Is Heritage Phonology Conservative?: Evidence from Toronto Heritage Cantonese

Some researchers and language teachers have observed that HL (Heritage Language) speakers sound “native-like” in their phonology (Polinsky and Kagan 2007) in contrast to their “incomplete acquisition” of morpho-syntactic features (Montrul 2008). Yet, these impressions of conservatism in phonology have largely been anecdotal. Only a handful of studies have specifically examined HL vowel systems using acoustic data (cf. Godson 2003 for Western Armenian; Chang et al. 2011 for Mandarin; Ronquest 2013 for Spanish). \ud \ud In this talk, I present data from a study of intergenerational (GEN 1 vs. GEN 2) vowel change in Toronto Heritage Cantonese to address the question of conservatism in HL phonology. Can we find evidence for conservatism through acoustic analysis of vowels in sociolinguistic interview data?\ud \ud The data examined comes from the Heritage Language Variation and Change (HLVC) in Toronto Project (Nagy 2011), which includes recorded and transcribed hour-long sociolinguistic interviews and responses to an Ethnic Orientation Questionnaire (EOQ). The F1 and F2 of 75 vowel tokens were measured and normalized across two phonetic contexts (pre-velar-coda and open-syllable) for each of four vowel categories (/i:/, /u:/, /ɛ:/, and /ɔ:/) from each of 17 speakers (1,020 vowel tokens). Rbrul (Johnson 2009) was used to run mixed effects models with phonetic context, generational background, sex, age, and Ethnic Orientation (EO) continuum scores (Nagy et al. 2014) as independent variables to model variation in the continuous dependent variables F1 and (in separate analyses) F2. \ud This presentation will focus on the following results:\ud \ud (1) There is a significant lowering effect (higher F1) of /i:/ and /uː/ in pre-velar context (compared to open syllables) across the speech community (p < 0.001). This lowering effect has also been described in Homeland Cantonese (Yue-Hashimoto 1972; Bauer and Benedict 1997). \ud \ud (2) There is a split pattern based on generation in the realization of /ɛː/ across phonetic contexts (p < 0.01 for F2) and a similar split based on generation and sex for /ɔː/ (p < 0.01 for F1). Unlike GEN 1 speakers, GEN 2 speakers produce allophonic splits in these two vowel categories. These allophonic splits have not been described as a feature of Homeland Cantonese (Yue-Hashimoto 1972; Bauer and Benedict 1997).\ud \ud (3) Both sex (p < 0.01) and Ethnic Orientation scores (p < 0.05) are significant predictors of variation in the F2 of /ɛ:/, but only for GEN 2 speakers. Female speakers produce significantly more centralized variants of /ɛ:/ than male speakers, while higher EO scores (more Chinese-oriented) correlate with more fronted variants of /ɛ:/. The fronting of /ɛː/ also occurs along with the alveolarization of syllable-final /ŋ/, which is an innovation that has also developed in Homeland Cantonese (Matthews and Yip 2011). \ud \ud Overall, the results show more evidence for innovation (likely due to contact with Toronto English) than for conservatism in the vowel system of Toronto Heritage Cantonese. Still in its infancy, the study of HL variation and change has much to contribute to developing a better understanding of the various mechanisms motivating linguistic change.\u