Generalization Across Segments in Second Language Consonant Identification

Abstract

This article examines the extent of differences between second language (L2) learners in their abilities to identify L2 consonants and provides evidence for linguistic generalization from one consonant to other consonants. It distinguishes among different sorts of models of the relationship between segments: (a) segmentally specific models in which each segment is treated separately and (b) generalized featural models in which segments that share features are treated as a group. It also examines (c) ordered models in which one segment might rely on another previously existing one. The article presents the results of an experiment with 40 Korean learners of English identifying English obstruents in different prosodic locations, and it examines the pattern of variation in perceptual accuracy across the listeners. Results indicate that manner perception varies from person to person as a unit, regardless of the segment, suggesting a single generalized featural skill. Voicing identification, however, does not vary as a unit across the different prosodic positions, suggesting different skills that are applicable to different prosodic positions and are acquired separately. There is also evidence for an implicational relationship between pairs of segments that differ in markedness, providing some evidence for an ordered model; however, such patterns are not pervasive across different contrasts.