Robert Stalnaker


The inverted spectrum is one of the oldest philosophical chestnuts - a puzzle that children sometimes think up for themselves. What if red things looked to you the way green things look to me, and vice versa? More generally, the hypothesis is that the qualitative character of my experiences is a permutation of the qualitative character of yours. Your experience, when presented with something of a certain color, corresponds in its intrinsic character to the experience I have when presented with something of the complementary color. If this were true, how would we tell? We would both still call red things "red," and might respond to the colors of things in the world in all the same ways. The only thing that would differ is our respective inner experiences, experiences that no one is in a position to compare.


Verificationists took the inverted spectrum hypothesis to be a paradigm of meaningless metaphysics. Because, they argued, no one could be in a position to test the hypothesis, even in principle, one should conclude that no sense has been given to the statements that your inner experience is the same as, or different from, mine. But verificationist arguments and Wittgensteinian skepticism about inner experience are not taken to have the force that they were once thought to have, and philosophers with a wide variety of different views about the nature of mental phenomena are now inclined to take the inverted spectrum hypothesis seriously. More generally, they are inclined to take seriously the hypothesis that there are facts about whether one person's conscious experiences of colors, tastes and smells have the same or different qualitative character as those of another. But it is clear that such hypotheses raise problems for materialist and functionalist theories of mind, and that the account philosophers give of the relations between the phenomenal experiences of different people will have consequences for their general philosophical account of mind - on their accounts of consciousness, and of intentionality, and of the relations between them.


Sydney Shoemaker is one philosopher who has taken on the burden of reconciling a broadly functionalist and materialist conception of mind with what he calls "the common sense view" of the inverted spectrum.[1] My plan in this paper is to look critically at Shoemaker's articulation and defense of the common sense view, and at the conception of the content of qualitative experience that lies behind it. I am interested both in the issue itself, and in the argumentative strategy that Shoemaker uses - a strategy that trades on the relation between intrapersonal and interpersonal comparisons of qualitative content. The dialectic of intrapersonal and interpersonal comparisons of qualia is two-edged sword: it can also be used (and is used by Shoemaker) to raise problems for the coherence of the inverted spectrum as well as to defend it.


Shoemaker contrasts the common sense view that he defends with what he calls "the Frege-Schlick view," a view that rejects the meaningfulness of interpersonal comparisons of the character of phenomenal experience. It is a version of this old-fashioned view that I want to defend against Shoemaker's alternative. My criticisms of the common sense view, and defense of the Frege-Schlick alternative, will build on challenges that Shoemaker himself presents to the thesis he defends, challenges that I don't think he succeeds in meeting.

Before looking at the specific issue and arguments, I will (in section 1) make some general and impressionistic remarks about what is at stake in the way one responds to this puzzle. In section 2, I will sketch the specific problem posed by the inverted spectrum for the possibility of a functionalist theory of the mind, and the account of qualia that Shoemaker thinks can reconcile functionalism about mental properties with the common sense view of the inverted spectrum. I will look at what Shoemaker calls the intra-inner argument[2], which begins with a defense of the meaningfulness of an intrapersonal spectrum inversion, and argues from this to a defense of the coherence of interpersonal qualia comparisons. In section 3, I will sketch the Frege-Schlick view, and use some analogies to try to make clear what it claims, and that it is at least a coherent view. In section 4, I consider how the Frege-Schlick view should respond to the intra-inner argument, and what this kind of view ought to say about intrapersonal comparisons over time. Finally, in section 5, I will look in some detail at a counterargument (which I call "Shoemaker's paradox") that Shoemaker uses to raise a prima facie problem for the view he is defending, and at his response to it. I will argue that when Shoemaker's account of qualia is developed in response to the paradox, it loses its intuitive appeal, and its claim on the label, "the common sense view."


1. Qualitative and Intentional Content


Philosophers have often distinguished two components of the mind-body problem, two different distinctive features of the mind that are difficult to reconcile with a materialist account of mental life. First there is the problem of intentionality - the problem of explaining how the physical states of a human being can represent the world. How can the physical states of a physical object be about something? How can they be states that are capable of being evaluated as true or false? Second, there is the problem of phenomenal experience - the task of explaining how a physical being can be in states with the kind of qualitative character that our conscious states seem to have. How is it possible to explain, in purely physical terms, "what it is like" to see colors, or to experience the taste of wine, or the smell of turpentine. Different philosophers have had contrasting attitudes toward the relation between these two problems, and the two kinds of content - representational and qualitative content - use to characterize them. One strategy, best exemplified in the phenomenalist tradition, takes qualitative content as more basic, and aims to explain representational content in terms of it. We project our experience onto the world, or we construct a conception of the world out of the materials provided by phenomenal experience. Phenomenalism is not a doctrine that has many adherents these days, but I think the ideas underlying it persist in some internalist accounts of intentionality. A contrasting strategy is to begin with an account of the capacity to represent, and to explain qualitative experience in terms of it. We project our conception of the world back into ourselves; Qualitative states are to be explained in terms of the way things (including our own bodies) seem or appear to be - as hedged and self-conscious versions of the way we take things to be. Uncompromising proponents of this second strategy (Fred Dretske, Gil Harman, Bill Lycan, Michael Tye) have recently argued for the thesis (dubbed "representationism" by Ned Block) that phenomenal character just is a kind of representational content.[3]


One motivation of the second strategy is that from a functionalist or materialist point of view, the problem of intentionality seems more tractable than the problem of consciousness. Functionalist analyses of intentional states such as belief and desire are prima facie more plausible than functionalist accounts of phenomenal consciousness, since it is intuitively more natural to hold that there are conceptual connections between intentional mental states and the environmental causes and behavioral effects that they normally have. So if the two problems are tied together by an account of phenomenal character in terms of the ways our experience represents the world, the prospects for an explanation of consciousness that is compatible with materialism will be improved.


