Black and White and Color

Presenter: Kathleen Akins, Simon Fraser University

Commentator: Pete Mandik, William Paterson University



  1. Dr. Bill

    That’s a very interesting article — and while my theory does not explain findings, it is certainly consistent with them. I’m not sure how closely your follow the neurophysiology of colour vision, but there has been a recent dispute over how chromatic contrast is gained by the M – L system. As you know there are two chromatic systems in humans, the newer M-L or ‘red-green’ system and the older S- (M+L) or blue-yellow system. There seems to be good evidence that the Blue-Yellow system is selectively wired, a wiring that creates the opposition between the S and the M + L cones. That was always the story, so that is not so surprising. But there is an open question about the red-green system. Do you get chromatic contrast because there is selective wiring that separates the 2 populations of cells, L vs. M? Or does this occur as a result of chance, random combination, with inevitable differences in the two populations, center versus surround, just as a result of the patchy distribution of the two cone types, the probabilities and chance wiring? I favor this latter ‘chance’ theory and this new result is consistent with it. That is, in the colour blind individuals who do not have an M cone, the L cones will be wired together to make opponent signals. Now, add in, at random, some new L cells, a result of the experiments. This means that the very same wiring would produce an M-L signal because an M-L cell does not require selective wiring. This makes the fact that no new retinal wiring is needed far less surprising. What would also be required, presumably, is some sort of cortical plasticity which makes use of this new chromatic contrast information, which allows the monkey to see the difference between the stimuli. And presumably that is why over a number of months, the MONKEY becomes better at these chromatic discrimination tasks.

  2. Response to Pete Mandik

    First, I want to thank Pete Mandik for an excellent summary of the view—and congratulate him on promotion to Full Professor (YES!!). Barring very small and probably unimportant details, I think that Pete does have down the views of my paper—and this is the case, even though he also believes that the views are true. (Couldn’t resist that: Pete wondered whether he was reading his views onto mine. The answer: no.).

    Let me start with perhaps the only detail that I would have emphasized over Pete’s summary (and I call him ‘Pete’ because we have known each other a very long time) before turning to the specific philosophical questions that Pete asks. If you look at luminance vision alone—and here I mean any sort of luminance system that one actually finds on earth—you are looking at a system that is primitive. It is primitive not because it lacks ‘colour’, but because it lacks ‘black and white’, i.e. intensity contrast. When you look at a black and white photograph, you are presented with a representation of light intensity, and an image with uniformly high intensity contrast across an image gives you lots of information. With enough hardware and enough time, it is possible to figure out most of what the human visual system figures out about the environment. But luminance systems don’t have intensity information.

    There are really several problems with a ‘pure’ luminance system. One is pure range: whether you start with rods or cones, every photoreceptor is responsive to a very narrow spectral range of light (relative to human visible light). Unless you’re the sort of critter that didn’t get any broad-band sunlight to start out with (e.g. deep sea marine critters), that doesn’t give you much to work with.

    The second is that every photoreceptor conflates intensity with wavelength. So you don’t have either intensity information or wavelength information about the image. This means that many of the laws of optics are just not available for use, not in any sort of reliable way. You’d know that you’re looking at a transparent substance if only you could see that substances effect on intensity—that everything within a certain window was reduced in intensity the same percentage. But you don’t know that, because you’ve got the conflation with wavelength mucking up the equation. And so on for the rather huge number applications of the laws of physics that would allow you to reverse engineer your way to understanding the distal scene.

    Finally, you’re also missing ‘colour’. Well, that is really the same problem as the one above—missing wavelength—but with a twist. If you haven’t got wavelength information, you can’t use any of those laws of optics that apply to wavelength, sure enough. But in addition there is information in the image to which your luminance system(s) just do not respond. This is a new piece of information in the neuroscience world. There is a standard assumption that always ruled the vision science, one that essentially discredited wavelength as a good source of information for human vision: wherever you find a chromatic edge, you also find a luminance edge. And you already have the luminance edges. So why bother with the chromatic ones? That view made a lot of sense at the time. Every object that differs in wavelength with its background will also differ in intensity. So an object that stands out in virtue of its ‘colour’ also stands out by being brighter or darker. E.g. A red strawberry is red while it’s foliage is green; but a strawberry is also darker than the leaves. So it would seem sensible to think that a critter can get by with only a luminance system. Of course, if you look at those critters, you will see that in fact, none do—or none that are active in full sunlight. So there were at least two things wrong with that largely a priori argument. Again, luminance is not equal to intensity. And second, no one asked whether, given the actual chromatic and luminance systems that creatures have, for example, that humans have, whether both types of edges would be visible. Answer: No. Some edges would simply go unseen by human luminance vision. (Note: you have to answer this question one species at a time,, as no two luminance or chromatic systems are alike.) In addition, those two types of edges do not usually produce the same signal strength. Even when the luminance signal is present, it may be so low as to be entirely unreliable. In fact, the difference between a monochrome photo of strawberries and a colour print, is a huge difference in visual contrast . Colour gives you both a new source of contrast and a very strong signal. (Strawberries are a bit darker than their foliage, but not much.) Now add that to the luminance contrast signal and you have LOTS and LOTS of ‘visual contrast’. Seeing the strawberries with only luminance vision is usually possible, of course, but unfortunately possibility is a long way from survivability. Take your time eating those strawberries, or miss a significant amount, and well, you don’t want to think about it.

    Any system with chromatic information thus has three different advantages over a ‘pure’ luminance system. And that is why you will not find any single cone critters that are active primarily during the day.

    So, that was ‘information’, now on to phenomenology. Here is what I am thinking. We have no idea—or at least no real knowledge only a few guesses—about which visual representations underlie conscious human sight. Yes, most of what you see is the product of processing we can not yet even imagine. YOU don’t get a front row seat on your own visual processing. But in virtue of some visual processes, we do see various properties in the distal world. Heretofore, the general sense has been that all of these properties are processed without chromatic information. But now we think not. Yes, all chromatic processing except surface spectral reflectance—object colour—could be unconscious. Yes, it is possible that all properties discerned using chromatic information could be ‘rendered’ in black and white. But why would think that? If you are proposing that luminance information enters into consciousness in a way that chromatic information does not, and the two systems do work in the same way, the onus is on you to say why chromatic vision is the ‘silent partner’. You can’t say “Because it just looks that way to me!”. Or well, you can. But what exactly looks ‘that way’ to you? Do you perceive the world as like a black and white photograph with the colour applied on top? No. You don’t. Nor do we have any conscious access to the individual contributions of chromatic and luminance processing to your visual awareness. (If we did, neuroscience would not be nearly as much fun.)

    Let’s say you look at the cat, now lying on the bed, a calico cat with lots of fur colours. You see where the cat and the bed meet (everywhere in the case of a cat). And we know that the processes involved in the visual differentiation of the cat from the bed are both chromatic and luminance ones. Lo and behold, you see that the cat differs in colour from its background. But this is perfectly consistent with the cat’s body differing in terms of two types of chromatic contrast and one or more kinds of luminance contrast. In other words, it need not be an inference based on the colours of the cat and the bed that makes you see the cat and bed as different. Rather, you see the cat as different (low level chromatic and luminance encoding) and then, looking more closely, you see the colour of the cat (a high level conceptual process).

