In article hc8 at pinot.ext.jussieu.fr, tij at valadon.espci.fr (tareq i. j. albaho) writes:
> hello;
>> please forgive a theoretical physicist for invading your privacy, but
> i am currently working on connectionist computation, which is supposed
> to be at least partially inspired by neuroscience.
>> i have a question concerning hebbian learning, which has long been
> accepted in my domain as the acceptable model of learning via neural
> plasticity. it was proposed by donald hebb in 1949.
>> in this model, the connection between two "neurons" is strengthened if
> both neurons are firing, and this leads to learning.
>> however, only recently i read that there is no direct experimental
> support for the hebbian model of neural plasticity.
>> is this true?
>> can someone please give me references on what mechanisms
> neuroscientists currently believe are behind learning?
>> thanks so much,
>> -----------------------------------------------------------------------
> tareq. tij at pmmh.espci.fr>http://pmmh.espci.fr/~tij/personal.html>> *--->laboratoire de physique et mechanique des milieux heterogenes<---*
> *------>ecole superieure de physique et de chimie industrielles<------*
Boy, what a minefield you have just wandered into!
It is partially true that there is no direct support for hebbian
plasticity, in that no-one has yet seen spontaneous changes in the
connections between neurons as a result of a learning experience (at
least not incontrovertibly). There are lots of reasons why this might be
the case even if hebbian-type changes *do* occur: it is quite hard to
find pairs of synaptically connected cells in a behaving (i.e. learning)
animal, to name just one.
Some other people have mentioned LTP as a model of hebbian plasticity.
LTP occurs in a laboratory under somewhat artificial conditions, not
directly related to learning, but the end result is an increase in
synaptic strength that comes about when activity in a pre- and
post-synaptic neuron are associated artificially, which resembles Hebb's
postulate. The physiological mechanisms underlying LTP are now quite well
understood and a lot of people believe that it's hard to see what all
that very complicated machinery could be for, if not for producing the
same plastic changes in a normal animal. Many labs have tried very hard
to determine whether there is any connection between LTP and real
learning, but so far there is no hard and fast evidence (though there is
plenty of soft evidence, and I could send you some refs if you want). The
issue is still unresolved at present. One of the main problems is that
both learning and LTP are the end result of very complicated series of
events, and it is quite easy to find correlations which may or may not be
spurious. However, no-one has really produced a convincing alternative
hypothesis for how memories might be formed so we keep plugging away,
hoping it will all become clear in time!
Kate Jeffery
kate at anatomy.ucl.ac.uk