The puzzle of the inverted spectrum has played a prominent part in recent discussions of the relation between representational and qualitative content, since if we take spectrum inversion to be a possibility, we seem to have a case in which qualitative character and intentional content come apart. If your experience when you see a ripe tomato is (and always has been) like mine when I see an unripe pepper, and vice versa, then the same experience that in you represents the tomato as being red would, in me represent the tomato as being green. We have contrasting experiences when we look at a ripe tomato in normal conditions, but the tomato appears to be red to both of us (the representational content of our two experiences is the same). So if this hypothesis is coherent, it seems that we cannot explain the qualitative character of the visual experience in terms of the properties that things appear, to the experiencer, to have.


It is not my aim in this paper to defend representationism, since I do not really understand the thesis that the representational contents of experiences "completely capture the phenomenal character of the experiences."[4] But I am sympathetic to the general strategy of trying to explain qualitative content in terms of representational content, and skeptical about the coherence of thought experiments such as the inverted spectrum that attempt to pull them apart. I have to grant that common sense supports the coherence of the inverted spectrum, and that one who rejects it has a compelling cognitive illusion to explain away. But I think the illusion can be explained away, and that common sense does not speak with a single voice about the comparison of qualia over time, and between different persons. While common sense may support the coherence of the inverted spectrum, I think it also supports the idea that there is a conceptual connection between the qualitative character of a person's experience and the way things seem, to that person, to be.


2. Reconciling qualia with functionalism


If we assume that the inverted spectrum hypothesis makes sense, why is this a problem for a functionalist theory of mind? Functional accounts of qualia explain them in terms of a relational structure. We have certain perceptual discriminatory capacities, we are disposed to make certain judgments of similarity and difference in our experience, and to associate certain kinds of experiences with others. Discriminability is the intrapersonal criterion of identity for qualia: experiences we can't tell apart have the same qualitative character, and those we can tell apart differ in some qualitative respect. But suppose (as the inverted spectrum hypothesis does) that the relational structure has some symmetry in it. Suppose we could define a systematic permutation of types of qualitative experience that preserved all judgments of similarity and difference, and that more generally preserved the whole relational structure. If the experiences of two different people were permuted with respect to each other in this way, then we would have functional identity (because of the symmetry) with qualitative contrast (because qualia are mapped onto qualia discriminable from them). If this is possible, it seems to imply that no functional account of qualia could be right.


One response to the problem is to deny the symmetry. One may argue that even if there is some symmetry in the structure of color experience - in discriminatory capacities and judgments of similarity and difference - the whole relational structure in terms of which qualitative color experience is defined is more complex, including cross-modal associations (red is hot and loud, blue is cool, etc.), and facts about the way contrasting colors interact that will not be preserved in a permutation of colors with their complements. I will follow Shoemaker in setting this response aside. All we need assume is that symmetry is at least possible for some possible creatures with qualitative experience - that whatever the case with the actual capacities of human beings for qualitative experience, there could be creatures capable of qualitative experience with a symmetrical structure - capable of permutation that preserved all the relations between qualia that are available for a plausible functionalist account of qualitative experience.


Since a functionalist analysis of qualia individuates them by intrapersonal discriminatory capacities - relations between experiences that have no interpersonal analogues - one might expect a functionalist to embrace the Frege-Schlick view that rejects the intelligibility of interpersonal qualia comparisons. But Shoemaker resists this easy way out of the problem. His aim is to reconcile interpersonal qualia comparisons, and the coherence of the inverted spectrum, with a functionalist account of qualia. The strategy of reconciliation is to grant that "we cannot functionally define particular qualitative states," but to argue that "there is a sense in which we can functionally define the class of qualitative states - we can functionally define the identity conditions for members of this class, for we can functionally define the relationships of qualitative (phenomenological) similarity and difference."[5] This kind of functional analysis will determine an equivalence relation on the physical states of the person - equivalent states will be states that are realizations of the same qualitative state. The qualitative states are then identified with their physical realizations (or with the set of functionally equivalent physical realizations). To say that a ripe tomato looks to me just like it looks to you is to say not only that we are in functionally equivalent states, but also that your qualitative state is realized physically in a way that is also one of the possible realizations of the state I am in.


The main reason that Shoemaker resists the Frege-Schlick response to the problem is that he thinks that no one can plausibly deny the coherence of the hypothesis that there has been an intrapersonal spectrum inversion, and he thinks there is an irresistible argument from the intrapersonal case to the coherence of the interpersonal case. It is this argument, and more generally the intrapersonal/interpersonal dialectic, that will be my main concern. Shoemaker begins with a passage from Wittgenstein in which he considers a situation in which a person reports that "I see everything red blue today and vice versa." In such a case, we would, Wittgenstein suggests, "be inclined to say that he saw red what we saw blue." Neither verificationists nor functionalists would have any basis for rejecting the intelligibility of such a case, since the inversion is reflected in the behavioral dispositions and capacities of the person. But how does one move from this to the interpersonal case? The argument seems to be something like this: if an intrapersonal spectrum inversion is possible, then if Fred undergoes this change while Mary does not, then it cannot be that Fred's color experiences are the same as Mary's both before and after the change. Fred and Mary must be spectrum inverted with respect to each other either before or after the change.


