Presented at the John B. Pierce Laboratory, Yale University, 24 February 1997.
Department of Philosophy U-54
University of Connecticut
Storrs, CT 06269-2054
I address this talk to anyone who believes in the possibility of an informative empirical science about sensory qualities. Potentially this is a large audience. By "sensory quality" I mean those qualities manifest in various sensory experiences: color, taste, smell, touch, pain, and so on. We should include sensory modalities humans do not share, such as electro-reception in fish, echolocation in bats, or the skylight compass in birds. Those pursuing empirical science about this large domain might pursue it in the halls of experimental psychology, psycho-physics, psychometrics, psycho-physiology, sensory physiology, neuroscience, neuro-biology, comparative psychology, neuro-anatomy, and so on and on. These days even molecular genetics has kicked in with some notable recent contributions to the sequencing of genes for photopigments and for olfactory receptors. But to all those investigators in all those halls I bring bad news. Your discipline is a priori impossible. Philosophers whom you do not know have uncovered a priori proofs that empirical investigation which proceeds along the lines currently underway, or which will proceed along lines that are currently imaginable, does not, will not, and cannot explain the sensory qualities of experience. Or at least so they say. You might as well give up now.
The proofs are a priori because they do not rely on any data; they are prior to experience. They are simple--seductively simple--thought experiments, which require only that one pay careful attention to some straight-forward conceptual distinctions. If they succeed, their conclusions are conceptual in scope. That is, they conclude that no account of a certain kind can explain the qualitative character of sensory experience. And the kind turns out to be broad enough to encompass all of the disciplines I mentioned, as well as to include any explanatory scheme that one can imagine such disciplines might produce.
The philosophers who produced these evil conundrums did it while working on a problem that they call the "qualia problem" or the problem of understanding the "qualitative character" of sensory experience. Qualia are understood to be those properties of sensation in virtue of which things appear as they do. So if I see something which is green, and something else which merely looks green, the idea is that both stimuli cause in me a state with a particular "qualitative character" (or quale), in virtue of which the things in front my eyes look green. Similarly for other modalities: an ache has a qualitative character which distinguishes it from an itch or a cramp; a salty taste presents a quale distinct from that of a sweet one, and so on. Philosophers borrowed the term "qualia" from the psychological writings of William James, who used it in this fashion. The same is true for the term "raw feels" in my title. Herbert Feigl used that term to characterize the problem in his 1967 book, The Mental and the Physical, borrowing it from E. C. Tolman. Philosophers who think that qualia pose an intractable problem in understanding consciousness sometimes refer to themselves as "qualio-philes" or even "qualia freaks". I like to think of them as philosophers on raw feels. What I plan to do is describe three of the thought experiments that have caused the ailment, present a diagnosis, and then describe a response--a treatment regimen, which I hope will at least serve as an inoculation, if not something which can heal the ravaged minds of those who have spent too many days--too many years--on raw feels.
All joking aside, I do think the three arguments I will discuss are intriguing and significant intellectual achievements. They lead simply and plausibly to implausible conclusions. As in so much of philosophy, the intellectual challenge is to point out, just as simply and plausibly, where the reasoning goes wrong. This is not so easy to do.
The first of these thought experiments is the best-known: spectrum inversion. The first premise is that "it makes sense, or seems to make sense, to suppose that objects we both call green look to me the way objects we both call red look to you." (Block 1980, 287-8). If we are both looking at green grass next to a red stop sign, perhaps the qualitative character of my sensation of the grass is just the one that you have when you look at the stop sign. "It makes sense" in that there is no logical or conceptual impossibility in the supposition. This is a point about our concepts: nothing in our concepts forbids this combination of events.
Furthermore, it seems to make sense to suppose that this "switch" or "inversion" could proceed systematically through all the colors. Perhaps to you every color looks like its complement does to me. If you balk at the possibility, you are asked to imagine that the inversion takes place within you, overnight. You wake up one morning and the world suddenly looks like a color negative did the day before. The sky presents the quale which yesterday you called "yellow", the grass looks red, the stop sign looks green, and so on. Yet every pair of stimuli you could discriminate yesterday you can discriminate as well today. If such "intra-subjective" inversion is conceivable, then inversion across individuals is as well. Furthermore, if the inversion is sufficiently systematic, then it will not be revealed by any of the discriminations among colors that either individual makes. We will both call the grass green, the sky blue, and the stop sign red, even though all of those things present different qualia to you than they do to me.
One final bit of terminology and we can get to the conclusion. The "functional role" or job of a sensory state is described by describing its place within the causal organization of the system: its causal relations to stimuli, to other mental states, and to behaviors. Two individuals are functionally isomorphic if every state in the psychology of one has exactly the functional role of the corresponding state in the other. The same psychological theory is then true of both individuals; in fact they both satisfy all the same causal laws. Now spectrum inversion seems to be conceivable even in two individuals who are functionally isomorphic. Whatever job in you is served by sensations with a green qualitative content is served in me by sensations of red, and conversely, and throughout the spectrum. But then the difference between us could not be detected by any psychological tests. Indeed, there would not be any experimental or causal test at all that could show that one of us is qualia-inverted relative to the other.
