Acquisition of aspect in self-organizing connectionist models

Abstract

Two connectionist networks, DISLEX and DevLex-II, were used in this study to model the acquisition of lexical and grammatical aspect. Both models use multi-layered self-organizing feature maps, connected by associative links trained according to the Hebbian learning rule. Previous empirical research has identified a strong association between lexical aspect and grammatical aspect in child language, on the basis of which some researchers argue for innate semantic categories or prelinguistic predispositions. Our simulations indicate that such an association can emerge from dynamic self-organization and Hebbian learning in connectionist networks, without the need of a priori assumptions about the structure of innate knowledge. Our modeling results further attest to the utility of self-organizing neural networks in the study of language acquisition.