Brendan Lalor. Synthese 121, 249-289, 1999 (pre-publication version).
ABSTRACT. I defend a species of wide intentionalism about the phenomenal appearance of color and qualitative features of experience generally; I’ll lay it out in section 2. In section 3, I explain some of the main challenges to my intentionalism and some preliminary qualms about these arguments for nonintentionalism. Then in 4, I introduce an intentionalist form of internalism about qualia, and explain its inadequacies. Finally, section 5 consists of an extended defense of my theory of the phenomenal in which I make it plausible that qualitative content is not locally supervenient, give pointed replies to the arguments in section 3, and discuss the implications (for my theory) of empirical data pertaining to color memory and imagination. Not only shall it turn out that the legitimate needs of a theory of qualitative content do not support internalism about qualia, but, more than that, a positive explanation of what many have predicted will prove inexplicable on any physicalist account will be explained: the way experience seems to each of us qualitatively.
Consciousness is born supported by a being which is not itself. (Jean Paul Sartre 1958: lxi)
What is perceptually experienced is solely the information presented to the senses by the world. (Robert Shaw and John Bransford 1977: 16)
Ernie is admiring his home-grown tomato: it is shiny, red, plump, and obviously juicy. But to say that, of course, doesn’t capture the ineffable way the experience seems to him, qualitatively speaking. It’s not clear these features of experience — the so-called qualia — could ever be ‘captured,’ scientifically speaking.
This matter has exercised philosophers considerably. Historically, they have been attached to the idea that in perception agents such as Ernie ‘directly’ experience something other than the properties of perceived objects, that these experiences are properly analyzed as acts of apprehension of internal objects which, according to Frank Jackson (1976), resist physical explanation — mental images, sense data, things which account for the intrinsic phenomenal character of experiences. But the contemporary dispute is not over whether qualia are nonphysical. The current terms of the debate about qualia are due to Christopher Peacocke (1983), who characterized them as nonintentional features of experience — as mental contents which lack the property of being about anything, object, property, or relation. Hence, they can be characterized as introspectively accessible things or properties which determine phenomenal character and yet can vary independently of intentional content. Peacocke (1983 ), Ned Block (1990, 1996), and Sidney Shoemaker (1993) have argued that there are such, thus conceived; and others, such as Michael Tye (1993), Daniel Dennett (1988, 1991), and Gilbert Harm an (1990), have denied them.
Many contemporary philosophers, intentionalist and nonintentionalist, have suggested that if qualia are to be captured in the physicalist’s net at all, they must be internalistically constituted. Any account of the red look with which a tomato presents a gents like Ernie will have to proceed by reference to structures local to the nervous system, or to symbols’ internal functional roles, or to some such internal factors. In any case, the view goes, fixing the local physics must fix the phenomenology. Even if we grant that physically type-identical agents can have different thought contents (a la Twin Earth), when it comes to the qualities of twins’ experiences, contextual differences couldn’t make a difference. Or could they?
Yes, I shall argue. I defend a species of wide intentionalism about the phenomenal appearance of color and qualitative features of experience generally; I’ll lay it out in section 2. In section 3, I explain some of the main challenges to my intentionalism and some preliminary qualms about these arguments for nonintentionalism. Then in 4, I introduce an intentionalist form of internalism about qualia, and explain its inadequacies. Finally, section 5 consists of an extended defense of my theory of the phenomenal in which I make it plausible that qualitative content is not locally supervenient, give pointed replies to the arguments in section 3, and discuss the implications (for my theory) of empirical data pertaining to color memory and imagination. Not only shall it turn out that the legitimate needs of a theory of qualitative content do not support internalism about qualia, but, more than that, a positive explanation of what many have predicted will prove inexplicable on any physicalist account will be explained: the way experience seems to each of us qualitatively.
2. An Intentional Theory of the Phenomenal
It’s my conviction that the mental is intentional through and through. Even qualitative content is exhausted by aboutness; it is constituted by properties objects are represented as having. So qualitative properties of experience cannot vary differentially with respect to intentional ones.[ 1 ] I call my version of qualia intentionalism qualia interactivism, since, according to it, perceptual qualia are emergent properties of organism-environment interactivity, and depend especially on the fine-grained information-theoretic properties of perceptual systems, such as a system’s ability to discriminate different properties in the world, and when those properties are determinable, to make informational distinctions. In the form of a motto: ‘Same cognitively accessible informational properties, same qualia; different cognitively accessible informational properties, different qualia.’
To be sure, this intentionalist thesis is about the metaphysical constitution of qualitative aspects of experience, for instance, color experience. It holds that just as water = H2O, and temperature = mean molecular kinetic energy, experience of color C in S = the introspectible activation of structures in S which signal C, which carry C-information.
Let me mention a few considerations which support intentionalism about the qualities of experience in a general way. First, it allows a nonmysterious account of how qualia could relate to the physical world, alleviating the need to find supra-scientific entities to serve as phenomenal aspects of experiences, such as qualia of unique red experiences, or, say, the duck-look of the duck-rabbit figure. If we can naturalize other intentional properties, as we’re already committed to doing, there’s no reason to withhold intentionalist naturalization from these. Second, this allows intentionalists to account for misrepresentations such as phantom pains, after-images, hallucinations, and perceptual illusions simply by explaining why states with those intentional contents are tokened in the relevant conditions. Finally, it makes for a unified account of mental phenomena as intentional, as famously articulated by Franz Brentano (1874/1973). So the semantics of expressions embedded in linguistic contexts such as, ‘This looks — ,’ advert to the intentional contents of visual experiences, just like the semantics for expressions in such contexts as, ‘This is believed to be — ,’ advert to belief contents.
Qualia intentionalism would be nice, but, prima facie, it might be argued, no explanation we can presently imagine would do justice to Ernie’s experience of color. That is, it seems it will always be possible for him to say, ‘Yes, I understand the explanans, but what about my rich experience of color?’ This sense gives rise to what’s been called the ‘explanatory gap’ between qualia and the physical world. There are two reasons I won’t be preoccupied with closing this gap. First, these gaps come and go as intuitions change and concepts become more sophisticated; so, in a sense, I don’t think this problem per se is too important for any physicalists now. It may fix itself. Second, I accept a kind of irreducibility thesis anyway, according to which phenomenology is an emergent product of interactivity between an organism and the world, not supervenient on merely local physical constitution. So I will not be concerned if my completed account doesn’t give one the sense that qualia have been reduced. Properly appreciated, however, interactivism changes intuitions about the significance of the gap.
The explanatory gap may nevertheless account for why some of the subtlest anti-individualist philosophers have suggested that we might vary the intentional while holding phenomenology constant. Indeed, this is often considered part and parcel of famous externalist thought experiments. Tyler Burge (1979), for instance, speaks of ‘Holding … non-intentional phenomenal experience … constant’ and varying intentional properties (section IIb.f). Davidson (1987) allows this as well (448). Burge (1988) even speculates that
at least some aspects of one’s mental life are fixed by the chemical composition of one’s body. One might call these pure phenomenological feels. (653)
I see no reason for anti-individualism to permit this, and interactivism shall show why not.
Before going on to consider the perception of color, let me say, first, that I envision what I say here about visualia extending to the other modalities giving rise to phenomenal experience: audition, olfaction, gustation, kinesthesis. Second, I am all for granting intentional status to the phenomenal feels of tickles, pains, orgasms, moods, and sudden insights,[ 2 ] as well. My view is that these stand to states of the body, as visual and other typically outward-directed form s of perception stand to states of the environment.[ 3 ] Not that there is a hard and fast line between inner and outer perception: For instance, haptic perception of deformations of the skin and articular stresses provides information about both the weights of objects and the position of the body’s limbs.
I take all of these kinds of qualia to be introspectible activations of our perceptual systems as they register the complex information-theoretic properties they are designed, evolved, or trained to indicate. The process by which we introspect these qualia is just our testing the degree to which we can gain access to the information carried by our activated perceptual systems. I fulfill the burden of making this plausible in section 5 below.
We now turn to the ‘hard case’ of techni-color qualia.
3. Attacks on the Intentional Theory of the Phenomenal
Critics of the intentional theory of qualia claim that qualitative or phenomenal contents and intentional contents come apart under at least various counterfactual circumstances — that experience is essentially more than content which has an aboutness to it: the redness of Ernie’s experience has a nonrepresentational quality. The most forceful nonintentionalist of late has been Block, on whose attack I’ll dwell.
Block has devised arguments to substantiate his thesis that even if in fact qualitative content and intentional content are always correlates, it is metaphysically possible to decouple them, as when one gets inverted with respect to the other. I’ll focus here on the two arguments which have proved to be of lasting significance. One of them fosters the intuition that intentional properties can vary even though phenomenal ones remain constant; and one cultivates the intuition that phenomenal ones can vary even though intentional ones remain constant. They work together to counter all brands of intentionalism on the market. The alleged possible property difference would show, by Leibniz’ Law, that there is more to this aspect of the mental than intentionalism captures.
Before presenting the arguments, let me mention two requirements a case of property inversion must satisfy in order to be effective for the anti-intentionalist’s purposes. First, to be convincing, it ought to be a case in which either all the relevant intentional color properties of experience shift while the phenomenal ones remain constant, or vice versa. So phenomenal inversion should not show its effects in behavior-driving intentional states, like belief, since that would indicate that intentional properties had shifted along with phenomenal character. An inverted subject’s remark, ‘Peculiarly, the grass now looks red,’ would show as much. In effective inversions of this kind, the beliefs and reports of the inverted subject regarding the character of experience should have the same intentional contents as normal subjects’. It is likewise necessary for intentional inversion that subjects be unable to report the phenomenal sameness, since their ability to do so would indicate an incomplete intentional inversion. In these ways, inversions must be behaviorally undetectable. Second, it must be possible to imagine in one world both such an inversion and all the nitty-gritty details which intentionalists claim are relevant to the constitution of color experience. Block must at least comply with these requirements to show that intentionalists are mistaken.
I’ll present his arguments, and then two kinds of ultimately unsatisfying intentionalist responses appealing to technical difficulties which attend the engineering of the inversions.
3.1 Block’s One-Two Punch
3.1.1 Inverted Earth[ 4 ]
You and your twin are separated at birth. You’re left on Earth, your twin is brought to Inverted Earth, which is as much like Earth as possible, except for this: each thing’s color is the complement of its counterpart on Earth. Ripe tomatoes there are gre en, although called ‘red’, and the sky is yellow, although called ‘blue’, and so forth. (So your twin’s skin pigmentation is appropriately altered.) However, your twin has magic lenses inserted into his eyes so that ripe tomatoes look red to him, and the sky blue. Thus, your brain states are relevantly similar in your respective tomato- and sky-viewing situations.
[I]f you and your twin are looking at your respective skies, saying ‘Still blue’, you have experiences that have the same qualitative content, but inverted intentional content, and they are functionally inverted. Conclusion: the distinction between intentional and qualitative content is vindicated … (Block 1990: 67)
In a variant of the example, Block asks us to imagine that one night mad scientists insert color inverting lenses into your eyes and substitute you for your counterpart on Inverted Earth. Upon awakening, unsuspecting, you might look out your window and say, ‘How green the grass.’ Thanks to the inverting lenses, you would enjoy the same overall janglings in the visual cortex and other relevant parts of the visual system as used to be produced in you by Earth’s green grass. Your statement would be false, of course, since ‘green’ in your mouth means green, not red, the color of grass on Inverted Earth. But with time and conversance, you would assimilate to your physical environs and social community, after which, your utterances of ‘green’ would come to mean red, and your beliefs and remarks about the colors of things would once again be systematically correct. Block again concludes: Same qualitative content, different intentional content. Distinction vindicated. QED.
