to different perceivers; analogously, the same pain state may (in the sense of logically may) feel different to different observers. [93]

If so, then the way experiences of red feel is not explicable solely in terms of the nature of red perceptions or solely in terms the nature of our awareness of those perceptions. That is why the question “What is it about pain states that makes them feel the way they do?” does not have an answer, and also why we cannot imagine what an answer would look like. The gap between brain states and why they should feel like anything is as real as the gap between roses and why they should look like anything, but does not automatically create any great mystery about brain states, any more than it creates a mystery about roses.

5.5 Further Issues

The account of sensory qualities I am putting forward here faces some prima facie damaging objections, the most serious of which it is important that I address here.

5.5.1 Empirical issues

Güven Güzeldere (1997, p. 794), sceptical of the idea of higher-order perception, writes that “the claim that there are self-scanners in the brain that are responsible for introspective consciousness…has no solid physiological or anatomical basis in the neuroscientific literature.” Rosenthal (1990), quoted in Lormand (1996), and Carruthers (2000) express similar doubts about the existence of a perceptual mechanism.

As it happens, Wolf Singer (1998 and 2001), a neurophysiologist, has recently been speculating that one of the roles of the now-familiar 40-Hz synchronous firing of neurons in parts of the cortex may be to represent perceptual processes. He goes so far as to propose that “the aspect of consciousness that we address as phenomenal awareness results from an iteration of the same cognitive operations that support sensory processing.” (p. 127)

Crick and Koch’s (1990) original paper on the firing of certain neurons in unison proposed that it is involved in perceptual binding, which they in turn proposed may be necessary for consciousness. More recently, Crick and Koch (1998) have downplayed the role of synchronous firing, but Singer and his lab have taken it up as a vehicle of representation. [94] Whether Singer is right that synchronous firings are, in particular, the vehicle of an iterated, perceptual representation is another matter. But at any rate Singer’s papers show that there is some reason to at least keep the empirical question open for now. That is the most that my proposal here needs.

5.5.2 Perceiving a thing vs. perceiving its nature

The red appearance of a rose in normal light is explained partly by the fact that in normal light we perceive it to be red. We represent it, somehow, as being red. But in doing so, what do we know about the rose? Locke argued that a key feature of our ideas of secondary qualities is that they lack any resemblance to a discoverable feature of objects – “There is nothing like our ideas [of secondary qualities] existing in the bodies themselves” (II;VIII;15). What he is alluding to is the fact that our sensations of colour don’t seem to tell us anything about the natures of the objects we see to be coloured. Now of course this itself can not serve as an argument to the conclusion that colour perception is not a perception of a real property of objects. [95] Even Locke’s characterisation of secondary qualities allows that they convey some information about their bearer, namely, the power to induce in us certain sensations. Two red objects, by virtue of being red, have the same power with respect to inducing colour sensations. Since it is possible to perceive some similarity without perceiving the features that account for it, there is no philosophical problem in the idea that in colour perception we represent sameness and difference without representing sameness or difference in any particular respect.

Like the perception of colour, our perception of what colour sensations feel like does not seem to reveal much about the nature of sensation states. It does tell us at least this much, that the sensation of colour is a visual rather than an auditory sensation. I will not speculate here on the usefulness of that kind of information but I do want to point out that there is no reason to think that phenomenal qualities are not inner sensings simply because they do not reveal anything of the nature of visual processing. [96] Colour perception is a good example here – to perceive that two objects are the same colour is not really to know what properties they share, if any, apart from simply their colour. Likewise, to notice that the perception of red and the perception of orange feel similar is not to know what properties they have in common. Their feeling similar just is one’s perception that they are similar. We cannot articulate the respects in which they are similar because it isn’t part of the intentional object of the perception, and it needn’t be in order to count as one.

5.5.3 The knowledge argument

An important result of this thesis is that a good theory of representation is needed by the study of consciousness. Within the framework provided here, which is consistent with the hard-line representationism of Harman, the views of even the most stubborn anti-representationists like Block can be accommodated. To misquote a certain American president, there is nothing wrong with representationism that cannot be fixed by what is right with representationism. Our awareness of the quality of our sensory experience does not threaten the idea that perceptions are representational, since that awareness is itself representational. What it represents is to some extent unimportant, and may only be bare similarities and differences. Like all perceptual systems, they represent the world in a way that is evolutionarily useful, which does not necessarily cut nature at its joints—a quality we expect rather of our scientific representations.

In the light of this it is also useful to take a somewhat novel approach to the so-called “knowledge argument” against physicalism, which has not yet been addressed here. In Jackson’s (1982 and 1986) argument Mary has full knowledge of an ideal neuroscience but has never seen colours, to which the question is put, does Mary learn anything on having her first visual colour experience?

As a first pass across this question, it is worth asking whether if Mary were a physicist she would learn anything about red objects on first seeing something red. In one sense she would, since she would learn what red objects look like. But that is of course slightly unfair to physicists, since the way objects appear is not given purely by any characterisation of the objects themselves. As was pointed out earlier, how the objects are perceived is also an essential part of the story. Nevertheless the red appearance of post-boxes, whether or not is a question answerable within physics, is a contingent fact for which we can seek an explanation. The part that cannot be answered by Mary the physicist must presumably be answered by Mary the neuroscientist. Still the question remains a question about post-boxes rather than mental states—did Mary the neuroscientist learn anything about red things on her first red experience? Certainly her neuroscience should have told her how red things are perceived, and in that sense how red things appear. If ideal neuroscience includes a representational theory of perception then Mary ought to have known what properties she would perceive the red objects to have—that is, redness. That a certain class of objects would appear red ought not to have surprised her.

A first-order representationist such as Harman. would stop there, but I think a second pass at the question is necessary, and is warranted by the second-order representationism defended here. Mary is supposed, in the original argument, to learn something about her mental life on having her first red experience. This idea can be approached similarly to the previous question about post-boxes. Merely knowing about the process by which red objects are perceived to be red should not be sufficient for knowledge of the “what it is like” kind that the knowledge argument is really about. That kind of knowledge is not concerned with our awareness of red objects – it is concerned with our awareness of a kind of mental object that we are often aware of when we are aware of red objects. A theory of ‘external’ perception would have nothing to say about the mental object. For an account of our awareness mental objects, a theory of how we perceive them may be needed – a theory of ‘internal’ perception. If Mary had such a theory, she would at least know what properties she would perceive her perceptions of redness to have. In at least this sense she could predict what her colour experience would “feel like”, just as she could predict what post-boxes look like: what properties will be represented in their perception. The properties we are most directly aware of our perceptions as having have a fair claim to be what are known as “the way sensations feel”. A representationist can therefore admit the most of the force of the knowledge argument as originally posed without needing to admit its conclusion.

What science currently lacks, then, is an uncontroversial theory of representation – a theory using which we would be able to interpret a bat’s echolocation as representing parts of the world as having certain properties, and furthermore whether or not there is a system in the bat’s brain which represents its echolocation as itself having certain properties. If there are any representations of the latter sort, then we will know that whatever is being represented of echolocation is what echolocation is like for the bat.

Nagel’s (1974) and Jackson’s argument can then serve as a useful criterion for an adequate theory of perceptual representation. Whether the current theories of representation pass this test is not something I have evaluated here, but at any rate there is no good reason for thinking it is an impossible one. If the view defended here is on the right track, then using a theory of representations science ought to be able to describe what sensations feel like. Answering the question as to how anything could possibly represent anything else may well be just as difficult a question as the “hard problem” is thought to be, but following G.E. Moore it is enough for now to hope that I have posed the right question.

5.5.4 Sensory qualities and the problem of consciousness

It is important to emphasize that I am not proposing a dispositional account of consciousness. I happen to believe that something close to Bernard Baars’ “global workspace theory” (as, for example, in Baars 1997) is the most plausible theory of consciousness. In this respect I differ from Lycan and Armstrong. But the issues of consciousness and sensory qualities have often been, if not conflated, then at least mixed up—Chalmers’ “hard problem” of consciousness is precisely the problem of sensory qualities. [97] I think Lycan is right that they are distinct issues, though I think the postulation of an inner sense can solve the problem of sensory qualities but not problem of consciousness considered more broadly as the state that distinguishes your waking moments from your sleeping ones. The dispositional account of sensory qualities offered in this chapter is intended to answer the question, What are the sensory qualities and how are they related to the way they appear? It proposes that the question, and therefore the answer, is similar to the question, What are the secondary qualities how are they related to the way they appear? Since in both cases we can fail to be conscious of the appearance, or the feel, the question of consciousness is separate, and not dealt with here. Nevertheless, since sensory qualities are often taken to be the most intractable aspect of our mental lives, perhaps once the question of their nature is solved the problem of consciousness will be that much easier.

