It seems possible that soon there may be a cure for colourblindness. The Mary's Room thought experiment attempts to pin down something about the nature of qualia in a contrived but similar situation, but my feeling is that the actual result of such an experiment would not be obvious. Would we consider the experiment valid if it was performed on somebody familiar with blue and green, but not red?

New to LessWrong?

New Comment
7 comments, sorted by Click to highlight new comments since: Today at 4:48 PM

People have already been cured of complete blindness from birth (e.g. http://www.newyorker.com/tech/elements/people-cured-blindness-see ). Not much seems to be revealed by this, especially since in the Mary's room experiment, she is supposed to know everything there is to know about color in advance, and obviously this is not true in such cases.

I think this topic deserves a fair bit of clarification.

First of all, there are many forms of colorblindness (6 major types, plus several rarer types). This is important because this has direct implications for the discussion of qualia, and probably not in the way you think. Let me explain.

The different kinds of colorblindness are informally described as 'red-green' or 'blue-yellow' or somesuch, but it's better to describe them according to their medical terms. The prefixes 'prot-', 'deuter-', and 'trit-' are used for red, green, and blue photoreceptor molecules, respectively, while the suffixes '-anomaly' and '-anopia' are used of modified or absent forms of these molecules, respectively. So we have, for instance, deuteranomaly, which means 'modified form of the green photoreceptor', or protanopia, which means 'no red photoreceptor'. Deuteranomaly is the most common form of colorblindness.

So what does 'modified form' mean? It means that those people with deuteranomaly have the green photoreceptor, it's just a mutated/defective version which is closer in color sensitivity spectrum to the red photoreceptor than the green photoreceptor in normal people. What this means is that people with deuteranomaly actually do have the 'qualia' for all three colors, it's just that since birth the signals from the red and green photoreceptors to the brain have been mostly identical with very little difference, so the brain never learns to make a distinction between the two. In this case, it seems that just by altering the spectrum so that the difference becomes more pronounced, you can 'cure' colorblindness. In fact there are already special glasses that do this.

But in the case where the person totally lacks one photoreceptor, this isn't possible, because as far as I know they fundamentally lack the requisite neural wiring to get three color channels to the brain (but see below).

Now, these people claim to have a gene therapy that gives people a new set of photoreceptors. That's great, but at the end of the day, all that's being accomplished is that some cone cells are being suppressed or excited. Imagine if you took a photo and added some kind of grainy filter over the photo so that the color spectrum of some of the pixels were altered, but only within the range of colors you can already see - no new colors. Now imagine you permanently saw the world through a camera setup that did this to everything. That's basically what's happening here.

It may really cure colorblindness; I don't know. The brain is weird like that. It could definitely help colorblind people learn to distinguish between colors they previously couldn't. But I don't immediately see how this could cause such people to have a novel color sensation/qualia pop out for them. One way I could be wrong is if the neural circuitry for trichromacy exists but is dormant in such individuals.

One way I could be wrong is if the neural circuitry for trichromacy exists but is dormant in such individuals.

If they have grown up without proper trichromatic neural signals flowing from the retina to the LGN into V1, then V1 will have already developed a set of low level gabor features that are bichromatic. To actually see in color, the visual systems of these patients will need to do some relearning.

Based on other known plasticity/rewiring experiments, it seems fairly reasonable that this type of rewiring will occur automatically as a result of new explained spike signal components flowing into V1.

As a reference example, MIT did these famous experiments way back in the day where students put on these special goggles that flip vision completely (vertically I believe). Apparently the patients reported that at first everything was upside down, confusing, and even nausating. However after some period of time (a week or two?) there is a sudden aha moment and their vision 'flips' and they can see normally - with the goggles on. Removing the goggles then results in a similar process, but with faster relearning.

All of this can be explained in an ANN type model with continuous incremental online (gradient descent style) learning dynamics. The aha moments even correspond to the rather sudden phase transitions seen in the evolution of ANN weights.

I have no doubt that rewiring like that can and will happen. But then there's the question of why introducing new photoreceptors is special in this regard. And if this type of stimulus can't be produced by other means (like seeing the world through a special camera setup, like I mentioned), and, if so, if such other means could in fact produce novel color qualia. After all, all we're doing is making some modifications to the most superficial part of the visual neural system.

But then there's the question of why introducing new photoreceptors is special in this regard.

Because introducing new signals from new photoceptors changes the network dynamics and leads to new concept learning.

And if this type of stimulus can't be produced by other means (like seeing the world through a special camera setup, like I mentioned), and, if so, if such other means could in fact produce novel color qualia.

From the information you provided, I suspect that the deuteranomaly case is functionally similar or equivalent to the bichromatic case. What really matters is the actual connectivity structure of the gabor filters in V1.

if so, if such other means could in fact produce novel color qualia.

Well, theoretically you could add even more chromatic signals - for infrared say - and if V1 rewires to include those signals, then the patient would report a new infrared color qualia, of a type no human had experienced.

On a related note, there is a wierd experiment involving a device that encodes images onto the surface of the tongue, allowing blind patients to 'see' the output of a camera through their tongue. That could be considered a new 'qualia', as the resulting visual pathway is undoubtedly quite different than normal.

Why is the experience of color not physical knowledge? Why is Mary's experience of learning the science physical? What would the science of color vision look like if nobody experienced color?

One problem with the experiment is that Mary will not see like other people see when she gets outside of her BW room. As experiments with cats raised in specially colored (striped if I recall correctly) rooms show the visual system can be largely rewired. Mary may experience something unusual but that will be entirely different from what a normal person perceives.