Now it should be clear that this argument has no force against the Frege-Schlick view, since it presupposes that Fred's and Mary's color experiences are comparable, both before and after the change, which the Frege-Schlick view denies. What the argument shows is only that if it makes sense to say that two people's qualitative experiences are the same, then it must make sense to say that they are different. I don't think Shoemaker disagrees here - at one point he explicitly includes as a premise of the intra-inner argument the assumption that "the relations of similarity and difference that are well-defined intrasubjectively are also well-defined intersubjectively (i.e., . . . the Frege-Schlick view is false)."[6] So the intra-inner argument has a narrow target: those who accept the intelligibility of the claim that our color experiences are qualitatively similar, while rejecting the intelligibility of the claim that they are qualitatively different in a behaviorally undetectable way. But the argument still has some intuitive force for those who think that, at least in the special case where two people are exactly alike, both functionally and physically, it is hard to resist the conclusion that their experiences are qualitatively identical.[7] For there could be a person who was a functional and physical duplicate of Fred before his inversion, and another who was a duplicate of Fred after the inversion. (Or better, a duplicate of what Fred would become if his inversion were later followed by amnesia about the earlier experiences). These two twins, it seems reasonable to conclude, would be spectrum inverted with respect to each other.


Shoemaker notes that Wittgenstein, in the passage cited, seems to commit himself to interpersonal comparisons, at least in some situations, for he says that in a case such as the one he described, we would "be inclined to say that we saw red what he saw blue." But even a proponent of the unqualified Frege-Schlick view can make sense of the supposition that red things appear blue to someone. Before his mysterious alteration, Fred, like the rest of us, could tell what colors thing were by the way they looked. After the change, red things looked to him to be blue. Even when Fred comes to believe that it is he who has changed, and not the colors of things in the world, it remains true that red things look blue to him - that his visual system represents them as being blue. He just knows that things don't look (to him) the way they are. The Frege-Schlick view has no problem with this description of the situation since what that view rejects is only a notion of qualia that is both independent of representational content, and also comparable across persons. So I think this passage gives no reason to think that Wittgenstein is not in complete agreement with Frege and Schlick in rejecting this conception of qualia.

To counter the intentionalist interpretation of the case of the intrapersonal spectrum inversion, Shoemaker elaborates the story by imagining that the subject accommodates, semantically, to the change. After a while, Fred no longer reports that red things look blue. It is not that they have changed the way they look; it's just that Fred has adjusted to the fact that that is (now) the way that red things look, and so describes that look as "looking red." After the accommodation it can no longer be said that Fred's experience misrepresents the colors of things. So, it is argued, the qualitative contents that remain switched can not be identified with the intentional content.


Shoemaker also bolster's the argument by imagining that the spectrum inversion takes place gradually, in a series of partial inversions. The subject reports that "nearly everything looks the way it did before, but that the shades of color with a very small range have 'switched places' with their complementaries." The partial change results in a difference in the relational structure of the person's quality space, but a sequence of such changes (accompanied by semantic accommodation) might result in a full symmetrical inversion, a situation in which the relational structure, and the discriminatory capacities and behavioral dispositions were all as they were at the beginning. But, it is argued, we cannot plausibly say that the qualitative character of our subject's experience is the same at the end of this sequence of changes as it was at the beginning. "It would be ridiculous to suppose that these (final partial inversion plus semantic accommodation to it) would at one fell swoop, restore the character of the person's experience to what it as originally - for what the subject reports, and what her behavior would confirm, is that the final change affected the appearance of only a tiny fraction of the shades of color."[8] I will express some skepticism about this argument below, but first I want to try to get clearer about exactly what the Frege-Schlick view is, and how it should respond to the story of the intrapersonal spectrum inversion.


3. The Frege-Schlick view


The Frege-Schlick view is sometimes thought of as an eliminativist thesis - a thesis that denies the existence of phenomenal character, or qualia. I don't think this is quite right. The proponent of this view may say that there really are qualitative properties - it is just that the common sense view of them distorts their nature. The common sense view is that qualitative properties are local and intrinsic properties of experience - in fact they are the paradigm of an intrinsic property. The Frege-Schlick view holds that qualitative properties are relational - part of a purely relational structure. To try to clarify this kind of view, I will use some analogies with other purely relational conceptual structures: first, with a relational theory of space, and second with utility theory.


The kind of relational theory of space I have in mind a simpleminded theory - more Leibnizian than Einsteinian in that it assumes the conceptual independence of space and time. This kind of theory may not be plausible, but I think it is coherent, and that the analogy with it can throw some light both on the form of the Frege-Schlick view of qualia, and on the structure of some of Shoemaker's arguments. There are, on this view, no absolute locations. One can still talk meaningfully about the spatial location of things, but this is just a framework for talking about the spatial relations between things. The framework is conventional. Of course even an absolutist about space will say that the framework with which we assign triples of numbers to spatial points is conventional - there is no nonarbitrary origin or axes. The issue between the absolutist and the relationist is whether the identification of spatial points across time is conventional. There is no absolute motion - only change over time in the relative positions of things. Claims about the locations of things, and about motion, or change of location over time, are perfectly meaningful, according to such a theory, but such claims have to be understood in terms of a frame of reference.


Now suppose someone set out to refute the relational theory, or at least to show it to be highly implausible, by using an argument analogous to Shoemaker's gradualism argument cited above. Analogous to the denial of the meaningfulness of an interpersonal spectrum inversion is the denial of the meaningfulness of the hypothesis that the entire universe was shifted in space, three feet to the left. A possible world in which everything is three feet to the left is (according to the relationist) just a conventional redescription of the world as it is. "But even the relationist must grant," the critic argues, "that this chair might be shifted three feet to the left, and then this table. If one such change is possible, there is surely no impossibility is a series of such changes that all together would amount to a shift in everything in the universe. It would be ridiculous to suppose that the last of these changes would, in one fell swoop, return everything to the way it was originally when it is obvious that the last small change affected only the location of a single object." I hope that no one will take this argument to have any force against the relationist. The effect of breaking the alleged shift into small steps is to make salient, at each stage, a certain frame of reference. The relationist can grant that, and can explain why, it would be quite unnatural to describe the first change, or the last one, as a change in which one small thing remained fixed while the rest of the universe shifts three feet to the right. But he has no reason to grant that the alternative descriptions of these changes falsifies the facts, or to give up his claim that the overall effect of the sequence of hypothetical changes is to leave things exactly as they were.