    Actually, I don’t pretend to know how things actually work here, whether it is empirically possible to see, first, the cat as ‘differing’ with its background, and then set the cat’s colour. But certainly the sort of processing involved—their relative difficulties—suggests that this is possible. And given that you do have concepts to keep the world in place, even when you don’t see each of its properties, it is not as if you would notice this difference. It’s not as if you would first notice that the cat was ‘contrast-y’, and then, saw the cat fill in with its’ proper colours. It’s a fact about the world that if the cat differs in contrast, it differs in colour. So the brain can assume this, and not be even vaguely disconcerted that it does not yet know the colour of the cat. The visual brain knows that the cat must have a colour—and this is so even though the only information ‘in hand’ is that the cat is more R than G and darker (it being a calico cat, as opposed to cat dressed up for St. Patrick’s day).

    I should add here that work on cerebral achromatopsia has yielded just this sort of result. (And here let me advertise a conference this summer on ‘anomalies’ of cortical colour vision to be held in Vancouver this summer Colour Conference). Cerebral achromatopsics, who have a normal trichromatic retina but who suffer from cortical lesions, can see shapes and motion defined only chromatically. However, given all conventional tests for colour, they test as colour blind (e.g. the Ishihara plates). Thus it seems that there some actual human subjects in whom chromatic information and colour perception ‘come apart’. The cerebral achromatopsic is able to see certain properties in the world in virtue of colour, but not see the colours—red, green and blue—themselves. More pointedly, the cerebral achromatopsic can distinguish differences in intensity from differences in wavelength. So, for example, such a subject can see that two squares that differ in intensity are different from two squares that differ in wavelength. In fact, this was one of the first bits of experimental evidence that made me wonder about ‘chromatic’ phenomenology.

    Pete Mandik has raised a number of other questions in his commentary that need to be addressed. But I hope that those questions and issues come in out discussion, or in Pete’s response to this response. For now, I hope that this ‘short’ response is enough to get the conversation rolling.

  3. Thanks, Kathleen, for the further remarks. And thanks to Richard Brown for all the work in putting this terrific conference together.

    I want here to continue to press on the issue of what kind of sense, if any, can be made of chromatic phenomenology thought of as separate from color phenomenology. To help put a point on it, I’ll here phrase things in terms of a much stronger opposition between me and Kathleen than reflects reality. Hopefully, drawing out a strong dialectical contrast is of some use to making collective progress on these issues. 

    I’ll press the case here in a series of steps.

    First step, recall the sorts of contrasts I used to articulate the Orthodoxy that Kathleen opposes, contrasts like non conceptual v conceptual, low-level v. High, etc. Kathleen’s unorthodoxy can be swiftly characterized as giving an account of visual phenomenology that eschews the first item in each contrasting pair.

    Second step, consider the kind of theory of consciousness and phenomenology you get when all that’s left over are the second items from those contrast pairs. Moving very quickly here, we can identify two key necessary conditions for some representational content’s appearing in consciousness and thus figuring in a subject’s visual phenomenology. the first necessary condition requires that one posses certain concepts. In order to consciously perceive George as astride a horse, one has to have a concept of horses, a concept of being astride something, and so on. Second, one has to be “wired up” in certain ways so that these concepts are automatically, noninferentially applied as the result of certain external stimulations. The gist of this second necessary condition is to distinguish really seeing the animal as a horse versus seeing it as a large animal and deciding after awhile that it must have been a horse.

    Third step, note how few concepts anyone has that merit being regarded as chromatic concepts. Consider your average poorly educated teenager. For starters, they have concepts of specific colors, probably no more than the number of crayons they had as young children. They might also have a few concepts for comparing colors, likely pretty coarse grained concepts like *bright* and *brighter than*, that I call coarse for they are unlikely to distinguish intensity from saturation. And that’s it. As far as conceptual repertoires go, this is a pretty impoverished one. And even if the teenagers in question read up on the subject, that might give them a few more concepts, but may not suffice to get them wired up in the ways needed to satisfy the second necessary condition in the toy theory of consciousness rehearsed in step two.

    Fourth step, let us wonder out loud about what it would be like to be one of these teenagers under conditions in which either the first or the second necessary conditions weren’t met regarding color concepts. What would it be like to be one of these teens if the teen either didn’t have a concept of redness or didn’t have the concept “wired up” in the above described way? Clearly, no part of their phenomenology includes things being red: things won’t seem red to these conceptually deficient teens. And if one or both of the nec conditions fail across all color concepts, then these people won’t have any color phenomenology. But if color concepts are the only ones they are likely to have that are even remotely chromatic, then these imagined subjects who lack color phenomenology also lack chromatic phenomenology.

    So, let me wrap this up for now and state this as a puzzle. As Kathleen notes toward the end of part 2 of her paper, it is very strange to think in terms of a chromatic phenomenology as opposed to a color phenomenology. But note that it’s even harder for people to think this way if they don’t even have chromatic concepts besides their color concepts. Further, if what I’ve said about the role of concepts in phenomenology is correct, then it looks like we’re forced to say that unless a person has a fully Churchlandish mastery of some pretty sophisticated concepts, then for that person there just won’t be any chromatic phenomenology aside from color phenomenology.

  4. Hi everyone. This is really interesting. But there’s a lot going on, so I thought I’d just say one thing in favor of the ‘orthodox’ view that there are conscious color sensations.

    I guess I’m puzzled, as Pete puts it in his commentary, about what the difference is “between seeing the colors and using information about chromatic contrast.”

    In discussing how one might distinguish a cat from its background, Kathleen (I hope it’s OK if I call you Kathleen–I’m saying ‘Pete’ because I know him too) replies that it’s possible that “you see the cat as different (low level chromatic and luminance encoding) and then, looking more closely, you see the colour of the cat (a high level conceptual process).”

    But on a commonsense view of color sensations, the sensory properties of color sensations are just whatever mental properties enable a creature to discriminate visually objects in respect of their colors. So visually discriminating the cat from its background is enabled by a color sensation–and there’s no extra step needed to see the cat’s color.

    I agree that cases like subliminal perception suggest that color sensations can occur in the absence of consciousness, but that isn’t evidence that they aren’t ever conscious (for example, on higher-order accounts, all it is for one to be in a conscious sensation is to be aware of oneself as being in a state with a sensory property, which plausibly happens).

    And it would seem that, if color sensations just are whatever mental properties they are that enable a creature visually to discriminate objects in respect of their colors, a sensation makes one aware *of* a color, but perhaps not aware *that* something has that color. This is arguably a kind of nonconceptual awareness, so no concepts of the color must be deployed in the experience–as Pete seems to be worrying in his last post.

  5. Response to Pete Mandik, Round Two.

    Well, fortunately Pete doesn’t ask difficult questions.

    The question Pete has asked is this: if you are a good Kantian and think that consciousness requires concepts, exactly what sorts of concepts could we possibly have that give us ‘chromatic’ consciousness? The average teenage knows about red and blue; most teenage males won’t know the concepts for ‘chartreuse’ or ‘teal’ or ‘aubergine’. And if they don’t know that, then how could they possibly know about the sort of information that separates a shadow from an object, luminance edges from chromatic edges?

    First let me be clear about one point, about in Pete’s initial description of my view in his commentary (a difference that didn’t seem to matter before but may well be at the heart of the matter here). I am very leery of the contrasts that Pete has outlined in his commentary—low level versus high level, conceptual versus nonconceptual, phenomenal versus cognitive, and so on. And because I am leery of (some of) them as legitimate distinctions, I wouldn’t want to say that I endorse one half or the other, to say that consciousness is high level but not low level, etc.

