Browse Bar: Browse by Author | Browse by Category | Browse by Citation | Advanced Search
Author(s): Grossberg, S. | Pepe, J. |
Year: 1970
Citation: Behavioral Science, 15, 359-362
Abstract: The hypothesis has been advanced thatcertain schizophrenic patients are in acontinual state of overarousal, leading topoor attention, and perhaps to schizophrenicpunning (Kornetsky and Eliasson, 1969;Maher, 1968). Physiological factors that canyield overarousal include a pathological re-duction in spiking thresholds of cells thatsend signals to many other cells, or alterna-tively a reduction in the strength of lateralinhibitory interactions between these cells.Even low energy but persistent peripheralexcitation can under these circumstancesbuild up high internal noise levels, becausesignals between the cells will not be suffi-ciently damped by the threshold or inhibi-tory mechanisms.
This note announces the occurrence ofanalogous phenomena in a rigorously definedlearning network having a suggestive psy-chological, neurophysiological, anatomical,and biochemical interpretation (Grossberg,1969a-c). We study in this network theserial learning of a long list of behavioralevents as it depends on spiking threshold. Innormal subjects, one characteristically findssuch phenomena as bowing (middle of thelist harder to learn than the ends) and pri-macy dominating recency (beginning easierto learn than the end)(Grossberg, 1969d).Altering the spiking threshold of the net-work systematically alters these effects.For example, as the spiking threshold de-creases, recency gradually gains strengthover primacy until finally recency prevails.This phenomenon is due to the buildup ofbackground noise as a result of the persistentpresentation of events in the serial paradigm.As ever more serial events are presented, theassociational strengths near the list's be-ginning are eventually competitively in-hibited by incorrect associations that formdue to inadequate threshold cut-off s. Corre-sponding difficulties in "paying attention"to long behavior dependencies also occur,and "punning" based on low-order associa-tions becomes plausible.
Topics:
Mathematical Foundations of Neural Networks,
Models:
Other,