Conclusion

It is commonplace these days for metaphysicians to say that metaphysics is continuous with science. That is easy for metaphysicians to say —scientists do not offer opinions on the problem of universals. For philosophers of mind, the negotiation between empirical and “philosophical” research is of direct significance. If the question of “what consciousness is” is an empirical one, why is it that scientists are handballing it back to philosophers?

In the first part of this thesis I argue that empirical work on sensations does not need philosophical intervention until it becomes research on sensations qua the referent of the word “sensations”. At that point it is important that scientists be able to defend that they are studying what are known as “sensations”. In order to do that they need, as at least a working definition, the ordinary conception of what a sensation is. Whether those ordinary conceptions can be known to hold of sensations a priori, in some strict sense of that term, is irrelevant to the necessity of their use. The causal and descriptivist theories of reference do not differ in this respect, though they are taken to differ by Block, Stalnaker and Byrne.

However, it is not at all obvious what, in the case of sensations, the ordinary conception is. Very often it is taken to simply be a mental state that we recognise; moreover, particular sensations are taken to be individuated according to the particular property that we recognise. Chalmers, Block, Loar, Tye and others all hold this view. In the second part of the thesis I argue against it, mainly on the grounds that whether or not we recognise sensations when we are in them (I agree that we do), we do not recognise them as the referents of the general term “sensation”. The fact that we recognise when we are having a sensation may well be part of the ordinary conception, but not all of the respects in virtue of which we are able to recognise them—in particular, their sensory qualities—because we currently have no way of knowing whether or not other people are recognising the same sensory qualities.

Nevertheless, our recognition of particular sensations, say pain, follows a pattern. We recognise ourselves typically to be in it at the same time that we often also notice damage to have been inflicted on us. Furthermore, part of the sensation of pain is the inclination to think of it as “residing” at a certain location. When we combine these facts we arrive, I claim, at an indexical conception of pain—roughly, pain as the state that I recognise myself to typically be in when I do damage to myself. This conception leaves open the possibility that different people might typically recognise different types of states when they do themselves damage. It thereby allows, at least logically, for there to be a common conception without needing to cross the epistemological gap between your and my inner states.

Sensations are states that it feels like something to be in. The way they feel is one of the respects in virtue of which we tend to recognise them (unlike our recognition of beliefs, for example). But since, as I claim, particular sensory qualities are not part of the common conception of sensations, it follows that it is not what scientists need to identify in order to defend the claim that what they are studying are in fact sensations qua the referent of that word. What they need to identify is, in the case of pain, (1) the state that we are typically in when we are damaged in some way, (2) which we recognise, and (3) which inclines us to think that it is located in a certain place on our bodies. It is worth noting that finding (1) is a matter finding the so-called neural correlates of pain — but the other two are functional.

(3) is the representational aspect of sensations, and noted by almost every philosopher in the field. (2), however, is the aspect of sensations that Block argues goes beyond the representational—their sensory qualities. Pace Block, I claim that this, too, is an intentional object. But pace Harman and others as well, I argue that it is the way our perceptions themselves are represented. What is doing the representing is, I claim, something very similar to Lycan and Armstrong’s “inner sense”—a perceptual mechanism that has primary perception as its object. But against Lycan (and for that matter Armstrong), I claim that the inner sense is directly responsible for sensory qualities, in that sensory qualities are created partly by the fact that they are perceived. Sensory qualities are in this way like colours, whose relation to a disposition to be perceived marks them out as qualities of objects. In both cases the qualities are there independently of their being perceived, but as properties they may be profoundly unnatural, in the sense of indicating a merely tenuous resemblance between their instances. Whatever clump of properties underlie a sensory qualities is entirely a matter of which ones are disposed to result in a perception being perceived to be one way rather than another.

It may yet be, therefore, that sensations are brain processes and could not be anything else—that sensations are, as a type, biological in nature. When I report that I am experiencing a sensation of redness, I am reporting that I am in a state similar to the state I am normally in when I look at tomatoes. I recognise that state by its perceptual content and its sensory qualities, the former of which I can describe (“red”) though not, it seems, the latter. If a biological theory is capable of explaining the content “redness” in me at the times when I say I am experiencing a sensation of redness (which, granted, may be a big “if” – see, e.g., Millikan (1984 and 1989) in favour and MacDonald (1989) against), then to the extent that it is also a plausible explanation, it is thereby plausible that sensations of redness are biological processes. Any such theory must also be compatible with the true explanation, when it comes along, of the sensory qualities. If I am right, however, this condition is easily satisfied since an iteration of processes responsible for the perceptual content ‘redness’ should suffice. The question of ‘what sensations are’ seems to depend, therefore, on whatever can best explain perceptual content. If a biological type of process can do that, then sensations are brain processes.

Bibliography

Almog, Joseph, John Perry, et al. (Eds.) (1989), Themes From Kaplan. Oxford: Oxford University Press.

Armstrong, D. M. (1961), Perception and the Physical world. London: Routledge & K. Paul.

Armstrong, D. M. (1963), “Vesey on Sensations of Heat.” Australasian Journal of Philosophy 41; 359-362.

Armstrong, D.M. (1963), “Is Introspective Knowledge Incorrigible.” The Philosophical Review 72; 417-432.

Armstrong, D.M. (1968), A Materialist Theory of the Mind. London: Routledge.

Armstrong, D. M. (1973), Belief, Truth and Knowledge. London: Cambridge University Press.

Armstrong, D. M. (1980), The Nature of Mind and Other Essays. St. Lucia, Qld: University of Queensland Press.

Armstrong, D. M. and Norman Malcolm (1984), Consciousness and Causality : a Debate on the Nature of Mind. Oxford: B. Blackwell.

Armstrong, D.M. (1997b), “Smart and the Secondary Qualities.” The Philosophy of Color. A. Byrne and D. Hilbert (Eds.) Cambridge, Mass., MIT Press: 1; 33-46.

Austin, John (1964), Sense and Sensibilia. Oxford: Oxford University Press.

Baars, Bernard (1997), “In the Theatre of Consciousness.” Journal of Consciousness Studies 4; 292-309.

Block, Ned and Jerry Fodor (1972), “What Psychological States are Not.” Philosophical Review 81; 159-181.

Block, Ned (1980), “Troubles with Functionalism.” Readings in the Philosophy of Psychology, Vol 1. N. Block (Ed.) Cambridge, Harvard University Press.

Block, Ned (1990), “Inverted Earth.” Philosophical Perspectives, 4, Philosophy of Mind and Action Theory. J. Tomberlin (Ed.) Atascadero, Ridgeview Publishing Co.; 53-79.

Block, Ned (1992), “Begging the Question Against Phenomenal Consciousness.” Behavioral and Brain Sciences 15; 205-206.

Block, Ned (1995a), “On a Confusion About a Function of Consciousness.” Behavioral and Brain Sciences 18; 227-247.

Block, Ned (1995b), “Mental Paint and Mental Latex.” Philosophical Issues, 7, Perception. E. Villanueva (Ed.); 19-49.

Block, Ned, Owen Flanagan, et al. (Eds.) (1997), The Nature of Consciousness. Cambridge, Mass.: MIT Press.

Block, Ned and Robert Stalnaker (1999), “Conceptual Analysis, Dualism, and the Explanatory Gap.” The Philosophical Review 108; 1-46.

Boghossian, Paul and David Velleman (1989), “Colour as a Secondary Quality.” Mind 98; 81-103.

Braddon-Mitchell, David and Frank Jackson (1996), The Philosophy of Mind and Cognition. Oxford: Blackwells.

Byrne, Alex (1997), “Some Like it Hot: Consciousness and Higher-Order Thoughts.” Philosophical Studies 86; 103-129.

Byrne, Alex (1999), “Cosmic Hermeneutics.” Philosophical Perspectives, 13, Epistemology. J. Tomberlin (Ed.) Oxford, Blackwells; 347-385.

