research



Style in Time: Online perception of sociolinguistic cues
My dissertation examined how socioindexical cues like high-rising terminals (also known as 'uptalk') shape listeners' real-time perceptions of speakers. Using a series of eye-tracking experiments, I investigated how different cues individually and jointly contribute to participants' judgments about speakers' social identities.

click for slides [CUNY 2021] click for slides [Dissertation Defense] email for the diss!

Speakers use styles - combinations of socially meaningful sounds - to construct and project different social identities, or personae. How do listeners recognize these sounds as belonging to a particular linguistic style, spoken by a particular kind of person? Sociolinguistic work has shown that the contribution of individual speech features (or cues) is highly mutable, and context-dependent - suggesting that listeners must integrate the meaning contributions of individual cues with all the other social impressions that arise when hearing someone talk. In this dissertation, I ask how and when they do so. I present results from a series of web-based experiments that ask how two cues - -in' (vs. -ing) and HRT ('uptalk') - modulate listeners' implicit, online inferences, and explicit, offline judgments about two social personae; a Tough and a Valley Girl speaker.

I first show that the contribution of these cues on explicit, offline judgments is dependent on the contextual information available to listeners: sociolinguistic cues don't vary just in the range of meanings they elicit, but also in the degree to which they give rise to specific meanings. I then show that listeners rapidly reconcile the meaning contributions of sociolinguistic cues when making inferences about a speaker's persona, and that the extent to which listeners' online and offline beliefs are modulated by a given cue is broadly proportional to the cue's socioindexical informativity. I then ask whether these results generalize to an additional set of four voices: I show that they do, and that listeners weigh the meaning contributions of a cue against their existing expectations about the speaker. Finally, I show that when faced with conflicting information about a speaker's persona, listeners will variably weight the contribution of a given cue based on voice-specific detail. Together, these results point to a probabilistic account of sociolinguistic perception in which listeners integrate various sources of contextual and linguistic information, but prioritize and place the greatest weight on the most informative aspects.

This dissertation draws on insights from the cognitive sciences in systematically investigating problems and phenomena that are central to the study of sociolinguistic style. In doing so, it provides a methodological and conceptual framework for future experimentation and theorization.

Expectations of speaker performance and social meanings in Tweets
Sociolinguistic work has found that listeners make social evaluations about speakers based on the linguistic features they use. I wanted to understand whether previously-established sociolinguistic effects would replicate when people didn't hear a speaker - but read a Tweet including the feature instead.
I ran a series of experiments, and found that they did - but that also, the highly performative modality of Twitter might have been shaping 'listener' expectations of 'speaker’ performance - thereby generating distinct sets of social evaluations for particular variants: the realizations of (ING), and the use of um and uh.

click for slides [NWAV 2018]

Effects of phonetic distance and social evaluation on vowel convergence
I ran web-based experiments where people recorded themselves repeating other speakers' sentences - to try and understand vocalic convergence behavior (i.e., why people might start talking a bit more like someone they've just heard).

click for slides [VALP 2018] email for manuscript

The frequency and distribution of um and uh in acquistion
In this corpus study, I investigated how young children's production of the delay markers um and uh changes over time, to meet changing conversational needs as their linguistic skills develop .
BTW: For this project, I annotated a large portion of the Providence Corpus for turn-type. Email me if you want the data.


click for poster [IASCL 2017] email for manuscript