WhenOctober 26, 2017 03:30 PM - 05:00 PM
WhereRosenwald Hall, Room 015
Contact InformationLinguistics Department
DescriptionHeather Burnett. The National Center for Scientific Research (CNRS), Paris Diderot University

Scaling Up: Meaning-Based Approaches to Social Stratification

In this presentation, I present a new methodological approach to analyzing quantitative patterns of social stratification in large sociolinguistic corpora within a Third Wave approach to the meaning of variation (TW; see Eckert 2012). TW has yielded important insights into the meaning of variables and the identity construction process; however, as discussed in Schilling-Estes 2002, Kendall 2011, a.o., it remains an open question "whether the interpretations gleaned through close-up analysis of individual initiative shifts can be generalized in any way to the larger community" (Schilling-Estes 2013: 382). `Scaling up' TW analyses is challenging because analyses set in this framework are usually based on data associated with speakers' beliefs and ideologies which have been obtained through substantial ethnographic research with a limited number of participants (Eckert 2000, Podesva 2007, Zhang 2008, a.o.). 'First Wave' macro-level studies (i.e. Labov 1966, Wolfram 1969, Trudgill 1974, Sankoff & Vincent 1977, a.o.) are generally based on around 1hr long sociolinguistic interviews with a large number of participants (often 50-150). As such, it is unclear whether sufficient information concerning speakers' identities/ideologies can be extracted from such short interviews, and the sheer size of the corpora can render the development of detailed analyses concerning indexicality for this many agents daunting.

This presentation addresses these challenges by building on Bourdieu & Passeron (1970), Bourdieu (1979)'s observations that aspects of individuals' performed identities can be elicited from even very short interviews, provided the right kinds of questions are asked. I argue some sociolinguistic corpora contain the appropriate ideologically oriented questions which allow the analyst to develop some assessment of how the speakers in the corpora situate themselves in their social world. Using the stratified Montréal 84 corpus of spoken Montréal French (Thibault & Vincent 1990), I construct a measure of speakers' values from their different responses to the ideologically loaded survey question: Qu'est-ce que c'est bien parler? Mal parler? `What is speaking well? Speaking badly?'. I then incorporate this measure into a computational model that formalizes TW in terms of game-theoretic pragmatics (Burnett 2017), particularly in terms of signaling games with a Bayesian approach to speaker-listener reasoning (as in Franke 2009, Frank & Goodman 2012). I present a new quantitative study of variation in negative polarity items du tout vs pantoute `at all' in the corpus, and I show how we can use the model to determine the indexical fields (Eckert 2008) associated with each variant from the observed patterns of variation combined with speaker ideologies (as instantiated in our measure).
CategoriesConferences/Lectures, Conferences, Seminars
Persons with disabilities who need an accommodation in order to participate in this event should contact the event sponsor for assistance.