Do animals have an internal physics engine?

When you throw a ball to me (provided you’re not totally skilless), I catch it. This involves a sequence of events (weighing back your arm, checking the distance from you to me, levering your shoulder etc.) that need to be executed on cue in order to be accurate.

I can understand that over time we develop a better “sense” (for lack of a better word) for these motions. After all, when we are children, we have not yet developed and mastered the requisite skills needed to perform these actions (e.g. throwing, catching, …).

However, why is it that after performing these actions on a regular basis that we do become better at them? I mean if there were absolutely no means of mathematical measurement (or approximation), I ought to miss the ball more often than I catch it. So how is it that I am able to master the art of catching (and you the art of throwing)?

Is there some sort of calculating mechanism inside the human brain that can determine the order and precise sequence of physical motions accurately enough that we can rely on it (as we so often do)? Do we have some sort of finely tuned “physical engine” - much like computer games that use physics engines in order to simulate realistic movement(s)?

And if so, are there some who have more innately finely tuned “physics engines” than others? Would this explain the “natural” affinity that some people display when undertaking certain tasks requiring skill (such as driving, throwing, running etc.)?

I think it’s generally agreed that we do have innate computational powers of the kind you describe, but how that computing is being done is not well understood. We’ve got some idea about the parts of the brain that are responsible for things like kinesthetics, proprioception, and hand-eye coordination. But how those parts of the brain are really “programmed” remains to be discovered. What’s good for day-to-day functioning, but bad for reverse-engineering the system, is all these fantastic calculations and adjustments are executed at a subconscious level. We simply know how to throw a ball to hit a target without having to think about it much, but programming a robot to do the same thing is necessarily a carefully thought-out task.

It’s a very good question. I think the answer is yes, we do have something akin to a “physics engine”, but it it’s still quite a mystery how it works.

You might be interested in reading this recent discussion on the matter.

My memory is failing rapidly, but I seem to recall that in The Computer and the Brain (Mrs. Hepsa Ely Silliman Memorial Lectures) John Von Neumann raised the possibility that the brain may use a different form of mathematics than we use consciously.

Looking at the remarks on that Amazon page doesn’t back me up very well, I know. There’s a possibility I’m thinking of something said by Norbert Wiener at around the same time.

I think it’d be just about impossible to have a mathematical physics engine in the sense that games would. It’d be very hard to do all that. If we had such an engine, then unusual circumstances like throwing a ball to someone else when you’re both on a merry-go-'round would not be so surprising (from your point of view the ball curves drastically and inexplicably).

It seems liklier to me that we notice paterns in what we do and get good at timing muscle signals. So, throwing a ball would require the same kind of coordination as, say, speech - getting all those muscles to react in the right rapid sequence to form syllables.

Or that we have hardwiring for some of these things. Maybe walking and speech and throwing are learned, but swallowing is hardwired - and for caterpillars, even walking is hardwired.

Well, yes… of a sort.

The cerebellum mediates the “learning” of such activities, whether it be catching a ball, walking, brachiation, manipulating a clam knife, or typing. Through trial and error, the brain learns to coordinate visual input, prediction of trajectory, proprioception, and conscious intent in order to subliminalize the detailed motions, balances, and adjustments necessary to perform even complex physical activity.

It’s not numerical modeling of the physical world, as if it were a giant game of Halo, but a tuned response based on many attempts that result in failures and successes whose feedback fine tunes the response. It’s a lot closer to a Rube Goldberg machine than a mathematical model.

Electronics engineers and computer scientists mimic this sort of response with a class of circuits called “neural nets.” These can be implemented with analog circuits, as the brain does, or with numerical approximations.

If you’re interested in learning more about how this works, take introductory Psychology in college, and then take upper division classes on the subjects of central nervous system, neuroanatomy and neural networks.