Thanks to everyone for all the responses! A quick word or two to report what I know about synaesthesia. (Blame Kriston for asking!)
First, there are indeed lots of different kinds. Tone-color synaesthesia, of the sort that minniemarx reports for Scriabin, is often found in musicians. Not surprisingly, there is a higher incidence of perfect pitch among tone-color synaesthetes than among normals. (Perfect pitch amounts to something like being able to identify colors for them.) There is also general grapheme-color synaesthesia - association of either letters or numbers or both with colors. There doesn't seem to be agreement among synaesthetes about what color the various graphemes are (hence Emm's family squabbles!). But there are patterns. It's much more likely for instance, if you're a synaesthete, that you'll see the letter A as red than as blue or yellow. In addition, there is also a relatively common but little-studied kind of synaesthesia that gives shape - sometimes a very convoluted shape - to the number line. And Oliver Sacks writes famously about taste-shape synaesthesia in his essay "The man who tasted shapes."
But perhaps the most interesting work on synaesthesia was done about 8 years ago by Ramachandran. He established pretty conclusively that at least one type of synaesthesia (what he calls "lower synaesthesia") is a strictly low-level sensory phenomenon. The synaesthete isn't just remembering the color of the refrigerator magnets he had as a child - he's actually
seeing colors. Ramachandran's group established this with a pop-out task. Look at the picture on the left in the link
here. It is a mix of 2's and 5's, but for the non-synaesthete it can be very difficult to pick out the one from among the other. For the synaesthete, by contrast, the 2's pop-out very quickly, since they are experienced as a different color than the 5's. The synaesthete's experience is modeled with real colors in the picture on the right. The synaesthete answers
much more quickly than normals the question "What shape do the 2's make?" (Answer: a triangle.) Pop-out phenomena like this are well-established as low-level sensory phenomena, so the fact that the synaesthetes can perform this task so much faster than normals shows that they are really sensing colors, not just remembering them or associating them. Not all synaesthetes are lower synaesthetes, however, so failing the pop-out task is no guarantee that someone doesn't have synaesthetic experience. Or so Ramachandran argues.