One important goal in linguistic research is to explain why some patterns are very common across languages whereas others rarely or never occur. For instance, why are [a], [i], and [u] very common in vowel inventories? Linguists have long hypothesized that part of the explanation lies in languages optimizing the resources available to speakers and hearers to ensure an efficient communication. For instance, among all vowels which can be produced by the human vocal tract, [a], [i], and [u] are the most acoustically distinct. If languages select their vowel inventories in a way which maximizes the distinctiveness of their component vowels, it is expected that [a], [i], and [u] should occur frequently
In my research, I take this approach and extend it in two directions: (i) beyond static phonological inventories, to account for the distribution of sounds in context (e.g., patterns of allophony and contextual neutralization ) and, (ii) beyond phonology, to account for the distribution of morphemes in context (e.g. syncretism, phonologically conditioned allomorphy, homophony avoidance). I am also interested in how the optimization observed in phonological and morphological patterns interacts with grammar, i.e. whether it is part of grammar as a substantive bias shaping speakers’ productions or only arises as a by-product of transmission across generations.
I address these questions using a range of tools and methods including theories developed in formal linguistics (constraint-based grammars), experimental methods (using judgment data, perception data, production data, artificial language learning data), and computational modeling (Bayesian hierarchical modeling, probabilistic constraint-based grammar). French plays a prominent role in my work as a testing ground for linguistic theories and I have experience working with a range of French varieties and creoles.
I have extensive teaching experience in all main subfields of general linguistics, going from phonology to formal semantics, as well as in quantitative research methods.