Remember we started by talking about color perception, but this thought experiment now seems to have shown that psychology is in principle incompetent to address the question. Here is how Ned Block puts it
What passes for the "psychology" of sensation or pain, for example, is (a) physiology, (b) psychophysics...or (c) a grab-bag of descriptive studies. Of these, only psychophysics could be construed as being about qualia per se. And it is obvious that psychophysics touches only the functional aspect of sensation, not its qualitative character. Psychophysical experiments done on you would have the same results if done on any system Psycho-functionally equivalent to you, even if it had inverted ... qualia. If experimental results would be unchanged whether or not the experimental subjects have inverted ... qualia, they can hardly be expected to cast light on the nature of qualia. (Block 1980, 289)
Block admits that the resources available to the science might change, but he goes on to say
on the basis of the kind of conceptual apparatus now available in psychology, I do not see how psychology in anything like its present incarnation could explain qualia. We cannot now conceive how psychology could explain qualia. ... on the basis of what we have to go on, it looks as if qualia are not in the domain of psychology. (Block 1980, 289).
This conclusion is a recurring theme in all three thought experiments that I will discuss. Current science is conceptually inadequate; it cannot explain qualia. The claim isn't merely that we don't know enough; it is rather that there is a conceptual gap, so that even if we filled in all the antecedents in explanations that we can imagine, those explanations would still fail to explain the qualitative character of sensation. To explain qualia we need more than just these kinds of facts; we need a new kind of science. This theme is reiterated in various guises by Ned Block, Thomas Nagel, Frank Jackson, David Chalmers, and Colin McGinn.
It suggests a stratagem for a response. The thought experiment concludes with a claim about the viability of current explanations. Psychology in its present incarnation cannot explain qualia. Yet it starts solely with considerations of what we can or cannot conceive. This is small footing for such a large conclusion. I suggest that you can cheerily concede to the philosophers whatever they want to claim about what they can or cannot conceive. Yes indeed spectrum inversion is conceivable. If like the mad queen in Alice in Wonderland you can conceive it three times before breakfast, good for you. But then as empiricists, dig in your heels when the inference is drawn that therefore the concepts and methods of empirical science are inadequate to the task of explanation. You can just say to the philosophers "these explorations of our concepts are very interesting, but we are in a different line of work. We do not study the nominal essence (or concepts of) qualitative states. We study their real essence--what in fact they are. And nothing you say about our concepts logically implies anything about that real essence."
That will work. It is the inoculation I promised; it works for any of the thought experiments I will talk about today. You simply challenge the philosopher to explain how these conceptual considerations impugn the validity of any actual explanation.
But one can do more. As a philosopher myself, I would like to provide some positive account of our talk about qualitative states--an account which coheres with what we know about their real essence. It would be nice to be able to move into the interior of the thought experiment and disarm it from within. Even better would be an account which yields coherent replies to the entire gamut of thought experiments. In the time that remains I hope to start on this project.
The empirical notion that I need for this account is the notion of a multidimensional spacing of sensory qualities--what I and others have called a "quality space". From here on I will talk more or less exclusively about color vision, since that is the modality which in these terms is best understood. The crucial move in developing an adequate empirical theory was made by psychologists many years ago--to abandon the notion of intrinsic properties of experience, and move instead to studying and scaling their similarity relationships. Instead of worrying about the quale blue, study the capacities which enable one to discriminate blue from other colors. A quality space is a representation of the similarity relationships obtaining among the qualities of sensory experience. It is constructed in one way or another to obey the principle that distances between qualities in the space are monotonic on their relative similarities. If P is relatively more similar to Q than to R, the distance between P and Q in the space should be less than the distance between P and R.
There is a delightful empirical story to be told about the task of ordering sensory qualities: how one can use various relationships of indiscriminability, matching, or relative similarity to construct an ordering relation, how one can determine the number of dimensions required to represent the structure of similarities, and how in some modalities it is possible to commence to search for the neuro-physiological mechanisms yielding dimensions of variation in terms of which discriminations are actually made (see Clark 1993). So for example the qualities of colored light that one can present to a normal subject through an optical viewfinder require three dimensions of variation (such as hue, saturation, and brightness) and there is a good consensus that the nervous system actually registers those three dimensions in a rather surprising way: an axis through the gray scale from white to black, an axis from red to green, and an axis from yellow to blue. These are the three "opponent processes". Ties to neuro-physiology are such that one can begin to understand the facts of color mixing and matching, the cancellation of complementary hues, adaptation and constancy effects, and various abnormalities of color vision.