Byrne, Alex (2001), “Do Colours Look Like Dispositions? Reply to Langsam and Others.” The Philosophical Quarterly 51; 238-245.

Carruthers, Peter (2000), Phenomenal Consciousness: A Naturalistic Theory. Cambridge: Cambridge University Press.

Chalmers, David (1996), The Conscious Mind : in Search of a Fundamental Theory. New York: Oxford University Press.

Chalmers, David (1999), “Materialism and the Metaphysics of Modality.” Philosophy and Phenomenological Research 59; 473-496.

Chisholm, Roderick M. (1957), Perceiving: A Philosophical Study. Ithaca, N.Y.: Cornell University Press.

Churchland, Patricia S (1985), “Reduction, Qualia and the Direct Introspection of Brain States.” The Journal of Philosophy 82; 8-28.

Clark, Andy (1997), Being There: Putting Brain, Body, and World Together Again. Cambridge, Mass.: MIT Press.

Clark, Andy (2000), “A Case Where Access Implies Qualia?” Analysis 60; 30-38.

Cole, Peter (Ed. (1978), Syntax and Semantics, vol 9: Pragmatics. New York: Academic Press.

Crane, Time (Ed. (1992), The Contents of Experience. Cambridge: Cambridge University Press.

Crick, Francis and Christof Koch (1990), “Towards a Neurobiological Theory of Consciousness.” Seminars in the Neurosciences; 263-275.

Crick, Francis and Christof Koch (1998), “Consciousness and Neuroscience.” Cerebral Cortex 8; 97-107.

Danckert, James and Melvyn A. Goodale (2000), “Blindsight: A Conscious Route to Unconscious Vision.” Current Biology 10; R64-R67.

Davidson, Donald (1970), “Mental Events.” Experience and Theory. L. Foster and J. W. Swanson (Eds.) Amherst, University of Massachusetts Press; 79-101.

Davies, Martin and Lloyd Humberstone (1980), “Two Notions of Necessity.” Philosophical Studies 38; 1-30.

Dennett, Daniel C. (1978), “Why You Can't Make a Computer that Feels Pain.” Synthese 38; 415-449.

Dennett, Daniel C. (1986), Content and Consciousness. London: Routledge & Kegan Paul.

Dennett, Daniel C. (1988), “Quining Qualia.” Consciousness in Contemporary Science. A. Marcel and E. Bisiach (Eds.) Oxford, Oxford University Press; 43-77.

Dennett, Daniel C. (1991), Consciousness explained. Boston: Little Brown and Co.

Donnellan, Keith (1966), “Reference and Definite Descriptions.” The Philosophical Review 75; 281-304.

Dretske, Fred (1993), “Conscious Experience.” Mind 102; 263-283.

Dretske, Fred (1999), “The Mind's Awareness of Itself.” Philosophical Studies 95; 103-125.

Edelman, Gerald M. (1992), Bright Air, Brilliant Fire : On the Matter of the Mind. New York, N.Y.: BasicBooks.

Evans, Gareth (1982), The Varieties of Reference. Oxford: Clarendon Press.

Farah, Martha J. (1997), “Visual Perception and Visual Awareness after Brain Damage.” The Nature of Consciousness. N. Block, O. Flanagan and G. Güzeldere (Eds.) Cambridge, Mass., MIT Press; 203-236.

Farrell, B.A (1950), “Experience.” Mind 59; 170-198.

Fodor, Jerry (1998), In Critical Condition: Polemical Essays on Cognitive Science and the Philosophy of Mind. Cambridge, Mass.: MIT Press.

Foundation, Ciba (1993), Experimental and Theoretical Studies of Consciousness. Chichester, New York: J. Wiley & Sons.

Gazzaniga, Michael (1999), “What Are Brains For?” Mind and Brain Sciences in the 21st Century. R. Solso (Ed.) Cambridge, Mass., MIT Press; 157-189.

Gertler, Brie (2001), “The Explanatory Gap is Not an Illusion: Reply to Michael Tye.” Mind 110; 689-694.

Gold, Ian (1999), “Dispositions and the Central Problem of Color.” Philosophical Studies 93; 21-44.

Graham, George and Terry Horgan (2000), “Mary, Mary, Quite Contrary.” Philosophical Studies 99; 59-87.

Guttenplan, Samuel (1994), A Companion to the Philosophy of Mind. Oxford: Blackell Reference.

Güzeldere, Güven (1997), “Is Consciousness the Perception of What Passes in One's Own Mind?” The Nature of Consciousness. N. Block, O. Flanagan and G. Güzeldere (Eds.) Cambridge, Mass., MIT Press; 789-806.

Harman, Gilbert (1990), “The Intrinsic Quality of Experience.” Philosophical Perspectives, 4, Philosophy of Mind and Action Theory. J. Tomberlin (Ed.) Atascadero, Ridgeview Publishing Co.; 31-52.

Heil, John (1998b), Philosophy of Mind: A Contemporary Introduction. London: Routledge.

Henderson, D. K. (1994), “Accounting for Macro-Level Causation.” Synthese 101; 129-156.

Hill, Christopher (1988), “Introspective Awareness of Sensations.” Topoi 7; 9-22.

Hill, Christopher and Brain McLaughlin (1999), “There are Fewer Things in Reality than are Dreamt of in Chalmers' Philosophy.” Philosophy and Phenomenological Research 59; 445-454.

Hume, David (Ed. (1955), A Treatise of Human Nature. Oxford: Oxford University Press.

Jackson, Frank (1977), Perception: A Representative Theory. Cambridge: Cambridge University Press.

Jackson, Frank (1982), “Epiphenomenal Qualia.” Philosophical Quarterly 32; 127-36.

Jackson, Frank (1986), “What Mary Didn't Know.” The Journal of Philosophy 83; 291-295.

Jackson, Frank (1997), “Finding the Mind in the Natural World.” The Nature of Consciousness. N. Block, B. Flanagan and G. Güzeldere (Eds.) Cambridge, Mass., MIT Press; 483-492.

Jackson, Frank (1998a), From Metaphysics to Ethics : a Defence of Conceptual Analysis. Oxford: Clarendon Press.

Jackson, Frank (1998b), “Reference and Description Revisited.” Philosophical Perspectives, 12, Language, Mind, and Ontology. J. Tomberlin (Ed.) Oxford, Blackwells; 201-218.

Jackson, Frank (1998c), Mind, Method and Conditionals. New York: Routledge.

Jackson, Frank (2001), “Précis of 'From Metaphysics to Ethics'.” Philosophy and Phenomenological Research 62; 617-624.

Jackson, Frank (2001), “Responses to Commentaries.” Philosophy and Phenomenological Research 62; 653-665.

Jarrett, Greg (1996), “Analyzing Mental Demonstratives.” Philosophical Studies 84; 49-62.

Johnston, Mark (1997), “Manifest Kinds.” The Journal of Philosophy 94; 564-583.

Kaplan, David (1978), “Dthat.” Syntax and Semantics, vol 9: Pragmatics. P. Cole (Ed.) New York, Academic Press; 221-243.

Kaplan, David (1989a), “Demonstratives.” Themes From Kaplan. J. Almog, J. Perry and H. Wettstein (Eds.) Oxford, Oxford University Press; 481-564.

Kaplan, David (1989b), “Afterthoughts.” Themes From Kaplan. J. Almog, J. Perry and H. Wettstein (Eds.) Oxford, Oxford University Press; 565-614.

Kim, Jaegwon (1993), Supervenience and Mind : Selected Philosophical Essays. New York: Cambridge University Press.

Kirk, Robert (1994), Raw Feeling: A Philosophical Account of the Essence of Consciousness. Oxford: Clarendon Press.

Kirk, Robert (1995), “How is Consciousness Possible?” Conscious Experience. T. Metzinger (Ed.) Schöningh, Imprint Academic; 391-407.

Kripke, Saul (1980), Naming and Necessity. Oxford: Blackwells.

Kripke, Saul (1982), Wittgenstein on Rules and Private Language. Oxford: Basil Blackwell.

Langsam, Harold (1997), “The Theory of Appearing Defended.” Philosophical Studies 87; 33-59.

Levin, J. (2000), “Dispositional Theories of Color and the Claims of Common Sense.” Philosophical Studies 100; 151-174.

Levine, Joseph (1983), “Materialism and Qualia: The Explanatory Gap.” Pacific Philosophical Quarterly 64; 354-361.

