CATEGORIZATION DYNAMICS IN A HYBRID CONNECTIONIST MODEL

点击翻译按钮获取中文摘要

The proposed research explores the potential of hybrid systems to account for normal human category-learning behavior. Category learning refers to any situation in which a person must learn to classify stimuli into discrete groups, such as a medical student who must learn to classify collections of symptoms into disease categories. The proposed research combines a hybrid, rule-plus- exemplar model of categorization developed by Erickson and Kruschke with a model of perceptual choice reaction time behavior developed by Usher and McClelland. This new, dynamic model will be tested by applying it to data from four category-learning experiments. The first three experiments involve a rule plus exceptions. In the first two, the way people learn to classify separable- and integral-dimension stimuli in this setting will be evaluated. In the third, a response-signal paradigm will be used to interrupt people while they are making classification in order to evaluate how processing priorities change over time between rule- and exemplar-based representations. The final experiment evaluates whether dimensional attention is subject to priming as is visual spatial attention. The proposed research is relevant to our knowledge of mental health insofar as it will help identify underlying mechanisms in normal adults. Moreover, to the degree that evidence is obtained showing that rule- and exemplar-based processing components are dissociable, this research will have broad implications for teaching methods and therapies for both normals and those with mental challenges or brain damage.