Bespoke Probabilistic Modelling Reveals Subtle Effects of Autism on Pragmatic Optimisation in the Expression of Quantification

Autism is often associated with difficulties with pragmatic language use. However, several studies on the interpretation of quantifying expressions (e.g., ‘some’) failed to observe differences between autistic and neurotypical participants. In this study, we present data from a more naturalistic free-production study comparing how French-speaking autistic and neurotypical adults use quantifying expressions. To capture the complexity of this approach, we employ a bespoke probabilistic model that estimates a theoretically motivated pragmatic optimality parameter, which we take as a measure of pragmatic ability. Our results show that, as a group, autistic participants are less likely to produce optimally informative quantifying expressions than their neurotypical peers. Nonetheless, the difference between the two populations is relatively small, pointing to a difference in degree rather than kind. These findings contribute to ongoing debates about pragmatic ability in autism, and highlight the value of bespoke probabilistic modelling for testing hypotheses about subtle differences in latent traits and their complex empirical signatures.