You are absolutely right. Neither colour nor a colour is spatially extended, and a colour like pink is exactly not like a gas that fills up a volume or spreads itself, perhaps very very thinly, over a surface. That is a category mistake. Nor do colours have thicknesses. I am delighted to see a recognition of this important point, to which I have found very little attention paid in the literature on the philosophy of colour. Though I tried to get started sorting out the tricky logic of "pure" colours, as Wittgenstein calls them (to be contrasted not with impure colours but with things coloured the colours), in Chapter 7 of Colour: a Philosophical Introduction, published in 1987 and 1991, but that was only a beginning.

Consider two patches of the same red, patch A surrounded by green and patch B surrounded by red. We can say that the colour of patch A is identical with the colour of patch B. But then as G.E. Moore pointed out (in his early paper on "Identity") it follows, or seems to, that the colour of patch A is surrounded by red as well as by green. The right conclusion is that the patches do not have the same "surrounds", but that the colours they are coloured do not in logical grammar have surrounds at all. This means of course that the way that we describe simultaneous contrast (just for example) will not be as one colour being affected by the surrounding colour. Quine's conclusion was that a colour is a scattered object, like water, and that the colour of patch A is surround by red here, and green there, just as these waters are in France and those in England.

You are also right to say it seems to us that colour is spread out in space, and colour scientists have incautiously spoken about surface and volume colours. What we are talking about when we talk about volume colours are strictly not colours, I think, but the way colours look. The inky blue has the look of something into which one could reach, like a volume. Or "the colour" looks "spongy", say, when the ability to see surface colours is lost. And sponges are three-dimensional.

Perhaps there is nothing to worry about here. Numbers obviously are non-spatial, but one can have an impression that it is wrong not to describe as one of numerousness or multiplicity or something. If I see a lot of cows in a fiedl, I can see that there is more than one, though I may have to count them to see exactly how many. Before I do that, I might have the impression that - it looks (in the phenomenological sense) as though - there are a number of cows around the field. But the number itself is not spread out around the field, like slurry, for example. Appearances can be deceiving. It may be though that all that is needed is a sensitive treatment of the problem of universals as it applies to colours, and a recognition of the differences that exist among the characters of universals in different sensory modalities. Colours do not have origins, like sounds, for example, and so their relationship to space is not the same as the relationship of sounds to space.

I would hesitate to call the impression that the colour is in space any sort of illusion. It is something that the way colours manifest themselves gives us some temptation to believe, but I am Wittgensteinian enough to believe that such temptations should be thoughtfully resisted. Colours have only three dimensions, and they are not spatial. If something is scaled in some other dimension, then it is not a colour.

Colour relationalism tells us that colours are relations between perceivers and the objects that they perceive. (This gets a bit tricky of course if what they perceive is a colour, because then what they perceive is a relation between themselves and a relation between themselves and a colour, or rather, a relation between themselves and a relation between themselves and a relation between themselves and a colour.) "Colour phenomenology" doesn't mean an awful lot: just the (apparently) obvious facts about colour and classifications of things having to do with colour based on these facts. Phenomenology in this sense has seemed to oppose colour relationalism. We don't seem to see relations between ourselves and tomatoes when we see red tomatoes. What we see is a property flattened on to the surface of the tomato! So colour relationalists have to work a bit to square their view with the phenomenology. Without meaning to be rude, might I inquire whether this is an essay topic you have been set? Why are you interested in this question at all if the works you have read are not user-friendly? I mean why are you interested in the question if you don't know what it means?

It would not be right to raise a child in a very controlled environment where she is cut off from all that wonderful, colorful nature you get to experience every day. But we can think about such a child. So let's do this. Suppose this child is raised in such a very controlled environment where she can ever see only six basic colors: blue, white, black, red, yellow and violet. I think we would be able to explain to this child -- in fact, to any child -- that the color violet lies between red and blue, is really a mixture of red and blue. And I think we could then go farther and say that red and blue can also be mixed in different proportions, so that the mixture contains less blue and more red. In this way, I think we can get the child to imagine the color purple. So the answer to your question is yes, it is possible.

