Motor skill, coordination & learning
|||newspam||| at nezumi.demon.co.uk
Mon Jun 6 04:01:18 EST 2005
Glen M. Sizemore wrote:
> MB: The following claim about how we learn motor skills was made in the
> context of a more general discussion on AI and intelligence. I am not a
> biologist but I believe it to be New Age gibberish:
>>Let's take something simple. Learning how to type. When
>>you start out entering the word "the", your brain thinks
>>about t, then sends a signal to your left forefinger, tells
>>it to push, and repeats this process for each character.
>>As you practice, you stop sending a message/character but
>>you begin to send a message/word. Eventually, the cognitive
>>piece of the brain forgets all about the "recipe" of what
>>actions and their order that need to be taken to type the
>>I've heard some people calling this muscle memory. Whatever
>>it is, it's no longer resident in the brain.
> I believe the last claim to be wildly inaccurate for humans. Although it
> might be more nearly true for insects with distributed ganglia.
> GS: Well, ummm, there's some facts here probably, mixed with nonsense.
All the best pseudoscience uses this formula - genuine facts spun and
carefully twisted with BS.
> First, some of what is written could be construed as a description of
> certain behavioral facts. Karl Lashley, for example, pointed out decades ago
> that arpeggios could not be a chain of individual responses where feedback
> from each response controlled the next response and so forth, because the
> inter-press intervals were to short. But many times we learn extended
> sequences as a series of individual units but these come to be fluid
> movements, probably best described as a single behavioral unit.
But isn't it just possible that the arpeggio is done as an open loop
sequence of muscle commands sent by the coordination centre without it
bothering to wait for the feedback from each individual strike?
> As far as "muscle memory" goes, I doubt that even most of those that use
> this somewhat silly term think that the nervous system does not mediate such
I think this one stems mainly from sports coaches. It describes how it
feels to the practitioner rather than what is really happening.
> MB: My understanding is that in humans all of the higher functions for
> coordination and motor control are inside the brain. And that once
> learned motor skills are devolved to other subconcious parts of the
> brain like the cerebellum and motor cortex. And that the muscles have
> relatively simple local nerve cells for fine grain feedback control and
> reflex and a nerve connection back to the brain via the spinal cord.
> GS: There is no question that most behavior involves the brain. However,
> some aspects of coordinated movement are probably organized below the level
> of the brain. When animals behave in molar ways, these circuits are, no
> doubt, turned on and "blended together" by the brain.
Sorry I don't understand the term "molar" in this context. I guess my
question is how much devolved muscle control is there outside the brain?
I would guess there could perhaps be something like local neural PID
type control loops on all the important muscles but anything more than that?
> I tried a web search and most of what I found supports my viewpoint. But
> also I found some weird apparently genuine references about using the
> blink reflex to measure intelligence and decided to ask here for
> guidance. eg
> I'd appreciate suggestions on where to look for a reliable up to date
> online review of what is known about how we learn new skills from a
> biologist's perspective. Thanks for any pointers to reference material
> that would clarify this position.
> GS: You should keep in mind that how the brain mediates behavioral function
> is not well understood, despite the arm-breaking, self back-patting engaged
> in by neuroscientists after the "decade of the brain."
Oh. I was hoping that the biological viewpoint might be a bit more
concrete now after the much publicised brain decade.
> I teach, I usually tell students that we have a near complete understanding
> of, say, habituation of the gill-withdrawal reflex in Aplysia. We are not
> even close to describing all that is going on when, for example, a rat
> acquires a food-reinforced lever-press response class. In most cases, we
> aren't really asking the right questions.
So is it possible then that advanced neural net computer simulations may
soon be preferable test subjects for trying to understand some of these
complex issues in a more controlled white box environment?
Interestingly the threshold for simulating mutually interacting
gravitating particles has just crossed 10^10.
I'd appreciate any pointers to current biological neural net research in
the open literature. Most of the interesting stuff I found mentioned on
the web was in expensive pay per view journals.
More information about the Neur-sci