An artificial neural network stimulating performance of normal subjects and schizophrenics on the Wisconsin card sorting test

Mental diseases such as schizophrenia are being modeled by artificial neural networks in an attempt to understand the underlying neuropathological processes. We studied hospitalized psychiatric patients that met the DSM-IIIR criteria for schizophrenia (N=19), and normal subjects with no psychiatric history (N=18). Performance on the Wisconsin Card Sorting Test (WCST) by schizophrenic patients was poorer than normal subjects as estimated by various scoring measurements. We then modeled an artificial neural network, motivated by biological considerations, that is able to simulate performance of normals and schizophrenics on the WCST. In order to model the complex nature of the WCST, we designed novel learning rules based on non-associative learning paradigms. We found that there must be a minimum amount of noise, or inherent synaptic instability, for our model to perform similar to schizophrenics.