Already got a response to my primary colours question from T.J. Jankun-Kelly.
Some great points:
TJK also cited http://hyperphysics.phy-astr.gsu.edu/hbase/vision/colper.html which is a nice summary of colour science and makes a point that is key for me:
"It is found that many different combinations of light wavelengths can produce the same perception of color."
So there is a definite distinction between the dimensionality of colour perception and the actual space of light. I feel more confident now in asserting that the actual colour space of light is infinite dimensional but that it is projected onto a three dimensional non-linear space of perception.
So does that mean that the number of cones is the reason for the dimensionality of the (perceptual) colour space?
UPDATE (2004/06/08): TJK say yes. The number of cones is the reason for the dimensionality of the perceptual colour space. Chickens, which have 12 cones, would have a 12-dimensional perceptual colour space. Makes me think of a name one could use for an article on this topic: "If Munsell Were A Chicken".
The original post was in the category: colour_science but I'm still in the process of migrating categories over.