My second analogy - one that provides a richer parallel with the case of qualia comparison - is with Von Neumann-Morgenstern utility theory. One begins with a preference relation for a given person - an ordering of some set of alternative possibilities that represents the relative evaluation of the desirability of the alternatives. O'Leary, let us say, prefers the duck to the salmon, and the salmon to the haggis. The purely ordinal structure is extended to an interval measure by introducing uncertainty, and representing the person's preferences over lotteries that may result in better or worse outcomes. So we ask O'Leary which he would choose between the salmon and a fifty-fifty chance of the duck or the haggis. If he prefers the salmon, that means that the interval between the duck and the salmon is less than the interval between the salmon and the haggis. Further questions, with different probability values for the lottery, can determine exactly where the middle item falls on the scale between the best and the worst option. More generally, preferences between lotteries determine cardinal utility values for all the options in the set. But the scale is a conventional one: any linear transformation of the scale yields an equivalent scale that represents the same facts about the subject's preferences.

Does it make sense to compare utility values across persons? There is an obvious analogy between the dispute between the Frege-Schlick view and the common sense view on the one hand and controversies about the interpersonal comparison of utilities on the other. And if one took utility as a measure of pleasure, or of the felt intensity of a desire, or of the subjective satisfaction that would be brought by attaining the option in question, then perhaps the question about utility would be an example of a question about comparing the qualitative character of experience. But this notion of utility is not intended as a measure of the feelings of satisfaction that one would receive if the option in question were realized - of the extent to which one would enjoy the duck if one were to eat it. Nor is it intended to be a measure of the intensity of feelings one has when one contemplates one's preferences.[9] If such things could be measured, there is no reason to think that it would correspond in the right way to one's preferences between lotteries. This concept of utility is supposed to systematize a person's motivational structure - the way her attitudes toward the alternative outcomes that might result from choices she faces or might face are disposed to affect her actions. If we can contrast intentional mental states and properties from phenomenal states and properties, it is clear that utility, on the intended interpretation, belongs on the intentional side.


The questions about the possibility of interpersonal comparisons of utility in a theory that extends the Von Neumann-Morgenstern theory are complicated and interesting,[10] and I think the parallels and contrasts with our issue are worth developing, but for my purposes here, I want simply to take the simple theory as an example of a theory that presents a purely relational structure, one that gives content to certain intrapersonal comparisons without providing content to interpersonal comparisons - an example that helps to clarify, by analogy, the kind of claim that the Frege-Schlick view is, and to show the coherence of the view. It would not be right to describe the utility theory as eliminativist - as denying the reality of utilities. According to the theory, there really are facts about people's preferences that are represented by utility numbers, but since the zero point and unit of the scale are arbitrary, no significance has been given to comparisons of the utility values of different persons.[11] And it would not be right to say that this theory is essentially incomplete - that there must be some kind of underlying facts that provide answers to the questions about interpersonal comparisons. There are, of course, underlying facts about the brain (and the environment) on which utility functions presumably supervene, but there is no reason to think that they will provide a basis for treating utility values as a measure of anything other than intrapersonal comparative relations.


The analogies support, I think, the coherence of a purely functionalist account of qualia that takes phenomenal character seriously, but treats it as relational feature of our experience. Such a conception of qualia need not be based on outmoded verificationist assumptions, but might be grounded in the kind of account of the identity conditions for qualia that Shoemaker gives, an

account that is like the Von Neumann-Morgenstern utility theory in that it grounds the concept of qualitative character in intrapersonal relations - discriminatory capacities and judgments of similarity and difference. But the intuition that it makes sense to ask how my phenomenal experiences compare with yours is a compelling and persistent one. Even if it is coherent to reject those questions, one may think that an account of qualia that makes sense of them will better accord with our common sense notion of qualitative character. And even if an intra-inner argument cannot refute the Frege-Schlick view, that view does have to give some account of the possibility of an intrapersonal spectrum inversion.


4. Memory and intrapersonal inversions


So what does the proponent of the Frege-Schlick view - of a purely functionalist theory of qualia - say about Wittgenstein's example of the intrapersonal spectrum inversion, and more generally about intrapersonal comparisons of qualitative experience over time? Everyone should agree that intrapersonal comparisons over time essentially involve memory. One is comparing an experience one is having with what one remembers a previous experience to have been like. For the functionalist, memory is constitutive - part of the account of the identity conditions for the qualitative properties. If I am disposed to judge, on seeing a red tomato today, that its color looks to me exactly like it looked yesterday, then this will be the basis for saying that my qualitative color experience is the same (and that the relevant brain states I am in at the different times are realizations of the same qualitative state), provided that my memory is accurate. But one can correctly say that the memory is accurate only if the two experiences are in fact qualitatively the same. The Frege-Schlick view should be skeptical, not only about interpersonal comparisons of qualia, but also about the independence of intrapersonal comparisons and assessments of the accuracy of qualitative memories. The common sense view would say that a memory report that the tomato looks now just like it did before is an accurate report if and only if the way it looked then really is just the way it looks now. But on the Frege-Schlick view (and on Shoemaker's view as well), the memory mechanisms that dispose the subject to make such judgments will be part of the functional theory that says what constitutes sameness or difference of qualitative experience. Shoemaker is explicit in agreeing with this point: "The functional role of a quale must surely include the ways in which the instantiation at one time combines with instantiations of the same or different qualia at later times to produce certain effects - e.g. recognition or surprise. This means that the total realization of a quale will have to include the memory mechanisms by which qualia have the appropriate "downstream" effects."[12]


