Is there a paper or website or something illustrating this?
All right, I’m getting some conflicting information. Hamster King says that the red(-green) receptors also activate a bit on violet. mister nyx is saying the exact opposite: that the blue(-violet) receptors activate a little bit on red.
I do note that I can at least find a CIE color matching chart that matches what Hamster King says, and the Wikipedia article says that color matching was created based on human response. However, that doesn’t mean that’s how it physically works in the eye. The two different ideas would be functionally equivalent, so it’s possible that digital cameras (or the corresponding color spaces) work one way but our eyes work the opposite way.
I cannot seem to find a response chart that matches what mister nyx says about the blue(-violet) color receptors. Can anyone else? All the ones I can find that describe eye response have contiguous color ranges with no trough in the middle.
This paper (pdf) seems to be a pretty definitive study, and forms the basis for a lot of the graphs you see. One point about the graphs is that many are linear, and tend to hide some of the nuances at the low levels that are revealed in the log scale graphs seen in this paper. They eye is sort of linear when it comes to colour, but nothing is simple. The paper goes into excruciated detail, including CIE response.
Yeah, I think mister nyx is mistaken in suggesting that there is a trough in the response of the S-cones (if that is what he meant), but he is right to point to the fact that there is some quite complex, and very non-linear, processing of the cone responses going on in the retina (not to mention in the brain, after that). You can’t think of subjective color and color relations as being a direct product of mixing together the signals from the three cone types. It is much more complicated than that, and (so far as I can make out) still not fully understood by anybody.The leading theory these days is opponent processing, which deals with the retinal computations (or some of them), but it is almost certainly not the whole story.
Another thing that you need to take into account is the very different distribution of different cone types across the retina, with relatively few S cones in the foveal area (where color vision is concentrated), and relatively few L and M cones outside it (and even within the fovea, the L and M cones are distributed very unevenly, in random clumps). Thus the color information getting to your brain depends, to some extent, on whereabouts on the retina the image of the colored object is falling. In practice though, that image will always be moving about, as our eyes are never still. It is salutary to realize, though, that our subjective sense of the color of an object does not generally change as its image moves off the fovea and into the periphery, even though very different populations of cones will be being stimulated by it during the course of that movement. In the periphery of vision, we cannot really discriminate colors at all (there are only rods and a few S cones out there, and the rods are non-functional in conditions bright enough to see color), but, for most people, if they have already seen the color of something, they will still have some subjective sense of its color when it is in the periphery, even though they are no longer really getting any color information from it.
The Spanish Castle Illusion is a nice illustration of how subjective color is not tied directly to the stimulation state of the cones, and of how eye movements can affect color perception (in ways that are still, I think, not fully understood).
Thing is, I’d be more inclined to believe that the “error” was in the perception part rather than the encoding, unless there is proof otherwise. It seems more natural for something about how violet is perceived to cause it to be encoded as red+blue, rather than two different perceptions resulting in the same encoding.
The reason for this is stated in that page I linked. Someone who has cataract surgery, which temporarily allows larger amount of ultraviolet light to reach the retina, also sees ultraviolet as having a purplish hue. If it were an encoding quirk, you’d expect the results not to be the same as violet.
I have a hypothesis based on what is given: Violet is getting awfully close to double the frequency and half the wavelength of red. Perhaps there’s some sort of sympathetic response in our red-green cones to the octave, similar to the sympathetic vibrations you get when you hold down one key on a piano and then briefly sound the note an octave below.
And, of course, it could just as easily be the blue receptor with sympathetic vibrations from an octave below–as stated above, light doesn’t have to work like sound. Heck, it could even be both.
EDIT: I always thought the Spanish castle illusion was based on pigment bleaching–the fact that the receptors in our eyes take a bit to readjust after they’ve been stimulated for extended periods. Are you claiming it’s an encoding error instead? (I could see that–the same encoding error that allows us to see an object as the same color under different colors of light.)
The CMFs rely on metamerism. The task is usually like: there are two lights. One consists of three colors mixed, e.g. R+G+B, and the relative levels of these three can be adjusted. The other light consists of just one wavelength, and throughout the experiment would be changed to a large range of wavelengths. The task is to adjust the first light until it looks identical, even though it is physically not. So at 450 nm, the person would have to add a lot of blue, a little bit of red, and almost no green. I think (?) that this means making it similar to extraspectral “purple.”
