Title:  Thinking “Um…” out loud: Children’s inferences from speech disfluencies

Abstract: Conversation is profoundly structured by an expectation of timeliness. As a result, adults derive a range of inferences when a speaker pauses or produces disfluencies (e.g., “um”) in speech—inferring a speaker’s knowledge, willingness, and more. In this presentation, I will discuss 3 studies (total n = 305) exploring when and how children ages 4-9 use disfluencies to infer a speaker’s knowledge and preferences. Our mental states leak out in all manner of cues during conversation, and across these studies, I hope to paint a picture of how children begin to make use of that information to become smooth conversationalists, and also to learn about their social worlds.

Title: Connecting the Language Computed and Text Collected, a glance at the usage of computational content analysis in social language study

Abstract: Computational content analysis forms a highly potential yet unexplored method for the social study of language. Although there are quite a lot of projects and papers that leverage this method and make remarkable explorations, this newly emerging and somehow alienated method is far from mature. The general group of researchers lacks an understanding of what it can do, what it can’t, and what should be done in the seeable future. In attempting to address this topic, my presentation will provide a glance at the possible directions of leveraging computational content analysis, drawing on examples from my current projects and recent remarkable publications. I tend to dive into how these methods are configured (in a comprehensible way) and how they could be adjusted for a few practical or theoretical needs illustrated. What’s more, I would also like to articulate my own understanding of these methods and pose hints on how we could enlarge our imagination on research projects with the possibilities brought by these methods.

Title: The fate of the unexpected: Downstream consequences of prediction violations

Abstract: Amid increasing interest in the role of prediction in language comprehension, there remains a gap in our understanding of what happens when predictions are disconfirmed. Are unexpected words harder to process and encode because of interference from the original prediction? Or, because of their relevance for learning, do expectation violations strengthen the representations of unexpected words? In two experiments, we used event-related potentials to probe the downstream consequences of prediction violations.

Title: Two Models of Metalinguistics

Abstract: In this presentation, I will introduce two innovative cognitive-driven, corpus-based models for investigating “metasemantic presupposition” and “metapragmatic presupposition,” integrating cognition and computation with metalinguistic theory in sociolinguistics and linguistic anthropology (Silverstein 1976, 1993; Urban 2006, 2018). The first model develops a distributional metric in hyperbolic coordinates, mapping metasemantic presupposition as co-textual relatedness and metapragmatic presupposition as co-textual dependency. The second model adopts a dynamical-system approach (Niyogi 2006) to map metasemantic and metapragmatic presuppositions conditioned upon meta-level awareness of synchronization of linguistic categories over time. Empirically, the first model is applied to relatively long-term (hundreds to thousands of years) shifts of semantic categories with case studies including color, kinship, smell, and shape-based classifiers based on ancient and modern Chinese texts. The second model is tested against diachronic change data from the Google Ngram corpus of written American English, modeling shorter-term changes spanning decades to centuries in semantic categories related to color, kinship, smell, and political and religious identities. Together, this research advances cognitive-driven computational models informed by the contemporary literature of sociolinguistics and linguistic anthropology, and also proposes a novel framework for generating empirical hypotheses that contribute to bridging micro-level psycholinguistic and macro-level sociolinguistic phenomena.

Title: Must speakers perceive to achieve?: Investigating the role of feedback perception in audience design

Abstract: Speakers tailor communication to a particular listener based on what they believe that listener knows, a skill called audience design (AD; Clark & Murphy 1982). For example, a UChicago student who is giving directions to a campus visitor may refrain from referring to Regenstein Library as “the Reg” because they assume that a visitor may not understand that abbreviation. In general, speakers are very good at AD (Horton 2005), but sometimes they make mistakes: For example, Navarro et al. 2020 found that speakers under high cognitive load were more likely to make egocentric AD errors in a referential communication task; speakers with higher fluid intelligence and working memory capacity were overall less likely to make errors, however. Though not as well studied as the aforementioned cognitive factors, social factors—especially the ability to make use of feedback cues from listeners—have been shown to be another important part of successful AD (i.a. Horton 2005, Krahmer & Swerts 2005). In this presentation, I will outline a study design that aims to investigate the extent to which failing to perceive feedback cues contributes to failures in audience design. This is joint work with Monica Do and Serpil Karabuklu.