    To see the problems of being a Kantian of this stripe, let’s take Pete’s questions about chromatic concepts and apply them to seeing “in black and white” — to the achromat’s experience, an example of conscious visual experience that arises out of ‘pure’ luminance processing. Here the achromat—baby achromats, teenage achromats, and adult achromats—sees many things, and yet, unless the achromat is someone like Knut Nordby (who was an achromat, a physiologist and philosopher), this person is very unlikely to have any of the concepts we use to talk about luminance processing at all. Yet such people clearly see the world consciously: they see dogs and cats, microwave ovens and so on. And what they see is not merely that a cat is present, but the cat in all of its glorious furry detail. But what do they see exactly? How does what they see relate to the information in the system at various levels, information such as ‘this shape is defined by its luminance contrast’.

    Note that the Kantian is particularly pressed by these questions because his or her notion of a concept must be wide enough to capture the fine-grained nature of conscious perception. If x is a part of phenomenal experience, then we have a concept of x. And that seems pretty unlikely if grasping a concept requires that we can linguistically express it, for example. So whether a Kantian view of consciousness is workable or not hinges here upon the theory of concepts that Kantian uses—whether having a concept also requires that one be able to express that concept, linguistically (a very strong criterion), or have demonstrated reflexive knowledge or understanding of that concept, or whether it is sufficient merely to act on that concept (you are able to make the discriminations appropriate to the given the concept).

    In any event, the obvious move, for a Kantian, in the case of the achromat, is just to say ‘one sees that the object is different from its background’. As long as you have the concept of difference, presumably you don’t have to say in what the difference consists. Later on, when the achromat comes to understand the concept of albedo, of surfaces being inherently darker or lighter, then one can say that the achromat sees this surface as being darker or lighter than that surface, because now the subject has the concept needed. Similarly, once the subject has the concept of shadowing, the subject could also say something like “the berries look dark but that is just because they are in shadow.” (A good Sellarsian would begin to talk about black and grey ties at this juncture.)

    Now the question is how to ‘add in’ chromatic information to consciousness if chromatic information is used in just the same way, to see other properties. Imagine an infant gradually developing visual capabilities. At birth, the infant has very ‘fuzzy’ luminance contrast as input to the visual brain; at 2 months the infant has high-resolution luminance and chromatic contrast retinal input to the brain. Very gradually (or rather with astonishing rapidity), the infant begins to see the world, using chromatic and luminance information. I think that by age 3, we would want to say that most children see much of the world (although visual development goes on into the 20’s). And they are conscious of much of the world. But most 3 years do not know their colours, if by that one means being able to reliably apply colour names to colour surfaces, or to sort colours into categories. (The age at which children can apply colour terms has been steadily decreasing since the 1800’s when the average age of colour term acquisition was age 7. And if this was the average age, then in all likelihood there were children of 8, 9 or 10 who did not have this knowledge. In other words, almost teenagers.)

    Now suppose that you give a 3 year old child an Ishihara display, a pseudo isoluminant display of randomly darkened dots, and ask them whether the child can see, say, a (coloured dot) animal figure . (You don’t want to use digits for the 3 year old). And then you use a pseudo-isochromatic display, and ask the child if she can see the same animal shape in this new display, a shape now made visible by luminance information (i.e. a whale composed of darker dots but of many hues). In other words, you have a child who sees two tokens of the same type, two whale shapes, by means of two different kinds of information. This is perfectly plausible; the child will see both figures, a whale defined chromatically and a whale defined by luminance. And these whales would look different from one another (or so this theory claims). Call one ‘Jimmy’ and the other ‘Sam’ and I expect the children would tell you that they are not the same whale. So here the chromatic information ‘informs’ consciousness; the child sees the whale as differing from its background on the basis of shape (a chromatic/luminance difference) and can see the whales as different (a chromatic/luminance difference as well).

    This is what I am trying to get at when I have written about chromatic information entering consciousness, not in virtue of a concept of chromatic contrast, but in virtue of seeing a chromatically defined object and seeing that there is a difference between a chromatically-defined object and a luminance-defined object.

    One reason for preferring this sort of explanation is that it makes better sense of the acquisition of colour concepts (proper) and the phenomenology of childhood prior to that time. A child first starts by applying colour names randomly (with few enough colour names such that, in the basis of sheer random application, the delighted parents incorrectly ascribe colour term acquisition.) After the ‘random’ stage, the child begins to correctly use just one term, usually a colour associated with a particular beloved object. And after that, a period of expansion follows, with another colour name correctly applied, and so on. Clearly there is conceptual learning here about colour qua a stable surface property, although I don’t think it is clear what exactly is learned.

    So what happens as the child acquires, sequentially, these colour terms? Does the world acquire mental paint, one colour at a time? Or, one day, when the child has enough knowledge to satisfy Evan’s generality constraint for colour concepts does the child wake up to a coloured world? With all due respect to Evans, probably neither. If there is evidence that the child is using chromatic information to make first person decisions about the properties of the world, some chromatic concepts are in play, even if only in the form of ‘this object differs from that one in some unspecified way”. On this view, as the child gradually accrues new ways to use chromatic information to see the world, his or her phenomenology also accrues in complexity, not merely in terms of the properties he or she sees, but in terms of the similarities and differences between objects, etc. There is thus no single day when the coloured lights turn on, when the child gains the concept of surface colour. But one day the child will be able to say that the two whales differ in colour —one is a dark whale in many coloured dots, the other one is a green dotted whale.

    More generally, in both cases, the adult achromat or the toddler trichromat, the Kantian needs a view of concepts that allows each of these subjects to see, consciously, what they so clearly do see as young children, hapless teenagers, or just concept-poor adults. Now, I have said all of the above as if this was Pete’s problem, but I suspect that he and I are in the same (temporarily leaky) boat. Both of us are good Kantians, and both of us want a view that respects the Kantian requirement of ‘conceptualized’ content as a prerequisite of consciousness. But we also want to respect the empirical stakes in the ground. At least as I see them, those stakes suggest that we are using chromatic information to make first person judgments about the world prior to having colour concepts per se. So the question I pose to Pete is this: do you see a difference between the luminance and chromatic cases? And if so what?

  6. Thanks, Kathleen. That was enormously helpful. I feel like the view is sinking in much better now, and some of the things I was previously hesitant about strike me as much more convincing.

    Here I’ll try to say some about what I’m grasping for purposes of checking in: Am I indeed getting what I ought to be getting?

    So, the thing in your response that really triggers an “a ha!” for me is the paragraph about the 3 year old looking at Jimmy the chromatic whale and Sam the luminance whale. The following does indeed strike me as a fully coherent possibility: Armed with not much more than a concept of *looking different* or maybe just a concept of *difference*, Jimmy and Sam impact the youngster’s phenomenology like this: Jimmy differs from his background, Sam differs from his, and Jimmy differs from Sam. Since, as a matter of fact, the triggering conditions of this pattern of applications of DIFFERENT (to use the Fodorian capitalization scheme for denoting concepts) involve detections of chromatic contrast, this merits regarding the phenomenology as chromatic even though our young subject is not deploying any chromatic concepts.

    How, Kathleen, does the above paragraph strike you? Would you regard that as a fair characterization of your view of non-color chromatic phenomenology? If so, then I’m happy to count as an adherent. If not, then I’m still scratching my head a bit.

