Evolving another language: A selectionist and neural network approach to second language lexical memory
Abstract
This dissertation presents a neural network model of bilingual and second language lexical learning and memory developed within Carpenter and Grossberg's (1987) Adaptive Resonance Theory. This model was developed because there is no existing model of bilingual/second language lexical memory that is preferable to another in a principled way because none of these models are biologically plausible or embeddable within other accounts of psychological or physiological bases of human linguistic capacity. I hold that these two criteria, plausibility and embedability, can and should function as criteria in the selection of models. I describe an epistemological foundation, Campbell's (1974) Evolutionary Epistemology, within which I am able to identify desirable features of a model that can function as selective criteria—that is, criteria that can be used during a side-by-side comparison of models to select one model that is more promising than others. The model developed in this dissertation, the ART2-IF, accounts for a wider range of data than any other model, is biologically more plausible than any other model, and is the only model yet presented that is embeddable within larger theories of brain, language, and biology.
Degree
Ph.D.
Advisors
Berns, Purdue University.
Subject Area
Linguistics|Cognitive psychology
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.