Block need not have resorted to Inverted Earth to make the case. Imagine the perhaps more likely sort of scenario occurring on Star Trek: The Next Generation. Of two identical twin human babies, one is fit with lenses and raised on one of the Enterprise’s holodecks running a simulation of Inverted Earth, and the other raised lens-less in another holodeck simulating normal Earth conditions. Intuitively, both have the same qualitative experience, although different intentional contents. (Or, if you like, your eyes can be fit with lenses and you can be transported to the Inverted holodeck.)
Block’s variants on the Inverted Earth thought experiment provide two distinct arguments that qualitative aspects of experience stay the same while intentional aspects vary. In the first variant, the important thing is that the brain states of you and your twin remain relevantly similar. Block (1994) claims there is ‘a highly plausible principle that any materialist should accept: that qualia supervene on physical constitution’ (517), not intentional properties. This version of the thought experiment seems to support the principle: Where there is no difference in brain state, we don’t imagine there would be a difference in qualia, even if there is an intentional difference. In the second variant, in which you are unknowingly taken to Inverted Eart h, the important thing is that at no point do you notice a phenomenal change, which, Block then concludes, indicates that there was no such change, even though there was an intentional change.
Note that while the Inverted Earth example presents a challenge to what might be called strong intentionalism, which holds qualia are identical to certain intentional properties of experience, it leaves intact a weaker intentionalist thesis: No difference in qualia without a difference in intentional properties (see footnote 1 above).
But this is where Block’s Inverted Spectrum argument comes in.
3.1.2 Inverted Spectrum
Traditionally, the philosophers’ chestnut, the Inverted Spectrum hypothesis, involves imagining two persons who agree on the colors of objects and the use of color vocabulary, but whose color experiences are nevertheless inverted with respect to each other (see, e.g., John Locke 1690/1959: Book II, Chapter 32, section 15). As things stood, the trouble for the anti-intentionalist was that, given the inconclusiveness of reports, there was no way to settle the issue about whether the inversion was genuine — in principle, no third party could compare the qualia and let us know whether we weren’t just saying there was a spectrum inversion. Verificationists thought this version of the Inverted Spectrum paradigm nonsense.
Then Block (1990: 61) conceived the intrapersonal version of the hypothesis, which seems to alleviate the need for an ‘outside’ opinion. It unfolds in four phases. First, a normal subject enjoys normal color experiences at t1. Next, the subject undergoes the spectral inversion, and issues reports such as: ‘The sky looks yellow,’ ‘Ripe tomatoes look green,’ and ‘Lemons look blue.’ In the third phase, our subject brings her use of color terms and all other color-related dispositions into step with the community and the environment, and even issues reports like: ‘The sky looks blue,’ and ‘Roses are red.’ Still, she remembers that colors used to seem different to her until the final stage, which begins when the last memory of her colorful past vanishes (due, say, to amnesia) — label that moment t2. Here we have times t1 and t2 such that the relevant intentional properties are invariant across them while the qualitative properties shift. Presto: the phenomenal varies while the intentional doesn’t. And the subject reported it. QED.
3.1.3 Engineering Challenges
Both thought experiments are open to criticisms of an empirical nature: It might be denied that Inverted Earth affords an example in which qualia remain constant while intentional contents shift, or that our color experiences even could be inverted.
Evolutionary semantics, the view that much intentional content is phylogenetically fixed, is gaining a foothold in philosophy of mind today;[ 5 ] and adherents like Fred Dretske (1995) and Tye (1995) think Inverted Earth is unable to do the job Block assigns it. They argue that intentional phenomenal contents given rise to by color detectors are fixed by natural selection. Whenever the green-detector is activated, it will give rise to an experience as of the color of Earth’s grass and Inverted Earth’s ripe tomatoes. Your twin’s green-detector will maintain this intentional identity, bestowed upon it by evolution; and thanks to the inverting lenses, it will be activated at the sight of Inverted Earth’s red grass. Even if the semantics of your twin’s language were in sync with Inverted Earthers, the intentional isomorphism would be incomplete: When you and your twin view your respective tomatoes, thanks to the inverting lenses, you undergo the same qualitative states; but thanks to the way evolution fixes the semantics of phenomenal representation, the relevant intentional contents are also the same. For all the example shows, then, intentional properties do fix phenomenal ones.
To engineer a case which would convince evolutionary semanticists, one in which the qualitative constancy is accompanied by the right intentional variation, one might try to contrive a being unconstrained by phylogeny, a being, say, who comes into existence by accident, unconnected with any species, but who is physically type-identical to one of us. Block (forthcoming, section 4), has done this (see also Rey 1998, sec tion 2.4). He suggests that evolutionary semanticists will have to admit that its phenomenology would be fixed by its local physical structure. So if it were fit with inverting lenses and sent to Inverted Earth, its brain states would be relevantly type-identical to ours. And since it has no phylogeny to muddy the water, even evolutionary semanticists should admit that the qualia it enjoys upon viewing Inverted Earth’s red grass are the same as those we enjoy viewing Earth’s green grass.
Block’s argument will not impress Dretske (1995), who bites the bullet and argues that a being without a phylogeny is without not only qualitative content, but any content. In my (1998), I take in this phylogenetic orphan and argue that Dretske’s once-an d-for-all evolutionary determination of content is too rigid. Why shouldn’t a green-detector become a red-detector through adaptation? And once we admit this possibility, we allow that phylogeny won’t prevent twins’ intentional contents from undergoing fu ll inversion eventually — which I am prepared to allow. The real question, then, is about whether this brings phenomenological inversion in its train.
But there are empirical worries other than these. Other philosophers have attacked the possibility of behaviorally undetectable inverted spectra on the grounds that the asymmetric structure of phenomenal color space prevents a mapping of one set of hues onto another appropriately. The degrees of freedom for inversion are significantly constrained by the facts about this space, such as that unique hues (green, blue, red, and yellow) must be mapped onto unique hues, and composite hues (e.g., orange) onto composite hues to prevent behavioral detection. It is natural, as we have seen, to suppose that the mapping might be determined by switching hues for their complements (red for green, yellow for blue, etc.). But the number of hue steps between yellow and green, 18, and red and blue, 31, differ.[ 6 ] So a complement-flip-flop will not work.
There are also the familiar asymmetries in reports of color experiences: Red and yellow are experienced as ‘positive,’ ‘advancing,’ ‘warm,’ and green and blue as ‘negative,’ ‘receding,’ and ‘cool.’ Unless green is experienced as warm after inversion this will give rise to different reports, and likely different patterns of inference,[ 7 ] which the requirements on effective nonintentional inversion rule out. Block (1990) suggests that compensatory inversions of such experiential qualities as warm and cool might iron out this wrinkle. But this move suggests a horrible possibility: if all qualitative content can vary independently of intentional content in this way, we should be able to imagine beings with pleasure and pain qualia inverted with respect to ours — beings who prefer not to be bludgeoned, even though bludgeonings are associated with the qualia of experiences we describe as pleasurable, and who prefer backrubs, even though those are associated with our pain-qualia. Is it coherent to suppose that a pleasure and pain qualia inversion might be behaviorally undetectable? At bottom, this worry is that if we grant Block’s stories, we concede that the experiences of hue — and all qualities — are epiphenomenal captives, essentially inert remainders of conscious experience, which can neither be expressed, nor reported, nor influence action in any way.[ 8 ] And this fails to do justice to the potency the qualities of experience seem to ha ve to motivate action, to influence behavior.
These considerations make the idea of the right kind of inversion of our actual qualitative states sound like nonsense, and thus cast significant doubt upon Block’s anti-intentionalism about them.
But, at least for the sake of argument, I propose to allow that these engineering problems don’t cut to the core of the philosophical issues. Shoemaker (1981) has made a good case that this is so. He argued that even if the empirical facts of the actual case render inversion absurd, a sensible counterfactual case of inversion can be made to work. That is, even if inversion is precluded in the human case, there could be beings like us, but whose color experience is symmetrical. And if inversion is possible and qualia aren’t exhausted by intentional contents for them, then why should they be for us?
So I grant that the possibility of phenomenal and intentional properties varying differentially with respect to each other cannot be ruled out on account of the lately mentioned technical details — although I do not grant that it can’t be ruled out by other considerations. I conclude that there is enough room in logical space for Block’s position, and that his arguments thus require other intentionalist rejoinders. To those I now turn.
4. Why Not a Narrow Intentional Theory of the Phenomenal?[ 9 ]
Some philosophers, such as Rey (1992a, 1992b, 1998), Tye (1994),[ 10 ] and Stephen White (1994), think we can have our cake and eat it, too — that contra Block, we can remain true to intentionalism while preserving nonintentionalist intuitions about the internality of experience. The proposed compromise is that while there are intrinsic, internal features of experience, all its features are nevertheless intentional. This mixed view individuates color -terms’ meanings partly by reference to which colors they represent (the externalist component), and partly in terms of head-bound syntactic properties of the vehicles that do the representing — modes of presentation (the internalist component).
These philosophers propose that the internal aspect of one’s color space can be determined by reference to both the above-mentioned linguistic descriptions (of reds as advancing, and so forth), and something like beliefs about the formal relations hues b ear to one another, saturations, and brightnesses, and about what objects of a given color one has experienced before. The position or role of the computational concept |RED| in such a descriptive internal network might pin it down enough to distinguish it from |GREEN|. In effect, these philosophers help themselves to modes of presentation supplied by putative narrow color contents to account for the ways things seem.
In this section, I’ll explicate and argue against the narrow intentionalist position. It won’t be until section 5 that Block’s arguments are effectively countered, and a positive interactivist view developed by reference to which the Inverted Earth and Inverted Spectrum stories can be correctly evaluated.
4.1 Intentionalist Internality?
It is difficult to turn inward to attend to the intrinsic features of our experience of say, a tree, without thereby turning outward to the tree itself. This is what is meant by saying perceptual experience is diaphanous, or transparent to its objects; and it has been remarked by philosophers at least since G.E. Moore (1903: 450). However, many have asked, is it not also obvious that there is some component of experience which resides wholly inside us?
Rey (1998) asks us to imagine a man who was red-green color blind, but ignorant of this fact for years, having pretty well mastered the application of color vocabulary by ‘discriminating instead a property that co-varies with hue’ (ms., p. 15 f.). Rey’s suggestion for this property is reflectance. Suppose the man then learns of his color blindness and undergoes a corrective operation. Surely, Rey suggests, our man can attend to the difference in his experience of the tree before and after treatment; he will notice an apparently ‘new property, not so much of the tree (he always knew it was green), but of his experience considered in itself.’
Particularly when we attend in the way I imagined our newly color sighted man … might attend to ‘the way red looks’, we seem to be attending not to what is represented, but rather to the way in which it is represented. The actually representing features recede: it’s, so to speak, the mode of presentation, not the mode of presentation that becomes important. (Rey 1998, section 3.2.3)
This mode, says Rey (1998), ‘is narrow, supervening only on the nervous system of the experiencer’ (ms., p. 17); hence, there’s more to color experience than wide content.
But this is too quick. As Rey knows, hue and reflectance, the property by reference to which our man used to determine color, do not covary. C.L. Hardin (1988) explains that many objects with dissimilar spectral reflectance profiles nevertheless belong to the same hue family (80-81). And objects with different relative spectral reflectance properties may look to be the same color in daylight, but different colors under incandescent tungsten light (47). Not to mention, there are chromatic phenomena such as holograms and color television which do not depend on reflected light; in these cases, there simply is no reflectance to vary with color.[ 11 ] Hence, before and after the operation, the properties signaled in the world differ. What then is to prevent wide intentionalists from explaining the phenomenological difference in terms of the informational difference?
