Sunday, August 3, 2008

A forgotten but (perhaps) important experiment in colour vision.

Many years ago, in 1983, I published a paper on colour vision, in which I described an experiment that seemed to me to be important. Although the paper has been cited many times, the particular experiment I am referring to has never, to my knowledge, been quoted by anyone. In fact, even I had forgotten all about it until very recently, when someone made what seemed to me to be an inaccurate remark about colour opponency.

Colour opponency refers to the fact that there are three pairs of colours which have been described as those that “cannot live with each other, and yet cannot live without each other”. They are red-green, blue-yellow, and white black. For, as most people know, when we look at a green surface for a brief period of time and then transfer our gaze to a neutral, blank, screen the colour of the after image is red. A yellow surface will produce a blue after image and a white surface a black one.

One explanation of these opponent effects – the one most often repeated – is that they are due to adaptation in the retina. The explanation here is something like this: that a green surface reflects more green light, leading to the adaptation of the “green” or middle wave receptors. Thus adapted, the activity in the opponent “red” receptors holds sway. The result – we perceive red. The same explanation applies to other opponencies.

The experiment I described was derived from, and an extension, of the colour experiments of Edwin Land. Land had shown that a green surface which is part of a complex, multi-coloured, scene can be made to reflect more red light and yet still look green (though a darker shade of green). This is because the brain undertakes a somewhat complex operation to discount changes the wavelength composition of the light reflected from that surface. This makes sense. After all, a green leaf looks green when viewed at mid-day or when viewed at dawn and dusk (when it actually does reflect more of the long-wave or red light). If the perceived colour of the leaf were to change with every change in wavelength composition reflected from it, then the surface would no longer be identified through its colour.

It seemed to me interesting to take this one step further and get humans to look at a green surface that was part of a complex scene and get that green surface to reflect more red light (twice the amount of red than of green light). It still looked green. Now, by the traditional explanation, the after image should look green, because the “red” receptors, having adapted, would defer to the “green” receptors, which had not been as vigorously stimulated and hence had not adapted. Not a bit of it. The after image was red! I repeated this experiment with other colours and got the same general result. The after image is not related to the wavelength composition of the light reflected from a surface. Rather, it is strongly dependent upon the colour of the surface viewed. Since the colour of a surface, when part of a complex scene, is independent of the wavelength composition of the light reflected from it alone, it follows that the colour of the after image is also independent of the wavelength composition of the light reflected from that surface.

Hence the colour of the after image is constructed by the brain after the colour is constructed. It has nothing to do with retinal adaptation.

But no one has taken the slightest notice of this experiment and, as I said, even I forgot about it. That is a pity. It still seems to me to be an important experiment, but evidently no one shares my view.

5 comments:

Michael Proulx said...

That is certainly an important experiment! One aspect that I think would be intriguing to follow up on is whether the after image arises from the surface colour or the perceived colour. I can think of a way this can be manipulated, and the result would be fascinating.

Steven said...

Thanks for the article. As with many great experiments, some don't get the attention required, only to be rediscovered decades later and presented as a major breakthrough.

For example this happened to the discovery of backpropagation in neural nets by Bryson and Ho one year after it was claimed by the famous and convincing Minksy and Papert that multi-layer NN wouldnt overcame the problems of single layers in 1969. It took 10 years to be rediscovered an brought under general attention.

Nowadays we have wikipedia which is often used as a starting point or at least used as a source to look at. Since it is easy and free for anyone to share knowledge (they need input from experts!) you are encouraged to bring your work under general attention. As mentioned in another post, wikipedia encourages to be a complete source not limited to a single scientific view/theory.

Oh, and in case you wonder, i am not affiliated. Thanks for your blog, i read your posts with much interest.

S.Z. said...

Thanks, Steven. So there is hope yet that this experiment will some day become more prominent. Your post encopurages me to help it along, which I will do.

As regards Wikipedia, I have been quite disappointed with it in my general field. Many of the articles are self serving, although I notice that here and there the editors do ask for a more balanced and scholarly revision.

Perhaps I should take up your suggestion and give it a go. I will certainly think about it. Thanks

Alex said...

Would this be explained by colour constancy? The high level of the long red wave-length would be subtracted out of the perception of green surface in V4? As demonstrated in Schein and Desimone (1990)

Reading your musings as an interested student.

S.Z. said...

Yes, of course. I mean that constant colours are generated first, and since constant colours are independent (within wide limits) of the wavelength-energy composition of the light reflected from the surface being viewed when it is part of a complex scene, so is the colour of the after image. Thus the colour of the after-image is better accounted for by colour constancy than by retinal adaptation.

What/where are you studying?

SZ