  7. Hi Jake!

    You raise some interesting points and I’d like to comment a bit on them. I don’t in any way presume to be answering on Kathleen’s behalf and look forward to what her own response will be. But here’s a quick reaction of my own:

    Your characterization of sensations seems insufficiently detailed. Maybe I’m missing something, but as I read it, “whatever mental properties enable a creature to discriminate visually objects in respect of their colors” casts a net so wide as to capture anything mental transpiring between the transduction of the electromagnetic stimulus and the verbal or nonverbal discriminating behavior. Even thoughts, then, or their properties, would count as sensations. However, I’m betting you’d like a characterization of sensations that doesn’t thereby make thoughts or their properties sensations.

    Also, I wonder what work “mental” is doing in your characterization. Of course, without it, retinal activity and even the EM stimulus itself count as sensation. And that seems very broad indeed. So naturally you’d want to rein it in with something like “mental”, but I wonder what, precisely, you think “mental” is bringing in. That is, what is the right characterization of “mental” whereby we could tell whether a theory of conscious color perception has committed itself to the existence of sensations?

  8. Response to Pete..

    Yes, that is exactly what I mean.

    When I first started looking at this empirical material on chromatic processing, I wondered whether the facts pointed towards non-conceptual content. If there are informational distinctions that play a role in our conscious perceptual lives, but for which we have no concepts, then it seems to follow that….

    But if, for example, you look at what cerebral achromatopsics actually say about their experience, then this other possibility presents itself. The cerebral achromatopsic tests as colour blind on all the standard tests for colour vision. Yet the cerebral achromatopsic can identify shapes, with continuous borders, on the basis of chromatic information alone. He or she cannot identify shapes given the Ishihara plates, for these do not have continuous borders, and require what is called signed chromatic contrast, R-G contrast. That is, to ‘connect the dots’ into a number, you must view each dot has having the same property, not merely segment the scene by means of border detection. So you could connect the dots if you knew that each ‘dot’ differed from its background in the same way, with a certain R-G contrast. But that information does not seem to be available to the cerebral achromatopsic. Note that this is not blindsight. The subject will remark on the figure without being cued, can trace the figure with a finger, etc. The figure is clearly visible, as different from its background. But it is not seen as coloured. Obviously, the cerebral achromatopsic has the concept of an object looking different from its background.

    On this view of chromatic processing, you get an explanation of what the cerebral achromatopsic can and cannot do in terms of information processing. But you also get an explanation of what the subject can and cannot SEE. As I said before, in the normal case, when you or I look at a red apple, these many chromatic processes have coherent conclusions, such that if there is a R-G border, the colours of the apple and its background also differ. But in the pathological case, that of the cerebral achromat, some of the normal processes are missing. Thus one catches a glimpse of the multiple processes that go into normal trichromatic perception—what seems to us, in the normal case, as the perception of a ‘simple’ property, colour, which requires a ‘simple’ sensation.

    What’s nice about this kind of explanation is that it can accommodate many of the properties that are used in visual processing, which seem to enter into consciousness at least on occasion but for which the subject does not have full blown concepts. I am thinking here of the ‘painterly’ properties of scenes, often about light, that one learns either in art school (my husband) or in visual neuroscience (me). Most people don’t know the term ‘specular reflectance’ but they do know about those annoying white glints that you get on bright days off of water. That’s why you wear sunglasses in kayak. It would be a senseless project to try to prove that no one without the concept of specular reflectance has ever seen instances of specular reflectance—i.e. been consciously bothered by them. Then again, most people are unaware that almost all objects have some degree of specularity, that there are no truly matte surfaces in nature. And hence if you explain the concept and point out instances, you can certainly increase what the subject sees. This is so even though specularity may be a property that informs many human visual processes, such as colour constancy or material recognition. Just because your visual system reliably uses some property of the world in order to see the world it does not follow that you, the subject, thereby have conscious access to that property. But sometimes you do and therein lies the problem of how to explain HOW you do, if you lack that concept.

    As I said in my last post, one of the questions about colour perception that has really intrigued me is the question of colour development. Here you have infants and small children who can gradually do use chromatic information for more and more tasks. Long after toddlers can recognize dogs and cats and the volume control on your stereo, indeed YEARS later, they finally ‘get’ colour. And that is very, very odd when you think about it. What sort of ‘perceptual primitive’ suddenly makes itself known after you can categorize all of the ‘furniture’ in your daily life by means of that very primitive? This new way of thinking about colour makes sense of a developmental process that eventuates in perceiving colours qua colours. By talking about the kinds of chromatic information that are used prior to colour perception proper, and that in fact needed for colour perception proper, this explains how our ability to use chromatic abilities accrue over time, without supposing that the colour sensations were simply there since birth, but for some reason, unidentifiable or unnoticed or… Well, it’s hard to imagine just what the ‘problem’ would be, why young children would not have access to their own colour sensations, or if they had access to them, why naming them would prove so difficult.

  9. Response to Jake Berger

    Thanks for those comments, Jake. I put aside your question for a few beats because my brain went ‘off line’ for a few days, and I wasn’t sure I knew exactly what you were asking.

    As I understand it, you are exactly right about how the story goes on the orthodox view: you see the colour of the cat, the colour of the mat, and then you infer that the cat differs from its background in colour. Hence you see the cat as being different. I take it, as Pete said, that you have in mind a more robust view of sensations than simply any state that makes possible sensory discrimination. Rather, we are talking about either phenomenal or potentially phenomenal entities—and presumably retinal events are neither. Because I’d assumed that on your part, I read your question differently. I took you to be asking why anyone would seek an alternative explanation, given that we already have a good explanation in terms of sensations, an explanation of why we see the cat as different from its background. Why up-end the apple cart?

    Here the answer is something like this. If this were the only puzzle to be solved, then it most certainly wouldn’t be advisable to glom onto chromatic processing as an ‘explanation’. Rather, when I started this project many years ago, it began with philosophical questions about colour, old and familiar (e.g. if qualia are simple, hoe do you compare them, see them as similar or different?), but also some new questions from colour physiology. For example, why do 50% of all visual cells in cortex (human) have chromatic sensitivity? Why is so much real estate needed for a property that has no obvious importance in vision? Why is trichromaticity in mammals so rare, while dichromaticity, which involves rather poor wavelength discrimination, is the norm? And so on. With the new work on chromatic processing, a story started to emerge—one that made far greater sense vis-à-vis the requirements of visual processing and the mammalian visual physiology. That is, physiologically and in terms of the psychophysics, it looked as if the chromatic and luminance systems play by the same rules (e.g. share the same physical resources and are subject to just the same illusions). But with that view in hand, you have to go back to the philosophical problems and the questions of conscious colour perception. All of which is to say that this is an empirically motivated view, one that is getting built by moving back and forth between the physiology and the philosophy, in the hope of finding a coherent theory. This paper is one small piece of a very large project. It does not give any real evidence in favor of the view, only a short explanation of why a chromatic system is needed by a luminance system in informational terms—and you do not get any of the evolutionary, genetic, and comparative physiology that would make the view sound half way plausible. So, yes, based on this paper alone, you’d be hardpressed to see why anyone would want to upset the standard order of things.

