ABSTRACT
Dual language (DL) programs in the United States are increasingly promoted as a promising model that serves both language majority and minoritized children. However, many researchers also question whether the programs are truly serving the language minoritized students, or are these programs only treating their language as a resource to serve the already privileged group. Using structural topic modeling and critical discourse analysis, this study employs a raciolinguistic perspective to unpack the ideologies underlying the discourses in over 200 DL programs in a new Latinx South state of the United States. Our findings show that DL programs are promoted as a model that brings cognitive and employment benefits to its students. Moreover, private corporate power has a strong existence and influence on the establishment, staffing, and promotion of DL programs. The benefits to English speaking students are highlighted, while the interests of language minoritized children are largely ignored in the discourse.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 In this paper, we adopted the term ‘language minoritized’ to refer to the linguistic groups who are historically marginalized in the society, following Cervantes-Soon et al.’s (Citation2017) justification: ‘We use the term minoritized, to indicate racial, ethnic, or linguistic groups that may be labeled minority by whitestream society (Urrieta, Citation2010) but who are by no means “minor” ‘ (p. 406).
2 The figure is generated from structural topic modeling using software R. It shows the frequency distribution of topics across the corpus. For example, ‘Topic 1: grade, kindergarten, spanish’ is about the general introduction of structures of DL programs, which is most often discussed by DL programs. The fourth most discussed topic, ‘Topic 8: global, world, cultur’ emphasizes the global citizenship of students through learning in DL programs. After data pre-procession, we ran several STMs using the searchK function in the stm package in R. We first ran the model with a range of topics from five to 55, then narrowed to five to 15. The goal is to select a number of topics to maximize held-out likelihood, semantic coherence, and lower-bound and to minimize the residuals (Roberts et al., Citation2019). We finally narrowed down to nine topics to analyze further.