Words of estimative probability (WEPs), such as ‘possible’ and ‘a good chance’, provide an efficient means for expressing probability under uncertainty. Current semantic theories assume that WEPs denote crisp thresholds on the probability scale, but experimental data indicate that their use is characterised by gradience and focality. Here, we implement and compare computational models of the use of WEPs to explain novel production data. We find that, among models incorporating cognitive limitations and assumptions about goal-directed speech, a model that implements a threshold-based semantics explains the data equally well as a model that semantically encodes patterns of gradience and focality. We further validate the model by distinguishing between participants with more or fewer autistic traits, as measured with the Autism Spectrum Quotient test. These traits include communicative difficulties. We show that these difficulties are reflected in the rationality parameter of the model, which modulates the probability that the speaker selects the pragmatically optimal message.

In anticipating upcoming content, comprehenders are known to rely on real-world knowledge. This knowledge can be deployed directly in favor of upcoming content about typical situations (implying a transparent mapping between the world and what speakers say about the world). Such knowledge can also be used to estimate the likelihood of speech, whereby atypical situations are the ones newsworthy enough to merit reporting (i.e., a nontransparent mapping in which improbable situations yield likely utterances). We report four forced-choice studies (three preregistered) testing this distinction between situation knowledge and speech production likelihood. Comprehenders are shown to anticipate situation-atypical meanings more when guessing content (a) that a speaker announces (rather than thinks), (b) that is said out of the blue (rather than produced when prompted), and (c) that is addressed to a large audience (rather than a single listener). The findings contrast with prior work that emphasizes a comprehension bias in favor of typicality, and they highlight the need for comprehension models that incorporate expectations for informativity (as one of a set of inferred speaker goals) alongside expectations for content plausibility.