Contrasting Age of Arrival and Length of Residence in dialect contact

Abstract

The speech of migrant populations experiencing dialect contact provides insight into one of the key mechanisms of language change (Trudgill, 1986; Dodsworth, 2017) and patterns of dialect acquisition (Chambers, 1992; Tagliamonte & Molfenter, 2007; Nycz, 2011; Otheguy & Zentella, 2012). Age of Arrival (AoA) and Length of Residence (LoR) in a new community are two factors that predict successful acquisition of new linguistic variants, but since it’s often the case that speakers who arrived earlier are also the ones who have lived the longest in the host community, their effect may have been masked in previous studies (Siegel, 2010). This paper reports on a systematic analysis of a corpus specially built to disentangle the effects of these predictors’–forty Brazilian Portuguese speakers from the Northeastern states of Alagoas and Paraíba living in the Southeastern state of São Paulo, balanced for sex, AoA (-19 y.o.; 20+ y.o.), and LoR (-9 or 10+ years), all of whom are between 20-45 years of age from rural areas with no more than a highschool education. We analyzed 18,500 tokens of two phonetic and two morphosyntactic variables: (i) coda /r/ (porta ‘door’) as aspirates or tap/retroflex; (ii) /t,d/ before i as stops or affricates; (iii) sentential negation (simple: não vi, double/post-verbal: não vi não/vi não ‘I haven’t seen’); and (iv) nominal agreement (os menino-s, os menino-ø ‘the boys’). While variables (i) and (iii) distinguish Northern-Southern dialects, (ii) and (iv) differentiate rural-urban dialects. Usage of “Paulista/urban” variants were compared to the rates of speakers of similar sociodemographic profiles both from the area of origin (48 speakers from rural Alagoas and Paraíba) and São Paulo (58 speakers). Proportion tests indicate that speakers have approximated Paulista speech regarding coda /r/ (NE: 0% tap/retroflex; migrants: 31%), /t,d/ (NE: 13% affricates; migrants: 38%; χ2 = 3,953(1), p<0.001) and negation (NE: 83% simple negation; migrants: 87%; χ2 = 38,393(1), p<0.001), but not nominal agreement. Results from multivariate mixed-effects logistic regression models in R, including speaker and lexical item (for phonetic variables) as random effects and AoA and LoR as continuous predictors, show that AoA correlates only with the phonetic variables (the earlier the arrival, the more Southeastern/urban variants), suggesting—contrary to previous studies (Siegel, 2010)—that the critical period for acquiring morphosyntactic variables may close sooner than for phonetic variables. LoR correlates only with coda /r/ (the longer the residence, the more Paulista variants), showing that phonetic variables are differently sensitive to traits that distinguish broad dialectal areas or the rural-urban divide. Interestingly, self-reported “identity” indices on a 0-10 scale of how “Paulista” and “Alagoano/Paraibano” the speaker feels also align with the different types of variables’ geographical distribution: the indices correlate with coda /r/ and negation (the Northern/Southern variables) and don’t correlate with /t,d/ and nominal agreement (the rural-urban variables). Thus, while AoA and LoR distinguish phonetic and morphosyntactic variables, dialect acquisition also involves a complex web of differently defined regional and individual identities.