Daniel J. Smith
In the speech of members of a Georgia (U.S.A.) community of immigrants from Latin America, as their bilingual sentences of Spanish/English reach a certain threshold of English content, Spanish morphosyntactic patterns begin to converge toward those of English. Data from naturally-occurring conversations by 56 children and adults of both sexes are analyzed within Myers-Scotton's (1993 [1997], 2002) Matrix Language Frame model. Eight language types were identified, including monolingual Spanish and English turns, codeswitched turns, and turns showing convergence (morpheme strings from one language with some grammatical structure from the other). An instance of each language type per turn was counted as a token of that language type. Each sentence of a multisentence turn was counted as a separate token. Tokens of each type were counted per informant. A rank ordering of the data by percentage of monolingual Spanish allows observation of how certain thresholds signal changes in the types of language mixing. Analysis reveals that percentages of monolingual and codeswitched utterances pattern in relation to percentages of utterances showing convergence, indicating that informants' Spanish does not begin to converge toward English until fewer than 70% of their utterances are monolingual Spanish. The data thus show that both codeswitching and convergence are mechanisms of language shift and change from dominance in one language to another. Social factors are also shown to be associated with the linguistic patterns.