One can accept the conceptual interdependence of memory and qualitative identity conditions and still be able to distinguish clear cases of memory error, or memory tampering. But a functionalist might allow that in some cases it will be arbitrary whether one says that two different visual experiences are qualitatively the same, or that the subject is systematically misremembering one to be like the other. This perhaps conflicts with the common sense view, but I think reflection on some examples suggests that it is not only what a functionalist should say, but is also intuitively plausible.


Consider another kind of inversion, the famous case of the inverting glasses. The glasses contain lenses and mirrors that turn everything upside down. After one has worn them for a while, one begins to adjust - one reflexively looks up rather than down when lightning flashes above, and one reports that, while things still look somewhat confusing, they are beginning to look a little more normal. And when the glasses are then removed, the subject reports, at first, that things look upside down. It is my understanding that no one has worn these glasses continuously for very long, but suppose one micro-engineered a device that could achieve this effect in a less cumbersome and intrusive manner, and that our heroic subject was able to wear it continuously for several years. It is at least a reasonable empirical hypothesis that the subject would eventually adjust completely, reporting that things look completely normal, and recovering his abilities to ride a bicycle, and to hit a curve ball. And when the subject who has completely adjusted to the glasses then removes them, let us suppose, he reports that things look upside down, just as they did when he first put the glasses on.


Now should we conclude that our subject is reporting accurately? Should we say that his visual experience after the adjustment really is just like it was, qualitatively, before he first put on the glasses, or should we conclude that he is systematically misremembering his previous experience? Perhaps the subject misremembers, or misinterprets, his phenomenal experience because of the way that, after the adjustment, his visual experiences are connected with other things, for example the way he is disposed, unreflectively, to respond to them. How should we tell whether this is a case of an adjustment to reversed qualia, or a case of return to the original qualia? Suppose physiological investigation reveals that various brain maps that are candidates for the core realizations of the qualitative visual experience remain inverted after the adjustment, just as they were immediately after the subject donned the glasses. Would that mean that the qualia remain inverted but that earlier qualia were misremembered because of way the adjustment changed the relations between the qualitative visual experience and other things, or should one instead say that the total realization of the qualitative experience involves all of the connections to one's responses (hand-eye coordination, etc,) that were involved in the person's adjustment to the change? On the second hypothesis, the two total physical states, with the reversed brain maps and corresponding changes elsewhere, would be alternative physical realizations of the same qualitative state. I don't think even naive common sense supports the judgment that there must be an answer to the question of which of these hypotheses is correct.


Here is still another kind of inversion: suppose it were possible to rewire nerves so as to change the felt location of pains. Just after the rewiring, when you stub your left toe, you feel a pain in your right toe. But in time, let us say, you adjust. You undergo something like the kind of semantic accommodation that Shoemaker assumes in his cases of intrapersonal color reversals, but the accommodation is more than semantic since one's reflexes and spontaneous behavior adjusts along with one's linguistic behavior. (The accommodation is more than semantic in Shoemaker's case as well, since the subject's spontaneous reactions to color experiences adjust. Are the reactive tendencies that change when one accommodates part of what constitutes the quality of visual experience, or are they just responses to it that may change without the quality of the experience changing?) It is only an empirical speculation that one would accommodate to such a change, but suppose the subject reports, after a certain amount of time, that the pain in the left toe now actually feels like it is in the left toe, just as it did before the rewiring. We can assume that, as in the inverted glasses case, the adjustment was not a sudden reversal, but that there was perhaps a transition period of confusion and disorientation during which one finds it hard to locate one's pain, or to describe its felt location consistently. Might the subject be misremembering what it was like to feel a pain in the left toe? Can we separate the representational content of the pain experience - the fact that it seems to be something happening in my left toe - from the qualitative character of the experience?


Or consider the intrapersonal spectrum inversion case itself: Suppose physiological examination of Wittgenstein's character reveals that the explanation for his unsettling experience ("I see everything red blue today, and vice versa") was a disturbance located in his memory/recall system. What is happening (physiologically, in the visual system) as he perceives the world is exactly like what happened yesterday. What has changed is what is happening, physiologically, when he remembers how things looked before. Would it be right to say to Fred, when we learn this, "Don't worry - things don't look queer to you at all - they look perfectly normal, just the way they always looked. What's queer is the way you remember things looking yesterday. At least the qualitative character of your visual experience is, despite the way it seems to be, just the same as it was." I would expect Fred to reply, with some justification, "I don't care what you say, the sky really does looks fiery red to me, and it didn't look that way yesterday. I don't mind your telling me why things look so strange - why the sky looks fiery red - but don't tell me I'm wrong about how things look to me." If Fred did reply in this way, I think he would have common sense on his side.