In Rey’s defense one might argue that, on many theories of function, even before the operation, the relevant subsystem of the color blind man’s visual system, C, had the function of picking out color information for the cognitive system in which it is embedded. To call C a reflectance detector is to assume — controversially — that its features taken in isolation would have been the only determinants of its function. But, arguably, since color was the distal property that mattered to the larger system, C was a color-detector.[ 12 ] Sure, when color and reflectance were pried apart, the color blind man ran the risk of misrepresenting, say, red as brown. But this possibility wouldn’t make for a difference in the wide content of color categorizations since the same distal properties are signaled in our man after the operation as were signaled before. Still, something is different; it must be the narrow content.
Stephen White (1994) gives a different argument that narrow content is vindicated by the alleged failure of broad content to account for an internal difference. We can conceive of a subject ‘whose similarity metric is atypical and who sees a significantly narrower band of the spectrum as red than does a normal subject’ (497); the narrow content of the subject’s word ‘red’ is thus different than the broad content of the community’s.
But these sorts of argument defeat a strawman, a naive externalist view about intentional content which fails to recognize that it’s the information-theoretic properties of visual systems that matter to experience, the actual and counterfactual performance of the perceptual system’s discriminatory ability. The wide representationalist need not hold that the property in the world which is represented is all that matters to the quality of experience, as Rey and White suppose. The information picked up by the newly color sighted man does shift, even if both his old and new visual systems count as color detectors. His new one performs differently: He no longer guesses hue on the basis of reflectance; he can finally reliably sort colors in all counterfactual situations, and make distinctions inside the classes specified by his former color categorization abilities.
Consider an analogy. Of two relatively reliable tests for strep throat, the ELISA test works by checking chemically for an antigen, which is a part of the offending agent itself (a bacterium called Group A Streptococcus), while an ASO TITER test checks for antibodies produced in response to the presence of the agent.[ 13 ] Note that it is the use to which the tests are put that makes them strep-detectors rather than antigen- or antibody-detectors. There is significant overlap or agreement in the tests’ results. But there are informational differences due to their different sensitivities: For instance, since there are rare disease states of the body, such as syphilis, which can fool the ASO TITER, it can produce false positive s in some cases in which the ELISA yields true negatives; yet the ELISA might produce more false negatives overall than the ASO TITER, while the ASO TITER yields more false positives; etc. A negative result from an ELISA test carries different information than the same result from an ASO TITER.[ 14 ]
Likewise, two thermometers both measure temperature, but the one mounted on the wall and the one designed for use in brewing are calibrated differently, and so detect different information — one has the ability to make discriminations the other doesn’t. In both these cases, the devices detect the same distal property, but have different information-theoretic properties.
If information-theoretic properties specify the grain and shape of intentional directedness, then detecting the same property in the world need not give rise to experience with the same qualitative character. Determining what distal properties devices like disease tests or perceptual systems have the function of signaling is not just a matter of what information they carry; it’s a matter of the use to which they’re put. But use, as we’ve seen, collapses information-theoretic distinctions. Hence, n o reason has been given to think the newly color sighted man entertains new, nondiaphonous properties, nor have we reason to think we need White’s ‘narrow red,’ since, in such cases, the obviously different qualities of experience can be identified with t he informational differences.
Tye (1994) proposes a different route to the internality intuition. His thesis is that there is something so intrinsic about looking square that something could have that phenomenal look even if that look failed to be the way square things normally looked (163) — and that look is quite independent of the environment (169). To see this, claims Tye, we need only imagine a possible world in which some x looks square without looking the way square things normally look. So suppose a world in which, due to odd atmospheric conditions, triangles normally look square, and squares octagonal. In these circumstances, if x were a triangle, it would be true that x looks square, but false that x looks the way square things normally look; if x were an octagon it would be false that x looks square, and true that x looks the way square things normally look. Now imagine that Ernie is looking at a square on Earth, and the recently imagined counterfactual conditions obtain on Twin Earth, where his twin’s brain is in a state physiochemically identical to his, only he’s looking at a triangle. ‘Intuitively,’ Tye says, having the same brain states guarantees that ‘visual experiences are phenomenally identical’ (168). ‘So, visual experiences have, in one clear sense of the term, narrow intentional contents’ (169).
If this is right, it’s hard to imagine what could account for the square look entertained by the Ernies if not that ‘looks square’ picks out an internal thought constituent, the tokening of which is sufficient to produce that intrinsic look. That rules out wide content as an explanation of the look, but not narrow.
It remains to be seen whether these modal claims about truth-values are correct. Of course it may be true that in the imagined world retinal images produced by viewing triangles are square. But would that mean triangles look square? For the time being I’ll leave the question open, and return to it at the end of section 5.2 below.
4.2 Computation and Intrinsic Looks
Rey and Tye argue that the way to specify intrinsic mode-of-presentation-properties of experience is by reference to the computational roles of module output representations — which are supposed to fix the narrow contents of sensory symbols.[ 15 ] Modular visual processes operate on retinal input in a fixed and autonomous manner (Tye 1994: 165), insensitive to holistic facts, such as the normal appearance of red things (they’re informationally encapsulated).[ 16 ] Rey (1998) explains, ‘to enjoy a red experience is to stand in a certain computational relation to a special, ‘restricted’ predicate in one’s language of thought,’ a module output which enters into processing in characteristic ways (ms., p. 6; cf . his 1992b). The relative simplicity of these outputs explains the feeling that qualitative aspects of experience, like color, are simple, apparently nonconceptual, and hard to describe — like Ernie’s experience of his tomato, with the account of which I began the paper.
This view rests on the empirical hypothesis that there are internal symbols with stable, peculiar, internal causal-functional roles (Rey 1998: section 3.2.4). Tye (1994), for instance, bets this will be true of only representations of things like colors, simple shapes, and relative distances (1994: 165 f.): only these are processed modularly; hence only they can give rise to intrinsic looks. The rest of the looks — the nonintrinsic ones — involve extra-modular processes which operate o n representations of shapes, colors, sizes, and textures, and compare them with background beliefs to produce new beliefs, such as that what’s being seen is a feline.
Narrow intentionalists have available a neat response to the Inverted Earth argument, namely, that the relevant intentional contents entertained by you and your twin aren’t completely different at all. Since the computational role of the relevant representations tokened in you as you look at Earth’s ripe tomatoes is isomorphic with those tokened in your twin upon viewing Inverted Earth’s ripe tomatoes, you share a narrow meaning, and thus both experience the same intrinsic look. This view actually predicts that you should have the same phenomenology. Rey (1998: section 2.5) even claims we have reason to think that your Inverted Earth twin’s intrinsic red-experiences will seem exciting and advancing. After all, fixing such computational facts as are definitive of intrinsic red might automatically instantiate the causal powers which bring about these introspectible effects. If so, it might be claimed that narrow intentionalism explains the laws of perceptual psychology which subsume these facts about color experience.
This point might be criticized by evolutionary semanticists. If the twin is your Earth blood-twin, then evolution explains the role of the red-detector’s outputs in cognition: Red is experienced as advancing and exciting, say, because in the evolutionary context it was associated with fire, blood, and danger. But if the twin shares an evolutionary history with Inverted Earthers, there is no non-question-begging reason to think your twin’s intrinsic red-experiences, rather than green-experiences, would be like yours in these respects. Since green played the corresponding role in that evolutionary context, thanks to the inverting lenses, your twin would instead experience the color of Inverted Earth grass as exciting.[ 17 ] On Rey’s view, however, these historical facts at best set up narrow content by helping fix computational roles; they aren’t constitutive of it.
The narrow intentionalist’s response to Block’s Inverted Spectrum argument is less clever, but palatable: Deny it is possible. The identity of the computational roles of sensory symbols before and after inversion guarantees identity of phenomenology.[ 18 ]
The computational take on intrinsicality holds out other possibilities, too. For instance, Dennett is fond of claiming there is no fact of the matter as to whether phenomenal content itself shifts or just our reactions to it. He (1988: 526 ff.) discusses a case in which professional coffee tasters dispute why they no longer enjoy the taste of Maxwell House: Has time taken its toll on the perceptual machinery (so it no longer tastes like it used to)? Or have their reactions to that very taste become more sophisticated? Dennett maintains that the question is misconceived, since it presupposes a false distinction between the apprehension of an inner presentation, and a judgment about it (cf. Dennett 1991: 392). However, the narrow intentionalist can insist that there is a fact of the matter: Either the taste module now outputs different types of representations than it used to, or changes elsewhere in the cognitive system have caused the changes in judgment, or both.
4.3 Phenomenology and Computation
The narrow intentionalist theory of qualia depends on a dubious harmonization of computation with phenomenology. We have no reason to believe consciously accessible phenomenological content will be so much as a good index of representations’ computational roles. It would be unsurprising if, from the phenomenological point of view, we are unable to discern a perceptual deliverance which is the direct output of a module[ 19 ] from one which has been operated on by intervening computational comparative processes, and only then becomes consciously accessible.[ 20 ] It’s a truism of cognitive psychology that many mental operations are cognitively impenetrable by consciousness. So the phenomenological and computational might turn out to be different levels of organization.
What we know about other higher level sciences, such as information science, suggests that such relations between levels of organization are ubiquitous.[ 21 ] Just as the multiple realizability of higher levels of organization allow s informational structures to be implemented in multiple ways, it might permit the phenomenological content, x looks red, to be tokened without any systematically corresponding computational token. And just as the context dependence of systems’ lower level features allows syntactically type-identical objects or properties to implement different informational properties, lower level computational features might implement different phenomenological properties on different occasions.[ 22 ] And since one cannot insist a priori that phenomenology be mapped onto computations, we have yet no reason to think phenomenological judgments involving ‘looks red’-type expressions will be telling of computational underpinnings.
If Tye and Rey insist that intrinsicality is a function of computational details, they must accept the counterintuitive result that, since we may often be unaware of computational facts, we ourselves might be unaware of how things intrinsically look to us. And to avoid this conclusion by assuming that module output representations supply modes of presentation in the phenomenological sense is to beg the question.
Of course, that there needn’t be a clean mapping between computation and phenomenology does not imply that there isn’t one. Even if we knew there were, the data we presently have merely hint at how a completed computationalist account would gerrymander experience into phenomenologically intrinsic and non-intrinsic features. Maybe looking red, square, and far away will turn out to be processed modularly; and maybe looking feline and expensive won’t. If language turns out to be processed modularly for the most part — as it’s widely believed to be — then the narrow intentionalist will count complicated things like the logical forms of sentences as intrinsic features of experience. The modularity of word recognition would imply that the facts that ‘fuse’ is a word, but that ‘fghes’ is not a word are intrinsic features of experience. Such pills are not too bitter, and I see no reason why narrow intentionalists shouldn’t accept them.
But it could also turn out that something’s looking red is no more intrinsic than its looking feline. Consider some of the ways in which extra-modular information about context often matters to experienced color. Just as we experience objects as having constant shapes in spite of flux of the retinal image (due to shifting perspective or saccadic movements), we experience color constancy of objects in spite of the chromatic flux on the retinal image (caused by shadows, refraction, and a host of other variables). It has been remarked that a piece of coal in sunlight appears darker than a piece of white paper in the shadows, even though the coal sends more light to the eye (Hardin 1988: 83; 1990: 561). A red piece of paper folded so as to cast a shadow on itself is perceived as uniformly red, even the shadowed region corresponding to the dark gray patch on the retinal image.
