On Segmental Factorability in Second Language Learning

This paper distinguishes between two models of learning a second language phonological classification system, segment-by-segment learning, which starts with individual elements and generalizes to broader featural classification skills later on, and featural classification, which starts with featural classification schemes first and fills in specific details later on. Evidence reported here comes in two forms. First, subjects who are more accurate in distinguishing one pair of segments are also more accurate with other pairs contrasting the same feature. Second, identification patterns for unfamiliar (‘new’) segments approximate those expected for featural independence, as estimated using Jfactor analysis. The tentative conclusion is that second language learners create generalized classification schemes first and develop more detailed and robust classification skills with increased experience.