    So, yes, the view I am putting forward here does try to upend the orthodox view. It places chromatic processing at the heart of ‘colour’ vision—as the raison d’etre of the system as a whole—and sidelines colour perception proper as one among many tasks that uses chromatic information. It also ‘farms out’ the tasks of ‘colour sensations’ to various sorts of chromatic processes, thus eliminating colour sensations as a kind of causal nexus that it occupies in the orthodox view. We don’t have to see colour in order to use chromatic information, and on the contrary, it makes good sense NOT to, given the difficulty of the task. Turn your attention to any object in the room, your cat or the couch or your new moleskin notebook, and you will indeed see its colour. But that does not prove that first there are colour sensations and then there are perceptions of the world. Rather, it says something about what attention and the resources the brain can marshall when it wants to examine an object very closely. (Colour perception is the hard trick; chromatic processing the easy one. So colour perception, on this view, goes ‘high’ and conceptual.)

  10. Thanks Kathleen. And thank goodness this isn’t a nonconceptual-content story.

    Since I’m satisfied now with the line about non-color chromatic phenomenology, I’d like to turn to press a bit more on what you think the right thing to say is in connection with point 5.3 in my commentary.

    The point there was to wonder whether there could be a revisionary view of sensations that shared enough with the orthodoxy to be worth bearing the label “sensation” but was still, well, Akinsian.

    This might be in the ballpark of the sort of thing Jake Berger is worrying about, so I’d be happy to bear what he has to say about it too.

    Focusing on red and so-called sensations of red, the gist of the revision I envision sees sensations of red as the causal antecedents of certain thoughts of red (I’m using thought- talk as shorthand for episodes in which red is conceptualized) that serve to make these thoughts count as perceptions of red and not just mere thoughts of red. The perception v mere-thought distinction seems worth drawing, and if it’s not going to be drawn in terms of sensation, then in what terms will it be drawn? Anyway, to continue sketching the revisionist tale: the imagined revisionist may be happy dumping lots and lots of the orthodoxy. Sensations of red need not be conscious, they need not be the bearers of what-it’s-likeness (maybe because there’s no such thing in need of bearering), they need not represent maximally determinate shades. Heck, maybe they’re not even mental. This revisionist view of sensations of red is a very minimalist view: only the barest minimum of the orthodoxy is preserved.

    Now, here’s a reaction against the revision, a reaction that strikes me as Akins-ish. But I do wonder whether you’d endorse this way of reacting.

    Anyway, here’s the reaction: As minimal as the revision is, given that it’s a story about sensations *of red*, it incurs an obligation to say what the relevant *of-ness* is. Unfortunately, this obligation will never be met, because there’s nothing in the causal chain between a red surface and high-level thinky-type things that is *of red* aside from the thinky things themselves. That none of the causal antecedents of the high-level stuff are *of red* is another way of putting the moral of the story about the massive underdetermination of spectral surface reflectance by the proximal stimulus.

  11. Hi everyone, really great discussion going on in here!

    I wanted to follow up on the line that Jake and Pete are perusing in their recent comments.

    So Jake has in mind a view where we represent physical properties, say color, via a non-intentional property that we can call a mental quality. A mental quality simply is whatever mental property the creature uses to ‘track’ the external perceptible property. So, mental red –red*– will be some property of a creature’s mind that preserves the characteristic similarities and differences between perceptible red (as determined by psychophysics).

    In addition to these properties there will also be intentional states (thoughts) to the effect that one is is various (groups of) starred states (there is a family of starred properties for each perceptible property including time and space).

    This looks like a version of the orthodox view and it looks like your work undermines this kind of view, but it is not clear to me that it does so. Pete suggests one way to keep the thought/sensation distinction by counting certain thoughts as sensations (if I got Pete’s proposal right) but another way of doing it might be available. Surely there must be some property which stands in these various similarity and resemblance relations, even if it is a very complex property; call that the sensation.

    Also, you ask “What sort of ‘perceptual primitive’ suddenly makes itself known after you can categorize all of the ‘furniture’ in your daily life by means of that very primitive?”

    A diehard defender of the orthodox view might defend themselves against this by arguing that the sensory quality becomes conscious when it finally ‘makes itself known’ but that is no bar from it being around to do all of the categorizing work that you point to. This kind of stuff often happens when metal qualities are not conscious.

  12. Thanks so much for your very helpful reply, Kathleen. And thanks for filling out the view a bit more, Richard–I do think that’s largely what I had in mind.

    The view that Richard’s presented, I think, retains the element of the orthodox view that states can exhibit either concepts or mental qualities (or both). But I don’t think the revised view regards mental qualities as simple in any problematic way. So I’m wondering with Richard whether you think there’s a problem with it (though I admit that Richard’s sketch is only a sketch).

    One nice feature of this view is that it provides a way to distinguish sensations (states that have only mental qualities) from thoughts (states that only have concepts), and even makes room for perceptions (states that have both). And it seems that there are examples of all of those kinds of states. You might be most skeptical about cases of conscious sensations, but I think there are examples of those–for instance, you might have a conscious sensation of after just waking up. You sense red, but do not conceptually represent that anything is the case.

    And if the sensory properties are representational in the nonconceptual way that Richard suggested, then we have a way of explaining away Pete’s worry about how people who lack the requisite concepts can nonetheless make the relevant discriminations.

    So I think I agree with you that “[c]olour perception is the hard trick; chromatic processing the easy one. So colour perception, on this view, goes ‘high’ and conceptual.” I think explaining the conceptual bit of perception is hard, but I think the sensory bit of perception is easy to explain, whatever the complex processes underlying it are.

    Moreover, another virtue of positing those representational mental qualities is that they might help explain why perceptions have the correct ‘high’ conceptual content that they do.

  13. Hi Jake, Pete, Richard,

    Okay, this response was written before Jake’s last one was posted. And I have to admit, that would be a nice and, well, cohesive way to explain things. I have to admit that today I wondered how seriously I have wandered off the philosophical map, into the neuroscientific undergrowth, simply because of the difficulty I was having in understanding the question. A bit of trouble switching hats.

    At any rate this is my response, pre-Jake response.

    Here I’ll take as my starting point Richard’s question: “A diehard defender of the orthodox view might defend themselves against this by arguing that the sensory quality becomes conscious when it finally ‘makes itself known’ but that is no bar from it being around to do all of the categorizing work that you point to. This kind of stuff often happens when mental qualities are not conscious.”

    Okay, this question finally tipped the balance—it made very clear to me how different our background empirical assumptions are. My view here is very ‘bottom up’: I think that it is unlikely that there are colour sensations because I think there is nothing for them to do, no role for them to play in colour processing or visual processing more generally. (Whereas, from where you start from, you see a good role for them to play in a theory of perception.) I don’t think that there is anything in this paper that suggests any wider conclusions — e.g. that the very notion of a sensation is logically incoherent or something of the kind. Rather, I am here treating the question of colour sensations as an empirical question, and I conclude, based on what I know of luminance processing (and of course low level chromatic processing) that it is empirically improbable that there are the sorts of entities or states that Richard suggests. That is, I don’t think there are any states that encode the sort of information that would be required to ‘under gird’ a colour sensation at low levels of vision — or indeed at any level prior to the representation of surface colour. What I find interesting is that you think either (a) such states must exist, or (b) it is likely that they exist because there is something for them to do, down in the dark, prior to conscious colour perception. That’s the crux of the response, so let me now explain how the read the neuroscience.