Even from an external theoretical point of view, it might not be clear from the details about the physiology and functional organization of our spectrum inverted character, whether to explain the switch in terms of memory error or in terms a change in qualia. Consider the following simpleminded model of a memory mechanism, inspired by Hume's picture of ideas as pale copies of impressions. Suppose that when you have a qualitative experience - seeing red (realized, say, by the stimulation of X fibers) a record of the experience is stored (in memory location M) The mechanism for recalling the memory is for the M state to cause the X fibers to be stimulated again (perhaps in a milder way than when one actually has a red experience). Now suppose (to draw on one of the kinds of cases Shoemaker discusses in his original paper on the inverted spectrum) our subject has a backup perceptual system. When the backup system kicks in, it is Y fibers rather than X fibers that are stimulated by seeing red. It seems that X and Y fiber stimulation are alternative realizations of the same quale, since the subject has no awareness of any change when the backup system takes over. If he is staring at a ripe tomato when the physiological change takes place, he reports no change in the way it looks. The theoretical explanation for the fact that he reports that the tomato looks the same (when the Y fibers are being stimulated) as it did earlier (when the X fibers were being stimulated) is that the memory of the earlier experience (which was caused by X fiber stimulation) is recalled by stimulating the Y fibers (in the pale way). Now if the qualia are really different, then this is then a case of misremembering, but if they are alternative realizations of the same qualia, then this is a case of accurate memory. On the common sense view, there is a fact here that is perhaps impossible to know (since we may suppose it is physically impossible for the two states to be realized simultaneously). But it is not only verificationists who should be skeptical of this: I think a functionalist like Shoemaker should be skeptical as well.


The case of the intrasubjective inverted spectrum shows that the relational structure that constitutes a quality space involves memory as well as capacities for simultaneous discrimination. Bringing memory into the functionalist account of qualia widens the intrapersonal relational structure, but does not contribute to providing a basis for interpersonal comparisons. Shoemaker will not disagree with this point; he agrees that we cannot ground interpersonal comparisons of qualia in functional analysis alone. His strategy is to explain qualia identity across persons in terms of identity of the physical properties that realize qualia. But there is a potential problem with this combination of two very different kinds of criteria for identifying qualitative properties - a problem that is brought out in an example that Shoemaker presented as a problem for his view in a postscript to his original paper on the inverted spectrum. In the next section, I will sketch the problem and Shoemaker's response to it, and say what I think it shows about what qualia must be if we are to make sense of interpersonal comparisons of them.


5. Shoemaker's paradox


The paradox presents a sequence of four scenarios involving the visual experiences of a person.[13] From assumptions about how qualia are identified, both within and across persons, that seem intuitively plausible, given Shoemaker's account of qualia, a contradiction is derived. Here are the stories:


Case A:


Alice is capable of being in either of two qualitative states (suppose they are visual experiences as different as red and green are for us). The qualia are physically realized by physical states Px and Py, respectively.


Case B:


Bertha is just like Alice, except that she has a backup visual system of the kind discussed above. Call the primary system (the system of case A) the α system and the backup system the β system. When the β system is active, there are two possible qualitative states realized by physical states Pz and Pw. Only one system is active at a time, and the subject does not notice the difference when the β system takes over. If Bertha was previously in state Px when looking at a ripe tomato, and is now in state Pz, she reports that things look just as they used to look. Similarly for states Py in the α system, and Pw in the β system. (No significance should be attached to the fact that the α system is called "primary." Either might have been active first, or more often.)


Case C:


Like Alice, Clara has only one visual system, but in her case, it is the β system rather than the α system.


Case D:


Dorothy is like Bertha. She has two visual systems, one a backup to the other, and they are the same two α and β systems. But in Dorothy, they are differently connected. If Dorothy was in state Px before when looking at a ripe tomato, and is now in state Pz, she reports that it looks very different - the way grass used to look. (When Dorothy earlier looked at grass, she was in physical state Py, rather than Px.)


In order not beg any questions about which qualia are the same and which different, I will label the qualitative properties with both the person and the physical realization state. So, for example Q(A,x) is the quale experienced by Alice when she is in physical realization state Px.


The following qualia identities seem reasonable:


1. Q(A,x) = Q(B,x)


since Alice and Bertha are in the same state, physically, at the time of the experience. It does not seem reasonable to suppose that the presence of an inactive backup system could affect the qualitative character of the experience.


2. Q(A,z) = Q(B,z) [Note from Ned Block: this is a misprint, which also appears in the printed version of this paper in The Philosophy of Sydney Shoemaker, Philosophical Topics, 26. 385-404.. This line should read Q(B,z) = Q(C,z)]


3. Q(D,z) = Q(C,z)


4. Q(D,x) = Q(A,x) for the same reason.


5. Q(B,x) = Q(B,z)


since there is no introspectible difference for Bertha between being in the Px state and being in the Pz state. This looks like a clear case of alternative physical realizations of the same quale.

It follows from 1-5 that Q(D,x) = Q(D,z), but it seems undeniable that


6. Q(D,x) Q(D,z)


since for Dorothy, there is an introspectible difference - as great as the difference between red and green - between being in the physical states Px and Pz.


The source of the problem is this: Shoemaker's account of qualia relies on two different criteria of identity for qualitative properties - two different equivalence relations between physical realizations that are supposed to make them realizations of the same qualia. The functional theory provides a criterion for intrapersonal identities in terms of discriminatory capacities, and identity of physical realization properties provides a criterion for interpersonal qualia identities. One cannot assume that the two equivalence relations can be coherently put together. x might be equivalent to z and y to w in the interpersonal sense, and x equivalent to z in the intrapersonal sense, even if y and w are nonequivalent in the intrapersonal sense. The mere abstract possibility of this kind of conflict between two equivalence relations is not enough to raise a problem, but the story makes at least a prima facie case that the particular equivalence relations needed for a plausible development of Shoemaker's account may clash in this way.


