A concurrent visuospatial working memory (vsWM) task has been shown to reduce predictive gaze behaviour in several visual world eye-tracking experiments. This reduction could be due to general aspects of prediction or an effect contingent on the additional demand of visually co-present objects. In an N400 event-related potentials study, L2 English participants heard the same sentences and completed the same vsWM task from a study which showed a reduction in predictive gaze behaviour, but without visually co-present objects. We observed robust effects of predictability but evidence against an influence of the vsWM task. These results suggest that vsWM does not interfere with the pre-activation of a target word without a co-present visual array. We conclude that previously observed effects of vsWM on prediction are instead likely specific to gaze behaviour and situations in which spoken language refers to a co-present relevant visual context.