    So, let’s look at what Richard’s question would be if applied to what I am contending is the parallel case, luminance processing. Everyone agrees that, for the luminance system, the ‘basic unit’ of information is luminance contrast. On this understanding, vision works best using relative measures, using luminance differences, and hence this is the form of encoding used for say, motion processing, shape processing, etc. At a much higher level, albedo processing, we want to know about a property of objects, surface reflectance, the percentage of light reflected from a surface. And here, after strenuous computational effort, object surfaces are assigned a surface reflectance value. To save the orthodox view, Richard’s question, applied to luminance processing, would be this: But isn’t it possible that there is a “gray area” in the brain, one in which intensity is represented? And here presumably one means image intensity, the intensity of light at each point in the retinal image (because you haven’t yet determined albedo). So isn’t it possible that, once we determine the albedo of a specific object, that the albedo is ‘presented in consciousness’ by an image intensity value, which then becomes conscious? Um, well, yes. I suppose that is possible. But at least from where I stand, it sounds like a strange story.

    For one, it’s hard to see what role the encoding of image intensity is to play, what it does, apart from being used, later, to represent surface reflectance, or provide the phenomenology of surface reflectance. That is, if you represent the intensity at each location in the visual image, these are not representations that ‘track’ any property of the world. It is true that each intensity value is the result of reflected light from the environment, but that is not ‘tracking a property’. Rather, what one is representing is a property of the image, and one can then use that information to discern and encode a distal property, say shape. The very same moral applies to the representing colour. That is, you might think that there is a colour area of the brain that takes chromatic output from the ganglion cells and represents colour along the standard dimensions of brightness, saturation and hue. But if this did occur, this would be image colour, the wavelength composition of some point in the retinal image. Let’s just assume this is so, that there is just such a colour area. What you would have would not be an encoding that ‘tracked’ any distal property, not even ‘more or less’. The problem of figuring out object surface spectral reflectance (object colour) is significantly harder than figuring out albedo (in fact, on one way of thinking about it, you need to figure out albedo—lightness or darkness—as propert part of the problem of surface colour.). So such an encoding would not serve the purpose of ‘tracking’ any colour property in the world.

    To get an intuitive feel for the problem, take out a monochrome photograph of your living room. Close your eyes and circle the photograph randomly with a pin. Now select a place and open your eyes. What shade of gray is that pinpoint? If possible, isolate that area by using a piece of white paper as a mask, with a small hole in it. What relation does the image intensity bear to the actual albedo of the object it represents? Answer: none. Or rather: random. If the white couch is in shadow, you may have pinpointed an entirely black part of the image. But the interesting thing is that the very same result will hold for a colour photo. You may have put the pin on the edge of your arm chair but the colour of the image could be from any area of the spectrum, depending upon shadowing, lighting and nearby objects. This is why those ‘colour pickers’ in photo alteration programs always seem to give you an entirely unintuitive colours, an unrecognizable colour relative to what you know the colour of your armchair to be (i.e. green). This is because, without the rest of the image, there is no direct mapping between the pinpointed colour and the colour of the actual object. Hence colour encoding for the image does not ‘track’ any colour property in the world. Rather, with a lot of hard work, you can figure out, from the image as a whole, the colour of an object. But that is as good as it gets.

    Once you look at it this way, the utility of a ‘gray area’ or a ‘colour area’ is not at all obvious. Even if you had this information, what would you do with it? Well, certainly not use it to discern the colour of objects. For that task, it is more or less useless.

    As you have probably foreseen, another problem is just the idea that one would use a representation of image intensity to somehow render object albedo conscious. Why would the brain do that? After all, they aren’t the same property. Once you have in hand the value of albedo, presumably you have in hand what you need for seeing the object as having a certain lightness or darkness. I think that in the luminance case, this seems obvious, and for this reason, I have never seen it suggested in any physiological literature. Yet somehow, when we think about the colour case, it suddenly seems like a good idea. Yes, there is a colour area in which image colour is represented. When you finally figure out the object colour, you then select the ‘right’ colour representation, from the colour area, and ‘apply it’ consciously to the object. Pardon? Why would the brain use the representation of an image property ‘as’ a representation of a very different object property, SSR? Why would that even seem like a coherent idea unless one already had a background philosophical theory—or very good evidence that such an area and a ‘colouring’ process existed?

    At bottom, however, are the hard empirical facts that suggest that no such area exists in the brain, for either luminance or the chromatic system. Take the chromatic system. At the retinal level, chromatic cells do not encode wavelength or even wavelength contrasts. Such cells react to wavelength contrasts, but also to luminance contrasts, plus combined luminance and chromatic contrasts. Further up in V1, ‘real’ colour cells (double opponent) have been found, but they do not form anything like a ‘map’ of colour space. Rather, such cells also seem to encode multiple image properties, properties that would be helpful in discerning a multitude of object properties.

    So, when I look at this sort of schema, I don’t know what you mean when you say that such states could ‘do all the categorizing work you point to’. If by that you mean ‘are used to discern shape, etc.’, then I agree. But such states do not have any of the right sort of informational properties—i.e. the dimensions of colour sensations—that would make them eligible to emerge as conscious colour sensations. But if you mean ‘used to categorize colours’ then the point above holds: there just are no cells or cell assemblies at all, no ‘gray area’ or ‘colour area’, that categorize object object colour prior to the computation of albedo or colour.

  14. At the risk of turning this comment thread into a comment bush, I want to address some side-issue- type things that have popped up. Of course, if this strikes anyone as too branchy or beside the point, please feel free to ignore. 

    For Richard:

    You write: “Pete suggests one way to keep the thought/sensation distinction by counting certain thoughts as sensations (if I got Pete’s proposal right) but another way of doing it might be available.”

    Actually, the sort of thing I had in mind (not that I agree with it, but just as devil’s advocacy for revisionary orthodoxy) was not so much about drawing a thought/ sensation distinction, but drawing a mere-thought/perceptual-thought distinction. What is alleged, on this imagined story, to distinguish the perceptions (perceptual thoughts) from all the other thoughts is that only the former are directly triggered by sensations. So, whatever a sensation of red is, it needs to be the sort of thing that is causally prior to the perceptual judgment or perceptual thought. And, if I’m reading the dialectic correctly, this priority thesis is the main thing that makes the revisionary orthodoxy vulnerable to Kathleen’s critique, since on her view you can’t get anything that’s mental and “of red” until a whole bunch of computations drawing on high-level conceptual-type stuff have been brought in.

    For Jake:

    You write: “You might be most skeptical about cases of conscious sensations, but I think there are examples of those–for instance, you might have a conscious sensation of after just waking up. You sense red, but do not conceptually represent that anything is the case.”

    I know, from our dealings outside of this thread (e.g. hanging out together in bars) that you’re likely to grant that if this bit of phenomenology is reported by you, then you must have conceptualized it. Why think such a thing can occur consciously yet unconceptualized? I’m betting that you’ll agree that no amount of first-personal introspection is going to turn up an unconceptualized item. So the case for mental thingies that aren’t conceptualized by the people that have them is going to have to be third-personal. And it’s a short step from admitting that we have to go third-personal to realizing that empirical explanatory justifications need to be brought in to make the case for sensations. So what are the empirical explanatory merits of the sensational story?

    More for Jake:
    You write: “Moreover, another virtue of positing those representational mental qualities is that they might help explain why perceptions have the correct ‘high’ conceptual content that they do.”