Others in this room could tell the story better than I. For my purposes the key claim is that the qualities presented by a particular sensory modality always stand in some such structure of relationships of qualitative similarity. They can be ordered. The order represents similarities among the qualities. Presumably what generates that ordering of qualities can ultimately be explained in terms of neural mechanisms of discrimination. The normal human is born with vision that is trichromatic, and that works in such a way that a mixture of lights appearing red and yellow yields an appearance of orange. These "facts of qualitative similarity" ultimately derive from the neurophysiological realities, and probably are largely innate. They are the facts in virtue of which, in any similarity spacing of color qualities, orange must be found between red and yellow.
Now if this is the real essence of the perception of color, how might we learn terms in a natural language for qualities such as red, orange, and yellow? I offer a three part hypothesis:
The semantics of our terms for qualia rely on the inborn sense of similarity which generates the spacing of qualities. Given that spacing, terms for sensory qualities could function in something like the way Hilary Putnam and Saul Kripke claim that "natural kind" terms function. Putnam and Kripke deny that we can define with necessary and sufficient conditions terms such as "lemon" or "gold". We do not know necessary and sufficient conditions that will pick out all and only the lemons. Even if we did, it seems implausible to claim that they are part of the meaning of the word "lemon". Instead there is some relationship of theoretical similarity--sameT--which collects lemons in a kind. Elucidation of this relationship may require prolonged empirical investigation and new theoretical terms. It may rely on the genetics defining a species, for example. But we assume there is such a relation, and with it we can pick out all the lemons by pointing to one of them and saying
Lemons are just those things which are the sameT as that thing there.
Similarly "gold" applies to whatever is the sameT stuff as that stuff there--some demonstratively identified paradigm.
Qualitative terms could work in roughly the same way, except with them the requisite similarity relation is built-in (based on the operation of our sensory systems) rather than something requiring scientific theorizing. Instead of trying to frame necessary and sufficient conditions for "x is red", we could successfully identify everything that looks red by picking out some paradigm, pointing at it, and saying
A thing looks red if and only if it looks like that.
"Looking like that" here flags the use of the inborn sense of qualitative similarity. To take into account the varying borders and varying precision of color categories, we can be a little fancier and pick out both a paradigm and a foil (something clearly outside the border we wish to draw), then say
A thing looks red if and only if it looks relatively more like this than like that.
With suitable choice of paradigm and foil one can draw the borders as wide or as narrow as one pleases. Because of all the pointing involved, I call this account "indexical".
I think our qualitative terms really could work this way. While it does not yield a conceptual definition, the scheme is workable. One point to make right away is that the paradigms could vary from person to person, as long as everyone in question shares the sense of what resembles what. Suppose that whenever Jack judges x and y to match, Jill does too. Jack might have learned the word "red" by some early experiences with geranium petals. Jill's initial exemplar of the quality might have been a favorite toy firetruck. Jack says "red" to anything which roughly matches those geranium petals, while Jill says it to anything matching that firetruck. But (given successful learning of the term in the prior generation) the geranium petals roughly match the firetruck, and both Jack and Jill would agree on that match. So the use of different paradigms does not matter; both Jack and Jill would apply "red" to the same class of things. As long as different speakers (and different generations) share a sense of qualitative similarity, the torch can be passed from person to person, and from generation to generation, even though the actual physical items used to make the transfer vary each time. It is like a wealthy family passing down its wealth over several generations even though the actual dollar bills involved change each time.
Paradigms (or any stimuli) should not be mentioned in any analysis or definition of qualitative terms. One reason has just been given: the paradigms can vary from person to person, and no priority can be assigned to any one of them. More importantly, we run into logical difficulties if the "baptism" for red is treated as a definition of the meaning of the term. Suppose we say that a thing looks red if it presents the same color as geranium petals. One problem is paradigm existence: if this is to be true by definition, it implies that geranium petals must exist if anything in the universe looks red. But geraniums might not have evolved. Second, treating the identification as an analysis would imply that geranium petals must be red. But that too seems a contingent matter: they might have evolved to be yellow instead.
The implication is that we cannot treat the mention of paradigms as any part of the analysis of qualitative terms. They function in a different way. They serve to secure the reference of those terms. This is my second condition on the successful native use of qualitative terms: there must be a successful demonstrative identification of some exemplar. We need both a built-in sense of qualitative similarity and ostension to paradigms.
A final claim: that's all we need. The first two conditions would suffice to give us workable qualitative terms. Provided that we share a built-in sense of qualitative resemblance, pointing to paradigms could allow the hearer to latch onto the appropriate class of stimuli. To learn to identify that class we do not need to learn some set of necessary and sufficient conditions, which, in virtue of the meaning of the term "red", are true of all and only the red things. Perhaps there is no such set of necessary and sufficient conditions. Instead the semantics of the term relies upon those neural mechanisms of discrimination which generate the structure of similarities. To engage those mechanisms, and use them to pick out some class of red things, one requires some history of actual encounters with red things, so that one can pick out the appropriate place in quality space.
