Neural mechanisms of learning

Stephan Anagnostaras stephan at
Wed Sep 20 03:39:12 EST 1995

It's generally agreed that the learning of motor skills (and all of
those things we do "naturally without thinking") is mediated by a
separate neural system than the learning of facts (and anything
you can verbalize).  This evidence dates back to the patient HM,
who received bilateral hippocampal damage. You can read this account
anywhere, but briefly, HM lost the ability to learn new factual,
semantic, verbal, or declarative information but still has (to this
day) the ability to learn things like motor skills, classical
conditioning, and a few other abilities.  The localization of
the learning which he can do is diverse;  many motor skills
appear to be learned in the basal ganglia, notably the dorsal
caudate.  The process by which that kind of learning occurs is
not available to central analysis (that is, it is unconscious);
although the "hippocampal" declarative system may also acquire
information about these kinds of tasks, it is often not
sufficient to generate the responses (i.e., for example, you
could learn and memorize and verbalize all of the steps
necessary to tie your shoe, but you wouldn't be able to do it
unless you practice the steps and there would be little or
no transfer from your verbal knowledge of the steps; for
many tasks, that kind of knowledge actually impedes acquisition of
the skill.)  

At the cellular level, the same kinds of synaptic changes thought to
mediate all of these kinds of learning are actually thought to be
similar.  Synaptic faciliation (e.g., LTP) is certainly a
suitable mechanism for all motor learning; given the abundance
of LTP in the hippocampus, and evidence that the blockade of LTP
in the hippocampus is relatively equivalent to a hippocampal lesion
(on learning), this is probably also a suitable substrate for
declarative knowledge (how this would work is up to debate
most people envision synaptic faciliation operating in some kind
of distrubuted correlational matrix system -- however, it is
still unclear how meaning is derived in the representations in these
systems.  Since motor learning doesn't have any meaning or
necessary verbal informational content this isn't a problem for those
kinds of learning.)  LTP is probably one of the mechanisms by
which memories are initially established, at least for certain
kinds of learning.  This primarily involves glutamate receptors,
and to a lesser extent gaba and opioid receptors. I'm not sure what
is meant by "50-100 chemicals" below -- there are certainly nowhere
near that many known neurotransmitters to be involved and certainly
at least 20 times as many "chemicals" (i.e., >1000) known to be involved in
at least some portion of these processes. 


In article <laszloDF6H2C.ABG at>, laszlo at (Laszlo) wrote:

> During the first few months of learning to drive a car (or learn martial 
> arts techniques, what have you) the brain is conscious of sets of actions 
> it wishes to teach itself. Through repeating these on a step-by-step 
> basis sooner or later trains the neural network itself (the brain, rather 
> than the mind) how it should respond to particular actions or intentions.
>   As for the actual physiological changes which take place in and amoung 
> the neurons, the closest one might come to understanding this [still 
> uncertain] area is to study how analogous man-made neural networks are 
> trained.
>   Very loosely (I don't know if this helps in the least) an output or 
> ultimate goal is established and steps are taken to achieve the goal. If 
> the goal is accomplished (you got the car out of 1st gear without causing 
> smoke or horrendous noises) you feel happy and releived, sending chemical 
> signals back through the network confirming that those last steps taken 
> were [somewhat] correct. After doing this umpteen times the motor neurons 
> have learned accurately enough that they, themselves can move your hand 
> without the training, leaving your frontal lobes free to think about signing 
> up for those martial arts classes <smirk>.
>   I won't even touch the neurotrasmitter names issue. I've heard guesses 
> that there are 50 to 200 chemicals used - somehow I doubt anyone has a 
> solid idea yet how individual neurons *really* work. <shrug> hope this 
> helps. 
>   Laszlo
> Hugh S. Selsick (127hugh at wrote:
> : Does anyone know what the latest consensus is on the neural mechanisms of 
> : learning, both with regard to memory and to motor skills? What physical and 
> : physiological changes occur during the process where an initially foreign 
> : reaction becomes almost(?) reflex? For example, when we first learn to
> : the control of the clutch was difficult and required a great deal of 
> : concentration. But after a few months it is so natural we do it 
> : automatically without thinking. 
> : Related to this question is the problem of whether or not all reflex 
> : reactions are hardwired into our systems or whether theyn can be learned or 
> : unlearned. eg Can a karate master override his reflex to cringe in the face 
> : of an attack and replace it with a reflex to sidestep and block?
> : Any help would be much appreciated.
> : Hugh Selsick
> : Wits Physiology Department
> -- 
>                        Government out of Internet!!!!
>       Laszlo, CNE     laszlo at     San Francisco    DoD#3434
>  "Here we are, born to be kings, we're the princes of the universe" -Queen


More information about the Neur-sci mailing list