Abstract

This thesis investigates the variable phonology that characterizes the grammar of learner speech (interlanguage) in the context of data produced by native speakers of Spanish in a classroom environment. Specifically, the study follows the development of English /s/ plus consonant clusters (sC henceforth) across three levels of proficiency: beginners, intermediate and advanced. Because these clusters are absent in Spanish, learners syllabify these constituents with a preceding epenthetic [e] ([e]-epenthesis); as seen in the word 'snake' / sn ejk/ [arrow right] [e[barbelow]s.nejk]. The production of sC onset clusters is a variable interlanguage phenomenon influenced not only by the L1 interference, as implied above, but also by linguistic and extralinguistic factors. The results of the GoldVarb 2001 statistical analysis indicate that sC clusters occur more frequently in the following contexts: (1) when the preceding segment is a vowel; (2) in /s/ plus nasal sequences; (3) in more formal stylistic environments; and (4) in higher proficiency levels. The analysis of the variable data is couched within a stochastic version of the framework of Optimality Theory (Boersma's 1998 Gradual Learning Algorithm) because it allows the encoding of variability and its frequency effects within a language by means of a single grammar. Moreover, this study promotes a multidisciplinary and integrative approach that combines theoretical and methodological tools from three linguistic disciplines: sociolinguistics, second language acquisition and formal phonology, in an attempt to develop a "socially realistic linguistics" (Wilson & Henry, 1998)