sturla.molden at svt.ntnu.no -- delete this (Sturla Molden) writes:
> The problem with "memory models" is that they are derived from invalid
> studies, e.g. studies of human patiens that have experienced cerebral
> ischemia. Cerebral ischemia produces effects that are different from
> those of selective hippocapal lesions, e.g. deficits on the delayed
> nonmatching-to-sample test which do not depend on hippocamal
> integrity. The animal experiments that have seemed to
> support this hypothesis have either been bady performed or
> insufficiently controlled. The problem with "spatial models", although
> more consistent with the data than memory models (both in animals and
> humans), is the empirical fact that hippocampal lesions seem to
> disrupt some tasks that are of "non-spatial" nature. These include
> trace conditioning and latent inhibition. The problem with
> "computational models" is that they are so premature and simplified
> that they seldom allow any testable predictions to be drawn.
Hmm. What about H.M.? His amnesia can't be attributed to ischemia,
can it?
Also, there is some difficulty in comparing the results of
delayed-nonmatch-to-sample experiments using humans versus animals.
In humans the tests usually have been done by giving the subjects
verbal instructions such as "choose the object that is different from
the one you saw most recently". Animals, on the other hand, can't be
motivated by verbal instructions, so they require an extended training
procedure using some kind of reward. It isn't clear, as far as I
know, that human amnesics given the same kind of training used in
hippocampal-lesioned animals would do any worse than the animals.
Certainly there are many situations where amnesics show no conscious
recollection of a previous experience but still show strong implicit
effects on their behavior.
I am personally convinced that both the memory and spatial models have
substantial validity, and that the great task of hippocampal research
is to come up with a "grand unified theory" that accounts for both
(rather than expaining one aspect away, as the current theories do).
-- Bill
