First and second language learners typically employ different repair strategies for novel consonant clusters. Deletion dominates among child L1 learners (e.g., /blu/-->[bu]; Dyson & Paden 1983), while epenthesis dominates among adult L2 learners (e.g., /blu/-->[belu]; Weinberger 1987, 1994). This paper argues that a bias toward these distinct repair strategies emerges automatically when an Optimality-Theoretic (McCarthy & Prince 1995, Prince & Smolensky 1993/2004) grammar is learned using the Gradual Learning Algorithm (Boersma 1998, Boersma & Hayes 2001). Within this framework, error patterns during L1 learning affect not just the immediate productions of the child, but also the final constraint ranking that is established. When sources of syllable structure and segmental markedness are acquired in tandem during the L1 stage, a covert MAX-C >> DEP-V ranking results; when this ranking is transferred as part of the initial state for L2 learning, the consequence is a preference for epenthesis. These findings are illustrated through learning simulations, and the contributions of various aspects of the grammar are discussed.
Selected Proceedings of the 5th Conference on Generative Approaches to Language Acquisition North America (GALANA 2012)
edited by Chia-Ying Chu, Caitlin E. Coughlin, Beatriz Lopez Prego, Utako Minai, and Annie Tremblay
Table of contents