Recent studies suggest that the social presence undermines performance in difficult tasks, because the presence of others would automatically capture the attention needed to achieve these tasks. Here, I will present a series of experiments that tested if this attentional capture (here caused by the experimenter presence) impacts working memory. Several models indeed suggest that maintenance in working memory relies on an attentional mechanism. Besides this mechanism, another non-attentional verbal rehearsal could also maintain verbal information. Our works showed that the effects of social presence were observed only in participants performing a concurrent articulation (a manipulation known to prevent the use of verbal rehearsal). We replicated these results in another experiment in which concurrent articulation was manipulated within participants. Finally, we compared the previous results obtained with the Brown-Peterson paradigm with new results obtained in the complex span task. The Brown-Peterson allow to rely more on long term memory and require less attention that the complex span task. in accordance with our prediction, the effect of presence of the experimenter was larger in the complex span task. Altogether provide the first evidence that social presence hinders attentional, but not non-attentional, maintenance in working memory. They have strong implications for understanding working memory and the impact of social presence, but also important methodological implications. In conclusion, I will argue in favour of a social cognition that relies on precise models issued from cognitive psychology rather than on more general metaphors such as the notion of "resources".