There is a brain mechanism that works to identify colour differences directly, without first identifying the absolute colour of each surface. So on my view there is no reason to suppose anything like ten million colour responses to surface viewed singly.

I think my view is rather more radical than Pete Mandik's. Both of us want to show that colour perception doesn't transcend what can be conceptualized, but I don't think he goes so far as to deny that it doesn't involve different responses to all the discriminable surfaces.

I don't think that we are capable of anything like this many possible colour responses. Instead I argue that the perception of colour differences between two surfaces viewed side-by-side is a gestalt phenomenon.

Even if no learning to speak of was involved in locking my mental term onto doorknobs, it is odd to say that therefore my possession of a doorknob concept is innate, just as it is odd to say that my head-injury-caused singing is innate.

The relevant features of scientific practice often have mundane explanations which don't point to any deep metaphysical moral. (Thus it would simply be messy and pointless for the chemists to essay physical reductions, or for the biologists to offer number-free explanations. It's a weird kind of science-worship that views these practical considerations as clues to the nature of reality.)

Of course our genes will make some capacities very much easier to learn than others, and of course our genes themselves are not learned. But the point remains that genes themselves are not cognitive capacities, and that anything worth calling a cognitive capacity will depend to some degree on learning and so not be innate.

If there is such a simple argument for physicalism, how come everybody hasn't always been a physicalist? That's a good question, and there is a good answer. The 'causal completeness of physics' wasn't widely accepted until recently.