The cone fundamentals tell you how the cones respond. As njtt says this can barely tell you everything. The CMFs measure human matching responses, which meant that the light is filtered through: the cornea, the lens, humors, macular pigment, the layers of the retina, the cones, and then on to LGN, striate, extrastriate cortex, etc. That said, the CMF can be converted to cone fundamentals, and it assumes linear cone responses. So it you are looking for the responses of the cones in isolation, color matching functions are a bit hard to read directly.
The fact that you see the complementary colors at all is indeed due to pigment depletion (“bleaching”). However, that is old news. What is interesting about the Spanish Castle, and the aspect of it that was relevant to what I was saying, is the fact that you continue seem to see the scene in its true colors for as long as you can continue to keep your eyes almost stationary (and it seems to have much better color definition than you would ever get from a regular negative afterimage, if you “bleached” your cones and then looked at a blank space), but as soon as you move them the color disappears altogether. While you are staring at the dot, most of that image is in peripheral vision, where actual color perception is poor or non-existent, due to lack of L and M cones, yet it seems to be vividly colored all over…until you move. On the other hand, if you can keep your eyes still for long enough, the apparent colors persist (I think) for much longer than an ordinary negative afterimage would remain vivid.
The more general point is, you can’t expect to get much insight into actual color experience merely by understanding the spectral response curves of the cones. That is only one small part of a very complex (and still incompletely known) story.
I guess I never noticed that because I can’t keep my eyes stationary very long. No matter how long I stare at it, the color only lasts a second or two, about as long as an afterimage.
BTW, of course there’s better color definition than looking at a blank sheet of paper. On a blank sheet, the only difference is color, but, in a black and white image, we also have the differences in brightness. And we are more sensitive to changes in brightness. The differences in brightness supplement the differences in color. Plus, since you can’t move your eyes, you can’t move them to elsewhere in the image to see if there are any aberrations.
Well, I don’t know what you are seeing. If you cannot follow the instructions and keep the dot fixated you will not see the effect, I guess. I see vivid color while I keep the dot fixated, but it disappears the instant my eyes move more than very slightly.
In fact, having looked at it a bit more, I do not think the Spanish Castle effect depends on a regular “visual pigment bleaching” negative afterimage at all. Afterimages of that sort move with your gaze, but this does the opposite. Anyway, I do not think the colors of the inducing picture (the reversed color one you start with) are bright or saturated enough to induce much of a regular afterimage. When I stared at the inducing image for a bit and then shifted my gaze to a blank, white area I saw only a very faint and brief afterimage of an orangey castle shape (and no blues or greens at all). By contrast, if I follow the instructions and keep fixated while mousing over the picture to change it to grayscale, I see vivid “true” colors, with the castle a whitish gay, green vegetation, and the sky a vivid blue with white clouds. (I also find I do not need to stare at the inducing image for anything like the suggested 30 seconds for it to work. Five seconds seems to be plenty.)
I suspect, though, that the illusion does depend on the fact that the true color of the area around the fixation point is in fact gray (the castle wall), so what your fovea is seeing in the grayscale image is the actual gray of the castle. The fovea (where most of the cones actually are) is not being deceived (if it were experiencing a negative afterimage, the castle wall wold look orange); it is only in the periphery (which, as I said, is actually largely insensitive to color) where the illusory colors appear.
Sorry if you can’t fixate for long enough to see this illusion. I guess, in that case, for you it was a poor example. But for those of us who do see it as intended, it is clearly something different from a regular negative afterimage, and, in particular, responds to eye movements in a quite different way.
What you say about definition is true, but, once again, when I try this illusion I do get a strong impression of well defined colored things in the periphery, even though the periphery is not in fact capable of sensing either fine definition or color,
Except that always travelling at c doesn’t apply to light, as c is referring to light in a vacuum. 
I can see the illusion, it just doesn’t last but a second or two. The other illusions that have a white screen last the same amount of time. Hence I thought they were the same illusion.