(Appropriate disclaimer: Although I absorb a good amount of neuroscience as part of my research, I'm not an expert; take this as you will.)

There's a classic experiment on motor learning and transfer (I forget the reference offhand) where participants were trained to throw an object to hit an underwater target. The target was then moved to a different depth. In one condition, participants were briefly instructed about refraction of the object to give them a conceptual model; in the other condition, participants received no such instruction. The result was that the instructed group trained up at the new depth much faster; in fact, many in the non-instructed group had persistent difficulty. Having a high-level conceptual model thus significantly improved transfer of motor performance between tasks. This goes to the point you raised about variations in sensory equipment or fast muscle speed not being sufficient to explain individual differences in performance. The network governing motor learning and transfer involves many different “skills” (and also brain areas), including planning, prediction, eye movement, and focusing of attention. Altering execution of any of these can sharply affect performance, leaving a wide scope for individual variation.

Once we have learned the basic skills of some activity by paying attention it seems as if the skill and knowledge is in our bodies rather than our conscious minds, that our hands know what to do. You're right. Automatization of practiced motor tasks is very effective, which is why it can be so difficult to slowly demonstrate a well-practiced maneuver and why it's so easy to choke on a high pressure shot. (As you suggest, there is a lot of evidence generally that experts at a motor-skill have problems when they think about their actions.)

Precise timing in organic systems seems to depend on synchronization of a large number of cells, none of which are individually precise but when together influence one another. The more the better. Precise direction, too. There's an old study on monkeys from the 1980s where it was shown that large groups of neurons in primary motor cortex encode movement direction. Each individual neuron is tuned to a rather broad range of preferred directions, and many neurons have similar preferred ranges. But in encoding the movement direction they average together vector-like, and thus achieve greater precision than any one neuron can alone. (I'm sure there's been much more work on this of late.) This is consistent with the larger cortical representation in the primary motor area (the so-called homunculus) associated with parts of the body requiring precise control.

A similar averaging probably also enables precise timing, though the specifics seem more complex to me. For a complex movement like hitting a moving target from a distance, the fastest response times to stimuli, even under the best conditions, are larger than the differences in timing between success and failure. Preparation and planning are also required and can take more time than the response itself. Thus, the movement must be initiated substantially before the throw. I can imagine (though this is only speculation) a population of neurons with varying timing windows; inputs related to target prediction and from feedback received during practice strengthen a subset of these whose combined firing averages out to a particular timing. As practice accrues, some parts of the task presumably become automatized.

I'm not sure this implies that more is better, however, at least beyond some point of diminishing returns. For instance, there are some inherent constraints on precision (e.g., stochasticity in neural firing, muscular response) below which it does not pay to go. And there is a trade-off between the cost of supporting that precision (in energy needs, opportunity cost, etc.) and the survival value of the task. Without taking too hard an adaptationist line, I'd guess that brain size is rather well tuned to this trade off in most species. The example of the chickadee comes to mind: every fall, it's brain volume increases by 30% through the addition of new neurons, only to decrease again in the spring. One working model is that this change is related to its increased need for spatial memory in the winter to find it's many caches of food that it makes in the fall. That the brain volume decreases again in the spring may suggest that its extra metabolic cost is not worth the added survival value during that period.

Thanks for the interesting comments.

Hi Chris,

It seems true; brains are as big as they need to be and no bigger. The sea squirt, a relative of ours that shares a remarkable number of features with us, no longer needs its cerebral ganglion once past the larval stage so it goes away. I'm sure you've heard all the jokes.

It's not clear what survival advantage in the EEA would have been served by an even bigger brain for humans. The price would have been high not only in the cost of running it but in the skeletal changes to carry it around. It would have been pretty hard on mothers too unless the trend to delayed maturity and neoteny also increased. That's not cheap either.

I'll think about this more, sleep on it, and maybe say something useful. Thanks for your help.