Empirical studies have demonstrated that when comprehenders are faced with informationally redundant utterances, they may make pragmatic inferences to accommodate the informationally redundant utterance (Kravtchenko & Demberg, 2015. Previous work has also shown that the strength of these inferences depends on prominence of the redundant utterance – if it is stressed prosodically, marked with an exclamation mark, or introduced with a discourse marker such as “Oh yeah”, atypicality inferences are stronger (Kravtchenko & Demberg, 2015; 2022; Ryzhova & Demberg, 2020). The goal of the present paper is to demonstrate how both the atypicality inference and the effect of prominence can be modelled using the rational speech act (RSA) framework. We show that atypicality inferences can be captured by introducing joint reasoning about the habituality of events, following Degen, Tessler, and Goodman (2015); Goodman and Frank (2016). However, we find that joint reasoning models principally cannot account for the effect of differences in utterance prominence. This is because prominence markers do not contribute to the truth-conditional meaning. We then proceed to demonstrate that leveraging a noisy channel model, which has previously been used to model low-level acoustic perception (Bergen & Goodman, 2015), can successfully account for the empirically observed patterns of utterance prominence.

Experimental pragmatics tells us that a metaphor conveys salient features of a vehicle and that highly typical features tend to be salient. But can highly atypical features also be salient? When asking if John is loyal and hearing “John is a fox”, will the hearer conclude that John is disloyal because loyalty is saliently atypical for a fox? This prediction follows from our RSA-based model of metaphor understanding which relies on gradient salience. Our behavioral experiments corroborate the model's predictions, providing evidence that high and low typicality are salient and result in high interpretation confidence and agreement, while average typicality is not salient and makes a metaphor confusing. Our model implements the idea that other features of a vehicle, along with possible alternative vehicles, influence metaphor interpretation. It produces a significantly better fit compared to an existing RSA model of metaphor understanding, supporting our predictions about the factors at play.

When humans communicate, they typically adapt to their conversational partner in how they speak, and in how they interpret what the conversational partner says. In the area of pragmatic language comprehension, there is so far little work that has studied the individual differences between listeners with respect to adapting to a given speaker. We investigated which individual cognitive factors correlate with listener's ability to associate speakers with humorous utterances. We found that working memory updating (as measured by the Keeping Track Task) was a significant predictor of adaptation to the speaker. These findings are in line with a recent related study (Schuster et al., 2023) which investigated speaker-specific adaptation to the use of uncertainty expressions. We furthermore observe a correlation between speaker adaptation and the Faux Pas Test. This task is used for measuring theory of mind abilities and is believed to specifically tap into intention recognition, an ability which is also very relevant to joke comprehension.

One of the advantages of using metaphorical expressions over literal ones might be that speakers can convey not only the intended property, but also its degree. For example, when hearing “John is a shark”, the listener might infer that the speaker aims to communicate that John is as mean as a typical shark. We present experimental findings supporting this hypothesis, along with a novel metaphor interpretation model, which is implemented within the Rational Speech Act framework. We compare our model's predictions to those of an existing RSA model of metaphor understanding, within which the listener infers just the presence or absence of a feature as opposed to its degree, and find that our model produces a significantly better fit.