My previous post, Weird lands of the tails, had some concepts implicit which I didn't elucidate in detail. For example, I assumed that the speed is a quantitative trait, and the many genes which control its variation have pleiotropic effects. That is, gene 1 has effect on phenotypes 1 through n. Gene 2 has effect on phenotype 1 through n. Speed may be just one of those phenotypes. More formally what I'm thinking about is a genetic variance-covariance matrix, or G matrix. If you keep the G matrix in mind I think it's kind of ludicrous to expect that speed was actually what was being selected for directly; but that's just me.
If you want to know more about the G matrix, Comparative quantitative genetics: evolution of the G matrix is a good paper. Also check out The multivariate breeder's equation. There's a lot of talk about going beyond one-gene models and the retarded concepts which arise from them, but if people are serious about that, well, start looking into quantitative genetics and refine your mental models....
I was hoping you'd make this more explicit.
Now, how about hitting the phenomena that a single or few quantitative loci of large effect are often identified... yet when those are held constant, other loci pop up with large effects. Wash rinse repeat.
My quantitative genetics is quite rusty, but the basic concepts weren't very difficult... just not exactly intuitive since the variable names tend to not quite to reflect what they actually measure.
Biomechanically, speed = stride frequency x stride length. Assuming direct genomic influences, speed = gene(s) associated with fast twitch (FT) fibers x gene(s) associated with height. Biomechanics research has shown that skilled sprinters differ from less skilled sprinters by having extremely high stride frequencies and short ground contact times, but non-significant differences in stridelength. Speed may thus be primarily governed by % FT fibers (but height is definitely a factor as revealed by Bolt's above average height for an elite sprinter).
Research concerning FT and ST fibers shows a normal distribution pattern; the majority of the population has an approximate balance of FT and ST fibers, with a small percentage showing greater FT fibers (who may become elite sprinters) or ST fibers (who may become elite distance runners).
An important question concerning the genetics of speed, is the extent to which epigenomic factors (e.g., training) influences % FT/ST fibers (quantitative genetics and research on FT/ST fibers appear to assume genomic factors). As an example, developmental research has shown cultural variations in motor development related to the way babies are handled by caregivers during early motor development. Before 1 year, the Kipsigis of Kenya and the West Indians of Jamaica handle their babies in a way that appears to speed motor development progression (e.g.,holding head up, sitting alone, walking) relative to North American infants.
An interesting study that would implicate an epigenomic influence on speed, would be one assessing whether the observed cultural variations in handling children prior to age 1, influences the proportion of FT/ST fibers over time. This would probably require some kind of developmental study involving an FT/ST fiber analysis in children prior to age 1 (not sure how feasible that is).
There is a third aspect: storage of energy. Sprinters don't have time to breathe, so most of their energy comes from sugars stored in the muscles. That is certainly a trait that can be trained.
> If you keep the G matrix in mind I think it's kind of ludicrous to expect that speed was actually what was being selected for directly; but that's just me.
I'm not as knowledgeable as you Razib, but for what it's worth I could almost agree with you - minus the word "ludicrous" which seems too strong. After all, isn't it one thing to appeal to pleiotropy, but another thing to actually "flesh out" by coming up with a list of phenotypes that definitely or plausibly covary with short-distance speed? It might be challenging to come up with copious material for that list, though I certainly don't deny that a greater fast-twitch lower body strength will help you with throwing a projectile and at least a few other important things.
Also I'll add - though I'm sure you know this and just haven't explicitized it - that for a really fulsome model we should be sure to include constraints (and releases from constraints) as well as adaptive benefits. This includes possibilities like the sprinters' ancestors having an excess of nutriment compared to other human populations, which would render it economical for them to bear more alleles favoring greater fast-twitch brawn.