Code-mixing in simultaneous language acquisition.

Abstract

This thesis is based on the recorded speech and field notes of the author's three-year-old child who was acquiring three languages simultaneously (Chichewa, Chitumbuka and English). Chichewa is his mother's first language, Chitumbuka is his father's first language and English is both the language of the preschool that he was attending and the official language in Malawi. This study was unusual in that it involved African languages that are under-researched in the field of language acquisition and dealt with two cognate languages (Chichewa and Chitumbuka) and a non-cognate language, English. The fact that Chichewa and Chitumbuka strongly resemble each other may have made movement between the two easier for the child. The analysis of the child's recorded speech shows that he mixed more at the lexical level (64.2%) and less at the phonological level (6.3%). The findings demonstrate that what the child had learnt at school in English fulfilled a booster function when either Chichewa or Chitumbuka was used. The results also reveal that the child's language mixing was influenced by the topic of discussion, the context and the interlocutor's mixed input. The interlocutor's discourse strategies also had an impact on the child's use of mixing. The results therefore provide support for the bilingual bootstrapping hypothesis, the modeling hypothesis and the discourse hypothesis. The results also demonstrate that Chichewa was generally the matrix or host language when mixing occurred. At school, however, where only English was permitted, the question of a matrix language did not occur. Furthermore, the combination of lexical and grammatical morphemes demonstrates that Chichewa was dominant in the child's speech, in terms of the dominant-language hypothesis proposed by Petersen (1988). This study challenges the Free Morpheme Constraint and the Equivalence Constraint in that they do not appear to be universally applicable. Instead, the Matrix Language Frame Model is supported as it applies to code-mixing involving English and Bantu languages. This model was relevant, as the speech analyzed in this study involved code-mixing between English and the two Bantu languages, Chichewa and Chitumbuka. However, it was difficult to apply the Matrix Language Frame Model to some of the child's mixed utterances because the MLU was low. It is hoped therefore that researchers will create further models that will allow for an analysis of the mixed morphemes in single word utterances, especially for the Nguni African languages, which are agglutinative by nature.