To the extent that we can give an analysis of the meaning a term like "red" or "green" it can mention only such structural facts as follow from the structure of similarities themselves. Red is a color that has a unitary hue (one that is not at all yellow and not at all blue); it is the complement of green, the color which with yellow is a component of orange, and so on. An attempt to define "red" must confine itself to detailing a pattern of relationships between the quality and other ones. Since we cannot mention any stimuli, that is all we have to work with.
Now to show (briefly) how this account can help re-arrange the innards of spectrum inversion.
To imagine spectrum inversion is to imagine a new mapping from stimuli to the color qualities they present. Suppose Jack and Jill are spectrum inverted relative to one another. If this inversion is to be undetectable, if it is to preserve "functional isomorphism", then all the judgments of matching and relative similarity made by Jack must be made as well by Jill. This implies a rather strong and surprising condition on color quality space: to be invertable, it must be symmetric. There must be at least two distinct assignments of stimuli to color qualities which preserve all the relations of relative similarity, matching, and indiscriminability among the colors. Otherwise in some region or other Jack will find two classes of stimuli to be relatively similar that Jill does not, and the game is up.
I think as matter of fact the ordering of human color qualities has various odd bumps, bulges, and anisotropies. So even though it is conceivable that there be a systematic inversion, in fact any such inversion would be detectable. For example of the unitary hues red is the one that can be the most saturated. Yellow is the one that when most saturated still most resembles an achromatic stimulus of the same brightness. Red and green are the hues that emerge first out of blackness, but then as the lights go up yellow and blue get brighter quicker. In these and other ways the quality space is asymmetric, so any actual inversion would be detectable. Hence the logical possibility of inversion does not impugn the actual success of our explanations.
Furthermore, I suggested that any analysis of qualitative terms can mention only such structural facts as follow from those similarity relations. Conceptual analysis does not yield much of a harvest when it focuses on a term like "red", I think because much of the semantics of these terms is implicit in the operation of sensory mechanisms. Basically you need to point to an exemplar and hope that your hearer shares your sense of what matches what. But if conceptual analysis yields anything at all, it will yield only facts intrinsic to the geometry of the quality space--those which a geometer confined on its surface could discover. These are facts about the structure of relationships which the qualities bear to one another. Such "analyses" do not and cannot mention any stimuli.
If that's our "concept" of red, we can explain why spectrum inversion is conceptually possible. No conceptual rule is violated when we suppose that what to you looks green looks red to me. In inversion we do not change the structure of relationships in which the various color qualities stand. Red is still the complement of green, orange is still between red and yellow, and so on. What changes is merely that new stimuli come to present those qualities. All those structural facts are unperturbed by an inversion. So contemplation of it does not cause any squawk of protest from our concepts. (Endnote 1)
In short, in spectrum inversion, we imagine the same psychological color space--the same structure of qualitative relations--associated with new stimuli. Essentially we assign new stimulus labels to the points within it. It is conceivable, but this bare conceivability fails to demonstrate any actual inadequacies in current science. The prospect has been tamed--disarmed from within.
On to two more thought experiments. I will first consider Thomas Nagel's worries about alien sensory modalities in "what is it like to be a bat?" (and subsequent publications), and then Frank Jackson's "knowledge argument" first published in "What Mary didn't know".
Nagel starts by pointing out that we all believe that bats have conscious experiences, since "after all, they are mammals" (Nagel 1979b, p. 168). Odd reasoning there, but let it pass. To believe that bats have conscious experiences is, he says, to believe that there is something it is like to be a bat. There is something it is like to fly around at dusk, tracking moths and other flying insects with echo-location, and catching them with your claws or teeth. Nagel's basic conclusion is that we humans cannot form concepts adequate to describe the character of that experience. We cannot form concepts that describe what it is like to be a bat. And, a second conclusion: Only a bat or something sufficiently like a bat can do that. (Endnote 2)
Why can't humans form such concepts? The basic difficulty is that bats have a sensory modality which is radically unlike any of our own:
bat sonar, though clearly a form of perception, is not similar in its operation to any sense that we possess, and there is no reason to suppose that it is subjectively like anything we can experience or imagine. (Nagel 1979b, p. 168)
And--here's a wonderful test--if you spend some time in an enclosed space with an excited bat, you will "know" that a bat perceives the world in a "fundamentally alien" way. The sensory apparatus and the "fundamental structure" of a bat is so different from our own that we cannot form concepts of what it is like to be a bat.