Levine, Joseph (1993), “On Leaving Out What It's Like.” Consciousness. M. Davies and G. Humphreys (Eds.) Oxford, Blackwells; 121-36.

Levine, Joseph (2001), Purple Haze: The Puzzle of Consciousness. Oxford: Oxford University Press.

Lewis, David K. (1963), “An Argument for the Identity Theory.” The Journal of Philosophy 63; 17-25.

Lewis, David (1972), “Psychophysical and Theoretical Identifications.” Australasian Journal of Philosophy 50; 249-258.

Lewis, David K. (1980), “Mad Pain and Martian Pain.” Readings in the Philosophy of Psychology, vol. 1. N. Block (Ed.) Cambridge, Mass., Harvard University Press; 216-222.

Lewis, David K. (1994), “Reduction of Mind.” A Companion to the Philosophy of Mind. S. D. Guttenplan (Ed.) Oxford, Blackell Reference.

Lewis, David K. (1994), “Humean Supervenience Debugged.” Mind 103; 473 - 490.

Lewis, David K. (1995), “Should a Materialist Believe in Qualia?” Australasian Journal of Philosophy 73; 140-144.

Lewis, David K (1997), “Naming the Colours.” Australasian Journal of Philosophy 75; 325-342.

Lewis, David K. (1999), Papers in Metaphysics and Epistemology. Cambridge: Cambridge University Press.

Loar, Brian (1997), “Phenomenal States.” The Nature of Consciousness: Philosophical Debates. N. Block, O. Flanagan and G. Güzeldere (Eds.) Cambridge, Mass, MIT Press; 597-615.

Loar, Brian (1999), “David Chalmers' "The Conscious Mind".” Philosophy and Phenomenological Research 59; 464-472.

Locke, John (1690/1898), “Essay Concerning Human Understanding.”. London: Ward, Lock & Co.

Lormand, Eric (1996), “Inner Sense Until Proven Guilty.” (Unpublished Manuscript). http://www-personal.umich.edu/~lormand/phil/cons/inner_sense.htm.

Lycan, William (1974), “Kripke and the Materialists.” The Journal of Philosophy 71; 677-689.

Lycan, William (1987), Consciousness. Cambridge, Mass.: MIT Press.

Lycan, William (Ed. (1990), Mind and Cognition: A Reader. Cambridge, Mass.: Basil Blackwell.

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

Lycan, William (1998), “In Defense of the Representational Theory of Qualia (Replies to Neander, Rey, and Tye).” Philosophical Perspectives, 12, Language, Mind, and Ontology. J. Tomberlin (Ed.) Oxford, Blackwells; 479-487.

Lycan, William (1999), “Dretske on the Mind's Awareness of Itself.” Philosophical Studies 95; 125-133.

Lycan, William (2001), “A Simple Argument for a Higher-order Representation Theory of Consciousness.” Analysis 61; 3-4.

MacDonald, Graham (1989), “Biology and Representation.” Mind and Language 4; 187-200.

Malcolm, Norman (1958), “Knowledge of Other Minds.” The Journal of Philosophy 55; 969-978.

McGilvray, James (1994), “Constant Colors in the Head.” Synthese 100; 197-239.

McGinn, Colin (1982), The Character of Mind. Oxford: Oxford University Press.

McGinn, Colin (1983), The Subjective View. Oxford: Clarendon Press.

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

McGinn, Colin (1996), “Another Look at Color.” The Journal of Philosophy 93; 537-53.

Merikle, Philip and Meredyth Daneman (1998), “Psychological Investigations of Unconscious Perception.” Journal of Consciousness Studies 5; 5-18.

Merleau-Ponty, Maurice (1962), Phenomenology of Perception. London: Routledge and Keagan Paul.

Mill, John Stuart (1869), System of Logic. London: Longmans, Green, Reader & Dyer.

Millikan, Ruth (1984), Language, Thought, and Other Biological Categories. Cambridge, Mass.: MIT Press.

Millikan, Ruth (1989), “Biosemantics.” The Journal of Philosophy 86; 281-297.

Nagel, Ernest (1961), The Structure of Science. New York: Harcourt, Brace & Co.

Nagel, Thomas (1974), “What is it Like to be a Bat?” The Philosophical Review 83; 435-450.

Nagel, Thomas (2000), “The Psychophysical Nexus.” New Essays on the A Priori. P. Boghossian and C. Peacocke (Eds.) Oxford, Clarendon Press; 433-471.

Neander, Karen (1998), “The Division of Phenomenal Labor: A Problem for Representational Theories of Consciousness.” Philosophical Perspectives, 12, Language, Mind, and Ontology. J. Tomberlin (Ed.) Oxford, Blackwells; 411-433.

Nelkin, Norton (1993), “The Connection between Intentionality and Consciousness.” Consciousness: Psychological and Philosophical Essays. M. Davies (Ed.) Cambridge, Blackwell.

Nelkin, Norton (1996), Consciousness and the Origins of Thought. Cambridge: Cambridge University Press.

Nida-Rümelin, Martine (1995), “What Mary Couldn't Know: Belief about Phenomenal States.” Conscious Experience. T. Metzinger (Ed.) Shöningh, Imprint Academic; 219-242.

Oppenhein, P and H Putnam (1958), “Unity of Science as a Working Hypothesis.” Minnesota Studies in the Philosophy of Science, Vol 2. H. Feigl, M. Scriven and G. Maxwell (Eds.) Minneapolis, University of Minnesota Press.

Papineau, David (1993), Philosophical Naturalism. Oxford: B. Blackwell.

Papineau, David (1995), “The Antipathetic Fallacy and the Boundaries of Consciousness.” Conscious Experience. T. Metzinger (Ed.) Paderborn, Ferdinand-Schoningh.

Papineau, David (1998), “Mind the Gap.” Philosophical Perspectives, 12, Language, Mind, and Ontology. J. Tomberlin (Ed.) Oxford, Blackwells; 373-388.

Parsons, Katheryn Pyne (1970), “Mistaking Sensations.” The Philosophical Review 79; 201-213.

Peacocke, Christopher (1983), Sense and Content : Experience, Thought, and their Relations. Oxford: Clarendon Press.

Peacocke, Christopher (1989), “Perceptual Content.” Themes From Kaplan. J. Almog, J. Perry and H. Wettstein (Eds.) Oxford, Oxford University Press; 297-329.

Pitcher, George (1971), A Theory of Perception. Princeton: Princeton University Press.

Place, U.T. (1956), “Is Consciousness a Brain Process?” British Journal of Psychology 47; 44-50.

Place, U.T. (1988), “Thirty Years On - Is Consciousness Still a Brain Process?” Australasian Journal of Philosophy 66; 208-219.

Price, H. H. (1954), Perception. London: Methuen.

Putnam, Hilary (1962), “The Analytic and The Synthetic.” Minnesota Studies in the Philosophy of Science Vol. 3. H. Feigl and G. Maxwell (Eds.); 358-397.

Putnam, Hilary (1963), “Minds and Machines.” Dimensions of Mind. S. Hook (Ed.) New York, New York University Press; 148-179.

Putnam, Hilary (1967), “Psychological Predicates.” Art, Mind and Religion. W. H. Capitan and D. D. Merrill (Eds.) Pittsburgh, University of Pittsburgh Press.

Putnam, Hilary (1975), “Philosophy and Our Mental Life.” Mind, Language and Reality: Philosophical Papers, Volume 2. H. Putnam (Ed.) Cambridge, Cambridge University Press.

Quine, W. V. O. (1960), Word and Object. Cambridge, Mass.: MIT Press.

Quine, W. V. O. (1985), “States of Mind.” The Journal of Philosophy 82; 5-8.

Roelfsema, Pieter, Andreas Engel, et al. (1996), “The Role of Neuronal Synchronization in Response Selection: A Biologically Plausible Theory of Structured Representations in the Visual Cortex.” Journal of Cognitive Neuroscience 8; 603-625.

Rosenthal, David (1976), “Mentality and Neutrality.” The Journal of Philosophy 73; 386-415.

Rosenthal, David (1986), “Two Concepts of Consciousness.” Philosophical Studies 94; 329-59.

Rosenthal, David (1990), “A Theory of Consciousness.” ZIF Report 40. Bielefeld.

Rosenthal, David (Ed. (1991), The Nature of Mind. Oxford: OUP.

