| |
Abstract:
Category learning deficits have been observed in a wide
variety of neuropsychological disorders. These deficits appear to
be of two distinct types: 1) deficits affecting the accrual of
associations between exemplars and categories (as in Knowlton,
Mangels, & Squire, 1996, Science), and 2) deficits affecting
the ability to apply logical strategies (as in the Wisconsin Card
Sorting Test). These results challenge current neuropsychological
theories, as well as current theories of category learning. A
recent neurocomputational model called COVIS (Ashby, Alfonso-Reese,
Turken, & Waldron, 1997) is tested against category learning
data from a variety of neuropsychological conditions, including
depression, frontal lobe pathology, Parkinson's disease, and
Huntington's disease. COVIS assumes there are multiple category
learning systems, and assigns key roles to the caudate nucleus, the
prefrontal cortex, and the anterior cingulate. A neural network
implementation of COVIS successfully accounted for deficits from
each of these populations in category learning tasks of both
types.
|
