ABSTRACT This project studies one subsystem of human language, inflection (the ability to derive walked from walk or wives from wife). The phenomenon is to be studied in intensive detail from a variety of disciplines, aiming at a unified theory and database of how the system works computationally, how it is learned, how it varies represented in the brain. A key hypothesis is that regular inflection (walk-walks) is computed by a very general mental rule ("add -ed") whereas irregular inflection (come-came) involves memorization of two separate words in an associative memory, allowing limited generalization only to similar-sounding words. This hypothesis, which bears on current study of connectionist and symbol-processing models of cognition, predicts that regular and irregular inflection should dissociate in a variety of ways. The project include tests of how speakers of English and other languages apply inflectional patterns to existing and new words, experiments where subjects produce and recognize inflected forms in real time, crosslinguistic surveys of universals and variation in inflectional systems, and model construction.