Rowlands, Mark (2001), “Consciousness and Higher-Order Thoughts.” Mind and Language 16; 290-310.

Ryle, Gilbert (1949), The Concept of Mind. London: Hutchinson.

Ryle, Gilbert (1951), “Feelings.” Philosophical Quarterly 1; 193-205.

Salmon, Nathan U. (1981), Reference and Essence. Princeton, N.J.: Princeton University Press.

Salmon, Nathan U. (1993), “Analyticity and Aprioricity.” Philosophical Perspectives 7; 125-133.

Searle, John R. (1992), The Rediscovery of the Mind. Cambridge, Mass.: MIT Press.

Sellars, Wilfred (1975), “The Adverbial Theory of Objects of Sensation.” Metaphilosophy 6; 144-60.

Shaffer, Jerome (1963), “Mental Events and the Brain.” The Journal of Philosophy 60; 160-166.

Shoemaker, Sdney (1975), “Functionalism and Qualia.” Philosophical Studies 27.

Shoemaker, Sydney (1994), “Self-Knowledge and "Inner Sense", Lectures I-III.” Philosophy and Phenomenological Research 54; 246-314.

Shoemaker, Sydney (1996a), “Colors, Subjective Relations, and Qualia.” Philosophical Issues, 7, Perception. E. Villanueva (Ed.); 55-66.

Shoemaker, Sydney (1996b), The First-Person Perspective and Other Essays. Cambridge: Cambridge University Press.

Shoemaker, Sydney (1999), “On David Chalmers' "The Conscious Mind".” Philosophy and Phenomenological Research 59; 439-444.

Singer, Wolf, Andreas Engel, et al. (1997), “Neuronal Assemblies: Necessity, Signature and Detectability.” Trends in Cognitive Sciences 1; 252-260.

Singer, Wolf (1998), “Consciousness and the Structure of Neuronal Representations.” Philosophical Transactions of the Royal Society of London: Biological Sciences 353; 1829-1840.

Singer, Wolf (1998), “Consciousness from a Neurobiological Perspective.” From Brains to Consciousness? S. Rose (Ed.) London, Allen Lane, The Penguin Press; 228-245.

Singer, Wolf (2001), “Consciousness and the Binding Problem.” Annals of the New York Academy of Sciences 929; 123-146.

Smart, J. J. C. (1959), “Sensations and Brain Processes.” The Philosophical Review 68; 141-156.

Smart, J. J. C. (1962), “Brain Processes and Incorrigibility: A Reply to Professor Baier.” Australasian Journal of Philosophy 40; 68-70.

Smart, J. J. C. (1981), “Physicalism and Emergence.” Neuroscience 6; 109-113.

Smart, J. J. C. (1987), Essays Metaphysical and Moral : Selected Philosophical Papers. Oxford: B. Blackwell.

Smart, J. J. C. (1995), “'Looks Red' and Dangerous Talk.” Philosophy 70; 545-554.

Smart, J. J. C. (2000), “The Identity Theory.” Stanford Encylcopedia of Philosophy. E. Zalta (Ed.) Stanford, Stanford University Center for the Study of Language and Information, http://plato.stanford.edu/entries/mind-identity .

Smith, Edward (1999), “Infusing Cognitive Neuroscience into Cognitive Psychology.” Mind and Brain Sciences in the 21st Century. R. Solso (Ed.) Cambridge, Mass., MIT Press; 71-89.

Smith, A. D. (2001), “Perception and Belief.” Philosophy and Phenomenological Research 62; 283-308.

Stalnaker, Robert (1978), “Assertion.” Syntax and Semantics, vol 9: Pragmatics. P. Cole (Ed.) New York, Academic Press; 315-333.

Stalnaker, Robert (2001), “Metaphysics Without Conceptual Analysis.” Philosophy and Phenomenological Research 62; 631-636.

Stich, Stephen and Jonathon Weinberg (2001), “Jackson's Empirical Assumptions.” Philosophy and Phenomenological Research 62; 637-634.

Stoljar, Daniel (2000), “Physicalism and the Necessary A Posteriori.” The Journal of Philosophy 97; 33-54.

Stoljar, Daniel (2001), “The Conceivability Argument and Two Conceptions of the Physical.” Philosophical Perspectives 15; 394-413.

Thompson, Evan (1992), “Novel Colours.” Philosophical Studies 68; 321-349.

Tichý, Pavel (1983), “Kripke on Necessary A Posteriori.” Philosophical Studies 43; 225-241.

Tuomela, Raimo (1978), Dispositions. Dordrecht: D. Reidel Pub. Co.

Tye, Michael (1986), “The Subjective Qualities of Experience.” Mind 95; 1-17.

Tye, Michael (1991), The Imagery Debate. Cambridge, Mass.: MT Press.

Tye, Michael (1992), “Visual Qualia and Visual Content.” The Contents of Experience. T. Crane (Ed.) Cambridge, Cambridge University Press; 158-177.

Tye, Michael (1994), “Qualia, Content, and the Inverted Spectrum.” Nous 28; 159-183.

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

Tye, Michael (1995), “A Representational Theory of Pains and their Phenomenal Character.” Philosophical Perspectives, 9, AI, Connectionism, and Philosophical Psychology. J. Tomberlin (Ed.) Atascadero, Ridgeview Publishing Co.; 223-239.

Tye, Michael (1998), “Inverted Earth, Swampman, and Representationism.” Philosophical Perspectives 12; 459-477.

Tye, Michael (1999), “Phenomenal Consciousness: The Explanatory Gap as a Cognitive Illusion.” Mind 108; 705-725.

Tye, Michael (2000), Consciousness, Color, and Content. Cambridge, Mass.: MIT Press.

Tye, Michael (2001), “Oh Yes It Is.” Mind 110; 695-697.

Van Gulick, Robert (1995), “What Would Count as Explaining Consciousness?” Conscious Experience. T. Metzinger (Ed.) Schoenigh, Imprint Academic; 61-71.

Van Gulick, Robert (1997), “Understanding the Phenomenal Mind: Are We All Just Armadillos? Part 1.” The Nature of Consciousness: Philosophical Debates. N. Block, O. Flanagan and G. Güzeldere (Eds.) Cambridge, Mass, MIT Press; 559-566.

van Inwagen, Peter (1994), “Composition as Identity.” Philosophical Perspectives 8; 207-220.

White, Stephen (1986), “Curse of the Qualia.” Synthese 68; 333-368.

Wittgenstein, Ludwig (1969), On Certainty. Oxford: Blackwell.

Wittgenstein, Ludwig (1972), Philosophical Investigations. Oxford: Blackwell.

[1] From Farah (1997).

[2] From Three Essays on Religion.

[3] Block (1995b) p. 19.

[4] Smart himself points out in his recent contribution to the Stanford Encyclopaedia of Philosophy (2000) that many physiological organs, such as the eye, are functionally characterised and even as a single type may differ between species. Brains—and therefore brain processes—should perhaps be no different. As long as the type of processes hypothesised to be sensations do not, as a type, allow for differences that are not biological, the Identity Theory survives. The essential point is that whatever type of processes they are, they involve nothing more than “complex arrangements of physical constituents” (Smart 1959, p. 142). It is only when the empirical characterisation of the processes is neutral as to their constitution that the Identity Theory, as originally proposed, is refuted. Putnam (1967) went on to propose a characterisation of exactly that sort, of course, but with no more empirical evidence than cross-species differences. The importance of biology in the characterisation of those processes in our brain that are responsible for our mental lives remains a matter of fierce debate; see, e.g., Gazzaniga (1999) and Smith (1999).

[5] He went on to emphasise this latter point in subsequent papers, such as Putnam (1975).

[6] They also argued that psychological states generally cannot be identified with Turing Machine states but that, on the other hand, empirical consideration of the kind mentioned by Putnam (1967) weighed against the “physicalism” defended by Smart and Place.

[7] E.g., Putnam (1967); Bishop and Stich (1998).

[8] E.g., Block (1995a), Block and Stalnaker (1999).

[9] Actually, I will qualify this, but it is the spirit of my claim.

[10] And is in any case not strictly relevant to its outcome, which I hope will become clear.

[11] I should say that I am intending “home” as it is used in “Home is where the heart is”—when it is used, as it often is, without the qualifying “my” or “your” or “his”, etc.

