A Rational Model of Incremental Argument Interpretation: The Comprehension of Swedish Transitive Clauses.

Thomas Hörberg,T Florian Jaeger

doi:10.3389/fpsyg.2021.674202

Abstract

A central component of sentence understanding is verb-argument interpretation, determining how the referents in the sentence are related to the events or states expressed by the verb. Previous work has found that comprehenders change their argument interpretations incrementally as the sentence unfolds, based on morphosyntactic (e.g., case, agreement), lexico-semantic (e.g., animacy, verb-argument fit), and discourse cues (e.g., givenness). However, it is still unknown whether these cues have a privileged role in language processing, or whether their effects on argument interpretation originate in implicit expectations based on the joint distribution of these cues with argument assignments experienced in previous language input. We compare the former, linguistic account against the latter, expectation-based account, using data from production and comprehension of transitive clauses in Swedish. Based on a large corpus of Swedish, we develop a rational (Bayesian) model of incremental argument interpretation. This model predicts the processing difficulty experienced at different points in the sentence as a function of the Bayesian surprise associated with changes in expectations over possible argument interpretations. We then test the model against reading times from a self-paced reading experiment on Swedish. We find Bayesian surprise to be a significant predictor of reading times, complementing effects of word surprisal. Bayesian surprise also captures the qualitative effects of morpho-syntactic and lexico-semantic cues. Additional model comparisons find that it—with a single degree of freedom—captures much, if not all, of the effects associated with these cues. This suggests that the effects of form- and meaning-based cues to argument interpretation are mediated through expectation-based processing.

Highlights

Language understanding requires comprehenders to integrate incoming information to form hypotheses about the intended structure and meaning of sentences
To illustrate the predictions of the rational model, we focus on the subset of transitive sentences as well as the subset of NP and verb semantic properties for which the rational model predicts the greatest variation in Bayesian surprise
We have provided evidence for the hypothesis that the effects of these cues to argument interpretation are mediated through expectations, based on their joint distribution over NP arguments in previously experienced language input

Summary

Introduction

Language understanding requires comprehenders to integrate incoming information to form hypotheses about the intended structure and meaning of sentences. One of the central components of this process is argument interpretation: determining how the referents of the verb’s arguments relate to the events or states expressed by the verb This determines, for example, whether an argument refers to the actor of the event described by the verb, i.e., the most agent-like referent, or the undergoer of that event, i.e., the most patient-like referent (see e.g., Dowty, 1991; Primus, 2006). Previous work has found that incremental argument interpretation is affected by a wide range of linguistic cues This includes both form-based (e.g., case-making) and meaning- or discourse-based properties of the arguments (e.g., animacy, givenness), as well their interactions with verb semantics (e.g., Ferreira and Clifton, 1986; MacWhinney and Bates, 1989; MacDonald et al, 1994; Trueswell et al, 1994; McRae et al, 1998; Kamide et al, 2003; Gennari and MacDonald, 2008; Bornkessel-Schlesewsky and Schlesewsky, 2009; Wu et al, 2010)

Objectives

Methods

Results

Discussion

Conclusion