Our own experience provides the basic material for our imagination, whose range is therefore limited. It will not help to try to imagine that one has webbing on one's arms, which enables one to fly around at dusk and dawn catching insects in one's mouth; that one has very poor vision, and perceives the world by a system of reflected high-frequency sound signals; and that one spends the day hanging upside down by one's feet in an attic. Insofar as I can imagine this...it tells me only what it would be like for me to behave as a bat behaves. But that is not the question. I want to know what it is like for a bat to be a bat. Yet if I try to imagine this, I am restricted to the resources of my own mind, and those resources are inadequate to the task. (Nagel 1979b, p. 169).
So for example it won't do simply to imagine that one's hearing becomes much more sensitive. We need to imagine that auditory localization becomes about as accurate as human visual localization. Even that is not correct structurally; the modality differs from any one that humans possess. We must imagine fundamental changes in our own internal structure--changes in mental organization that in the end might make us unrecognizable to our former selves.
I suggest that the real problem posed by the bat is that it has an alien sense of qualitative similarity--of what resembles what. Echolocation presents an alien structure of perceptual resemblances. Not only can the bat detect stimuli that we cannot, but it will discriminate stimuli that to us are indiscriminable (such as two minutely different locations of a fluttering moth), and it will fail to discriminate stimuli that to us are readily discriminable (such as the color of the moths). It can discriminate between edible and inedible insects in the dark. (See Nobuo Suga and Jagmeet S. Kanwal (1995).) The modality has a structure of qualitative similarities and differences for which we can find no analog.
One sense of the phrase "what it is like" is "what it resembles". What is it like to go bungee jumping? I have heard it described as a combination of skydiving and a roller coaster ride. That is what the experience resembles. This works fine if the subject we are trying to understand has the same sense of qualitative resemblance (the same quality space, or at least one that is sufficiently similar) so that we can anticipate new experience by mentioning some prior one that we shared. "What X is like" is then explained by saying: well, it resembles Y. "What is it like?" questions then rely on the antecedent meaningfulness of propositions of the form "x is like y".
To understand what it is like to be a bat chasing a moth is to understand what it resembles for the bat. We need to understand the class of experiences that to the bat would be qualitatively identical. The problem is precisely that the bat echo-location has a structure of resemblances which to us is alien. So mentioning X does not help us understand what it is like to experience Y. We cannot imagine what it is like, in the sense that our paradigms and our sense of qualitative similarity cannot get us to the point where for us (like the bat) x and z are qualitatively similar. We would need to acquire the bat's quality space.
Suppose something like the indexical account of qualitative terms is correct. To use the terms directly we require a built-in sense of qualitative similarity, and some actual demonstrative identifications of paradigms. You could then pick out later instances of the same quality by their resemblances to past exemplars.
By hypothesis the bat operates with a totally alien structure of resemblances. We cannot sense the matches that it does. Allow me to mangle Wittgenstein (who said "if a lion could speak, we could not understand it"). If the bat had words for its qualia, we could not learn them. More precisely, we could not learn successfully to apply the words in direct, first-person fashion. To be sure, with measuring instruments and calculators of sufficient speed and capacity, we could simulate the sensory system of the bat, and laboriously calculate the parameters for the application of a given term. Reading the meter, we could then apply the word. But our sensory systems are just not built that way, and the calculations do not correspond to anything gong on within our nervous systems, so without those instruments and calculators we would be at a loss. In that sense we could not learn the words for bat qualia. (Endnote 3)
Now if "to know what it is like" to be a bat is to be something other than merely being a bat, then it requires that one make some judgement. That in turns requires that one deploy concepts (or at least terms). But by the preceding, humans could not learn to use the needed terms in direct first person fashion. It follows that there is a sense in which we cannot know what it is like to be a bat. We could not natively deploy the concepts that a bat might use to describe its qualia. We do not share the perceptual resemblances that make those descriptions possible.
So once again we can agree with many of the premises, and assign an agreeable interpretation to Nagel's claims about our abilities to form concepts of alien minds. For example, he says:
these creatures have specific experiences which cannot be represented by any mental concepts of which we could have first-person understanding. (Nagel 1986, p. 24).
If by "have first-person understanding" Nagel means "acquire the ability natively to deploy" or "acquire the capacity natively to pick out its extension", then I agree. But the reason we could not have "first person understanding" of the concepts of phenomenal experience of an alien mind is simply that we do not have the same structure of qualitative similarities as the alien mind, the same quality space. As Nagel said in 1979, "our structure does not permit us to operate with concepts of the requisite type" (Nagel 1979b, p. 171). Since use of the terms relies on native abilities to sense similarities among stimuli, if we lack such abilities we cannot directly pick out the extension of the term.