[12] This is similar in effect to Johnston’s (1997) definition of a manifest kind as “a kind whose instances we identify and re-identify on the basis of their manifest properties”(p. 564), but includes the many apparent manifest kinds in Johnston’s sense that turn out not to be a kind at all. Although not everything with the same manifest properties is of the same kind –jade, for example– the same kind of justification is needed to assert that an apparent kind is not a kind as to assert which physical properties constitute it as a kind. As I am using the phrase, then, a manifest kind is simply something we ordinarily take to be a natural kind as indicated by our use of a natural kind word to refer to it. I take it that the existence of natural kind words as a linguistic category is not seriously in dispute, but in any case will not be arguing for it here.

[13] It is worth emphasising from the beginning that “H₂O” is an abbreviation. It is used here as shorthand for the description an ordinary chemist would give to uniquely identify the liquid it picks out. “Water” on the other hand is thought by no one to be literally shorthand (as opposed to being substitutable) for anything, a difference which makes it easier to see why “Water is H₂O” is not analogous to “Mark Twain is Samuel Clemens—a point that will have some importance in the next chapter.

[14] For this thesis I am conflating the categories of logical and analytic entailment, except where it is important not to do so. So, I will refer to statements such as “bachelors are bachelors” and “bachelors are unmarried males” both as logical truths. The differences, which obviously exist, will not be discussed here.

[15] That is to say, every possible world in which the person exists. Whether and to what extent the name refers in other worlds is a matter not even Kripke is very clear on, but for a good discussion on this point see Kaplan (1989a) and Salmon (1981).

[16] (1989a) and (1989b). Kaplan’s view differs in some respects from Kripke’s, but not in any respect that need concern us at the moment.

[17] On this question see also Evans (1982) and Lewis (1997).

[18] Clearly, perfect similarity is not necessary either—see Donnellan (1966) on definite descriptions. Donnellan makes the point (pp. 295-6), correctly in my view, that if a trick of the light were to make it seem to me that there is a person walking in the distance, my ostensive references to “that man over there” would actually constitute a failure of reference, rather than a mistaken reference to the trick of the light—at the same time, he points out, were it a man-shaped rock rather than the light, there are contexts in which “that man over there” might well refer to the rock. Acts of naming presumably operate under similar constraints.

[19] Davies and Humberstone (1980), Stalnaker (1978) and Kaplan (1978 and 1989a) are generally credited as the geneses of two-dimensionalism as it is currently defended by Jackson and Chalmers.

[20] Kripke’s reasons for rejecting the analyticity of statements such as “water is clear” are not terribly strong. Take the clearest statement of his rejection :

Note that Kant’s example, ‘gold is a yellow metal’ is not even a priori, and whatever necessity it has is established by scientific investigation; it is thus far from analytic in any sense. (p. 123fn)

Kant’s example may simply be too incomplete to be analytic. If his example was instead ‘gold around here (or of our acquaintance) is a yellow metal’ it would be at least a plausible candidate for a prioricity; certainly, it is not obvious what one needs to know beyond simply what “gold” means to know that gold around here is yellow. Furthermore, one does not have to know which properties if any gold has necessarily to know that it is in fact yellow around here.

[21] The point might be similarly put by saying that only rigid designators can have two dimensions in the sense being discussed here. Ordinary descriptions might simply be said to lack one of the dimensions that natural kind words have (proper names might be said to lack the other), rather than it being the case that the two dimensions coincide. But the point is not important here.

[22] It is worth noting that Jackson himself takes his disagreement with Kripke in these respects to be minimal, commenting (From Metaphysics to Ethics, p. 51fn) that it seemed to him that although Kripke does not explicitly use the two-dimensional framework, “the crucial point is implicit in his writings”. But for Kripke a demonstration (in the baptism) plays the role in reference-fixing that description plays for Jackson. It is perhaps possible to think of a demonstration having two dimensions - the object(s) picked out by a demonstration and the demonstration as such which might have picked out something else. But it is not clear that any such analogy between demonstrations and descriptions brings to light any real similarities between Kripke and Jackson on natural kind terms. Additionally, Chalmers (1996) says that

Kripke’s insight can be expressed by saying that there are in fact two intensions associated with a given concept. That is, there are two quite distinct patterns of dependence of the referent of a concept on the state of the world. (p. 57)

This seems to me to be a misstatement of Kripke’s view, which denies (on my reading) that natural kind words have any kind of Fregean ‘intension’ at all. Where Chalmers, who is a two-dimensionalist, and Kripke agree is that had the world been different then the natural kind words we actually use might have referred to different kinds of things. For Kripke, this follows from the trivial fact that any word might have been used to refer to something else; for Chalmers, however, it follows from the substantive fact, if it is a fact, that the very same concept might refer to H₂O for the inhabitants of the actual world, but XYZ for the inhabitants of a counterfactual world. Kripke’s insight and two-dimensionalism should not be confused.

[23] Wittgenstein (1969), §141.

[24] Evans makes the point that this is so even for proper names to some extent. In the case of natural kind words he cites several instances in which etymology does indeed reveal a change in referent; for example, “apple” was once used to refer to all fruits other than berries and “cobra” was once the word for all snakes (p. 390fn.).

[25] On this point I am following Jackson (1998a).

[26] See Chalmers (1995), p. 127.

[27] Almost. It is of course logically possible that there is no set of fundamental properties, that the world is complex “all the way down”. Should that in fact be the case, then physicalism might still be formulisable in terms of, say, physics being the science in which that set of properties is explicitly discussed which are simples of (or underpin) the properties discussed by all of the other sciences. Put this way, it does not matter if physics only talks explicitly about properties that are, in the end, complex.

[28] Self-evidently but not of course necessarily, and allowing for a certain amount of error. One can imagine its being universally believed that water melts witches without affecting the point being made here. On the other hand, if too many magical properties were believed to hold of water, it ought to be legitimate to wonder if there is really any such thing. This point is made by many writers but especially Lewis (1994) and (1995) when he talks of “good enough deservers of the name” in relation to mental properties in particular.

[29] This is similar to Lewis’s (1972) point, defended in a different way and with respect to mental terms, that “There is a strong odour of analyticity about the platitudes of common-sense psychology… If the names of mental states are like theoretical terms, they name nothing unless the theory (the cluster of platitudes) is more or less true. Hence it is analytic that either pain, etc., do not exist or most of our platitudes about them are true.” (p. 213)

[30] It must be said that Nagel (2000) seems to contain an affirmation of cosmic hermeneutics. Nagel contends that the theory that water is H₂O should entail that water has the “macro-level” properties it has. He argues that to reach the conclusion that a substance is water if and only if it is constituted by molecules of H₂O, “we must see that the behaviour of H₂O provides a true and complete account, with nothing left out—a strict entailment—of the features that are conceptually essential to water, and that this account is in fact true of the water around us.” (p. 439) But this is not a view held by either Jackson or Chalmers, with good reason. Nagel confuses explanation and entailment. Having a certain molecular structure is what best explains the viscosity of large collections of H₂O molecules, but does not strictly entail it. Indeed, it is very unclear in Nagel’s essay what role inferences to the best explanation are supposed to have in scientific identities. Because of this, the ‘explanatory gap’ is for Nagel to be blamed on a failure of entailment rather than explanation, which is therefore in turn to be blamed on the concepts involved. I criticise this conceptual reading of the explanatory gap in §5.1.

[31] If there turns out not to be anything at all fitting the ordinary conception then that may be grounds for some amount of revision, but that is the case for both the a posteriori and the a priori views – see Lewis (1995) for a good argument from the latter perspective.

[32] Moreover for Kripke, given that we lack access to anything like a complete history of any natural kind terms –in particular their baptisms– the “folk theory” serves as a basis from which we can infer which kind was baptised. The inference would be to the best explanation. Although the “folk theory” is in theory irrelevant to the question of reference for a Kripkean, in a practical sense there is no other way for determining reference in the absence of information specifically about the dubbing.

[33] In some imagined situations, we will be genuinely unsure of whether a certain word is appropriate. In that case, the conditions for our using that word will be indeterminate in that situation; this does not by itself, as some are inclined to say, threaten the possibility of an analysis of the concept expressed by the word.

