Outlier speakers and apparent effects: The case of variable subject placement in Spanish

Abstract

The current paper utilizes corpus data to examine variation and potential language contact effects regarding pronominal subject placement among first-generation immigrants in Atlanta, with particular attention paid to the individual speaker. The research questions that guide the study are the following: What linguistic and social predictors govern subject placement in Mexican Spanish spoken in Roswell, GA? What role, if any, does English contact play in the overall preverbal rates and/or constraints that govern subject placement? Do the data show any substantial individual differences with regard to rates of subject placement? If so, how might the inclusion versus exclusion of these speakers influence the overall interpretation of the results? This study employs a variationist sociolinguistic framework, and the speakers comprise 20 first-generation Mexican immigrants living in Georgia. Descriptive statistics are employed for purposes of overall as well as individual usage rates of postverbal subjects. In addition, linguistic (e.g., person/number, priming) and social predictors (e.g., sex, English proficiency, preferred media language) were incorporated into logistic regression analyses using Rbrul. Pronominal subject placement is most strongly influenced by person/number. Third-person singular pronouns favor postverbals while other pronouns favor preverbal position. Regarding social predictors, both sex and preferred media language appeared to play significant roles in the initial analysis; however, upon considering outlier speaker effects, the apparent effect of sex disappeared. The media language effect shows that those with a preference for English media favor preverbal subjects, suggesting a potential language contact effect. The current study is the first to analyze this phenomenon in southeastern U.S. Spanish. The findings have implications for the study of language contact and change, demonstrating the importance of considering how individual speakers could misrepresent the overall data, interpretations of results, and conclusions in sociolinguistic studies of bilingualism.