Are You Asking Me or Telling Me? Using Linguistic Diversity to Explore Foundations of Perspective-Taking in Language Comprehension

Data and code for the manuscript: Are You Asking Me or Telling Me? Using Linguistic Diversity to Explore Foundations of Perspective-Taking in Language ComprehensionDescriptionAlthough classic accounts of conversational perspective-taking claimed that common ground (shared knowledge) is essential for successful comprehension, contemporary work shows that, not only do core processes draw on multiple perspectives [1,2], but also that the relevance of shared vs. privileged knowledge varies according to utterance goals: e.g., whereas imperatives (“Pick up the …”) relate primarily to shared referents, questions (“What’s above the cow with the spots?”) require listeners to consider privileged referents [3]. Thus, recognizing the relevant speech act (e.g., question, command, assertion…) seems critical for calculating perspective. In this regard, it is fortunate that, in languages like English, lexical-syntactic and prosodic cues usually allow listeners to differentiate these speech acts from utterance onset. But are these cues essential, or could perspective be efficiently calculated without them? We address this question by leveraging a feature of Mandarin Chinese, namely that some utterances cannot be reliably identified as a question/statement until their final constituent. This is because a wh-word such as shénme(“what”) is expressed in situ, entailing sentence pairs with structures such as woman-wear-is-what (Engl: “What is the woman wearing?”) and woman-wear-is-shoes (Engl: “The woman is wearing shoes”). In other words, the Mandarin version of “what is the woman wearing?” (woman-wear-is-what) and “the woman is wearing shoes.” (woman-wear-is-shoes) are syntactically identical until the final sentence element. Further, prosodic cues also fail to differentiate the two sentences before the final element is heard [4]. Thus, syntax and prosody jointly entail that the speech act (question vs. statement) is formally ambiguous until sentence-end. We reasoned that, if speech act identification is key for the rapid framing of perspective, perspective-sensitive aspects of real-time comprehension should be delayed in these Mandarin utterances. However, because speech acts are themselves by-products of a broader communicative ecosystem (i.e., the context provides the motivations for interlocutors to provide or seek information), higher-level task goals may provide cues allowing listeners to correctly anticipate that the utterance is a statement vs. a question before hearing the disambiguating information at sentence-end. If so, this would allow for the earlier on-line identification of relevant referents for a question (known only to listener) vs. a statement (known only to speaker).To assess this, we used a two-player Visual World task based on [3], where displays contained cartoon animals wearing clothes and footwear. Critically, visual barriers entailed that different clothing and footwear items were occluded for a given player. At the end of each trial, players needed to answer verification questions posed by the experimenter (e.g., How many horses are wearing dresses?) to ensure they would query the other player about items unknown to them and describe items unknown to the other player. Importantly, one player was a confederate who produced scripted utterances on critical trials. These were initially ambiguous (i.e., in terms of speech act, with their status as a question/ statement being disambiguated by the utterance-final word) and entailed the anticipation of different referents depending on which speech act is incrementally inferred. The measure of interest was the participant player’s anticipatory gaze to the target (vs. a competitor), before hearing the utterance-final word. We manipulated task knowledge across two conditions (between participants). In the structured task condition, the participant player was tasked with answering verification questions about clothing, whereas the confederate player was responsible for footwear. Thus, each player could expect the kind of information the other player would seek vs. provide. This provides a higher-level cue that, in principle, could be used to infer whether a co-player’s utterance is likely a statement or a question based on the first few words. In the unstructured task condition, no information of this sort was provided, entailing that there was no way to infer the speech act before hearing the final word of a critical utterance. References[1] Heller, D., & Brown‐Schmidt, S. (2023). The multiple perspectives theory of mental states in communication. Cognitive Science, 47(7), e13322.[2] Dixon, P., & Bortolussi, M. (2024).  The analogical reader: A cognitive approach to literary perspective taking. Cambridge University Press.[3] Brown-Schmidt, S., Gunlogson, C., & Tanenhaus, M. K. (2008). Addressees distinguish shared from private information when interpreting questions during interactive conversation. Cognition, 107(3), 1122-1134. [4] Shyu, S. I., & Tung, T. L. (2018). Mandarin wh-phrases and prosody. Studies in Prosodic Grammar, 3, 32-76.