But it is possible only in those cases where the child knows colors that are close enough. You probably know from your experience in drawing that you can mix yellow, red and black to get brown. But I would be surprised if the child in our story could imagine brown with this instruction: imagine a mixture of black, red and yellow. Similarly, the color green can be produced by mixing yellow and blue, and again I think our little girl would not be able to imagine green from the instruction to imagine blue mixed with yellow. I am unsure about orange which, as you know, is a mixture of yellow and red. It would surely be harder for her to imagine than purple (where she already has another mixed color -- violet -- available to her). I am similarly unsure about turquois, which is a mixture of blue and green. Perhaps she would be able to imagine orange and turquois.

So the full answer is that which of the colors someone has never seen before she is able to imagine depends on the colors she has seen. New colors can be imagined if they are variations on, or combinations of, colors that one has seen before.

Great set of questions! Lots of literature for you to investigate (starting with Hardin's "Color for Philosophers"...) But let me just say briefly here that one typically begins by distinguishing clearly and purely physical properties (like "wavelengths") from "perceived color" -- for there are many demonstrable cases where a given perceived color can NOT be matched or mapped onto any given wavelength(s), and vice versa. Once you make this distinction then it is easy to hold that wavelengths (plus other factors) CAUSE perceived color, or at least are a causal factor therein, but are not identical to them. Then you will begin to ask whether perceived color can be identified with any clearly purely physical properties and will probably find out that the answer is no. (Or if so, it might end up being a brain property -- ie when you perceive color x you are always in brain state y etc. -- but that is a far cry from what we want to normally say about colors, namely that they are properties of surfaces of bodies, or of light ....) And once you've gone this far you will be tempted to see in the distinction between perceived color and all physical properties the basis of an argument that 'colors exist only in the mind' .... (the point about cameras doesn't seem to matter -- once perceived color is distinct from all physical properties ...)

best, ap

Let's agree that, at bottom, a thing's colour is a matter of how it affects us. Still, that rough thought can be sharpened in various ways. As a more careful second shot we might say: something is red if it is such that, in normal lighting, it will produce a certain visual response in normal perceivers. And for present purposes we can leave it open whether the response is best characterised in subjective terms (in terms of a certain look that red things typically have) or in neural terms (in terms of how our retinas and visual systems respond). For note that, either way, if we do analyse the notion of being red as a matter of being such as to produce a certain response in normal perceivers in normal lighting circumstances, it can continue to be such even when e.g. not in normal lighting.

Compare: the glass is still fragile even when it is dropped on a cushion and doesn't break, for being fragile is (roughly) a matter of being such as to tend to break when dropped on hard surfaces, etc. The rose is still a red rose, even though it looks a nasty murky shade under sodium light, for counting as red is determined by how it will look in daylight.

To begin with, why should we think that colors do not exist? Many philosophers have argued that colors are secondary properties, that is, they are properties that are perceived, and as such properly exist in the perceiving mind rather than external objects. But many of those philosophers also think that there are primary qualities which external objects have independently of the mind, and which, together with the perceiver, are responsible for the perceived secondary qualities. On that view, colors exist, they just are not “in” external objects. Some philosophers, such as Berkeley, deny the existence of the material world, but maintain that colors and other “sensible qualities” are real perceptions in the mind. But aside from the question of the reality of color, the question of whether total non-existence is conceivable is an important (and related) philosophical question. I recommend Descartes’ Meditations on First Philosophy as a starting point.