Nagel goes on to argue that there are "subjective" facts which are only accessible from a subject's point of view (Nagel 1979b, p. 172). Only a bat or something sufficiently similar to a bat could form the concepts necessary to understand what it is like to be a bat (see Nagel 1979b, p. 172). Here we should stop being agreeable. (Endnote 4) Nagel draws inferences similar to those of Ned Block: current science is conceptually inadequate. To understand the qualitative character of experience we need a new and currently unimaginable discipline. Nagel calls it "objective phenomenology'; it will proceeds with concepts radically unlike any of those we have today. They would enable you to describe to a blind man what it is like to see colors. Until such methods and concepts are developed, he says:
the status of physicalism is similar to that which the hypothesis that matter is energy would have had if uttered by a pre-Socratic philosopher. We do not have the beginning of a conception of how it might be true. (Nagel 1979b, p. 177).
By "physicalism" Nagel means the thesis that every mental state is just a physical state. I suggest that his talk of "subjective facts" and "points of view" is another way of asking whose sense of similarity is germane. To experience things from the bat's point of view is to experience the resemblances that the bat experiences; to have its quality space. This is why, to adopt that point of view, we must imagine radical changes in our own mental structure; the entire constellation of similarities and differences must change.
But I think this admission is a harmless one; it does not entail that there is some subjective fact which is perpetually beyond the reach of any physical science. Granted, we cannot be bats, nor could we learn first-person terms for bat qualia. It simply does not follow that only a bat or something sufficiently similar to a bat could form the concepts necessary to understand what it is like to be a bat. We can still study the resemblances that structure the bat's quality space. The subject matter of sensory resemblance is not one which is accessible only from one point of view; it can be studied in all those various halls of psychology, physiology, and neuroscience. In studying those resemblances we are studying the very same facts as are accessible to the bat from the bat's point of view.
Frank Jackson's "knowledge argument" is a variant of Nagel's worries. It again adopts the perspective of someone studying a sensory modality without enjoying its use. But instead of concentrating on an alien modality--one structured by resemblances that we do not share--here the modality is one that the student possesses, but has never exercized. This delightful thought experiment proceeds as follows. Mary is a brilliant neurophysiologist. By hypothesis she knows every physical fact there is to know about human color vision, about the reflectances of surfaces, about the wavelength composition of light, and so on. But she has been confined her entire life in a black and white room, and has never seen a color. Note that she is not color-blind. We suppose that she has a capacity to see colors, but that that capacity has never been exercized. One day she is let out of her room, and for the first time sees red. Premise: when she sees that first red item she learns something new. She learns what it is like to see red, and what it has been like for all the people living their lives outside the room. But by hypothesis she already knew all the physical facts there were to know. Hence knowing what it is like to see red is knowing something other than a physical fact.
Notice how radical this conclusion is. Even if you knew all the physical facts you would not know what it is like to see red. So it does not matter how long the investigations in all those halls scattered across campus proceed; whatever they produce will never be enough.
There have been many different responses to this argument. Most admit that Mary would learn something, but then urge that because the word "know" is ambiguous, accepting the premise that Mary learns something does not entail that Mary learns some non-physical fact. For example, Lawrence Nemirow and David Lewis have argued that what she learns is know-how rather than knowledge that. She learns a new ability to sort and classify objects. Learning a new skill is consistent with her prior knowledge of all the facts. So even though she already knew everything (in one sense), she can still learn something.
One worry about the Nemirow-Lewis line is that what happens when Mary is let out of the room does not seem to be akin to what happens when you learn a skill like wine-tasting or bird-watching. Granted, she thereafter has various new abilities, but they all seem to be consequent upon a judgement, of the form "that's red", whose content cannot be identified with a cluster of skills. So it seems that not all the knowledge gained is know-how.
I have argued that qualitative terms have an essential indexical component. To engage the mechanisms allowing unstudied first person use of the term, one must have some actual historical episode of a successful demonstrative identification. Otherwise you cannot use your innate sense of qualitative similarity to pick out other instances of the quality. When Mary steps out of the room, she makes her first successful demonstrative identification of a color. The content of this identification is expressed, naturally enough, with an indexical. She learns something to effect of "red is that".
Why can't Mary in her room learn what she learns when she steps outside it? One explanation is that what she learns is a non-physical fact. But there is simple alternative. In her achromatic chamber Mary cannot demonstratively identify any color. This is not because colors are non-physical properties, but simply because no actual samples are present. So no matter what text or other transmissions are sent into the room, we could not make it possible for her to demonstratively identify a color.
Now even within her room Mary might be able to calculate whether or not some item outside the room looks red. She could deduce: given its reflectance properties it will reflect a package of wavelengths which overall have the same effect as monochromatic light of 650 nm; hence to a normal trichromat it will look red. Even if she could complete such inferences, Mary would still be unable to say, just by looking at them, which objects are red. If she ascribes "red" at all, she must always proceed by inference and calculation. One way to put this: for her "red" is not an observation term. Its reference is always assisted, never direct. Suppose we let Mary out of the room, confront her with a novel x, ask her whether x looks red, and set the clock ticking. She starts madly scribbling away to calculate reflectance efficiencies, wavelength spectra, and the effects on her retina, but before she is done her time runs out.
