People all communicate in unique ways, yet we can notice when they use “the same words.” This suggests that linguistic information is special kind of shared knowledge. However, there is also a lot of information we get from language use that isn’t necessarily linguistic. We can get impressions of how a person is feeling, what size their body is, and how they might be raced or gendered. I study how “non-linguistic” features affect attention to linguistic information, impacting language learning processes. I theorize the joint social and linguistic significance of variation as emerging through the construction of meta-linguistic beliefs – that is, conscious beliefs about the social significance of specific language behaviors. I want to know when these beliefs emerge, how and why they predictably differ, and how they apparently shape representations of linguistic structure. I use a combination of computational modeling and behavioral experiments to explore these questions.

Key words: sociolinguistics, computational psycholinguistics, developmental psychology, person perception, sociophonetics

Impact of Social Beliefs on Assessments of Child Language Development

Trust and Infant Word Learning

As children age, they exhibit predictable changes in their preferences for social partners, acquiring preferences for social partners who are reliable, and for those who possess both physical and behavioral characteristics shared by the child’s own social group. I suggest that this process also applies to the selection of linguistic informants for language acquisition. In my dissertation work I proposed an explicit computational model to relate beliefs about social categories to infant behavior on tests of word learning and recognition. This affiliative learning model provides a theory of how developing systems of social category beliefs and language learning mutually support one another, and a higher-level explanation of why within-language sociolinguistic differentiation emerges. 

My model therefore posits the development of sociophonetic knowledge as an important benchmark in child language acquisition. Sociophonetic knowledge structurally connects knowledge of phonetic variation to psychological constructions of group membership, and suggests a developmental explanation of emerging cognitive differences between children of different SES levels.  Differences in the language skills of different populations may not be driven simply by availability of input, but also by variations in sensitivity to diversity of sources. Children from disadvantaged backgrounds have distinct sociolinguistic landscapes to navigate, and there is existing evidence that their attention to features such as race may develop at a different pace.

Social Perception and Child Vocabulary Scores

The failure of linguistic assessments to explicitly test sociolinguistic skills may obscure the strengths of children learning non-mainstream language varieties. Measures of development that rely on benchmarks in socio-phonetic development may actually advantage low SES children. To explore this possibility, as a Presidential Postdoctoral Fellowship at the University of Minnesota, I have been the lead investigator on a study combining a test of vocabulary, behavioral measures of children’s selective trust in language users, and computational models of performance to empirically evaluate the role of variable social sensitivity in determining assessment outcomes.

Impact of Social Beliefs on Assessments of Adult Language

The Role of Social Beliefs in Lx Learning

I am interested in how overtly assigning labeled social roles to language users affects perception of their language. In a collaboration with Rachel Hayes-Harb & Shannon Barrios (University of Utah), we examined affiliative social perception in adult language (Lx) learning. Our forthcoming paper in the Journal of Second Language Pronunciation presents the results of an online behavioral experiment wherein adult English speakers heard words in an artificial language. During an exposure phase, participants heard these words produced by two different voices, one of which was labeled and visually depicted as a “teacher,” while the other was a “student.” The voices assigned to each label and their phonetic behavior were manipulated to investigate whether the social context of labeling the teacher and student roles would predict participant judgements of pronunciation quality as performed by a third voice. As predicted, participants demonstrated a preference for productions matching the phonetic behavior of the teacher

Although the outcome of the study may be considered intuitively unsurprising, it represents important and novel empirical evidence that social factors affect how adults differentiate input in Lx acquisition. This data offers us insight into how learners of a language negotiate individual variability and acquire generalized knowledge of linguistic structure. In order to report that one pronunciation from a novel voice was “better,” participants must selectively recruit their experiences with individual voices. While this experiment targeted classroom language learning, the approach is readily and productively adapted to systematically investigate the impact of other social contexts on adult language learning, supporting the definition of perceptual mechanisms underlying the production of numerous discriminatory linguistic phenomena, including accentism, native-speakerism, and linguistic racism.

Perception of Race and Gender in Speech Materials

The design and perception of study materials may introduce implicit bias into experiments through the suggestion of ideological orientations towards social roles. I am also interested in how implicit biases held by participants shape their responses to language. Race and gender are demonstrably perceived differently in different modalities and are assigned significance in varied ways. Yet despite so much individual and cultural variation, race, gender and other social categories emerge as important organizing social concepts. The belief that that an individuals’ race or gender can be “perceived” reflects the way these social categories are discursively linked by common understandings of social power. I am interested in how manipulations of assigned labels and evaluative judgments of those labels may moderate sensitivity to linguistic information by directing effort and attention to non-linguistic signals.

Impact of Perceiving Race and Gender on Speech Intelligibility

I have been part of an effort funded by the NIH grant Race, Ethnicity, and Sentence Intelligibility in Normal Hearing and Hearing Impairment to (1) create and publish speech study materials which reflect racial and gender diversity and (2) conduct intelligibility experiments which examine the effects of this diversity on intelligibility in younger people, and older people with and without hearing loss. We have conducted an online experiment with a demographically diverse sample of 245 younger people, without identified hearing impairment. Participants were tested on-line and assigned to audio-only and audio-visual conditions. They were exposed to subsections of the recorded talkers and asked in an intelligibility task to report the speech they heard, then later to indicate what they believed the individual talkers’ racialized identities to be. Reports of racialization in audio-only and audio-visual conditions will be compared with prior results from norming studies. Data collection has been completed and analysis is ongoing. However, preliminary analysis reveals substantial differences in speech intelligibility across talkers of different racialized identities, despite the fact that all of the talkers learned English from birth in the U.S. or Canada. In-person experiments with groups of older adults with and without hearing loss are ongoing.

Philosophy of Linguistic Diversity

My research program investigates foundational assumptions about the nature of linguistic experiences as they are defined in relationship to social experiences. My work seeks to reconcile traditional goals of universalism, humanism and empirical rigor with the enormous gap in understanding associated with the underrepresentation of diversity. I apply DisCrit and intersectional lenses to advance a meta-epistemological approach to understanding linguistic perception.

A meta-epistemological approach supposes that an experience of meaning and the interior recognition and analysis of that experience are empirically distinct events. Given that both linguistic and non-linguistic experiences can be co-discursively indexed as having the “same” meaning, I seek to ground our understandings of social and linguistic variation in a theorized common faculty of social difference perception. This approach unites observations from language perception, language production, and non-linguistic social perception with the goal of systematically and equitably incorporating understandings of social and linguistic diversity into our accounts of normative language development.