Is the story really coherent? In particular, how could the two visual systems, α and β. be "differently combined" so that the correspondences between physical states are reversed? It is here that memory mechanisms become essential to the story. Recall our simpleminded model of a memory/recall system. When any of our four subjects has a visual experience (realized by states Px, Py, Pz or Pw), a trace of the experience is stored in a memory, and later recalled by recreating (in the pale way) the appropriate physical state. So it is the memory/recall system that connects the alternative realizations of a qualitative state. Px and Pz are realizations of the same qualia (in Bertha) because x states (experienced while the α system is in effect) are later recalled (when the β system is operative) by activating a z state, and vice versa. But the situation is reversed in Dorothy. This is, as I said, only a simpleminded sketch of a mechanism, but I think it helps support the coherence of the story. I don't see any reason to think that a more complicated and realistic account of the memory and recall of experience might not also be compatible with such differences in the way memory connects experiential states at different times.


There are two ways one might avoid the contradiction: one might reject identity 5, or one might reject identities 1-4. The first response moves away from functionalism and toward a more purely physicalist account of qualia. Subjective indistinguishability (over time) is no longer taken to be a criterion of qualia identity; phenomenal experiences may systematically seem to the subject to be just the same when they are not. Spelling out this response would require deciding at what level of generality to define the physical types that are to be identified with qualitative properties, and it is not clear on what basis this should be done. But to avoid the kind of clash that arises in Shoemaker's paradox it may be necessary to identify qualia with very specific, finely individuated physical properties, and this might have the consequence that our qualitative experiences are much more variable and idiosyncratic than they seem to us to be. Suppose, as seems to me not empirically implausible, that the precise location in the brain of the physical events that realize perceptual experiences is somewhat variable, depending on accidents of development and experience, and changes over time within a single person as the brain develops, deteriorates, and interacts with the environment, without the changes being perceptible to the person. Should we say, if this were discovered to be true, that our phenomenal experiences keep changing their qualitative character in ways that we are incapable of noticing? If we did say this, I don't think we would any longer be defending the common sense view, which should maintain some conceptual connection between the qualitative character of experience and the way things seem from the point of view of the experiencer.

The second response - to reject the identities 1-4, is the response that Shoemaker defends. He argues that the addition of the backup system will affect qualitative character, since it will change the memory mechanisms that are constitutive of the identity conditions for qualia. So we should conclude that Alice's color experiences are qualitatively different from Bertha's because of the differences in ways their experiences are remembered and recalled. This response avoids the contradiction, but does it save the common sense view? According to this response, later changes in a person's perceptual and memory system, even differences in unrealized counterfactual possibilities, can affect the qualitative character of one's experience. Compare two people who differ only in this way: Bertha has a flexible brain: should the part of the brain that realizes a certain qualitative experience be damaged, another part of her brain could take over the task. Alice, on the other hand, has a less flexible brain: should she suffer the brain damage, she would lose the capacity for that kind of qualitative experience. Even though the core realizations of their qualitative experience are the same, and even if the brain damage never in fact occurs in either person, it would seem that Shoemaker's response implies that their qualia would be different, because of the different connections with potential alternative realizations of the experience. This kind of difference might be purely intrapersonal: suppose Alice once had a flexible brain, just like Bertha, but as she aged the potential backup system atrophied, perhaps from lack of use. The Shoemaker response seems to imply that the qualitative character of Alice's color experiences changes with the change in her brain, even though the change is inaccessible to introspection.


To support his response, Shoemaker considers a version of the paradox in which all the comparisons are intrapersonal. Suppose we have a person - call her Ellen - with the same α and β systems, but in which they are connected first in one way (as in Bertha), then in the other (as in Dorothy). Since the same incoherence can be generated in a purely intrasubjective case, Shoemaker argues, it is "not to be avoided by retreating to the Frege-Schlick view and giving up on intersubjective qualitative comparisons."[14] But as we have seen, the purely relational account of qualia that grounds the Frege-Schlick view not only rejects interpersonal qualia identities, but also claims that intrapersonal comparisons across time may hold only relative to perhaps arbitrary assumptions about the accuracy of qualitative memories. Shoemaker's purely intrapersonal version of his paradox seems to me to add support to this kind of conclusion. Consider how things seem to Ellen through the changes she undergoes: At time t1, she sees a red tomato. By time t2 the backup system has taken over, but she reports that the tomato looks just the same. At time t3, when the two systems have been differently connected, she reports that the tomato looks exactly like it did at t2, but radically different than it looked at t1. Puzzled, she reports, "I know it seemed at the time (t2) that it looked just the same at it did at t1, but as I now recall it, they looked very different at those two earlier times." It is clear that Ellen must be misremembering something about her experiences at one time or another, but when (at t2, or at t3)? Need there be any facts that tell us which it is?


Reflection on the case of Ellen - the purely intrapersonal case - makes clear that at some point, with one or another of the changes that she undergoes, something momentous will happen in Ellen's experience, something similar to what happened to the character in Wittgenstein's story. At some point, when the perceptual systems switch places either before or after the relations between her perceptual systems and memory/recall system are realigned, she will report that things that previously looked red now look green, and vice versa. At which point will that happen? Our telling of the stories does not say, since that depends on how our subject's experiences are connected to the world - on what the external stimuli are that put our subjects into the physical states that realize their qualitative experiences. We have ignored this because, while it is relevant to the intentional content of experience, the qualitative content of experience is supposed to be independent of what causes, or tends to cause, our qualitative experience.


