Effects of working memory and language proficiency on second language predictive inference generation: An eye-movement study

In this study, the effects of language proficiency and working memory capacity on predictive inference generation during reading in a second language (L2) were investigated by analysing L2 readers' early and late eye-movements while they were reading predictive or neutral passages. The results suggested that while L2 readers can make predictive inferences on-line during reading regardless of their proficiency level or working memory (WM) capacity, these two factors and their interaction determine the time course of inference generation through different mechanisms. While WM capacity facilitates referent-antecedent resolution and readers with higher WM capacity can benefit para-foveal processing more, proficiency increases reading speed and makes lower level processes less resource consuming. As a result, high WM readers showed facilitation effects of prediction even before they encounter the to-be-predicted word, especially when the pre-target required referent-antecedent association. On the other hand, while high language proficiency readers can show the effect of prediction during early processing of the target word, low proficiency readers can show facilitation during late processing of pre-target word. During the late-processing of pre-target word, WM and language proficiency will have an interaction effect due to differences in mechanisms through which they contribute to predictive inference generation. Although all groups of readers showed facilitation effects relatively early, the greatest facilitation emerged during late processing of the sentence final word, where sentence wrap-up processes take place. This is in line with L1 studies and its implications for L2 reading were discussed.