The cognitive reality of ‘talking like’. Modeling linguistic stereotype formation in preadolescents’ roleplay

Background and aim | Previous studies in cognitive sociolinguistics (Kristiansen & Dirven 2008) have uncovered elements of how (socio)linguistic stereotypes (Labov 1972) emerge and develop throughout childhood. For instance, Kristiansen (2010) found that children become significantly better at locating speech samples of regional varieties of Spanish along with age. In addition, Buson & Billiez (2013) demonstrated 9- to 11-year-olds’ growing ability to reflect on stylistic variation in audio samples of the same French speaker. These studies on linguistic stereotypes, however, take the linguistic form as a stimulus rather than as a research object. This is reflected in research designs from linguistic form (language) to social meaning (world). The reverse perspective, taking the social as a starting point and the linguistic as a research target, is vital to fully understand how stereotypes are created in the young language learner’s mind. Method | Adopting this reverse perspective, this paper aims to introduce and pilot a new stereotype elicitation method. Through a combination of roleplay and stylization (Coupland 2007), the method instructs (young) respondents to perform a social role (e.g. Can you pretend to be a news reader?), hereby investigating what linguistic resources (e.g. the rise of standard forms) are drawn from in this imitation process. To facilitate the imitation, respondents follow a three-tiered protocol of (1) a persona description (new name, age and hobbies) for a new social role; (2) performance of that role (creating a play) and (3) concluding interview questions. Case study | The method is applied to a proof of concept targeting the stereotypical link between English loanwords and specific social persona in Belgian Dutch (cf. Schuring et al. 2023). In peer groups of 5, 26 preadolescent respondents (age range 7-13 y/o) follow the three-tiered protocol for 5 roles, 3 of which orient towards English (gamer, soccer player, rapper) and 2 away from it (farmer, minister). Data and Analysis | The resulting corpus consists of 9 hours of video data, corresponding to 6022 utterances. All utterances are mined for English lexical material, which is then analyzed in terms of differences in the amount and nature of English use between English-prone roles, on the one hand, and Dutch-prone roles on the other. Regression trees (Tagliamonte & Baayen 2012) are implemented to uncover patterns of respondent age and persona performed. Results | Our findings indicate a clear presence of English-inspired stereotypes, with performances of English-prone roles generating significantly more English elements than performances of Dutch-prone roles (see FIG 1). No age effect was identified, but instead high levels of idiosyncrasies are attested throughout the data, with wide variation in the type/token ratio and the type (role domain specific vs. non-role domain specific) of English produced. Implications | The proof of concept on English loanword use demonstrates the new method’s potential to investigate stereotypes starting from a social stimulus (here social roles). Examining the linguistic dimension of the social role imitations, this paper contributes to the cognitive investigation of stereotype formation and, therefore, sheds more light on how children develop the link between language, world and mind.