[34] One is in a somewhat tricky situation here, since the sense of a word goes hand in hand with the concept it expresses. If a word expresses a concept that is unanalysable then presumably the sense of the word will be inexpressible, which makes it very hard to explain precisely which sense of the word “consciousness” expresses the unanalysable concept. It is a large problem for those who hold this view, since they are open for a challenger to demand that they be explicit about the sense of “consciousness” they are using.

[35] There is an apparent objection to Chalmers’ approach worth mentioning here: in looking for fundamental laws, which is what Chalmers thinks we should be doing, are we not asking what it is about the nature of the world in virtue of which particular parts of it are conscious? But that is after all exactly what we would be doing if we postulated that conscious states are nothing more than brain states. But a detailed discussion on this point would be too much of a digression.

[36] It seems to me that this is basis of the apparent possibility of “zombies”, of which a great deal has been made since the publication of The Conscious Mind.

[37] The discussion here has some similarities with Malcolm (1958), though it avoids the Wittgensteinian conclusion.

[38] It has been suggested to me that this point might be put as a “radical translation” problem (see Quine 1960) with particular respect to sensory terms. However, since it involves only sensory terms it is not really the sort of problem that Quine raised, which is more to do with wholesale translation. My point here can perhaps be put in this way – that the “no conceptual analysis” view raises a problem of interpretation for sensory terms even if there is no such problem for the rest of the language.

[39] Leaving aside, of course, the possibility that it may one day be possible to hook myself up directly to your brain.

[40] It is worth pointing, as something of a digression, that Chalmers’ own theory would yield unintuitive results if blue sensations were perfectly correlated with retina damage. For in that case, Chalmers’ “fundamental laws” connecting brain states to consciousness could not ignore the correlation, since the laws are based entirely on correlations. So, for Chalmers, there would be a law connecting retina damage and blue sensations, a result at least as implausible, it seems to me, as the idea that retina damage is part of what “blue sensation” refers to.

[41] In adding this I am siding with Armstrong (1961 and 1993) and Smith (2001) over Dretske (1969) insofar as sensations are essentially perceptual – which I believe, with qualifications (in §5.2 I will argue that sensory qualities can be accounted for by iterated perceptions, where perceptions are construed as entirely representational). Having said that, the mere possibility of experiencing a sensation in such a way that it is disconnected from any belief does not affect the current point, any more than the possibility of opaque water is a danger to the analysis of water in the previous two chapters. The crucial point is simply that sensations are typically representational —representational in the sense that they are typically closely related to beliefs, which are representational if anything is—and that we ordinarily individuate sensory states partly by their content.

[42] There may be a lingering worry that the occurrence of “belief” or something like it (“temptation to say”) smuggles in mental concepts and thereby defeats the analysis. I think this concern is unfounded, since our analysis is of sensations, not beliefs or mentality simpliciter. It is not obvious, moreover, that our notion of belief involves the notion of sensation, or that in order to understand what a belief is, you must understand what a sensation is. On this point, even Chalmers is ambivalent (see The Conscious Mind, p. 21).

[43] The issue of what makes a state a representational state is something I leave aside. It is enough for present purposes that whatever it is, it doesn’t involve the notion of sensation. It is held in some quarters that intentionality requires consciousness (e.g., Searle 1994). That is a view that I think is implausible, though there is little room to discuss it here; for a good discussion see Nelkin (1993).

[44] Block, who agrees that sensations are representational but denies that they are entirely representational, tags the latter view “representationism”, which I will follow in this thesis.

[45] Indeed, the main thesis of Chapter Five is that the vehicle of our awareness of that ‘way’ is an iterated perceptual process, an “inner sense”, which, so I will argue, allows it to be denied that sensory qualities outrun representational content while conceding that sensory qualities outrun the representational content of external perception. The way sensations feel, I will argue, is the representational content of an inner sense.

[46] I am intentionally leaving aside the issue of whether the “is” is of identity or of constitution. Moreover, I am not presuming that it is a transitive relation (that if [A is B] and [B is C], then [A is C]). It may be that the analytic “is” of (1) is the identity relation while the empirical “is” of (2) is the constitution relation. In that case the argument will rely on the substitutivity of identicals salva veritate in extensional contexts rather than transitivity.

[47] Stoljar (2000).

[48] We are pretending that all heat is molecular motion. We know that to be false, of course, but its falsity does not affect the principle we are discussing, and it is convenient to imagine it to be true in this example.

[49] There is a sense in which it might be said that heat appears to us to be the cause of ice melting. “Appears” in this sense connotes the way things are prima facie, and clearly has nothing to do with any particular conscious experience. Similarly, the “appearance/reality” distinction is intended to distinguish our perception of the world and the world itself, and not the world according to our naïve belief and the world as it actually is. That Kripke, at any rate, intends the former is evident from the emphasis he lends to the effect heat has on us qualitatively.

[50] Kripke’s explanation of our pre-reflective assent to statements like “gold might have turned out to be a compound” is that it is a “loose and inaccurate” way of talking (p. 142). It is also a loose and inaccurate way of believing—that is, we tend to believe statements like that in their literal sense until its logical consequences are pointed out, which is of course what Kripke did in Naming and Necessity. Our pre-reflective assent is therefore a genuine, systematic mistake—a fallacy—the effect of which is an over-emphasis on our everyday evidence for the presence of a manifest kind. The cause of that effect is, as Kripke points out, a failure to notice the rigidity of manifest kind words in certain contexts.

[51] It may be that Kripke could deny that we do know that pains exist at certain times in other people, but this possible reply is tantamount to solipsism, or at the very least scepticism about other minds. I take it as a Moorean fact that we do know that others are in pain when they are the victim of an accident and displaying pain behaviour. The question is not Do we know? but How do we know?

[52] Fodor (1998) develops a similar critique of recognitional concepts generally, though he emphasises their lack of compositionality in particular. In the case of pain he abstains from judgement, but asks in a sceptical vein, “Does one recognise one’s pains when one has them? Or does one just have them? If I can recognise good instances of MY PAIN, I suppose it follows that I have the concept PAIN. Does it follow, as compositionality would require if pain is a recognition concept, that I can also recognise good instances of YOUR PAIN?” (p. 47)

[53] Block (1990), p. 55—an accusation to which Lycan (1996) responded, “Block has commit-ed the fallacy fallacy”.

[54] Papineau’s (1993) answer to the Knowledge Argument makes use of this distinction; he argues that while in the black and white room, Mary (that is to say, the Mary of Jackson (1982 and 1986)) has only the third-person concept and acquires the first-person concept when she has her first colour experience. She thereby acquires a new belief, but not new knowledge.

[55] “Currently” because he recommends the development, if that is the right word, of a new concept that would entail both pains and brain processes, and thus render the connection intelligible. At least, that is how I understand his proposal.

[56] Chalmers (1996) says that his “two concepts of consciousness” might be describable as two parts of the concept of consciousness, but they are clearly distinct for him, as they are not for Nagel.

[57] I should emphasise that I am not directly drawing on Wittgenstein’s “beetle in the box” argument here, merely his hypothetical example. I think the conclusion I draw is compatible with what Wittgenstein’s argument shows as I read it, but given the hazards of its interpretation I am not inclined to press the point here.

[58] By which I mean the claim that it is possible for there to exist a person physically and functionally identical to me but who lacks phenomenal experiences.

[59] Some, including “sense-data” theorists, have argued that our sensation on seeing a rose is literally red. The arguments against such a position are many, and not worth repeating here, but see e.g., Ryle (1949), Austin (1964) and Pitcher (1971).

[60] On this point see also Lewis (1997).

[61] But note that it is not possible, on this account, for someone to be talking of their sensation in the absence of any internal state that they recognise themselves to be in under the right circumstances. If there is nothing able to be identified under the character of “green sensation”—which would by definition be the case for a “zombie”, for example—then there is nothing that could count as its content. Moreover, as against the standard zombie objection, if there were an entity that is a physical duplicate of me, atom for atom, then that entity would have to be in exactly the sensory state that I am in. The only way for this to be false is for there to be some non-physical state that I am in—in the actual world—when I am having sensations, which I take to be empirically implausible. Because “the state I am in” is neutral as far as physicalism/dualism goes, according to account defended here it is not the ordinary conception that makes zombies impossible, but empirical science, which is as it should be.

[62] That is to say, I agree in this respect with Lewis (1980).

