Demonstratives, words like ‘this’ and ‘that’ in English, are a special class of function words: They appear in all documented languages, and are acquired early in ontogeny, together with pointing, which is another linguistic universal. Demonstratives are also supposed to have emerged early in language evolution, since historical linguists and typologists have not been able to trace back their origins. In a bid to explore these features of demonstratives, I will present cross-linguistic data from adults and computational models of demonstrative use, exploring the spatial and social parameters underlying demonstrative choice across languages. I will also present the results of a recent study investigating the role of demonstratives in early language acquisition. Combining head-mounted eye-tracking with machine-learning modeling of naturalistic interactions between mothers and infants, we show that demonstratives provide augmented inputs that scaffold early word learning. I will argue that demonstratives orchestrate attentional dynamics in multimodal social interaction, which is a fundamental pragmatic function that might explain their unique characteristics.

• Amal Hassen, Heiner Drenhaus, Ingo Reich & Robin Lemke
  Retrieval cues hierarchy in sluice resolution: The case of French. 📄 Abstract (PDF)
• Hannah J. Seemann
  On the emergence of expectations in discourse introduced by ambiguous cues. 📄 Abstract (PDF)
• Ana Celmare, Merel CJ Scholman & Vera Demberg
  Multimodal Investment in Epistemic Stance Expression: Gestural and Prosodic Indicators of Speaker Commitment
• Jun Lyu & Yuhan Liu
  So what happens next?: Reference prediction in implicit consequential events in L2 Chinese
• Yizhuo Zhang, Alexander Martin & Alexandra Lorson
  Expectations of Social Indexicality: Mandarin Chinese T/V distinction. 📄 Abstract (PDF)
• Timo Buchholz, Martin Villalba & Klaus von Heusinger
  German anaphoric demonstrative pronouns shape discourse expectations about contrast and animacy
• Katja Haeuser
  German-speaking 10–12 year-olds revise their noun predictions when reading prediction-inconsistent gender-marked articles
• Haochuan Wang
  Cross-lingual expectation failure in definitional clauses: structural variation and prediction across five EU languages
• Peter Furko & Eszter Szakács
  Pragmatic Markers as Signals of Discourse Expectations: A Corpus-Based Study of Political and Human–AI Interaction. 📄 Abstract (PDF)
• Jessie Jingxian ZENG
  Interpreter-mediated discourse predictability: A surprisal-based account leveraging large language models. 📄 Abstract (PDF)
• Jet Hoek, Klaus von Heusinger & Hannah Rohde
  Expectations while you wait? Not replicating anti-locality effects at the discourse level
• Marloes van Moort & David Rapp
  Supportively beneficial or convincingly problematic? Effects of supporting contexts on the use of (in)accurate information. 📄 Abstract (PDF)
• Ronja van Zijverden, Marloes van Moort, Karin Fikkers, Antal van den Bosch & Hans Hoeken
  The language of likes: A corpus study on linguistic predictors of text agreement. 📄 Abstract (PDF)
• Li Kloostra
  Discourse processing is subject to coherence illusions and the binding problem in memory: Eye-tracking and acceptability studies on presupposition resolution with ‘again’

The past two decades have seen a sea change in theorizing when it comes to language comprehension. From the highly incremental but fundamentally input-driven approaches of the 1980s and 90s, theories now propose that comprehenders generate probabilistic expectations in real time. In this talk, I ask how far this predictive view extends into discourse, and what kinds of representations support it. I adopt the view that sentence and discourse interpretation are not sharply distinct systems, but rather differ in the deployment of representational formats that support prediction under different conditions. Discourse coherence expectations arise when the evolving representation makes the relevant relations available and predictive. This perspective clarifies a broader question in predictive processing, asking not simply whether comprehenders predict, but instead what it is they predict and how different representational formats make such predictions possible in the first place.