But there are also experimental data suggesting that color phenomenology is fixed by more than just straightforward operations on retinal input, and may involve holistic cognitive factors.
Consider the phenomenon of memory color, in which subjects tend to see the colors they expect to see. G.S. Wasserman (1978: 110) reports that subjects see stereotypically green shapes as greener than shapes without a stereotypical color. Subjects asked to match test shapes of the same green color with target color samples of varying saturation match tree shapes rather than star shapes to the target green with the highest saturation — even though all the shapes are cut from the same paper. J. Beck (1972: 147-148) reports that subjects see orange shapes of characteristically red things, such as apples and hearts, as being closer in color to red than orange shapes associated with no characteristic color, even when they are cut from the same ora nge paper.[ 23 ] The ‘locations’ of objects in color space appear to be at least weakly sensitive to noncolor categories into which they fall.
Further, Eleanor Rosch (1975) has shown that the internal structures of color categories are organized around natural prototypes, and that more prototypical members of categories are recognized more quickly than poor category exemplars. Presumably reaction times are longer when representations of nonprototypical category members are being processed because classifying these requires more comparisons with representations of less probable instances or features associated with the category. Rosch also found that antecedent knowledge of category names significantly affects performance — indicating that the process of color recognition is cognitively penetrable, and to that extent, less modular. So colors might not even be candidate intrinsic qualities in the computationalist sense anyway.
Psychological data tell against intrinsic perceptual features in other sensory modalities as well, such as audition. For instance, phoneme perception, although categorical, seems to be sensitive to extra-modular visual information elsewhere in the cognitive system (H. McGurk and J. MacDonald 1976; D. Massaro 1987; D. Dekle et al. 1992), and higher level information about semantic context (R. Warren 1970; A. Samuel 1991).
We thus have reason to doubt that computation plays the role narrow intentionalists want it to in the production of color (and other) experience — thus, to doubt that modularized outputs are to be seen as equivalent to phenomenologically salient re presentations, internalist modes of presentation.
Block has his own argument against the narrow content view of qualia, and it’s worth mentioning. According to Block, narrow content about color is constituted by descriptive beliefs sufficient to individuate each color from the rest; but the right kind o f amnesiac would lack these beliefs, yet could experience color qualities; hence, narrow content cannot account for color experience.[ 24 ] One might even argue that, quite apart from the amnesiac possibility, you, your Twin Earth tw in, and an Inverted Earth twin without inverting lenses, all share narrow color contents. After all, if Inverted Earthers experience green as advancing, and so on, then the narrow causal role of the state your Inverted Earth twin occupies when look ing at local (green) tomatoes can be construed as narrowly type-identical to that tokened in you when you view local (red) tomatoes. That these are nevertheless different color experiences is shown by the fact that were you to view your twin’s garden, you would remark the color differences, and vice versa. So, if there is a phenomenal difference, it is not captured by narrow content. Block is right here: Inverted Earth shows that a qualia difference couldn’t consist in an internal functional difference.[ 25 ] If that’s what narrow content supervenes on, then it fails to do the trick.
White (1994) suggests that a way out might be found if appeal be made to an adaptation of narrow content which allows ‘discriminative skills and capacities,’ and not just internal functional roles, to help individuate contents. This gets things backwards though. It implicitly assumes that certain dispositions — nonintentionally characterized — can be said to be discriminative of something. But the way to individuate a clump of dispositions as a skill-for-discriminating-red is by reference to what in the world it detects, red — which adverts to broad content. In the next section, I shall argue, contra Block and the narrow intentionalists, that nothing short of adverting to our color-detectors-in-context will suffice to give the content of color experiences. While representations in the head may help implement qualitative content, such content isn’t specifiable — even in part — by reference to properties of the head, but requires reference to what in the world is intend ed, represented, signaled. So stop the sun and start up the Earth.
5. Interactivism and the Phenomenal
The interactivist picture of how informational links between organism and world support qualitative content shall provide a more subtle account of the relations between the phenomenal and other levels of organization than do either Rey’s and Tye’s computa tionalism or Block’s physiochemical view; there are also more empirical data which support it. The visual system keeps color intentionality up to date by adaptation to contingencies in the environment thanks to heterogeneous updated links between organism and world, at many levels of organization, and on many time-scales.
After situating my view in relation to a complementary program in psychology, I give my promised replies to Block, which, it turns out, require an interactivist theory of color memory, the outline of which, finally, I also provide.
5.1 Pure Experience of A World
The theoretical background against which narrow intentionalist arguments are most plausible is constructivism, according to which perception is so rich thanks to rich models of the environment built up by cognitive processes on the basis of a relatively impoverished sampling of stimuli. Helmholtz (1925), for instance, held that even the perception of objects as single bodies requires elaborate construction: ‘It is only by comparing the images of the two eyes … that the idea of solid bodies is obtained’ (23). It’s hard to get Humpty together again; on this sort of view, the cognitive resources allotted for perceptual processes are spent on performing these comparisons so as to reconstruct a picture of the world. It is thus tempting for constructivists to suppose that what consciousness beholds is a mode of presentation, a world-model, an epistemic mediator, ‘one of the links in the cause-and-effect chain that binds the perceiver to the distal cause of his percept in the world’ (Shaw and Bransford 1977: 12).
William James (1904a) remarked that these
theories of perception … violate the … sense of life, which knows no intervening mental image but seems to see the room and the book immediately just as they physically exist. (9)[ 26 ]
As Hegel would have it, there is an antithetical research program in the field, usually associated with James J. Gibson, which respects William James’ intuition, namely, ecological psychology. The bald claim of this approach is that perception as such consists in a direct epistemic link to the world — that the distal causes of perception, objects and properties in the world, and not retinal images or internal models, comprise the class of things with which our beliefs, actions, and experiential contents correlate.[ 27 ] An analogy will shed light on how this could be so. Consider the polar planimeter, a device which measures directly, without calculation or internal modeling, the area of any plane figure. It consists of an arm connected to a measuring wheel which skids and rotates as it traces the figure’s boundary such that the total angle of rotation consequent to a complete circuit of the figure is directly proportional to the area of the figure.[ 2 8 ] For this device, measuring qua measuring is not a matter of internal modeling and calculation, but of its interactivity with plane figures. It embodies abstract mathematical principles: It is built so that, given its manner of interaction, it will register that complex property.
Gibson’s insight was his recognition that our perceptual systems are like that: they don’t have to construct all of the properties in the distal environment they were constructed for detecting. Perceptual acts qua perceptual acts are a matter of interactivity with the perceived world, and not essentially of internal modeling. At the level of organization at which such acts occur, there is no epistemic mediation: perceivers don’t perceive internal constructs instead of the world; they perceive the world.[ 29 ] To contort the title of one of James’ essays, in this sense, we have pure experience of a world, unfiltered by internal constructs.
But that there is no epistemic mediation does not mean there is no causal mediation or implementation. Of course there are chemical, physiological, maybe computational, and other physically mediating events at lower levels of organization o n which perceptual experience supervenes. An example due to Shaw and Bransford (1977b) makes the distinction clear: In hearing a Beethoven piano concerto over the radio,
We should no more count the properties of these electromechanical media as an intrinsic part of the perception of the music than we should count the vibratory activities of our cochlear or neural transmissions. (32)
What is heard is the concerto, not the mediators. Experience itself involves seeing through the intrinsic properties of information-bearers to what is intended through them (cf. Harman 1990: 41).[ 30 ]
According to interactivism, the perceptual-phenomenological is constituted as a distinct level of organization due to agent-environment interactivity: What might be called the ‘subjective pole’ of this relation consists in agent skills for picking up the relevant kinds of information; the ‘objective pole’ consists in the adequacy of the environment as a source of information (given the kinds of information gathering skills brought to bear on it). Neither pole is sufficient in isolation. Together they generate a taxonomy of perspectival properties which falls out of the way organisms carve up the environment.[ 31 ]
Interactivism’s emphasis on the objective side of the relation prevents the underestimation of the role of the world in constituting phenomenal experience, of which nonintentionalist and narrow intentionalist views are guilty.
[Theories] of direct perception in no way belittle the richness of perceptual experience…. but they have sought the basis of that perceptual richness not in the elaboration done by cognitive processes but in the richness of the stimulation. (Michaels and Carello 1981, p. 9)
Dennett (1991) is in harmony with interactivism on this point:
When we marvel … at the glorious richness of our conscious experience, the richness we marvel at is actually the richness of the world outside, in all its ravishing detail. It does not ‘enter’ our conscious minds … (408)
Its ready availability to us constitutes the objective pole of consciousness.
Interactivism’s emphasis on the subjective side guards against the tendency of overzealous ‘externalists’ to neglect agents’ capacities. After all, it is agents that have to be able to pick up the information. It is only because of the way our visual systems are built that a high rate of symmetrical outflow of part of the optic array can signal collision for us (Gibson 1977: 73). It is only because our visual systems can represent distal edges that glimpsing top-right and bottom-left edges forming right angles in suitable opposition (and in an appropriate environmental setting) can indicate a door (D. Kirsh 1991: section 2). It is only because we can move our heads from side to side that our visual systems can enhance depth perception by exploiting the fact that the rate of optical flow of points in the visual array is inversely proportional to the distance of the environmental points from the observer; farther points translate more slowly (W. Mace 1977: 47 f.).[ 32 ] In a legion of such cases, the way we are built is what allows us to pick up such information about the distal layout of the environment as we do.
By design or use, our perceptual systems constantly exploit regularities of structures in the environment. Irving Biederman (1995) makes this especially clear in his discussion of the presuppositions of the visual system. For instance,
… a straight edge in the image is perceived as being a projection of a straight edge in the three-dimensional world. The visual system ignores the possibility that a (highly unlikely) accidental alignment of eye and a curved edge is projecting the image. Hence such properties have been termed nonaccidental… On those rare occasions when an accidental alignment of eye and edge does occur — for example, when a curved edge projects an image that is straight — a slight alteration of viewpoint or object orientation readily reveals that fact. (126)
Because of our manner of visual interactivity with the environment, curved edges almost never appear straight (for long).[ 33 ] These examples illustrate that the more our structure presupposes about the structure of the world , the less needs to be computed by the head; and the more our skills make the world’s structure readily available to us, the less needs to be explicitly stored in the head.
The deepest facts we know about color fit nicely into this perspective — namely, that color seems to be the upshot of an animal-environment agreement. Color researcher and founder of the Polaroid company Edwin Land (1977) has admitted that given our current knowledge, ‘there is no obvious physical quantity at a point that will enable us to specify the color of an object’ (115); and given our current technology, ‘only our eyes can categorize the colors of objects; spectrophotometers cannot’ (108). T his curiosity is due to the context-dependence and extreme multiple realizability of colors: ambient energy of the same composition may be perceived as different colors in different contexts, or by different animals; and light distributions composed of light of different wavelengths may be indistinguishable in terms of their color. This latter phenomenon, called metamerism, reflects the fact that for our visual systems, indefinitely many triples of properly adjusted spectral lights &# 8212; each absorbed by one of the three cone types on the retina — can give rise to one and the same color (see Hardin 1988: 26-28).
