Many studies have shown an advantage for high-frequency words in various language and memory tasks. Surprisingly, only one study reported a word-frequency effect in a working memory task (Engle et al., 1990). In the present study, we evaluated the word-frequency effect in complex span task in which the pace of a concurrent task was manipulated. By slowing down the pace of a location judgement task, attention could be switched to maintenance activities for a longer time and recall should be better (Barrouillet et al., 2007). As already reported in immediate serial recall tasks, the advantage for recalling high-frequency words increased across serial positions. This finding is congruent with the redintegration hypothesis, which stated that long-term knowledge helps at reconstructing degraded memory traces at recall (Hulme et al., 1997). Although the pace effect was replicated, it did not interact neither with the word-frequency effect nor with the interaction between frequency and serial position. This pattern of results favours the idea that pace and word-frequency affect distinct steps of processing, the maintenance during the concurrent task and the reconstruction at recall respectively.