A classic question, which has been MUCH discussed over the centuries -- especially with the rise of early modern philosophy and science (16th-18th centuries) -- rather than give 'the' answer let me mention some historical resources -- beginning at least wiht Galileo but especially prominent with figures like Descartes and Locke, it was recognized that colors don't fit easily/naturally into what were understood to be the genuine physical properties of things -- in Descartes's day it was thought that size, shape, and motion essentially were the only genuine physical properties, and if so, then colors -- which do not seem identifiable with those -- must be said to exist at best only in the mind, as perceivers' responses to those physical properties in objects. John Locke in particular offers numerous arguments in support of this view, you can find them easily by looking him up in the Stanford Encyclopedia of Philosophy or elsewhere. But now while our conception of the physical properties of bodies has changed over the centuries, the same basic concern exists: that one cannot identify colors with the properties of objects (molecular structures), nor of light (wavelengths), nor even with the way these things interact with the physical brain. Some useful sources for these views are Larry Hardin's "Color for Philosophers" which does a great job explaining why it's so difficult to identify color as perceived with any physical properties; and then maybe Frank Jackson's classic article "What Mary Doesn't Know" (now reprinted in an anthology filled with much discussion of it) arguing that such phenomena as perceived color can't be identified with anything physical ...

best,

Andrew

If you are talking about basic colors, then you are right: there are only finitely many of them, and to get beyond them one would then have to bring in "colors" beyond the visible spectrum, and this is indeed weird in the absence of beings that can actually perceive those "colors".

But here's an argument on the other side. Suppose we are willing to count as colors all the different shades on the visible spectrum -- between 360 and 750 nanometers, let's say. Suppose these are densely packed so that between any two wave lengths there's always another one. Then we'd have infinitely many different colors all of which we can actually perceive. (There's a serious questions about whether this account is consistent with the latest physics, but set this aside for a moment.)

Now you might object that two colors can be different only if (a) we are able to perceive both and (b) we are able to discern the difference between them. Our abilities of discernment are surely limited, and so there are not infinitely many distinct colors after all.

But there's a response to this objection. Suppose our discernment powers are limited so that we cannot distinguish colors when the wave lengths of the light hitting our retinas are less than 1 nanometer apart. (Take any other plausible number if you like, it does not matter.) Would it follow that we can only distinguish about 390 colors? No, because we are able to distinguish shades of color indirectly. For example, we may be able to distinguish light of wave length 444.44 from light of wave length 444.39 by the fact that we can directly distinguish the former but not the latter from light of wavelength 443.41. In principle, this method of indirect distinction might be carried on indefinitely (e.g. we may be able to distinguish light of wave length 444.4444444444440 from light of wave length 444.4444444444430 by the fact that we can directly distinguish the former but not the latter from light of wavelength 443.4444444444435). To be sure, you won't actually get very far with this in your life time. But this does not refute the hypothesis that there are infinitely many distinct and distinguishable shades of color.

Although Hume does not himself use the terms 'a priori' and 'a posteriori', those categories do, roughly, correspond to the distinction that Hume draws between relations of ideas and matter of fact in the Enquiry Concerning Human Understanding. (The Enquiry Concerning Human Understanding is also referred to as the 'first Enquiry', as I will do in what follows, to distinguish it from the Enquiry Concerning the Principles of Morals, the 'second Enquiry'.) Now, by the by, but interestingly enough, Hume doesn't draw the distinction between relations of ideas and matters of fact, at least explicitly, in the earlier Treatise of Human Nature, much of whose first Book was recast in the first Enquiry, although he does draw a related distinction in Book 1, Part 3, Chapter 1 of the Treatise, between relations that depend on 'intuition' and 'demonstration', and have only to do with ideas, in contrast to other relations, which do not so depend on ideas, and thus do not admit of the sort of certainty characteristic of intuition and demonstration.

Near the beginning of Part 1 of Section IV of the first Enquiry, Hume explains that relations of ideas "are discoverable by the mere operation of thought, without dependence on what is any where existent in the universe," whereas, by contrast, matters of fact "are not ascertained in the same manner...the contrary of every matter of fact is still possible, because it can never imply a contradiction, and is conceived by the mind with the same facility and distinctness, as if every so conformable to reality." Insofar as relations of ideas do not depend "on what is any where existent in the universe," they are akin to Kant's analytic truths, which depend, roughly, only on the meaning of their terms; matters of fact, which do depend on what is existent in the universe, depend on experience, and thus are akin to Kant's synthetic truths. As you quite rightly point out, there is no room in Hume's division between relations of ideas and matters of fact--which is sometimes called 'Hume's fork'--for the category of the synthetic a priori, which seems to straddle the realms of relations of ideas and matters of fact and thus to explode that supposedly exhaustive division.