Colors are complicated distal properties in the environment; but they’re not intrinsic to objects taken in isolation from any observer class, as properties couched in terms of objects’ objective physical features would be. There are no plausible candidate objectivist counterfactual supporting definitions of any color, such that all and only surfaces of objects which have the (possibly disjunctive) physical property specified by the definition are that color (Hardin 1988: 61-65). The bunches of objects which share their colors are curiously gerrymandered by reference to common effects they produce in us, which we call ‘colors’. As White (1994) nicely puts it, ‘the notion of fool’s gold is clearly intelligible, whereas the idea of an object’s being fool’s red — that is, of its being indistinguishable in color from red objects under normal circumstances without being red — is less obviously coherent’ (472). Again, this is not to say the colors exist merely in us. Things really do have their colo rs; it’s just that colors are context-dependent properties which emerge from the interactivity of the systems we have for detection and the information light provides us about objects.
The groundwork for a reply to Block is now laid; but before going on I want to suggest how the interactivist intentionalist account, when extended to qualitative aspects of experience quite generally,[ 34 ] provides a satisfying answer to Thomas Nagel’s (1974) what-is-it-like question. Nagel argued that there are some facts that exhibit point-of-viewiness which will never be captured by an objective physicalistic scientific theory; hence, materialism is false as a general theory about the world. If, however, the properties of experience emerge as a result of the interactivity of the organism’s discriminatory capacities and the world, then it should be no surprise that there are facts to which epistemic access cannot be gained by beings not constituted in the right way — by beings whose perceptual systems lack the relevant information-theoretic properties.
Information-theoretic differences are a commonplace. First, different organisms often segment the same informational continua differently. Not all color sighted animals carve up the wavelength spectrum the same way. Data suggest, for instance, that pigeons make hue discriminations within what constitutes a human hue category, yellowish-green (G.H. Jacobs 1981: 117 f.). And Michael Platt et al. (1996) showed that ecological considerations for different mammalian species, such as feeding patterns, were goo d predictors of performance in color memory tasks — their visuospatial memory abilities were specialized for tracking the spatial and color properties of their principal foods. Second, organisms are often sensitive to different kinds of information. Bats tap sound waves for information about the layout of the distal environment. Dogs can’t make the color distinctions we can, but many of them can pick out location by sound better than we can because of enhanced high frequency hearing and the design of their reshapeable ears. Finally, Dretske (1995: 95) points out that even if we did know what it’s like in a certain respect to perceive using bat sonar, there are many other (perspectival) properties, simultaneous awareness of which constitutes bat experience; and without appreciation of the whole mix, we fail to know what it’s like.
The interactivist might even agree with Nagel that knowing all the nonperspectival physical facts about the experience of, say, bats, will leave out what it’s like to be a bat. But that does not mean that all of the information-theoretic properties which comprise bat experience (and only those properties) leave out what it’s like. Just as the language of physics can’t capture properties referred to in, say, biology or economics (see Fodor 1974), such information-theoretic properties as humans are able to express might not be enough to allow expression of properties of bat experiences.[ 35 ] But interactivism shows that this does not refute materialism or intentionalism in any case.
5.2 Reply to Block
Internalists claim to have established that there are properties of experience constituted by purely inner factors, and that arguments to the contrary either fail or are unmotivated. My purpose in the rest of this paper is to show not only specifically ho w wide intentionalism about qualia is possible, but that it is well motivated, too — better than it competitors. I won’t be arguing that there could not be relevant internalist properties of experience, but rather that internalists don ‘t have a good attack on my independently motivated picture. And since my arguments lend wide intentionalism about qualia more plausibility than it has generally been granted, they are not merely defensive.
In this section, I’ll programatically stake out an interactivist position in logical space by means of interactivist reinterpretations of the Inverted Spectrum and Inverted Earth stories, distinct from nonintentionalist and narrow intentionalist interpretations. Then, in section 5.3, I’ll fill-in a theory which explains how this sketch could be right — how color phenomenal content could emerge from organism-environment interactivity, and how it might be that the thesis of the local supervenience of the phenomenal on the functional or physical is mistaken, contra our cast of internalists.
Recall Block’s 4-stage Spectrum Inversion. Stage 2 begins with spectral inversion, and in stage 3 the subject adjusts linguistic and other dispositions to the new look of the world; in 4, the subject forgets the relevant facts about stage 1. Block takes it that qualitative aspects of experience are switched while everything else is the same before and after. Interactivism suggests that this telling gets the empirical details wrong, that instead, everything is the same before and after, although there are qualitative and intentional switches in between. During stage 2 and part of 3, interactivity and feedback from the world about the colors of things allows pre-inversion green-detectors to adapt and become red-detectors.[ 36 ] At the point in stage 3 at which the informational properties of the perceptual systems reach isomorphism with stage 1, the subject would experience current perceptions as coherent with stage 1’s pre-inversion perceptions — both as representing fire truck s as red and clover as green. They would be seen as red and green again. By the time the intentional is completely back to usual, so is the phenomenal. Hence, contrary to what is often assumed, the subject would recall stage 2’s perceptions as different than stage 4’s, not stage 1’s.[ 37 ] (So Block’s stipulation that stage 1 memories are forgotten is irrelevant.)
At stage 2, while adjustments to inversion are in the works, things would seem strangely inverted. The color detectors which informed one of green now ‘go off’ in the presence of red (and vice versa), and fixing this takes time. The phenomenological oddity here is that one has to consciously compensate for one’s perceptual deliverances about color, much as one has to consciously deliberate about movements if one plays a video game with the joystick upside down, or attempts to copy a fine drawing while watching the mirror image of one’s drawing hand. But in these cases, if adjustment is achieved, phenomenology unnoticeably returns to normal.
This prediction is empirically supported by research on the ability of perceptual systems to recalibrate their information-conducing properties to the world. Sometimes in nerve damage patients who experience no sensations at the ends of their fingers pas t, say, the first knuckle, nerve tissue is transplanted from the thigh. Interestingly, when these patients are pin-pricked at the finger-tip immediately after the procedure they report sensations as of a pin prick at the first knuckle. Over time, however, the qualitative feels come to seem located where the stimulation is present. Austin Clark (1993) points out that although there is a spot on the retina, the macula, which yellows as age advances — and changes the information about wavelength received by the brain — there is no noticeable change in the quality of color-as-perceived, in qualitative content (170). Finally, in the much-discussed studies on visual field inverting goggles (reviewed in R.B. Welch 1978), the world comes to seem norma l to many visually inverted subjects to the point that they can go about their lives as usual. Subjects who adjusted most thoroughly did not think of their visual fields as upside down. In all of these cases, after recalibration, informational properties are restored or preserved, and with them, experiential properties.
The interactivist response to the Inverted Earth argument, then, is clear: deny Block’s intuition that ripe tomatoes would look red to his twin; they would come to look green, since that’s the color of tomatoes on Inverted Earth. Their looking green to him is just a matter of the information-bearing states in him being linked to the world so as to signal green under what have become normal conditions. There is a similar response available to Tye’s (1994) alleged possible world in which, when beings like us perceive triangles, they normally look square. The coherence of the described state of affairs is itself doubtful. If atmospheric distortions caused triangles to systematically project square shapes on our retinas, after adaptation, those retinal patterns would give rise to experiences as of triangles, not squares. That is what those signals would be informative about.
Interactivist intentionalism about the phenomenal is thus a distinct, competing theoretical possibility. Its correctness would show that Block mis-imagines both the Inverted Spectrum and the Inverted Earth scenarios — that he fails to imagine either all the relevant factors in one possible world, or at least one of the factors in sufficient detail. So, next, I locate the Achilles’ heel of Block’s arguments and provide an extension of and motivation for my account which leaves no doubt about the inadequacy of his case.
5.3 Interactivist Color Memory
Both of Block’s arguments depend on the principle that when subjects don’t notice a phenomenological difference there is none. In his intrapersonal Inverted Spectrum case, the subject reports the difference between stage 1 and stage 2, but doesn’t report any transition from stage 2 to 3; hence, Block concludes, there is no phenomenological change between the latter two. Block himself makes clear where the principle is invoked in the Inverted Earth example in his (1996) elaboration:
Imagine that on the birthday before you leave for Inverted Earth you are looking at a clear blue sky. Your visual experience represents it as blue. Years later, you have a birthday party on Inverted Earth and you look at the Inverted Earth sky . Your visual experience represents it as yellow… But the phenomenal character remains the same, as indicated by the fact that you can’t tell the difference. So, there is a gap between the representational content of experience and its phenomenal charac ter. (43)
The principle assumes that memory will faithfully preserve the colors of things and color relationships.
I shall challenge this assumption and argue for an interactivist view of color memory, on which the accurate retrieval of color memory contents depends on the subject’s relationship to the environment: The subject never notices a change in qualia because of slow compensatory adjustments to the mechanisms which recall memories of how things used to seem.[ 38 ] Agents’ inability to detect these changes in the color qualities of their experiences should be seen in the same light as their inability to detect content switches after they’ve lived on Twin Earth a while.
Block is not impressed by this move, as he thinks it pathetic to trot in externalism about memory of phenomenal character to defend wide intentionalism about phenomenal character. According to Block, intentionalism about phenomenal memory content
says that the memories of representational contents are wrong, so the memories are wrong too. But that is just to assume that as far as memory goes, phenomenal character is representational content. For the argument to have any force, there would need to be some independent reason for taking externalism about phenomenal memory seriously. (1996: 45)
Likewise, Rey (1998) complains that ‘such accounts would need to be independently motivated in order not to seem ad hoc’ (ms., p. 20). What follows is the requested motivation.
5.3.1 Intentionality and Perceptual Memory
The picture I defend of how color memory actually works is based on a model of memory widely accepted among cognitive psychologists. Barbara Tversky (1997) captures some of its relevant essential features:
… [M]emory … is not … like files of information or in this case, fading photographs, of stimuli encoded or representations established of particular events, tagged as such and stored away…. [E]vents are encoded and represented in multiple ways that are not necessarily integrated. At retrieval, there is a search for information and reconstruction of an event constrained by the situation at retrieval and facilitated by inferential processes. (272)
My point shall be that recall of color is no exception: it does not involve consulting internal, stable records, but a process of generation. It’s not too misleading in this context to think of memories of episodes of color perception in terms of computational metaphors — as schematically encoded, encrypted, compressed, zipped, whatever, for efficient storage in bins.[ 39 ] Recalling a memory then requires the right decoder. In this way, color memories are like juice concentrates: For efficient storage, something replaceable is taken out and re-added on an as-needed basis. Needless replication of elements common to many memories would violate sound principles of cognitive economy.
There is evidence that, as the cognitive division of labor has it, the decoder for color information borrows visual color codes from a current coding mechanism, which is hooked-up to the color sensors of the perceptual system. Correct color memories depend on the sensors’ proper calibration. So recall is a ceteris paribus affair. If nothing has changed, then when memory reconstruction mechanisms reverse the process of encoding, one re-entertains that same content. If encoding rules have been re-written, recall will undergo a corresponding shift. In the Inverted Earth and Inverted Spectrum cases, I claim that feedback from the world into the visual system results in such a retuning of the coding mechanisms.
This theory of color memory does constrain possible cognitive architectures, but, in many respects, in ways we already know they’re constrained, as I shall now show. (So this is not the ad hoc special pleading on behalf of color memory against which Bloc k railed.)