Then we tell her: that's red. She learns: ah, red is that. She could not have made this judgement prior to her release, not because red is a non-physical property, but simply because demonstrative identification requires an actual sample.
Second trial. We confront her with an object y which is a metamer to x; it reflects a very different spectrum, but that spectrum has the same effect on Mary's cones. We set the clock ticking. She glances at y and says immediately "that's red too".
The indexical episode has given Mary new abilities. She can now rely on her built-in sense of resemblance, instead of calculating the eventual verdict. "Red" soon becomes an observation term; she can apply it in an unstudied, first person fashion. She can now tell by looking whether an item is red. The baptism gave Mary her bearings in quality space, so now she can apply color words in a new way. Like the Nemirow/Lewis line, this account agrees that Mary gains know-how. But on this account those abilities derive from a successful demonstrative identification. The latter cannot occur before Mary's release because it requires an actual sample to be present.
Perhaps an analogy would be helpful. We need some domain in which you could learn everything propositional there is to learn, yet still be totally at sea until a successful demonstrative identification has been made.
Fortunately, there is a delightful analogy to be had: the Ozma problem. This is roughly the problem of defining "left" and "right". These terms are part of a family (including clockwise and counter-clockwise, east & west, north pole and south pole), any one of which can be defined in terms of the others, but all of which seem to rely ultimately on some successful demonstrative identification. In particular, suppose we begin receiving transmissions from a planet on the far side of the galaxy (so far away that no stars are mutually observable, or at least we cannot tell from the descriptions that they are mutually observable). The aliens have terms "lana" and "rana" which we know mean left and right, but we don't know which is which. Similarly they also have rotational terms kana-wise and counter-kana wise, directions eana and wana, planetary poles nana and sana, and we know that these stand for one of the other of our cognate notions, but we don't know which is which.
The Ozma problem is: can you think of any possible transmission which would allow an unambiguous translation of these terms? We might learn that if on their planet you face the sun at sunrise, you are looking towards eana, but unfortunately their planet might be rotating in the reverse direction from ours, so "eana" is west, not east. We need to know whether "kana-wise" means clockwise or counter-clockwise. We learn that if you curl the fingers of your lana hand, they are curled kana-wise, but "lana" might be the right hand, so "kana-wise" would be counter-clockwise. They could transmit pictures, but we don't know if they normally scan them from left to right or from right to left (and of course there is no way for them to tell us). So perhaps we're printing all their negatives backwards; or, to use the correct printer's term, flopped. Call their planet "Flopped Earth". (By the way, it is called the Ozma problem because "Oz" was flopped: the planet (and clocks) rotated the other way, the sun rose in the west (which they call east), and so on.)
You learn the latitude and longitude of the visitor's center on Flopped Earth, and boldly volunteer to fly there under suspended animation. You awake many centuries later after your NASA rocket malfunctions and crash-lands. You are about to step out of your rocket. Which way should you start walking to head to the visitor's center?
I submit that your position is precisely analogous to that of Mary's when she is about to step out of her black and white room. You have a mass of propositional knowledge but an inability to employ the terms demonstratively. But all it would take is meeting a native who would say "This is your lana hand. That is your rana hand". Or even more simply, the native needs simply point in the correct direction ("The visitor's center is that way"). From your knowledge of the latitude and longitude you could then deduce "Ah, so "nana" is the south pole, "eana" is east, and "kana" is clockwise". All the terms would lock in place and you would know how to orient yourself. (Endnote 5)
You have learned something, but you haven't learned some funny new fact. You already knew all the facts--anything which could be conveyed linguistically was already in that packet of transmissions. But those transmissions did not suffice to fix the reference of "lana" and "rana". A successful demonstrative identification finally does the job. You have learned something, something to the effect of "the lana hand is this hand". But this does not point to some spooky new kind of fact.
In an exactly analogous way, Mary does learn something when she steps out of the room, but she does not learn a new non-physical fact. She learns a second, demonstrative, route to the identification of objects she already knows about. She can then use her inborn faculties of perceptual resemblance, rather than merely mention them in her calculations. This gives her the abilities to use color terms demonstratively (and as observation terms). But the indexical element is essential to the function of those terms. It is therefore not surprising that Mary must actually confront a sample and proceed through the baptism of demonstrative identification before she can use the terms in the normal way. With that we can dispatch thought experiment number three.
