Bilingual phonological development across generations: Segmental accuracy and error patterns in second- and third-generation British Bengali children

Introduction: While developmental norms for speech sound development have been widely reported for monolingual children, and increasingly for bilingual children, little is known about speech sound development across different generations of children growing up in heritage language settings. The purpose of the present study was to gain a better understanding of intergenerational differences in the phonological development of British Bengali children. Methods: Typically-developing second-generation and third-generation Bengali heritage children living in Wales (n=19), aged between 4 and 5 years, participated in a picture-naming task in Sylheti and English. The single-word speech samples were transcribed phonetically and analyzed in terms of consonant and vowel accuracy measures, and error patterns. Subsequently, logistic mixed-effects regression models were fitted to identify the factors that predict accurate speech patterns in the children's productions. Results: The results revealed high levels of accuracy in consonant and vowel production by both sets of children, particularly in English. On Sylheti consonants, second-generation children significantly outperformed third-generation children, however only on language-specific sounds. In contrast, generation was not a significant predictor for accuracy on English consonants, but all children performed better on shared sounds than on English-specific categories, and on stops than affricates. The third-generation children exhibited a greater number of error types in Sylheti than the second-generation children, and more common replacement of Sylheti dental stops with alveolars. Conclusion: The results suggest that third-generation children have less developed pronunciation patterns in the heritage language, but not the majority language, than their age-matched secondgeneration peers, however only on language-specific sounds. These findings indicate that differentiating between the phonological norms of monolingual and bilingual children may not be clinically sufficiently sensitive, at least in the minority language, and that more fine-grained