If we bring the connections with colors of things in the world into our stories of Alice through Ellen, this will break the symmetry: either Bertha or Dorothy, but not both, will undergo the intrapersonal spectrum inversion when the perceptual system switches between α and β. Which it is will depend on whether Px in α subjects is caused by the same color as the one that causes Pz states in β subjects, or by the color that causes Pw states in β subjects. Does this give us a third way of resolving the paradox by giving a reason to reject two of the identities, 1-4, but not the others? This asymmetry can resolve the paradox only if we allow the intentional content of experience to play a role in the individuation of qualia. And while this could provide a basis for interpersonal comparison of the quality of experience, it won't do it in a way that makes sense of the interpersonal spectrum inversion.


Shoemaker's intra-inner argument began by assuming that, at least in normal cases it makes sense to say that the qualitative content of my color experience (when I see a ripe tomato) is the same as yours. The force of the argument, and the counterintuitive character of the Frege-Schlick view, rest on the plausibility of this assumption. The assumption is plausible, but I think that is because the common sense view of qualitative content does not separate it so sharply from intentional content. It does seem to be a common sense view to think that the qualitative character of experience has something essential to do with the way things appear to us to be.



Sydney and I were colleagues at Cornell for seventeen years. Throughout this time and since I have continued to learn from his work, and his example, what it is to be a philosopher. My debt to him will be evident from this paper, which even though it defends a view he disagrees with, draws most of its ideas and arguments from his illuminating and incisive papers on the topic. Sydney is a resourceful critic of his own views, and what I do here is mainly to try to put a different spin on an argument that he has presented against his own view, an argument that I find more persuasive than he does.



1. Shoemaker's most detailed discussion of the inverted spectrum is in "The Inverted Spectrum," in Shoemaker:1984, Identity Cause and Mind. Cambridge: Cambridge University Press, ch. 15 (first published in The Journal of Philosophy, 74 (1981). See also "Functionalism and Qualia," ch. 9 of the same book (first published in Philosophical Studies, 27 (1975). There are later discussions of the issue in "Intrasubjective/Intersubjective," in Shoemaker:1996, The First-Person Perspective and Other Essays. Cambridge: Cambridge University Press, ch. 7.

2. Shoemaker:1996, p. 141. I am not sure why it is "intra-inner" rather than "intra-inter" since it is short for intrapersonal/interpersonal, but I will follow Shoemaker's terminology.


3. See G. Harman:1990, "The Intrinsic Quality of Experience" in James Tomberlin(ed.), Philosophical Perspectives 4, Action Theory and Philosophy of Mind (Atascadero, CA: Ridgeview Press), pp. 31-520, F. Dretske:1995, Naturalizing the Mind (Cambridge, MA: MIT Press), W. Lycan:1996, Consciousness and Experience (Cambridge, MA: MIT Press), M. Tye:1995, Ten Problems of Consciousness (Cambridge, MA: MIT Press). The label, "representationism" comes from N. Block (forthcoming), "Mental Ink."

4. Block "Mental Ink", p. 1.

5. Shoemaker:1984, p. 195.

6. Shoemaker:1996, p. 144.

7. Though I think it is the assumption that physicalism is true - that everything is globally supervenient on the physical - that supports this intuition, and not the common sense view of qualia. I think that naive intuitions about qualia would tend to support the conceivability of a situation in which even physical duplicates might be spectrum inverted.

8. Shoemaker:1996, p. 144.

9. At least some reasons for thinking that utilities can be compared across persons derives from the assumption that utility is a measure of an introspectible feeling. John Harsanyi, for example, writes that "we have direct introspective access only to our own mental processes (such as our preferences and our feelings of satisfaction and dissatisfaction) defining our own utility function, but have only very indirect information about other people's mental processes. Many economists and philosophers take the view that our limited information about other people's minds renders it impossible for us to make meaningful interpersonal comparisons of utility." (Harsanyi:1987,"Interpersonal Utility Comparisons," in Utility and Probability, ed. by J. Eatwell, et al (The New Palgrave, New York: W. W. Norton), p. 128. But this is a misconception of the basis of the rejection of the meaningfulness of interpersonal comparisons. It is not that this kind of comparison is meaningless because it is too hard to get information about it. The view is rather that utility numbers are used to represent a structure of the preferences of a person, and are not intended to represent facts, available or not, about relations between persons. The critic of interpersonal comparisons views the question as like the question whether grams are bigger or smaller than centimeters. The reason we can't say is not that the facts are hidden from view.

10. Those who reject the meaningfulness of interpersonal comparisons of Von Neumann-Morgenstern utility need not deny that one might extend the theory to provide for interpersonal comparisons for the purpose of making welfare judgments - comparative judgments about the value to a whole community of various alternative social arrangements. Specifically, one might calibrate utility scales in one or another way, and argue that maximizing total or average utility, calibrated in that way, should be a goal of social policy. For example, one might argue for a scale that assigned 0 to each person's least favored outcome, and 1 to each person's most favored outcome. Alternatively, one might propose that the scale should be normalized by assigning 0 to the worst outcome, and 1 to the sum of the utilities of all the alternative possible outcomes. There might be arguments for or against such principles for calibration, or interpersonal comparison, and one might be clearly better than another. But such an extension of utility theory would be based on social and political principles - assumptions about the relative claims of different individuals on available resources - and not on factual judgments about the comparative intensities of people's experiences.

11. I emphasize the point that a purely relational theory is not eliminativist because I suspect that one of Shoemaker's motivations for resisting the Frege-Schlick view is that he takes it as a view that, as he puts it, "quines qualia," and he argues, against eliminativist views (such as representationism), that we need qualia to account for facts about our experience, and about secondary qualities. But I think a purely relational account of qualia can do the work that Shoemaker wants qualia to do.

12. Shoemaker:1996, p. 147.

13. I have told the story my own way, but the structure of the argument follows Shoemaker exactly. See Shoemaker:1984, 353-354.

[14].Shoemaker:1996, p. 149.