    It would be unfair of me to saddle you with a specific view of how that’s going to work, but just for discussion’s sake, consider a Hume kind of story, whereby ideas, by being copies of impressions, inherit their representational contents right from them. So red thoughts get their contents from the red* impressions of which they are copies. Crucial to this copying story is a commitment to a causal priority thesis. And like I was just saying to Richard a few pgraphs back, that’s going to be its main vulnerability. But of course, since you haven’t committed here to a specific view, you may very well have one in mind that is invulnerable to the priority objection.

  15. Kathleen,

    Coming back into the conversation, I think that what I see under discussion is at what stage of processing we ever have anything resembling a pure color quale. It would seem by your theory that the retina’s processing and even the initial visual cortical processing is concerned so much with edges and contrast that those who talk about retinal processing making quales are not talking about anything that is even close to conscious sensation.

    I should point out, though, that occipital electrode stimulation and focal seizure auras are reported to cause immediately percieved abstract flashes of color and other non-edged color sensations, not merely shapes, and these are claimed to originate from the part of the brain at or close to the primary visual cortex, not association cortex.

    This suggests that my quale of color may correspond to a a primary visual cortical event. It is not what my retina sees, but it is not an association cortex afterthought either.

    Does this follow from your new theory at all?

  16. Hi Drbill,

    I guess I would have to look at the experiments to see what was done, monitored, etc. in order to comment on this kind of case. Is there any evidence that these experiences occur as a result of ONLY the area of seizure? (That is, that the synchronized bursts don’t reverberate beyond the focal area?) I know you must be thinking of some quite specific results here, because obviously the proverbial poke in the eye can also cause similar sensations—and no one wants to conclude, therein, that the locus of colour sensation is in the retina.

  17. Very good point. So if an abnormal state can override retinal edge processing so could one elsewhere. Hmmm.

    Do you think then that we need to process any given (normal) image in association cortex in order to “abstract” a color perception from what the primary visual cortex does?

    The equivalence would be someone with perfect pitch identifying a particular note in the pitch of a speaker’s voice– it would require a deliberate breakdown of what he ordinarily would hear?

  18. Professor Akins, let me say first that I’ve learned an enormous amount about vision from your paper, and I am thoroughly grateful for that.

    And I think I’ve profited philosophically too – although not, perhaps, in the way you intended. What I’ll do here is to suggest a way of looking at the material you’ve given us, and then ask for your reaction.

    In brief, I think you’ve done a great service to epiphenomenalism. That’s because I think one can sum up what you’ve shown this way: Sensations are caused by a set of processes that have an underappreciated complexity, that have, at least until quite recently, been importantly misunderstood, and that are completely unconscious. Moreover, sensations are not needed as causes for discriminating objects or guiding our behavior – indeed, they cannot be so needed, at least in some cases, since some of the tasks for which it might be thought sensations are required have to be done already in order to cause the kinds of sensations we have in the first place.

    Evidently, I am reading you as *not* denying that there are sensations; you’re just denying them a causal role that others have attributed to them. So, in a nutshell: There are sensations, they are caused by unconscious processes, and they do not play the causal role they’re often assigned. I guess you could go on to find some other causal role for them to play, but you’ll see why I think that your paper can be taken as congenial to epiphenomenalism.

    Perhaps I should add that, while epiphenomenalism is an accusation in the mouths of some, I take the support for epiphenomenalism that I see in your paper to be an important and very positive aspect of the latter. But it may be that you will not be so welcoming of approval from epiphenomenalists; and thus my question: Is there anything that would block this way of looking at what you’ve provided? and if so, what?

    I’ll just add a comment about conceptualism, of which Pete well knows I am no friend. I don’t want to start a full blown discussion of the issues here, but I do want to make a distinction that seems relevant to some of the remarks in your and Pete’s back and forth.

    To get at this distinction, I’ll just suppose we agree that, in the unconscious processes that are needed to cause our color sensations, there must be uses of concepts (e.g., DIFFERENT or even SOLID or CURVED). That supposition would not, I believe, imply that the redness (for example) of a sensation was conceptual. I.e., to causally depend on a use of one concept (say, C1) does not imply that the possession of a (different) concept (say, C2) is required for the effect (e.g., a sensation quality) to occur. So, e.g., two year olds might perfectly well have color sensations without concepts of colors, even if they have to have unconscious representations of space, objecthood, etc. in order to come to have color sensations.

    From your discussion with Pete about conceptualism, I wasn’t sure what you’d think of this distinction between a sensation’s causally depending on an (unconscious) use of certain concept(s) and its being conceptual. So, again, a question: Does this distinction seem acceptable to you? (And Pete – Hello! – what about you?)

  19. Thanks again for your detailed reply, Kathleen. I think I’m going to have to think a bit more about the ramifications of the physiology on whether we would want to grant there’s anything like color sensations where it, as it were, all comes together. I take it one of the aims of theorizing is to take some folk phenomenon–like color sensations of the sort I mentioned–and see if there’s anything we can equate in the brain like that. But if we find there’s nothing in the brain that can do the trick, then maybe we’ve just been wrong all along. But as I said, I have to mull this over some more.

    As for some quick replies to Pete’s questions for me, Pete, you ask “you’re likely to grant that if this bit of phenomenology is reported by you, then you must have conceptualized it. Why think such a thing can occur consciously yet unconceptualized?” I agree that for a sensation to be conscious, the sensation’s properties must be conceptualized (that is, one’s higher-order awareness must deploy concepts of the sensation’s mental qualities), but I disagree that that means that the conscious sensation itself conceptualizes, say, the color its a sensation of. The color sensation itself deploys no color concepts–and so I take it these sorts of states, conscious or not, are not conceptual, despite being themselves conceptualized when conscious (e.g., if I think about a table, I conceptualize the table, but that doesn’t mean the table deploys any concepts).

    But you also ask about my comments about how color sensations can be of (that is, represent) colors without also deploying concepts of them; you write, “It would be unfair of me to saddle you with a specific view of how that’s going to work” and then go on to float a Hume-type account. So I am developing that story–it’s basically the substance of my dissertation–but it’s kind of complex and not particularly relevant here. Suffice it to say, I think mental qualities represent things in a completely different way than concepts represent things, and the representational character of the latter is at best obliquely related to the representational character of the former.

  20. Hello Bill (aka William S. Robinson)!

    I’m glad to know you’ve been following this thread. As for the distinction you ask about, I think I can accept it, but I am just a bit hesitant. My hesitation has to do with being a bit unsure what we must mean by “sensation of red” here. A lot of my earlier remarks about sensations presupposed that they were by definition nonepiphenomenal, but clearly here we’ll want a characterization that’s neutral between epiphenomenalism and its opposition and of course likewise neutral with respect to conceptualism/nonconceptualism.

    So, how does “sensory consciousness of red” sound as a neutral characterization? If that sounds all right to you, then I’m understanding the distinction you’re offering as acceptable, that is, it describes a coherent possibility: There may very well be a lot of concepts unconsciously deployed that serve as causal antecedents of the red sensation without the sensation itself being, in any relevant way, conceptual.

    Of course, admitting this as coherently possible is not to concede that things ever actually work that way.

    Bill, does it seem to you that I’m understanding your question correctly?

