Alignment between Thematic Roles and Grammatical Functions Facilitates Sentence Processing: Evidence from Experiencer Verbs

How do comprehenders process grammatically marked structures? And how is markedness represented? Previous syntactic, semantic, and psycholinguistic work has studied this question extensively in the domain of experiencer verbs. We consider different psycholinguistic approaches to markedness as it relates to alignment on various grammatical scales, including ones that posit effects of alignment in comprehension (Bornkessel & Schlesewsky 2006), grammatical constraint-based approaches (Wagers et al. 2018), cognitive constraint approaches (e.g., Ferreira 2003), and frequency-based approaches (Gennari & MacDonald 2009). We evaluate these approaches with two experiments focused on the question of thematic roles’ alignment with grammatical functions for experiencer verbs, using self-paced reading with comprehension questions and speeded grammaticality judgments. We find evidence that misaligned configurations are harder to process and understand, and less often considered grammatical compared to aligned configurations. We compare Bayesian models using grammatical predictors only, frequency only, and a combination of the two to determine how frequency influences our results. While there was little predictive difference between models using only grammatical predictors and models using both kinds of predictors, models using frequency only generally fared worse. In models using both kinds of predictors, no evidence for effects of frequency was found. We propose an approach called the Grammar Underlies Production and Processing (plus frequency) (GUPPy) framework where grammatical constraints directly account for alignment effects in production and processing. In addition, our results show the potential importance of directly comparing the distinct causal models of psycholinguistic frameworks.