Memory is relatively fragile. Psychologists have long known about what’s called retroactive interference with episodic, or event, memory, which occurs when later experiences affect a distortion of an original memory, as when subjects are fed misinformation which leads them to reconstruct memories incorrectly. Psychologists have become pretty good at implanting false event memories about, for instance, being lost in a shopping mall (E. Loftus and J.E. Pickrell 1995), or knocking over a punch bowl at a wedding (I.E. Hyman et al. 1995).[ 40 ] Memory for imagery and color is no exception. Psychologists have been documenting subjects’ tendency to inaccurately ‘fill-in’ imagistic aspects of remembered experiences for a long time (see F. Bartlett 1932). Elizabeth Loftus’ (1977) research on shifting color memory shows how easily color properties of event memories can be tampered with: Witnesses to a staged automobile accident in which a green car crashes can be made to ‘remember’ that it was blue.
It might be thought that these subjects must have had vague memories, but that others, whose memories were more vivid would perform better. But Friderike Heuer et al. (1986) have shown that people who report having vivid visual memories perform less accurately on tasks of color memory than others. And John Yuille and Judith Cutshall (1986) have shown inaccuracies in color memory in a real-life case of eyewitnesses to a murder, even when most other aspects of memory were highly accurate.
Color memory also degenerates relatively quickly. Charles White and David Montgomery (1976) showed that people begin to lose the ability to remember colors as soon as color stimuli are removed: When asked to adjust a colorimeter to match afterimages that resembled either a red-white-and-blue American flag or a simple striped pattern, subjects matched the afterimages of stripes in a flag pattern as redder than afterimages of the same stripes in a simple striped pattern.[ 41 ] Especially interesting, given my appeal to information-theoretic properties, is the recent work of Dominic Massaro and Geoffrey Loftus (1996) on the degeneration of vivid iconic memories; they claim that the identity of the informational and phenomenological properties of these memories is suggested by the precise coordination of their decay.
Still more evidence for the possibility of color shifting in memory comes from research which shows that color aspects of memory are in fact separable, taintable, and less reliable than spatial memory. Patricia Siple and Robert Springer (1983) amassed evidence that color is an independently accessible feature of memory representation, not bound up with shape or texture information. Denise Park and Charles James (1983) found evidence that in many cases spatial information is stored automatically in event memory, although color information is not; and Park along with David Mason (1982) found a superiority of spatial over color memory: color memory was most effortful, and it was at chance level unless subjects were specifically instructed to encode the information.
The explanation for the relative fragility of color memories is that, as emphasized by Tversky above, they are psychodynamic products which draw on various sources for their elements — which, in turn, gives rise to various sources of possible retro active interference.
This is the same general reason why the subjects in Block’s stories may fail to notice changes in the phenomenal appearance of color. In the Inverted Earth case, after adjustment, there is a kind of retroactive interference with memories of how things us ed to seem; and in the Inverted Spectrum case, subjects don’t notice the change (from the oddity of objects perceived during stage 2 to the normalcy of those perceived late in stage 3), because the interference sets in gradually. In both cases, the interference is due to the environmentally-sensitive nature of the memory reconstitution process. Memory presupposes the reliability of many information sources, including a properly tuned perceptual system.
In the Inverted Spectrum, then, stage 2 memories are odd since clover-perceptions will all of a sudden have begun to be coded as red and fire-truck-perceptions as green during that stage. Even after the visual system adjusts to the greenness of clovers and redness of fire trucks, when stage 2 clover-memories are retrieved and unzipped, they will represent clovers as having the complement of their currently perceived color — that was the way their look was coded.
What about the memory of your Earth sky and the look of the sky you enjoy as an assimilated Inverted Earther? Block thinks these will be qualitatively the same (see the Block quote at the beginning of section 5.3 above); but there is no reason to suppose this if intentional adjustment is complete. It is true that you cannot tell the difference between the phenomenal content of your memory of Earth’s blue sky and your current perception of Inverted Earth’s yellow sky. In one sense, your memory represents a blue sky; that’s what was encoded. But, in another sense it does not represent the sky as blue, since, as you now decode the ‘blue’-memory, you can’t help but unzip it as yellow, and thus misremember. Given the connection between color perception and memory, the adaptation of your perceptual system has limited your ability to correctly recall and elaborate the remembered colors of objects. Yours is a memory of a blue sky which represents that sky as being yellow. This is the shade of difference Nelson Goodman (1976) had in mind when he distinguished a picture of a black horse from a black-horse picture, the former expression referring to what is depicted, the later to how it is depicted.
We can now catch Block on the horns of a dilemma, bearing in mind that his arguments sought to show qualia intentionalism wrong on the basis of what phenomenal differences his protagonists notice. Agents either fail in his cases to notice qualitative changes in their color experience, or not. Either way, the intentionalist has the upper hand. To the extent that subjects do notice experiential differences — due, say, to subtle differences in the ways certain colors are perceived (cf. the candidate d ifferences discussed in section 3.1.3) — Block’s conclusion is blocked. To the extent that they don’t notice, I have told a plausible intentionalist story about why they don’t.
For the most part this concludes the reply to Block.
But I want to add a few touches to my intentionalist picture of qualia. In the next subsection, I suggest how the interactivist account extends to cover the qualia of mental imagery. In the final subsection, I broaden the picture in another way, by showing how interactive processes in color perception occur not just over relatively long periods of time (the time scale on which we’ve seen the visual system adjusts its settings), but, ubiquitously, on shorter time-scales as well.
5.3.2 Intentionality and Imagery
Even if we explain the qualia of perception and memory, there are still those of novel imagination. What of them? Incidentally, there is a case for a similar dependence of imagery in general — not just remembered image colors — on perceptual systems. Stephen Kosslyn (e.g., 1995) holds that perceptual recognition and encoding systems can be run in the other direction to form images. There is good evidence that channels and mechanisms used to recognize various visual properties like shape , letters, and color, are also used to imagine those properties. First, a now widely-replicated effect, S.J. Segal and V. Fusella (1970) discovered that visual imagery impairs visual signal detection, suggesting that imagery engages part of the mechanism of the perceptual system. M.J. Farah (1985) has also argued for common mechanisms which underlie both perception and imagery on the basis of other chronometric data. Second, curious properties of perception also show up in imagination, suggesting t hat image inspection is accomplished by some of the very mechanisms used in perceptual recognition. For instance, in perception humans can make out vertical stripes better than oblique ones: as distance increases, oblique ones begin to fade before vertical ones. Surprisingly, the same is true when the stripes are imagined (Kosslyn 1981: 239). In the McCollough Effect, subjects who alternate between viewing red and black vertical stripes and green and black horizontals for a few minutes, and who are then p resented with test patterns of black and white horizontal and vertical stripes, see green on the vertical white stripes and a pinkish red on the horizontal white stripes. Remarkably, R.A. Finke and M.J. Schmidt (1977) showed that if subjects are given gre en and red patches and merely imagine the black stripes, they experience the same effect when presented with the test patterns.[ 42 ] Finally, there are also neuropsychological data supporting the shared resources and mechanisms of imagery and perception.[ 43 ]
In broad strokes, this is the key to a naturalistic understanding of the so-called ‘intentional inexistence’ of mental phenomena. They are intentional because not only perception, but even imagery-generation runs through world-directed perceptual mechanisms so as to confer intentionality — informational content — on what is imaged. Even philosophers’ ‘color patches’ have aboutness in this less obvious way. This also explains why imagination is constrained by the informational distinctions available to perception, and to the dimensions of perception. And these mental phenomena have a kind of inexistence in that imagined objects need not exist. On the one hand, in color mental imagery we ‘see’ color in the absence of the appropriate sensory input; but on the other, the production of such imagery is sensitive to actual sensory input in that it is produced by some of the same mechanisms as are generally responsible for perception in the first place.
This points to the derivative nature of intentional inexistence and the primacy of intentionality as directedness on objects and properties in the world.
5.3.3 Here-and-now Visual Intentionality
I have given an account of the way the visual system adjusts itself to the world over time to keep the intentionality of visual perception up to date. In this section, I point to ways the world plays a role in the here-and-now intentionality of visual perception, as well, by helping synchronize the visual system with the world. Certain facts about visual memory suggest that its mechanisms often ought to be conceived as running outside the head in the here-and-now. The recognition of this interactivity on short time-scales will help place the longer time-scale interactivity we’ve been discussing in proper perspective — as an instance of more general phenomenon.
Consider the asymmetry between recall and recognition. Schematic codes appear to guide reproductive recall of words (W. Kintsch 1970) and imagistic features of experience, like shape and color, so that the more typical an instance of any of these categories, the more reliably and quickly it is recalled (Tversky 1997: 264). In recognition, however, the atypical is better recognized than the typical. Subjects are more reliable and quick at recognizing unique shapes, scenes, faces, and words, than they are at recalling or reproducing them (Tversky 1997). That recall of atypical instances should be most effortful is not surprising, given that recall involves pulling together standardized elements from various sources (e.g., perceptual mechanism s, schematic face and shape models) (see section 5.3.1 above). Since elements of atypical items will be nonstandard, and thus not included in the set of on-board schematic data usually available to memory reconstitution processes, recall of them will often be formidable. In recognition, however, the world is the mechanism that plays this role — that stores the data used by reconstitution processes. In these cases, the world, as Rod Brooks (1991) says, is its own model. What better way to cognitively represent something too obscure to remember than to have it right there in front of your eyes?
A second instance of the efficiency of reliance on the world — resulting in asymmetry between recall and recognition — comes specifically from the case of color recollection. Dana Ballard and his colleagues (1998) showed that people rely on the world for incremental access to information about color whenever possible, and prefer not to store it in on-board memory. In effect, the world becomes an extension of short-term color memory, part of a system of deictic representation. Subjects were given a task in which, on a computer screen divided into three sections, they had a simple model made up of variously colored tiles, a work space in which they were to promptly reconstruct the model, and a resource area from which to draw tiles. Using an eye-tracking device, Ballard and his colleagues found that subjects often consult the model for information about tile color even when they remember the location information f or that tile. This indicates that fixations on the tile in the model area can provide access to location information without color information. Other times, subjects remember color information and fixate solely for location information (section 3.1). While they could store both forms of information, they prefer to let the world model itself — presumably because, in spite of temporal cost (the task takes more time with more fixations), carrying the information in working memory is expensive compared to acquiring it serially on-line (section 3.2).
In another experiment, Ballard’s team changed the color of a tile in the model during a saccade to that area from the workspace (section 3.3). This had a minuscule effect on fixation time (50 msec.), even when that tile was the target of the saccade. In another condition, the change was made during a saccade from the resource area to the model, indicating that the subject needed location information. Fixation time increased by 129 msec.; but most subjects were unaware of the change, and fixated a nearby tile of the right color for its location information, indicating their reliance on the model for location and color information. Ballard et al. conclude that the color of tiles in the model area is not retained in visual memory from previous fixations. This suggests minimalism about internal representations of distal layouts — that perception is task-oriented, involving minimal processing of task-irrelevant features of the world. There is less we take with us from experience and more that we leave in the world than we might have guessed.
Yet conscious awareness of the world and its properties seems rich, full, seamless, and integrated. It seems that everywhere we direct our attention, there is a rich representation awaiting us. Any time we want to attend, say, to the hue of an object, it ‘s there available to us, on demand. Ballard et al. suggest a metaphor which sheds light on the apparent inconsistency between our rich, full experience, and minimalism about inner representations of the layout of the environment. It appears to application users at computer workstations that infinitely long programs can be seamlessly run; but system programmers know that chunks of the program in peripheral storage have to be shuffled in and out of memory as needed, which is not a seamless process f rom the programmer’s point of view. Both points of view are correct, and reflect different levels of description of the same system. Maybe our conscious awareness is like that: it simply may not have access to the synthesizing events which the body performs at short time scales, like that at which eye fixations take place — the scale at which ‘the human sensory-motor system employs many creative ways to interact with the world in a timely manner’ (section 5).
