[en] Verbal working memory (WM) is strongly influenced by semantic similarity and association between items. Serial recall of lists of semantically related words is better than of unrelated words. In this study, we assessed the possibility that semantic associations support WM maintenance via increased item-context binding strength. We manipulated semantic relatedness in a serial order reconstruction task (Exp. 1), a mixed immediate serial recall and serial order reconstruction task (Exp. 2) and an item-spatial association task (Exp. 3). These three experiments led to a clear convergent pattern of results: semantic relatedness did not affect performance in any condition that required only memory for word-context bindings. Therefore, it did not increase item-context binding. Using a computational modeling approach, we restrict the plausible range of mechanisms underlying semantic associative effects. Specifically, we show that semantic associations could benefit WM via stronger availability in activated long-term memory, or by efficiently using semantic information to restrict the set of candidates at a redintegration stage.
