L3 Portuguese by Spanish-English Bilinguals: Copula Construction Use and Acquisition in Corpus Data

Abstract

Previous research on third language (L3) acquisition has shown that the source language for transfer to the L3 can be either an L1, an L2, or both (Bardel & Falk, 2007; Flynn et al., 2004; Rothman, 2014). It has been hypothesized that either typological similarities between languages previously acquired and the target language (Rothman, 2010), or language status (L1 vs. L2) of previous acquired languages (Bardel & Falk, 2007) determine cross- linguistic influence. This dissertation investigates the acquisition of copula structures in L3 Portuguese by Spanish-English three groups of adult bilinguals: L1 English L2 Spanish, L1 Spanish L2 English, and L1 Spanish/English (i.e., heritage speakers of Spanish for the purposes of this dissertation). Language use by both native speakers (L1 Spanish, L1 English, and L1 Portuguese) and learners (L3 Portuguese) is analyzed using word embeddings and logistic regression modeling. The goal of these methods is to reveal patterns of copula use and acquisition. Copula constructions were chosen because they allow for the combined investigation of form, syntactic frame, and concept/meaning, as proposed by third language acquisition scholars. The main goal of this dissertation is to shed light on both transfer patterns from previously acquired languages (i.e., Spanish and English) on L3 Portuguese, and establish L3 Portuguese developmental patterns across bilingual groups. Results show evidence of L3 Portuguese development for all three groups of Spanish-English bilinguals. However, transfer patterns from Spanish and English onto L3 Portuguese are not the same across all groups, varying in degree depending on the copula construction. These results conflict with the Typological Primacy Model, which predicts that L3 acquisition in adulthood starts o from a wholesale transfer of the pre-acquired language system that is most typologically similar to the target language (Rothman, 2014). This dissertation offers support instead to L3 acquisition models that take into consideration structural characteristics of individual constructions, and how similar or different these are between source and target languages, including models such as the Parasitic Model (Hall et al., 2009).