Experiments on Linguistic Profiling of Three American Dialects

Abstract

When we listen to someone speak, we make “informal, on-the-spot, and largely nonconscious speech judgments”, which are linked to social stereotypes. Purnell et al. (1999) showed that such judgments can affect housing access for racially accented speakers in that the likelihood of receiving an appointment to view a property is influenced by the accent of a caller. To determine if linguistic profiling remains observable in the contemporary housing market, I conducted an audit study of 90 publicly-listed rental properties in three targeted neighborhoods (Black and White working class; middle class) of Knoxville. In each call, I used one of my three native varieties, African American Language (AAL), Mainstream US English (MUSE), and Southern American (SA), and asked each property manager a series of questions about the unit and neighborhood. The outcome of each call was assigned a ranking along an appointment commitment scale ranging from traditional appointment (high ranking) to no appointment (low). A given voice receiving a higher commitment level in a given neighborhood is seen as an indication that the speaker’s voice is experiencing local (neighborhood-specific) prestige. Using the Fit package in R, the appointment commitment results were modeled to determine the likelihood that appointment type for a given neighborhood and voice will fall above a certain commitment level. The Figure gives the model predictions and shows that, for example, in the WW neighborhood the SA voice receives the highest level of commitment whereas in the BW neighborhood the AAL voice appears to receive the highest commitment. A subsequent ordinal logistic regression model reveals results consistent with local prestige, showing that a non-Standard voice fairs better (10 or even 25 times so) in a neighborhood with demographics which match the indexed, social characteristics of that speech variety. Comparison of local prestige effects reveals that SA enjoys more robust prestige effects than AAL in demographically similar neighborhoods. This is attributed to the Southern variety’s indexation to Whiteness and the social privileges that entails. More generally, the strong local prestige effects in these data may be due in part to calling managers of smaller rental properties, who may have higher stakes in the choice of a tenant. While we cannot be sure of the listener’s exact personal and professional stakes, I will also share qualitative details of several calls that illustrate the presence of steering, a discriminatory behavior which is a direct violation of the Fair Housing Act. The nuance exemplified in the qualitative results speaks to the inequities faced in the housing market, or indeed when interfacing with any institution, by non-Standard speakers. Qualitative assessment of the audit calls is also included, along with an in-depth analysis of speaker character assessment from the listening populations.