I have argued that human agents rely on the world for their color perception and color memory over the short and long hauls. From a perspective which recognizes the general phenomenon of interactivity at various time-scales, the entire human color-detect ion system — considered on the time scale of auto-calibration earlier discussed — might be conceived as a kind of deictic device for representing color. After all, it keeps representing these colors by adaptive tuning to them. This essential thatness of the color detection system — its pointing beyond itself — makes it a model of diaphaneity. Color qualia, then, are exhausted by aboutness. In this way, consciousness of the visually perceived properties I have discussed is purely diaphanous.[ 44 ]
It might be thought that this constitutes an elimination, or at least a reduction. William James (1904a) apparently thought this; and he embraced it:
‘consciousness,’ when once it has evaporated to this estate of pure diaphaneity, is on the point of disappearing altogether. It is the name of a nonentity … (4)
I agree with James that consciousness is not a reifiable thing, a pure Ego, but do not on that account take consciousness to be nothing. It is a phenomenon, a process essentially constituted by subjective and objective poles: It draws its being from objec tive things and properties, on the one hand, and information-theoretic properties of (various systems of) the agent, on the other.[ 45 ] As Sartre said, ‘Consciousness is born supported by a being which is not itself.’ Its bei ng is negated only if its support is.[ 46 ]
It’s open to noninteractivists like Block and Rey to find ad hoc ways of accommodating the data which support color interactivism. But so far as I can see, there is no good reason, apart from the bald claim of either nonintentionalism or narrow intention alism, to think an inverted subject or a subject in an inverted world would not adjust to see things as they are.[ 47 ]
Anderson, M.C. and J.H. Neely, 1996: Interference and Inhibition in Memory Retrieval, in E. Bjork and R. Bjork (eds.), 1996, pp. 237-313.
Bartlett, F.: 1932, Remembering, Cambridge: Cambridge University Press.
Beck, J.: 1972, Surface Color Perception, Cornell University Press, Ithaca.
Berlin, B., and P. Kay: 1969, Basic Color Terms: Their Universality and Evolution, University of California Press, Berkeley.
Biederman, I.: 1995, Visual Object Recognition, in S. Kosslyn and D. Osherson (eds.), 121-165.
Bjork, E. and R. Bjork (eds.): 1996, Memory, San Diego: Academic Press.
Block, N. (ed.): 1981, Imagery, MIT Press, Cambridge, Massachusetts.
Block, N.: 1990, Inverted Earth, Philosophical Perspectives 4, Ridgeview Publishing Company, Northridge.
Block, N.: 1994, Qualia, in S. Guttenplan (ed.), Companion to the Philosophy of Mind, Blackwell, Oxford, 514-520.
Block, N.: 1996, Mental Paint and Mental Latex, in E. Villanueva (ed.), Philosophical Issues 7, Ridgeview Publishing Company, Northridge.
Block, N.: forthcoming, Is Experiencing Just Representing?, Philosophy and Phenomenological Research.
Block, N. and J. Fodor: 1972, What Psychological States are Not, Philosophical Review 81, 159-81.
Brentano, F.: 1874/1973, Psychology from an Empirical Standpoint, A. Pancurello, D. Terrell, and L. McAlister (transs.), Humanities Press, New York.
Brooks, R.: 1991, Intelligence Without Representation, Artificial Intelligence 47, 139-159.
Buck, R.: 1993, What is This Thing Called Subjective Experience?: Reflections on the Neuropsychology of Qualia (Special Section: Neuropsychological Perspectives on Components of Emotional Processing), Neuropsychology 7, 490-499.
Burge, T.: 1979, Individualism and the Mental, in P.A. French, T.E. Uehling, and H.K. Wettstein (eds.), Midwest Studies in Philosophy, vol. 4, University of Minnesota Press, Minneapolis, pp. 73-121.
Burge, T.: 1986, Individualism and Psychology, Philosophical Review 95, 3-45.
Burge, T.: 1988, Individualism and Self-Knowledge, Journal of Philosophy 88, 649-663.
Churchland, P.S. and P.M.: 1983, Stalking the Wild Epistemic Engine, No?s 17, 5-18. (Pagination to reprint in W. Lycan (ed.): 1990, Mind and Cognition, Basil Blackwell, Oxford.)
Clark, A(ustin).: 1993, Sensory Qualities, Clarendon Press, Oxford.
Cooper, L. and J. Lang: 1996, Imagery and Visual-Spatial Representations, in Bjork and Bjork (eds.), 129-164.
Davidson, D.: 1987, Knowing One’s Own Mind, Proceedings and Addresses of the American Philosophical Association 60, 441-458.
Dekle, D., C. Fowler, and M. Funnell: 1992, Audiovisual Integration in Perception of Real Words, Perception and Psychophysics 51, 355-362.
Dennett, D.: 1988, Quining Qualia, in A. Marcel and E. Bisiach (eds.), Consciousness in Contemporary Science, Oxford University Press. (Citations accord with pagination of the reprin t in W. Lycan (ed.): 1990.)
Dennett, D.: 1991, Consciousness Explained, Little, Brown, and Company.
Dretske, F.: 1995, Naturalizing the Mind, MIT Press, Cambridge, Massachusetts.
Farah, M.J.: 1985, Psychophysical Evidence for a Shared Representation Medium for Mental Images and Percepts, Journal of Experimental Psychology: General 114, 91-103.
Farah, M.J., F. Peronnet, M.A. Gonon, and M.H. Girard: 1988, Electrophysiological Evidence for a Shared Representational Medium for Visual Images and Visual Percepts, Journal of Experimental Psychology: General 117, 248-257.
Finke, R.A. and M.J. Schmidt: 1977, Orientation-specific Color Aftereffects Following Imagination, Journal of Experimental Psychology: Human Perception and Performance 6, 126-139.
Fodor, J.: 1974, Special Sciences, Synthese 28, 77-115.
Fodor, J.: 1984, Observation Reconsidered, Philosophy of Science 51, 23-43.
Gibson, J.J.: 1958, Visually Controlled Locomotion and Visual Orientation in Animals, British Journal of Psychology 49, 182-194.
Gibson, J.J.: 1977, The Theory of Affordances, in R. Shaw and J. Bransford (eds.): 1977a, 67-82.
Goodman, N.: 1976, Languages of Art, Hackett, Indianapolis, IN.
Guttenplan, S. (ed.): 1994, A Companion to the Philosophy of Mind, Basil Blackwell, Oxford.
Hardin, C.L.: 1988, Color for Philosophers, Hackett, Indianapolis, IN.
Harman, G.: 1990, The Intrinsic Qualities of Experience, Philosophical Perspectives 4, Ridgeview Publishing Company, Northridge.
Heuer, F., D. Fischman, and D. Reisberg: 1986, Why Does Vivid Imagery Hurt Colour Memory?, Canadian-Journal-of-Psychology 40, 161-175.
Hyman, I.E., T.H. Husband, and F.J. Billings: 1995, False Memories and Childhood Experiences, Applied Cognitive Psychology 9, 181-195.
Jackson, F.: 1976, The Existence of Mental Objects, American Philosophical Quarterly 13, 33-40.
Jackson, F.: 1982, Epiphenomenal Qualia, Philosophical Quarterly 32, 127-136.
Jacobs, G.H.: 1981, Comparative Color Vision, Academic Press, New York.
James, W.: 1904a, Does ‘Consciousness’ Exist?, Journal of Philosophy, Psychology, and Scientific Methods 1. (Pagination to reprint in W. James 1971.)
James, W.: 1904b, A World of Pure Experience, Journal of Philosophy, Psychology, and Scientific Methods 1. (Pagination to reprint in W. James 1971.)
James, W.: 1971, Essays in Radical Empiricism and A Pluralistic Universe, R.B. Perry (ed.), E.P. Dutton, New York.
Kintsch, W.: 1970, Models for Free Recall and Recognition, in D.A. Norman (ed.), Models of Human Memory, New York: Academic Press, 331-373.
Kirsh, D.: 1991, Today the Earwig, Tomorrow Man?, Artificial Intelligence 47, 161-184.
Kosslyn, S.: 1995, Mental Imagery, in S. Kosslyn and D. Osherson (eds.), 267-296.
Kosslyn, S. and D. Osherson (eds.), Visual Cognition: An Invitation to Cognitive Science, Volume 2, Cambridge, MA: MIT Press.
Kripke, S.: 1980, Naming and Necessity, Basil Blackwell, Oxford.
Kuenhi, R.G.: 1983, Color, Essence, and Logic, Van Nostrand Reinhold, New York.
Lalor, B.: 1997a, It Is What You Think: Intentional Potency and Anti-Individualism, Philosophical Psychology 10, 165-178.
Lalor, B.: 1997b, Rethinking Kaplan’s ‘Afterthoughts’ about ‘that’: An Exorcism of Semantical Demons, Erkenntnis 47, 67-88.
Lalor, B.: 1998, Swampman, Etiology, and Content, Southern Journal of Philosophy 36, 215-232.
Land, E.H.: 1977, The Retinex Theory of Color Vision, Scientific American 237(6), 108-128.
Loftus, E. and J.E. Pickrell: 1995, The Formation of False Memories, Psychiatric Annals 25, 720-725.
Loftus, E.: 1977, Shifting Human Color Memory, Memory and Cognition 5, 696-699.
Lycan, W. (ed.): 1990, Mind and Cognition. Blackwell.
Mace, Wm. M.: 1977, James J. Gibson’s Strategy for Perceiving: Ask Not What’s Inside Your Head, but What Your Head’s Inside Of, in R. Shaw and J. Bransford (eds.): 1977a, 43-65.
Marr, D.: 1982, Vision, W.H. Freeman, New York.
Massaro, D.: 1987, Speech Perception by Ear and Eye: A Paradigm for Psychological Inquiry, Erlbaum, Hillsdale, NJ.
Massaro, D. and G. Loftus: 1996, Sensory and Perceptual Storage: Data and Theory, in E. Bjork and R. Bjork (eds.), 67-99.
McClamrock, R.: 1995, Existential Cognition: Computational Minds in the World, University of Chicago Press, Chicago.
McGurk, H. and J. MacDonald: 1976, Hearing Lips and Seeing Voices, Nature 264, 746-748.
Merleau-Ponty, M.: 1962, Phenomenology of Perception, C. Smith (trans.), Routledge and Kegan Paul, London.
Michaels, C.F. and C. Carello: 1981, Direct Perception, Prentice-Hall, Englewood Cliffs, NJ.
Moore, G.E.: 1903, The Refutation of Idealism, Mind 12, 433-453.
Nagel, T.: 1974, What is it Like to be a Bat?, Philosophical Review 83, 435-50.
Park, D.C. and C.Q. James: 1983, Effect of Encoding Instructions on Children’s Spatial and Color Memory: Is There Evidence for Automaticity?, Child Development 54, 61-68.
Park, D.C. and D.A. Mason: 1982, Is There Evidence for Automatic Processing of Spatial and Color Attributes Present in Pictures and Words?, Memory and Cognition 10, 76-81.
Peacocke, C.: 1983, Sense and Content: Experience, Thought, and Their Relations, Oxford University Press.
Platt, M.L., E.M. Brannon, T.L. Briese, and J.A. French: 1996, Differences in Feeding Ecology Predict Differences in Performance Between Golden Lion Tamarins (Leontopithecus rosalia) and Wied’s Marmosets (Callithrix kuhli) on Spatial and Vis ual Memory Tasks, Animal-Learning-and-Behavior 24, 384-393.