  21. Hi Dr. Robinson — Bill,

    So, you’ll have to forgive me for taking a few beats to answer your post. I was incredibly pleased that you found the paper interesting and, for me, that is a far more positive result than having you agree philosophically with me its consequences. Obviously, epiphenomenalism is not my theory of choice. But I am interested in why, for you, it seems to follow naturally from the theory, whereas it didn’t occur to me at all. (Actually, that is false. Occasionally, epiphenomenalism creeps into my views, and when I notice it, I gasp and start beating it with a large mental stick. But I like to think that there are more rational explanations for this reaction than mere phobia.) (Dream on.)

    One of the things that has always puzzled/interested me about colour ‘sensations’— i.e. when you think about colour as presented under the standard psychophysical conditions, a combination of lights against a grey background—is the nature of colour dimensions. The standard story from time immemorial is that colour has three dimensions: brightness, hue, and saturation. And these dimensions roughly correlate with the physical properties of intensity, predominant wavelength, and percentage of the predominant wavelength relative to other wavelengths present. When you think in terms of the physical properties of reflected light, these dimensions are hardly intuitive. Saturation? Why saturation? Why is that a dimension of perceived colour? However, if you are thinking in terms of a system that is using chromatic processing for figuring out the world, these properties make more sense. For example, what separates shadowed from non-shadowed areas is a difference in light intensity between the two areas, not a difference in wavelength composition. Similarly, as light levels of the illuminant go up and down, the predominant wavelength will stay the same. And saturation could be read as luminance ratio—the luminance measure of the whole (i.e. amount of light overall independent of wavelength) versus the ‘amount’ (intensity?) of the predominant wavelength. In other words, perhaps these dimensions of colour are intuitive if viewed as measures of the stimulus dimensions (amplitude, wavelength) but understandable if viewed a ‘units of chromatic processing’ within human colour vision.

    This is a possibility I continue to think about simply because, if the human ‘colour’ system’s has utility primarily as a chromatic processing system, as opposed to a colour system, there is an interesting question about it’s evolution. How does one get colour out of chromatic processing? For example, many people have asked about colour constancy of the sort of basic kind, wherein one discounts the ‘colour’ or SPD of the illuminant in order to see the SSR of the object. Obviously, if you throw largely green light at an object, what comes back is largely green light, so you have to figure out how to ‘discount’ that uneven SPD of the illuminant. Here, my general answer is that there is not one mechanism of ‘colour constancy’ but several, at different levels of the system. At the retina, there is chromatic adaptation, such that, with predominantly green light, the system becomes less sensitive to the predominant wavelength. Such a mechanism does get you 90% of the way towards colour constancy. However, it is also a mechanism that vastly increases chromatic contrast as well. Remove the green, and you can see what the differences really are. This mechanism is quite primitive though qua surface colour constancy. So you will get chromatic adaptation whether the light source is green or whether you happen to be in a forest with lots and lots of green surfaces. So, given that we have colour constancy for object surfaces, some other mechanisms must be in play. Still, you can see what I am driving at here: this retinal mechanism may certainly be one of ‘mechanisms of colour constancy’, but its primary evolutionary function was to increase chromatic contrast. Bit by bit, parts of the processes that you need to know in order to discern surface colour are accumulated, and the system ‘approaches’ colour perception incrementally over time. Thus at a certain point the system acquires a rough and ready ability to process surface colour, a capacity that could be improved over time.

    The more general point here is that I doubt that the sort of complex computation that is required for surface colour merely popped up, de novo, in the evolution of human vision. We can see many ways in which such a competence could be improved, incrementally following general adaptation mechansims. For one, the dimensions of colour might arise as a result of disparate processes. For another, mechanisms that are required for colour constancy might have arisen in the service of quite different visual processes (e.g. basic contrast enhancing mechanisms)/ And then one has to wonder whether such ‘incremental’ mechanisms do or do not contribute to our phenomenology of a coloured world.

    Here again this would open up the possibility that colour processing is not an all of nothing affair. I am not suggesting that when you turn your attention towards an object, you perceive its surface as having only one dimension of another of surface colour. Rather, in the rough and tumble of chromatic processing, one or another of these ‘dimensions’ is processed. Thus you have some information about the surface of the object that, if understood as such, could provide you with partial information about object colour. Once you have the concept of colour, you do not need ‘full information’ about a particular surface in order to perceive as coloured. You saw that the object was dark, but not the kind of dark, as it were.

  22. Response to Pete Mandik’s post of 2/27 at 5:56

    Pete, yes, you’re right to clarify, and you are understanding me correctly. And I agree that accepting the distinction, and thus the possibility of my scenario, does not give us an argument in favor of that scenario – all that would follow is that one can’t rule out nonconceptual content in one particular way.

    (I’d better put in a caveat for other readers, though. I know from experience that there are some who would read the “of” in “sensation of red” in a way that implies commitments that I would not accept, or fails to imply commitments I would think essential (e.g., that sensations of red (or any other property) are conscious occurrences). And so I can’t rule it out that someone might read even “sensory consciousness of red” in a way that I’d find problematic. But I like that Pete’s phrase makes the consciousness explicit, and the context in which he puts it gives no reason to suppose he’s reading it in any way that I’d find problematic.)

  23. Response to Kathleen Akins’ post on 3/1 at 16:08

    Thank you for the response, Kathleen. I’ll just say that “follow . . . from” is stronger than I meant; “suggest” would do, and I think I said your view is “congenial to” epiphenomenalists.

    Beyond that, it seems to me from your sometimes noticing epiphenomenalism in your views that your unconscious processing is working very well. Perhaps you should stop trying to beat it into submission, and take a kinder attitude toward what it’s trying to tell you?

  24. For what it’s worth, I had a similar reaction to Bill to Kathleen’s extremely interesting article. The talk about the lack of role for sensations in the processing story naturally suggests either eliminativism or epiphenomenalism about sensations (or about sensations as traditionally conceived, or something like that) while leaving the choice between those views open.

    A third option that seems to be left open by the paper is a sort of globalist realism. The “rendering” view that is the target of the paper seems roughly to be a localist view on which sensations correspond to a certain step on the visual processing pathway. I don’t know how many theorists embrace this sort of view, but it’s certainly just one option concerning the neural correlates of consciousness. Rejecting that view leaves open global-NCC views (e.g. Tononi-style views) on which the relevant experiences (as they seem to introsopection) exist but depend more holistically on brain states.

    I’m not sure which if any of these three options is closest to Kathleen’s view, and which she takes her arguments to rule in or rule out, but I’d be interested to hear!

  25. David, David, David.

    The conference is almost over. Why NOW? I’ve nearly made it through, right through to the very end, without being unmasked before these good people—-a (Canadian) Eliminativist among a deep sea of Epiphenomenalism.

    Very unkind.

    More seriously? Thank you for putting the cards on the table. I think the answer to your question depends upon what you want the notion of sensations to do in your theory. If you want colour sensations to play the role of a causal nexus in a theory of perception, then I would say that, yes, I lean towards eliminativism. I don’t think such a causal nexus will be found (and I don’t think that is matter of whether the representations are local or distributed). But there are many uses for the concept of sensations to play in a theory of mind, and this is only one.

    As ever, my modus operandi is to inch forward, looking at one small piece of the empirical puzzle in which I have some confidence in the experimental results. But also as ever, the theory is massively underdetermined by the data. But, yah, for this one small corner of a very large problem, I am going to put my money on eliminativist view of colour sensations.

    (Curtain falls; terrorized philosophers flee towards the exits and large drinks at the New York Consciousness Collective!)

Comments are closed.