Colloquium: The social meaning of released /t/ among U.S. politicians: Insights from production and perception

June 4, 3:30-5pm, Cobb 201
Rob Podesva, Georgetown University

A great deal of work on the social significance of linguistic variation focuses on the indexical meanings associated with released /t/ (e.g., Eckert 2008, Podesva et al. 2006, Benor 2004, Bucholtz 2001). These studies have collectively suggested that /t/ releases are associated with intelligence and education, social meanings that can be recruited in the construction of articulate personas. While previous work has focused on speech production, little attention has been paid to the interpretation of released /t/. This paper examines the meaning of released /t/ among U.S. political figures – around whom discourses of ‘articulateness’ frequently circulate – from the perspectives of speech production, perception, and their interrelations.

The production study examines the speech of 6 prominent politicians (George W. Bush, Hillary Clinton, John Edwards, Barack Obama, Nancy Pelosi, and Condoleezza Rice). We examine an hour of speech from each politician, speaking to comparable audiences on similar themes. A VARBRUL analysis of 4,000 word-final /t/ tokens reveals that when taking into account the effect of linguistic factors, Clinton (0.724) and Rice (0.648) heavily favor the released variant in word-final position, while Edwards (0.310) and Obama (0.375) disfavor it (p≤0.000). An analysis of 1,500 medial /t/ tokens reveals first that the released variant appears much less frequently in this prosodic position than word-finally, and second that Pelosi releases stops most frequently (0.871) and Obama (0.188) least frequently (p≤0.007).

The perception study investigates the relationship between these production patterns and listener interpretations of variation. Pursuing a matched guise study using digital manipulation of stimuli, a between-subjects design was employed to test whether the addition of a /t/ release affects listener interpretations of otherwise identical utterances. 70 participants were asked to rate utterances spoken by the same 6 politicians on Likert scales for 9 adjective pairs. These pairs were selected based on results of focus group interviews conducted with undergraduate students about politician speaking styles. We draw three main conclusions based on the results of the web-based survey.

First, social meaning is sensitive to linguistic context (e.g. prosodic position). The presence of a release burst had a stronger effect in word-medial than word-final position (9 vs. 2 significant effects). It is argued that because medial releases are infrequent in production, they are relatively less predictable. They are therefore more salient when heard, making them better carriers of social meaning.

Second, listener interpretations of linguistic variants may depend on knowledge of how frequently variants are used by particular speakers. While most politicians were rated more favorably in their released guises, Obama was rated as sounding more intelligent in his unreleased guise (p≤0.029). Given that Obama was rated more favorably in guises that approximate his speech (he exhibited the lowest release rates in the production study), we conclude that listeners (most of whom indicated that they follow politics daily) may be influenced by their knowledge of how speakers habitually talk.

Finally, although conventional associations between linguistic forms and meanings exist, the social meanings attributed to variation are ultimately mediated by listener impressions of the speaker. For example, Edwards (p≤0.042) and Rice (p≤0.005) were unsurprisingly judged to sound more articulate in their word-medial released guises, but altering the phonetic realization of /t/ had no effect for other politicians. Previous studies emphasizing speaker agency rightly suggest that speakers can recruit /t/ releases to construct personas, but importantly, it must also be recognized that not all speakers can do so with equal effectiveness.