Date: February 17-18, 2017

Venue: Cordura Hall, Stanford University [map]

In recent years, the study of meaning has seen rapid advances in two still largely disconnected areas: probabilistic semantics/pragmatics (Goodman & Lassiter 2015; Franke & Jäger 2016) and psycholinguistics (Sedivy 2014). Both of these areas have drawn on traditional formal semantics/pragmatics for inspiration, especially Grice's original insights, while adding other perspectives from cognitive science. On the one hand, the burgeoning field of probabilistic pragmatics has been hugely successful in modeling a wide variety of phenomena as the outcome of iterated Bayesian reasoning between speakers and listeners -- including scalar implicature, M-implicature, figurative meaning, pronoun resolution; as well as the interpretation of gradable predicates, quantifiers, spatial relations, generics, and referring expressions (e.g., Frank & Goodman 2012; Goodman & Stuhlmüller 2013; Degen et al 2013; Lassiter & Goodman 2013; Carstensen et al 2014; Kao et al 2014; Qing & Franke 2014; Kehler & Rohde 2015; Potts et al 2015; Bergen et al 2016). On the other hand, psycholinguistic research in experimental semantics and pragmatics is painting an ever more complex picture of the interactions of multiple factors in the computation of speaker meaning, including literal meaning, perspective-taking, prosody, availability of alternatives, the Question Under Discussion, world knowledge, and speaker-specific idiosyncrasies (Heller et al 2008; Grodner & Sedivy 2011; Brown-Schmidt 2012; Kurumada et al 2014; Degen & Tanenhaus 2015; Pogue et al 2016; Yildirim et al 2016).

Despite the two areas' very similar goals -- to understand a) how listeners compute pragmatic inferences from the observed signal, and b) how speakers choose an utterance to produce in the first place -- there has been a surprising disconnect between the communities. This is partly attributable to different foci: the computational community has aimed at providing proof-of-concept models for the outcome of inference processes in various domains, testing these models on judgments of speaker meaning collected through offline measures. The psycholinguistic community, in contrast, has focused for the most part on online processing, trying to establish which types of information from the signal and the extra-linguistic context are processed at which point in time. 

The workshop aims to work towards an integrated cognitive science of meaningTo this end, researchers who study meaning from a variety of perspectives will come together with the goal of building mutually beneficial bridges between the communities. At least the following questions will be discussed:
  • How can rational, probabilistic models of meaning be extended to online processing? Should they be?
  • What insights from rational models of language use in other domains (e.g., syntactic or acoustic processing) do or do not apply to the study of meaning?
  • What well-known phenomena from psycholinguistics should modelers of meaning be more aware of and try to integrate (e.g., individual differences, resource limitations, incrementality)?
  • In building theories of meaning, what can we learn from data/insights from acquisition, language change, language variation, sociolinguistics, language disorders, brain imaging data, and NLP?
Invited speakers:
Sarah Brown-Schmidt Vanderbilt University
Vera Demberg Saarland University
Ashwini Deo The Ohio State University
Mike Frank Stanford University
Michael Franke University of Tübingen
Noah Goodman Stanford University
Chigusa Kurumada University of Rochester
Dan Lassiter Stanford University

Contact: Judith Degen

Important dates:
Submission deadline: November 1, 2016
Notification of acceptance: December 1, 2016
Conference dates: February 17-18, 2017