Rey, G.: 1998, A Narrow Representationalist Account of Qualitative Experience, in J. Tomberlin (ed.), Philosophical Perspectives, Ridgeview Publishing Company, Northridge.
Rey, G.: 1992a, Sensational Sentences, in M. Davies and G. Humphreys (eds.), Consciousness: Psychological and Philosophical Essays, Oxford: Blackwell.
Rey, G., 1992b: Sensational Sentences Switched, Philosophical Studies 68, 289-331.
Rosch, E.: 1975, The Nature of Mental Codes for Color Categories, Journal of Experimental Psychology: Human Perception and Performance 104, 303-322.
Samuel, A.: 1991, A Further Examination of Attentional Effects in the Phonemic Restoration Illusion, Quarterly Journal of Experimental Psychology 43a, 679-699.
Sartre, J.P.: 1958, Being and Nothingness, H. Barnes (trans.), Philosophical Library.
Schiff, W.: 1965, Perception of Impending Collision: A Study of Visually Directed Avoidant Behavior, Psychological Monographs 79, 604.
Segal, S.J. and V. Fusella: 1970, Influence of Imaged Pictures and Sounds on Detection of Visual and Auditory Signals, Journal of Experimental Psychology 83, 458-464.
Shaw, R. and J. Bransford (eds.): 1977a, Perceiving, Acting and Knowing: Toward an Ecological Psychology, Lawrence Erlbaum Associates, Hillsdale, New Jersey.
Shaw, R. and J. Bransford: 1977b, Psychological Approaches to the Problem of Knowledge, in their 1977a, 1-39.
Shoemaker, S.: 1981, Absent Qualia are Impossible — A Reply to Block, Philosophical Review 90, 581-599.
Shoemaker, S.: 1993, Lovely and Suspect Ideas, Philosophy and Phenomenological Research 53, 905-910.
Simon, H.: 1981, The Sciences of the Artificial, 2nd ed., MIT Press, Cambridge, Massachusetts.
Siple, Patricia and Robert M. Springer: 1983, Memory and Preference for the Colors of Objects, Perception and Psychophysics 34, 363-370.
Stevens, A. and P. Coupe: 1978, Distortions in Judged Spatial Relations, Cognitive Psychology 13, 422-437.
Strawson, G.: 1994, Mental Reality, MIT Press, Cambridge, MA.
Turvey, M.T.: 1977, Preliminaries to a Theory of Action with Reference to Vision, in R. Shaw and J. Bransford (eds.), 211-259.
Turvey, M.T. and R. Shaw: 1979, The Primacy of Perceiving: An Ecological Reformulation of Perception for Understanding Memory, in L-G. Nilsson (ed.): Perspectives on Memory Research: Essays in Honor of Uppsala University’s 500th Anniversary, Lawren ce Erlbaum Associates, Hillsdale, New Jersey, 167-222.
Tversky, B.: 1997, Memory for Pictures, Maps, Environments, and Graphs, in D.G. Payne and F.G. Conrad (eds.): Intersections in Basic and Applied Memory Research, Mahwah, NJ: Lawrence Erlbaum Associates, 257-277.
Tye, M.: 1993, Blindsight, the Absent Qualia Hypothesis, and the Mystery of Consciousness, in Philosophy and Cognitive Science, in Hookway and Peterson (eds.), Supplement to Philosophy, Cambridge University Press.
Tye, M.: 1994, Qualia, Content, and the Inverted Spectrum, Noûs 28, 159-183.
Tye, M.: 1995, Ten Problems of Consciousness, MIT Press, Cambridge, Massachusetts.
Tye, M.: 1998, Inverted Earth, Swampman, and Representationism, in J. Tomberlin (ed.), Philosophical Perspectives 12, Ridgeview Publishing Company, Northridge.
Warren, R.: 1970, Perceptual Restoration of Missing Speech Sounds, Science 167, 392-393.
Wasserman, G.S.: 1978, Color Vision: An Historical Introduction, Wiley, New York.
Welch, R.B.: 1978, Perceptual Modification: Adapting to Altered Sensory Environments, Academic Press, New York.
White, C.W. and D.A. Montgomery: 1976, Memory Colors in Afterimages: A Bicentennial Demonstration, Perception and Psychophysics 19, 371-374.
White, S.L.: 1994, Color and Notional Content, Philosophical Topics 22, 471-503.
Yuille, J.C. and J.L. Cutshall: 1986, A Case Study of Eyewitness Memory of a Crime, Journal of Applied Psychology 71, 291-301.
[ 1 ] There is room in logical space for a weaker form of intentionalism about qualia on which intentional properties fix qualitative ones, but the possibility is left open that agents in states which differ in intentional properties mig ht still enjoy the same qualia. In section 5, it will become clear why I deny there is any reason to accept this weaker doctrine.
[ 3 ] This extrapolation requires argument I won’t give. For that, see Dretske (1995) and esp. Tye (1995), Chapter 4. Also see Ross Buck (1993), a neuroscientist who argues that the qualia of affective experience are states whose functio n it is to provide the organism with knowledge of internal events relevant to self-regulation. (But cf. fn. 43 below on my hesitancy regarding neuroscientific data.)
[ 7 ] The categories in which things are classified affect judgments about relationships between them. For instance, A. Stevens and P. Coupe (1978) showed that people judge that Reno is east of San Diego (although it is not) because they think that in general Nevada is east of California. The pre-inversion categorizations of hues as advancing and receding would plausibly give rise to subtle inferential asymmetries which would have to remain intact post-inversion.
[ 13 ] Specifically, the ASO TITER requires that a blood sample be mixed with strep, and the presence of strep antibodies causes a typical but mistakable antigen-antibody complex to form, indicating the presence of the offending agent. The ELISA test, on the other hand, requires only a standard throat swab.
[ 15 ] Actually, I extend the idea that narrow content is fixed by computational role to Tye by courtesy. His stated view is merely that physical constitution fixes narrow content; but his appeals to mechanical perceptual processes and module output representations as determinants of intrinsic looks (see below) seem to require the computational reading (unless he thinks what matters is something about the material the output representation is made of — yet I don’t see how it can, especially if we accept multiple realizability). Indeed, if narrow content is determined by something more fine-grained than functional/computational role, it doesn’t seem intuitive to me that it could be intentional content at all — and it’s not clear that anything would separate such a view of qualia from Block’s. See fn. 18.
Since White has not explicitly wedded himself to this computationalist picture, he will be excused until the end of section 4.3 below.
[ 17 ] Of course, one might stipulate that the forces with which Inverted Earthlings had to contend in their evolutionary history were also inverted with respect to Earthlings; but then it gets very hard to maintain that they’re much like us. If we want it to turn out that they experience intrinsic red as exciting and advancing, then we might have to stipulate that what they would call ‘quickly spreading green-hot fires’ did not affect the survival of their species in the ways re d-hot fires affected ours. According to the evolutionary semanticists, only thus could the nuances of color-representations’ computational roles be preserved in the right way.
[ 18 ] Oddly, Tye (1994: 171) allows narrow contents to be inverted here, arguing that all inversion requires is some physical brain state difference. But this doesn’t square with the requirement that computational properties, not neuro logical ones, are what matter to narrow contents. See fn. 15.
[ 19 ] My hunch is that while module outputs behave systematically, they are typically subdoxastic, representing things like edges — and hence not things to which consciousness has the right kind of access.
[ 20 ] On Maurice Merleau-Ponty’s (1962) notion of phenomenological intrinsicality, the body pre-structures for consciousness even such nonintrinsic (so-called ‘non-observational’) properties as looking feline. This is not the co mputationalists’ notion: They must rest the intrinsic/nonintrinsic distinction on facts about whether and when representation-forming processes gain access to extra-modular information.
[ 22 ] The execution of a programmer’s command in a higher-level programming language is multiply realizable with respect to the details of machine code implementation, and, say, which buffers hold which digits. Type-identical chunks of machine code will perform different functions on different occasions, depending on embedding context. Thus, due to variations of all kinds — from hardware configurations to temporary memory locations to current variable values — there’s no re ason why the higher-level features of a program should be straightforwardly mirrored in machine code, much less lower levels of organization, like that of hardware.
[ 27 ] I don’t want to give the wrong impression with my appeals to Gibson. He has rightly been criticized for failing to recognize that detecting stable properties in the world is an information processing problem carried out la rgely by systems in the head; and he underestimated the difficulty of such detection (see, e.g., David Marr 1982: 29-31).
[ 29 ] Turvey and Shaw (1979) suggest that researchers might not be so tempted to think otherwise if they considered ‘perception as the act of representing rather than as a process of using a representation‘ (207).
[ 30 ] Compare Dretske (1995): ‘Conscious mental states … are states that we are conscious with, not states we are conscious of‘ (100 f.), and James (1904a): ‘In its pure state, or when isolated, there is no self-splitting of [the experience] into consciousness and what the consciousness is ‘of” (15).
[ 31 ] By calling these properties ‘perspectival’, I don’t mean to connote ‘unreal’. While they are context-dependent emergents of organism-environment interactions, they’re no less real on that account (see Gibson 1977: 69 f., and esp. 77 ff.).
[ 34 ] I take it that even if we haven’t specific reasons to generalize the account to qualia of, say, each sensory modality, the success of the embeddedness perspective in handling color qualia (which I shall show below), together with it’s illumination of the mental generally, constitutes a deep reason to do so — at least for argument’s sake.
[ 35 ] The significance of this point extends beyond philosophy of mind. It might, for instance, be extended by philosophers of social science into a defense of Verstehen theory, on which understanding of higher-level properties, like oppressiveness, requires having the right subjective experiences.
[ 38 ] I first found this idea in Daniel Dennett (1991: 395-397), who presents it only intuitively; after I began formulating my ideas, I came across a forthcoming (1998) paper by Michael Tye in which he works this out more systematical ly. I have benefited from both, although added much in the way detail.
[ 41 ] This suggests that even the apparent color of an afterimage depends on memory color. But this could be an artifact of memory color during the presentation of the stimulus (see the discussion of memory color in section 4.3 above).
[ 43 ] See, e.g., Farah et al. (1988). I am skeptical of the value of such studies since it’s not clear to me what janglings in brain areas known to process external visual information during the explicit formation of visual images shou ld mean. But it is suggestive. For a review of more cognitive neuroscientific evidence of substantial links between imagery and perception, see Lynn Cooper and Jessica Lang 1996, esp. 138-141.
[ 44 ] I make no pretensions to having given a full theory of consciousness (although I have given a theory of some central kinds of qualitative experience). Following Kripke (1980), I’ll call what I’ve given a picture rather tha n a theory — a sketch of how to approach the problems. A full interactivist theory of consciousness would involve, among other things, analyses of the ways in which integrative processes are interactivist (the lately discussed work by Ballard et al. suggests some ways), the ways in which control of attention is interactivist as opposed to just internal or just stimulus-driven, and the degree of access to tasks carried out by skill-processes.
[ 45 ] Interpreting James on these matters is difficult; but it seems he thought consciousness is merely an external relation (1904a: 16). He downplays the import of the taxonomy which falls out of the information-theoretic properties of the subject’s perceptual and conceptual systems — a taxonomy which specifies the nature of the subjective pole of consciousness. But see his (1904b: 44 f.) for a recognition of and struggle with this.
[ 46 ] I take it that the contrapositive of this point — that when the right kind of interactive agent-environment relation obtains, so does qualia consciousness — rules out the so-called Absent Qualia Hypothesis of a being with all the right functional or even intentional properties but no qualia.