CENTER FOR RESEARCH IN
COGNITION & NEUROSCIENCES/ConnexConnex

Individuals make social inferences about others based on the way they talk. Speakers project a so-called social 'persona' through their linguistic choices, activating social representations in their interlocutors (Eckert, 2008). These social 'meanings' -- traits about the speaker’s perceived identity -- are indexed and carried by linguistic variants (forms that can alternate in a given context), and they shape how messages are perceived and processed. While sociolinguistics has long studied (phonetic) variation in English, much less is known about how persona construction operates in French, and especially for higher-level linguistic structures such as morphosyntactic variants.

This study dives into the social persona associated with different forms of French inclusive writing (e.g., étudiant·e·s vs. étudiant(e)s vs. generic étudiants). Using a semi-open production task, participants described the perceived character and appearance of authors of social media posts using these variants. Results reveal different social personae depending on the writing (mid-dot, parenthesis, and masculine).


These findings are a first step in a broader research project which aims at refining noisy-channel models (Gibson et al., 2013), to account for social priors in language processing. In this perspective, if listeners infer social identity from linguistic cues, these inferences may act as Bayesian priors, biasing interpretation in noisy contexts. By formalizing how social persona interacts with linguistic processing, the project aims at refining models of linguistic interaction that capture not only what is said, but also who is perceived to be speaking, and how the two interact.

Eckert, P. (2008). Variation and the indexical field 1. Journal of sociolinguistics, 12(4), 453-476.

Gibson, E., Bergen, L., & Piantadosi, S. T. (2013). Rational integration of noisy evidence and prior semantic expectations in sentence interpretation. Proceedings of the National Academy of Sciences, 110(20), 8051-8056.