Now you want to bring these distinctions to bear on Hume's famous thought experiment of the 'missing shade of blue', which he himself took to constitute a prima facie counterexample to his claim that all ideas are derived from impressions, that is, that all thoughts may be traced back to experience. (It's worth noting that in both the first Enquiry and in the Treatise, Hume actually considers the missing shade of blue before he introduces the distinction between judgments based on experience and those not based on experience and hence admitting of certainty. He therefore must not have thought that his 'fork' was related to the missing shade of blue.) The missing shade of blue might seem to constitute a counterexample to this view, because in this case, the thought of a color is arrived at without one having had experience of that previous color. Now Hume himself, after propounding the example, says that although "this may serve as a proof, that not every idea is derived from correspondent impressions...this instance is so singular, that it is scarcely worth our observing, and does not merit, that for it alone we should alter our general maxim." Although Hume may have thought the missing shade of blue "scarcely worth our observing," his readers have given it considerable attention, and you propose that one might understand the example, in Kantian terms, as marking a distinction between imagined and experienced colors, with the latter belonging to the category of the 'analytic a posteriori' and the former belonging to the 'synthetic a posteriori'.

It's not clear to me that this suggestion can work. After all, analytic judgments, for Kant, are not supposed to depend on experience, so it does not seem to me that judgments about colors--when one uses the terms 'analytic' and 'synthetic', one must refer to judgments, what might today be called assertions--can be called 'analytic a posteriori'; synthetic a posteriori judgments, for their part, just are what Kant calls straightforwardly synthetic judgments, since they depend on experience.

But while it's not clear to me that your suggestion can be brought to bear on the particular case of the missing shade of blue, the notion of analytic a posteriori judgments is itself a very interesting one, which Kant himself was not able to envisage, but which may nevertheless constitute a category that merits some consideration. For consider an assertion such as 'Water is H2O'. This is a claim based on experience, but it's necessary. It can't, however, belong to the category of the synthetic a priori, at least as Kant conceives it, since synthetic a priori judgments are supposed to explain the possibility of experience, but not themselves to be objects of experience; the claim can't be analytic, since it's based on experience. So it might seem that, in virtue of its necessity and the fact that it is a claim about experience, that such a claim is best construed as 'analytic a priori'. If this is correct, however, what does this show about Kant's own division of judgments into analytic and synthetic? Might that not have to be rethought in much the same way that Kant himself may be taken to have rethought Hume's fork?

There are a couple of ways we might go here. Thinking about the sky, it's a pretty good bet that most people with normal color vision (roughly, people who can pass a color-blindness test like this one ) will say that the sky is blue and will agree that a good many other things are blue, even if they've never seen them before. And so while there is a certain amount of arbitrariness and convention in just where we draw the lines between colors and while the names themselves are certainly matters of convention, there are facts about human physiology here as well. Thus, one way to think about colors is in terms of the responses of "normal" individuals in "normal" conditions (where both of those "normals" are not altogether easy to pin down.)

Another way is to think about the physical characteristics of the things we apply the color words to. For example: light with frequency around 450 nanometers is blue light. If the light reflected from a body has this frequency, we call the body blue. But things are a bit complicated here. We can't identify colors with frequencies in any simple way because non-monochromatic light (light that's a mixture of frequencies) will typically appear to be a definite color.

So color is not a simple matter. There are objective elements having to do with our visual systems, objective elements having to do with the properties of physical objects and the way they interact with light, and there are conventions about where certain boundaries get drawn. A really good account would be even more complicated. For example: context makes a difference to what a color looks like. Here's one example and here's another, related one. What we can say pretty confidently is that it's not just a matter of convention, though the full story calls for a good deal more expertise than I have.