Evidence from the visual world paradigm raises questions about unaccusativity and growth curve analyses

Abstract

Many syntactic theories posit a fundamental structural difference between intransitive verbs with agentive subjects (unergative verbs) and those with theme subjects (unaccusative verbs). This claim garners support from studies finding differences in the online comprehension of these verbs. The present experiments seek to replicate one such finding using the visual world paradigm (Koring, Mak, & Reuland, 2012). We control for several factors that were uncontrolled in previous studies. We find no differences in the processing of unergative and unaccusative sentences in logistic regressions and cluster analyses. However, in growth curve analyses, modeled closely on the original paper, we find differences between the verb conditions that appear to be statistically significant but are unstable across experiments. A resampling analysis reveals that the growth curve analyses are highly anticonservative, suggesting that the earlier finding was a false positive. We conclude that there is no strong evidence that unaccusatives are processed differently from unergatives. We suggest that growth curve analyses only be used with visual world paradigm data when the underlying assumptions of the analysis can be validated via resampling.

Keywords: Argument reactivation; Growth curve analysis; Unaccusativity; Visual world paradigm.