| |
Abstract:
Abstract: The tail of the caudate nucleus is a region of the
striatum (part of the basal ganglia) that receives input from all
visual cortical areas (except V1). This input is convergent, in the
sense that one caudate neuron may receive input from as many as
10,000 visual cortical neurons. The caudate, in turn, projects to
areas in frontal cortex, including premotor cortex. Further, the
caudate receives input from midbrain dopamine areas that have been
implicated in reinforcement learning. On the basis of these
findings, it has been suggested that the caudate, and more
generally the striatum, is in a perfect position to link percepts
(i.e., visual cortical representations) to actions (i.e., motor or
motor planning representations). This view is supported by many
studies showing procedural learning deficits in individuals with
striatal dysfunction (i.e., Parkinson's Disease). Recent evidence
also suggests that the striatum (and especially the caudate
nucleus) may contribute to learning in complex categorization
tasks. In this paper, we use the known anatomical and
neurophysiological properties of the cortical-striatal system to
develop a biologically plausible model of implicit category
learning. We then report the results of simulations in which this
model is used to account for the performance of normal and
pathological human data in implicit category learning tasks.
|
