In successful communication, speakers tailor their language tothe context and listeners make inferences about the speaker’sknowledge. Several current accounts propose that both speak-ers and listeners accomplish this by rational analysis of thestatistics in the environment, including their partner. Here weexamine perspective-taking behaviour in a dyadic conversationtask, where the same individuals act in the role of both speakerand listener. We model perspective-taking in both productionand comprehension, taking into account the dyadic situation.Our findings suggest that conversational partners weight theirown perspective more than the partner’s when speaking, andthe partner’s perspective more than their own when listening.We also find that in both production and comprehension, con-versational partners change the weighting of perspectives overtime, moving towards relying more on the partner’s perspec-tive at the expense of their own perspective. Surprisingly, wefind little evidence that listeners or speakers adapt to the id-iosyncratic statistics of their partner’s linguistic behaviour.

This paper investigates the processing of linguistic elements whose interpretation depends on retrieving information that was available earlier in the situation. Using the visual-world paradigm, we examine the processing of the verb return, which requires that an object has previously moved. We manipulated whether the moved object (and the movement itself) was described using language, by its typical label or by its location, or whether it was seen moving without that movement being labeled. We also manipulated whether the instructions were positive (e.g., Return the X), therefore requiring the listener to perform an action, or negative (e.g., Don’t return the X), which required no action. Results reveal a sensitivity to how information was introduced. Most importantly, with positive instructions, the naming of the object did not have an effect, whereas with negative instructions, naming was important to interpretation. These results indicate that the way information is introduced affects the status of this information when it is retrieved; these findings also lead us to explicitly consider the hypotheses that link language processing and visual attention.

While speakers tailor referring expressions to the knowledgeof their addressees, they do so imperfectly. Our goal here is toprovide an explanation for this type of pattern by extending aprobabilistic model introduced to explain perspective-takingbehavior in comprehension. Using novel production data froma type of knowledge mismatch not previously investigated inproduction, we show that production patterns can also beexplained as arising from the probabilistic combination of thespeaker’s and the addressee’s perspectives. These resultsshow the applicability of the multiple-perspectives approachto language production, and to different types of knowledgemismatch between conversational partners.

Social factors, such as partner familiarity (e.g., talking to a friend vs. stranger) may affect some conversations but not others. While researchers do not always control the partner identity when conducting interactive studies, the current empirical report of a language production experiment conducted via Zoom presents effects of partner familiarity (friends vs. strangers) on the form and content of referring expressions in the mundane task of describing everyday objects. First, speakers interacting with a friend were less disfluent than speakers interacting with a stranger, showing that more effort is invested in interactions with strangers. Second, speakers interacting with a friend showed more sensitivity to prior context. Surprisingly, these effects reveal that speakers are sensitive to the partner identity even when describing everyday objects whose labels are shared across all language users. The current findings suggest that researchers should consider social factors as part of the experimental design of interactive tasks.

Across the world’s languages, some word orders are more common. We focus on noun phrases, where it is more common for adjectives to follow the noun than to precede it. Because the interpretation of adjectives depends on the noun they modify, we propose and evaluate the new hypothesis that the order N- ADJ is more prevalent because it is beneficial for semantic processing. In a silent gesture task, speakers of four typologically-unrelated languages (English, Mandarin, Arabic and Spanish) communicated noun phrase meanings to a partner. We find, first, that our task tracks the typologically- preferred orders of nouns, adjectives and numerals in the noun phrase. More importantly, we find support for our semantic processing hypothesis: size adjectives, whose interpretation depend more on the noun they modify, were more likely to be gestured after the noun than shape adjectives whose interpretation is less dependent on the noun they modify.

Conversational interaction involves integrating the perspectives of multiple interlocutors with varying knowledge and beliefs. An issue that has received little attention in cognitive modeling of pragmatics is how speakers deal with the choice of words like come that are inherently perspectival. How do such lexical perspectival items fit into a speaker's overall integration of conversational perspective? We present new experimental results on production of perspectival words, in which speakers have varying degrees of certainty about their addressee's perspective. We show that the Multiple Perspectives Model closely fits the empirical data, lending support to the hypothesis that use of perspectival words can be naturally accommodated as a type of conversational perspective taking.