|
Esther Brown & David Harper, University of Colorado at Boulder
|
||
|
Phonological evidence of exemplar connections: an analysis of Spanish / English cognates |
||
|
Many studies use evidence from phonological variation and change to substantiate theories of lexical representation. Very few studies [with notable exceptions: e.g. Torres-Cacoullos & Ferreira (2000)] use evidence from bilingual speech production as a way to test theories of phonological representation in memory. This study analyzes 3756 tokens of Spanish word-final /s/ in the speech of 22 bilingual speakers of New Mexican Spanish and English in order to test the exemplar model of lexical representation (Bybee 2001). The realization of word-final /s/ in the two languages contrasts in that Spanish is an aspirating language (/s/ is variably realized as [s], [h], [ø]) while English is not. In cases in which orthographic, phonological and semantic similarities exist between the two languages (for example: Spanish gas/English gas), the exemplar model of lexical representation would predict greater retention of word-final /s/ in Spanish due to strong exemplar connections with English. In instances where words do not share cognate status (for example: Spanish nosotros / English we), no such effect is predicted. Our analysis supports this assertion; word-final /s/ in cognates reduces at a rate of 41% (N = 568) while non-cognate tokens of word-final /s/ reduce at a rate of 59% (N = 3188). Results of variable rule analyses using Varbrul (Rand & Sankoff 2001) reveal that the greater the semantic and phonological similarities between the words and morphemes in English and Spanish, the greater the effect of English on the variable aspiration and deletion of word-final /s/. Further, based upon the morphemic status of the word-final /s/ (lexical, plural, or verbal), the ‘cognate-effect’ is differentially felt: lexical tokens show a strong correlation; plural tokens a weak correlation; and verbal tokens show no correlation. These results support the theory of an exemplar model of lexical representation (Bybee 2001).
|
||
![]() |
|