Pronouns Are as Sensitive to Structural Constraints as Reflexives in Early Processing: Evidence From Visual World Paradigm Eye-Tracking

Linear mixed-effects models (LMEMs) have become increasingly prominent in psycholinguistics and related areas. However, many researchers do not seem to appreciate how random effects structures affect the generalizability of an analysis. Here, we argue that researchers using LMEMs for confirmatory hypothesis testing should minimally adhere to the standards that have been in place for many decades. Through theoretical arguments and Monte Carlo simulation, we show that LMEMs generalize best when they include the maximal random effects structure justified by the design. The generalization performance of LMEMs including data-driven random effects structures strongly depends upon modeling criteria and sample size, yielding reasonable results on moderately-sized samples when conservative criteria are used, but with little or no power advantage over maximal models. Finally, random-intercepts-only LMEMs used on within-subjects and/or within-items data from populations where subjects and/or items vary in their sensitivity to experimental manipulations always generalize worse than separate F 1 and F 2 tests, and in many cases, even worse than F 1 alone. Maximal LMEMs should be the 'gold standard' for confirmatory hypothesis testing in psycholinguistics and beyond.

Participants' eye movements were recorded as they inspected a semi-realistic visual scene showing a boy, a cake, and various distractor objects. Whilst viewing this scene, they heard sentences such as 'the boy will move the cake' or 'the boy will eat the cake'. The cake was the only edible object portrayed in the scene. In each of two experiments, the onset of saccadic eye movements to the target object (the cake) was significantly later in the move condition than in the eat condition; saccades to the target were launched after the onset of the spoken word cake in the move condition, but before its onset in the eat condition. The results suggest that information at the verb can be used to restrict the domain within the context to which subsequent reference will be made by the (as yet unencountered) post-verbal grammatical object. The data support a hypothesis in which sentence processing is driven by the predictive relationships between verbs, their syntactic arguments, and the real-world contexts in which they occur.