[63] ‘Mad’ pain (again, from Lewis (1980)), incidentally, could be treated similarly—people who are in the brain state that I am normally in when I cut myself, when they exercise moderately on an empty stomach, would not call that state pain. Pain would, therefore, not be that state as they use the term. But as most of us use the term, pain is that brain state. If there is too much variety, then of course it becomes confusing to say which brain state pain is—it would perhaps be necessary to talk merely about which brain state my (or your, or their) pain is. Fortunately, it turns out that there is not that much variety.

[64] Lewis’ (1980) solution is in effect also an indexical solution, though he proposes that pain is indexed to a population (a species) rather than, as I propose, an individual.

[65] This is especially true of some other versions of the explanatory gap, such as Papineau’s. In Papineau (1998, p. 381) he argues that although it is benign, the reason for our lack of an available physical explanation for conscious states “is simply that there are no descriptions associated a priori with our everyday concepts of conscious states.” Whether or not this renders the explanatory gap benign, it is an undesirable view, as at any rate I argued in chapter three.

[66] Unless of course you happen to be a “non-rigid” analytic functionalist about sensations (i.e., you think that sensations are whatever plays a certain functional role, in any possible world), and consider these properties to be part of the various functional roles of the particular sensations. I have not argued against this particular view in this thesis, since my main goal has been to show that their contemporary opponents, such as Chalmers, are mistaken. Nevertheless, I have defended a view incompatible with that variety of functionalism, which is an argument against it inasmuch as my view is plausible.

[67] Levine (1993), p.133; quoted above.

[68] Loar (1997) argues in a similar vein, though he arrives at a different conclusion. Kripke also expresses this idea (Naming and Necessity, p. 152).

[69] As featured in Jackson (1982 and 1986).

[70] Even those that reject it, such as Loar (1997) and Nida Rümelin (1995).

[71] I am not here making the claim the Mary would actually know what the experience of red would ‘be like’, merely that it is not obvious that she wouldn’t, as is often assumed.

[72] I say “clearly” though there are those who may well disagree, such as McGilvray (1994) and Boghossian and Velleman (1989). I would like to say simply that they are clearly wrong, but cannot because they do give arguments which I have not the space here to refute. Nevertheless, my perceptions of red are clearly not themselves red.

[73] Another strategy is to challenge the representationist to spell out what is represented in intense bodily sensations such as pain or orgasm. Block (1995b) takes this strategy, among others, but I am not interested in pursuing them here.

[74] See Clarke (2000) for an argument to this conclusion.

[75] An “opaque” conception of content is one in which, for example, it is possible for me to appreciate the literary genius of George Orwell while at the same time believing Eric Blair to have been illiterate. Under a transparent conception, my belief in both cases would be about Orwell/Blair, making it harder to account for what would then seem to be a single inconsistent belief.

[76] In point of fact, Lycan (1998, p. 483) denies that there are qualities of experience in the specific sense that Block argues for, but not in a more general sense that motivates many of Block’s arguments.

[77] Following C.I. Lewis, Lycan defines a quale as “the introspectible monadic quality of what seems to be a phenomenal individual” (Consciousness and Experience, p. 69).

[78] Armstrong treats sensory qualities similarly to Harman. In Armstrong (1984, p. 186), for example, he writes that to account for ‘what it’s like’ for a bat to perceive through echolocation, “all that we need to admit is that the intentional objects of bats’ perceptions are utterly different from our own.” For this reason I have concentrated on Lycan’s rather than Armstrong’s arguments.

[79] This is circular because whether a perception is of redness will depend on whether it is caused by red things; but which things are to count as ‘red things’ is a matter of whether they tend to cause perceptions of redness.

[80] Neither can Lycan identify red with visual-red, since then red would be perceptible only visually, in which case the dispositional account collapses into a subjectivist account.

[81] Shoemaker (1994, p. 255) also makes this objection.

[82] As diverse, as, for example, Block and Fodor (1972], p. 172), Chalmers (1996, p. 28), Stoljar (2000, p. 47), Johnston (1997, p. 581) and Lewis (1980, p. 222).

[83] Lycan (1996), p. 65; see also Papineau (1998), p. 381.

[84] Thompson (1992) makes the related point that since (he argues) colours are defined relationally (as a position in a “colour space”), colours that are outside our actual range of experience are inconceivable, though a different range (i.e., a different colour space) is clearly possible.

[85] By “subjectivism” I mean the view that I take to be most directly derived from Locke, that colour is not a property of objects but rather part of the nature of our perception of objects, currently held by (among others) Boghossian and Velleman (1989) and McGilvray (1994).

[86] On this point see also Lewis (1997).

[87] We may sometimes be conscious of our inner sense, but I regard consciousness and sensory qualities as distinct problems. Lycan does too, though his inner sense—unlike mine—is supposed to address the nature of consciousness. Indeed, the question of which problems the postulation of an inner sense can solve is the most fundamental difference between Lycan and myself. Incidentally, most commentators on inner sense theories also discuss it as a proposed mechanism of consciousness rather than of sensory qualities, insofar as they also take the two issues to be distinct (for example, Levine (2001), and Carruthers (2000)). For that reason, many of the criticisms levelled at inner sense theories are not relevant to the version I am presenting here.

[88] By “primary perception” I mean simply the perception of the external world, to be contrasted with the proposed inner sense which has as its object primary perception.

[89] Consciousness and Experience, p. 64.

[90] Lycan’s view is not explicitly that the perception of sensory qualities are primitive concepts, but his discussion of the explanatory gap (Consciousness and Experience, pp. 62-66) demands that we construe it in that way. See particularly his discussion of the perception of “blutsiness” as a “private and perspectival word”.

[91] Appearances seem, indeed, to be three-place relations – in order to be said to have an appearance of a certain sort, a thing must be said to (at least potentially) appear to someone as something. At any rate they are not intrinsic properties.

[92] Carruthers (2000) objects to the explanatory power of a theory of this sort on the basis that introspection does not represent experiences of red as experiences of red—that is to say, we do not seem to be aware both of redness and an experience of redness (§9.1.2). Moreover, he argues (§8.1.2) that the contents of such an inner sensing would need to be enormously complex, implausibly given its doubtful evolutionary advantage. That is an objection against Lycan’s view, but not mine, since on my view the contents of inner sensings need not be nearly as rich as the contents of perception. My view can allow that because it denies that the inner sense is a mechanism of consciousness. Carruthers calls this (p. 212) a “mixed position” and “unobjectionable, so far as it goes”, since he is interested in consciousness.

[93] Moreover, there is no explanatory gap with respect to the second-order sensing, because it is improper to say of those sensings that they feel like anything. According to this account, only the object of a sensing can have a feel. I take it that this is consistent with the common-sense view that only our first-order perceptions feel like anything. Since the explanatory gap is thought, by those philosophers who believe in it, to be ‘created’ by the existence of feels, there is no question of it arising at the second-order level, on my view.

[94] See, for example, Roelfsema et al (1996) and Singer et al (1997).

[95] It may be argued here, as McGinn (1996 and 1983) and others have argued in the case of colours, that the reason our attention has focused on intrinsic properties is that we experience sensations as intrinsic properties, and that therefore they cannot be relational—or, specifically, that they cannot be dispositional. This is a mistake because, as Byrne (2001) points out in response to McGinn, what we perceive is typically the manifestation of the disposition—which itself need not be dispositional. Moreover, if a person were asked to point out the white objects in a room with red illumination, he or she would specifically not be relying on what the objects actually look like but rather what they would look like if the light were normal, and that is what makes red a dispositional property. Granted, then, that red is that property of objects such that they appear red in normal light, red appearances themselves must be relational simply because they involve two entities: the way some things appear (which are intentional objects)

and to whom they appear that way. The way of appearing need not itself be relational for the appearance to be relational (see Shoemaker (1994 and 1996a) for good discussions on this point). Moreover, the explanation for the way something appears must naturally involve the fact that it is an appearance. It is because ‘feels’ are standardly not taken to be the way sensory states appear that their relational aspect is missed and the explanatory gap arises.

[96] Shaffer (1963) brought an argument against physicalism along these lines—that introspection does not reveal anything neural, therefore sensations cannot be anything neural. Rosenthal (1976) refuted it partly along lines similar to those presented here.

[97] As he puts it, “The problem of explaining…phenomenal qualities is just the problem of explaining consciousness.” (The Conscious Mind, p. 5) What he means of course is hardest part of the problem of explaining consciousness.