The goal of this collaborative research project, led by Rena Torres Cacoullos at Pennsylvania State University and Shana Poplack at the University of Ottawa, is to uncover the strategies used by bilinguals as they switch seamlessly between their two languages. Bilinguals alternate between stretches of speech in the different languages even when speaking on the same topic to the same interlocutor. Beyond merely inserting isolated words from another language, such code-switching may involve juxtaposing entire phrases from different languages in the same speech event. Educators, journalists, and other public figures often disparage code-switching as haphazard mixing, yet it is now known that those who code-switch are in fact the most highly skilled bilinguals. What is still a conundrum are the rules governing code-switching. This project will integrate research and education by training students to analyze natural bilingual speech. The project will promote international inter-university collaborations, and by revealing the systematic patterning of bilingual speech, the results will help counteract the disparagement of what is often labeled Spanglish.
The New Mexico Spanish-English Bilingual speech corpus, constructed with prior NSF support, provides an appropriate database for the scientific investigation of code-switching because it contains copious instances produced during spontaneous bilingual conversations. The transcribed recordings will be comprehensively tagged for language and, based on this infrastructure, systematic quantitative analyses will be carried out to address two kinds of questions. One is the linguistic conditions on code-switching: At which syntactic and prosodic points in the speech stream do bilinguals prefer to code-switch? The other is the cognitive processes involved: Do contextual elements, such as cognate words and interlocutor language choice, act as triggers? By counting both instances of the presence and absence of code-switching, the researchers will be able to distinguish idiosyncratic tokens from community norms and rare occurrences from major patterns, estimate the propensity of code-switching at particular sites, and evaluate the actual role and scope of contextual elements that have been hypothesized to constrain patterns of code-switching.