Unfortunately there are many more than just three qualia-based objections to the viability of physicalism. Allow me in closing to make a somewhat speculative observation about all of the objections. Survey all of the qualia-based conundrums; all of the thought experiments and arguments including inverted spectra, absent qualia, ersatz pains, blindsight, super blindsight, homunculi heads, Great Brains of China, zombies, inverted Earthlings, echo-locating bats, and achromatically confined neurophysiologists. In every one of these cases I believe there is either a breakdown in the determinate application of a sense of qualitative similarity, or there is a failure to achieve successful demonstrative identifications. We are asked either to extend the application of the relation "x matches y" confidently into conditions and domains in which it may be undefined, or we are asked to accept successful identifications in conditions where normal indexicals fail. That our ascriptions of qualitative states show precisely this pattern of breakdown I take to be evidence in favor of the indexical account. I think all of these objections can be answered, but to do so in detail will keep philosophers happily employed for a while.
Back to Austen Clark online papers.
Back to Uconn Philosophy home page.
1. The only sort of inversion that would be conceptually problematic for my account would be one in which (allegedly) the same quale moves to a different place in the structure of qualitative similarities. One would need to imagine that orange does not change but no longer resembles its current neighbors in quality space. Orange--that same quale--would for example have to cease to a reddish yellow. Perhaps that quale comes most closely to resemble blue and green. It must continue to be the quale it is, though it could no longer resemble the qualia it does. These latter inversions seem to me to be incoherent. Orange, whatever it is, is somewhat reddish and somewhat yellowish, no matter what stimuli in turn present red or yellow. (Back)
2. McGinn says something very similar: "...our concepts of consciousness are constrained by the specific form of our own consciousness, so that we cannot form concepts for quite alien forms of consciousness possessed by other actual and possible creatures" (McGinn 1991, p. 27) and "We cannot form the concept of a bat's sonar experience, since we do not ourselves have such experiences (of [sic] anything like them) -- just as congenitally blind people (with no visual imagery) cannot form concepts of visual experience." (McGinn 1991, p. 72). (Back)
3. The success required is in using the terms in ways that humans in fact use the terms: in direct, unstudied, first-person, observational fashion. Reference is direct, not assisted. The terms are applied as a result of observation, not theoretical inference. A person can use to learn them without relying on any instruments. If one drops these qualifications, I think we could learn words for bat qualia. Similarly Mary before she leaves her black and white room can refer to red, and can deduce just those conditions under which the term applies. But her reference is assisted, laborious, and inferential in ways that the normal direct, observational use is not. She needs to do a lot more than merely look at something to apply the term successfully. (Back)
4. The critical move is from the claim that some facts "embody" a point of view to the claim that some facts are accessible only from that point of view. This in turn relies on the claim that "experience does not have, in addition to its subjective character, an objective character that can be apprehended from many different points of view" ((Nagel 1979b, p. 173). This latter premise is unsupported. (Back)
5. Like Mary, once you have arrived on the planet, you could, given enough time, probably deduce the correct application of the various terms from what you know. You face the sun at sunrise and then can yourself complete the demonstrative identification "eana is that way". Similarly, if Mary knew all the physical facts, she could presumably deduce on her own that the first colored object she encounters is red and not green.(Back)
Block, Ned. (1978). Troubles with functionalism. In C. Wade Savage (ed.) Perception and Cognition: Issues in the Foundations of Psychology. Minnesota Studies in the Philosophy of Science, Volume IX. Minneapolis: University of Minnesota Press, 261-326. (Reprinted with revisions in Block (1980b).
Block, Ned. (1980). Troubles with Functionalism. In Ned Block (ed) Readings in the Philosophy of Psychology. Vol. 1. Cambridge, MA: Harvard University Press, pp. 268-305.
Block, Ned. (1990). Inverted Earth. In James E. Tomberlin (ed), Philosophical Perspectives, 4. Action Theory and Philosophy of Mind, 1990. Atascadero, California: Ridgeview Press.
Clark, Austen (1993) Sensory Qualities. Oxford: Clarendon Press.
McGinn, Colin (1991). The Problem of Consciousness. Oxford: Blackwell.
Nagel, Thomas. (1979a). Mortal Questions. Cambridge: Cambridge University Press.
Nagel, Thomas. (1979b). What is it like to be a bat? (Originally published in Philosophical Review, 83 October, 1974.) Reprinted with some revisions in Nagel 1979a. Also reprinted in Ned Block, (ed.) Readings in the Philosophy of Psychology, vol. 1, Cambridge Massachusetts, Harvard University Press, 1980, pp. 159-168.
Nagel, Thomas. (1986). The View from Nowhere. Oxford: Oxford University Press.
Suga, Nobuo and Kanwal, Jagmeet S. (1995). Echolocation: creating computational maps. In Michael Arbib, (ed) Handbook of Brain Theory and Neural Networks. Cambridge, Mass.: MIT Press, 344-48.
Return to Austen Clark's online papers .
Return to the Philosophy Department home page.