Abstract Predictable words are processed more quickly than unpredictable words during reading, but how this predictability effect is modulated by working memory (WM) limitations is unclear. Two preregistered experiments tested whether the availability of visuospatial WM resources affects reading measures that are sensitive to predictability (gaze duration and skipping probability in Experiment 1, and self-paced reading time in Experiment 2) in native English speakers. The analyses of all measures revealed main effects of predictability and WM load. Their interaction was significant in Experiment 2 but not in Experiment 1. A follow-up analysis on relative reading times in Experiment 2 revealed main effects of predictability and WM load but no interaction between the two, replicating Experiment 1. The apparent interaction in Experiment 2 was likely due to the WM load speeding the overall reading times and masking the predictability effect. Taken together, these findings demonstrate a robust effect of predictability under a cognitively demanding task and suggest that visuospatial WM resources are not necessary for computing probabilities of upcoming words during reading.
Aine Ito (Thu,) studied this question.