Embarrassment arises when we reveal an apparent flaw of the self in front of others, for instance, in a faux pas situation. An audience is crucial for embarrassment, but the group membership of the audience has not yet been studied. According to the social identity approach, we assign more importance to evaluations by ingroup than by outgroup members, particularly when we identify highly, and the outgroup is of lower status. A pilot study (N = 30) showed that embarrassment correlated positively with group membership of the audience and with identification. Studies 1 to 3 presented participants with several faux pas scenarios. In Study 1 (between-participants design; N = 75), participants reported higher embarrassment in ingroup (Norwegian) and equal-status outgroup (Swedish) conditions than in a lower-status outgroup condition (Polish). In Study 2 (within-participants design; N = 135), participants reported higher embarrassment when they imagined the audience to be other Scots (ingroup) than Americans or Poles (outgroups), particularly when they perceived the outgroup to be lower in status. In Study 3 (between-participants design; N = 59), high identifiers but not low identifiers showed the expected ingroup–outgroup audience effect. Implications for intergroup relations are discussed. Key Message: Embarrassment following faux pas situations depends on the group membership of the audience, relative status of the audience and ingroup identification. Copyright © 2011 John Wiley & Sons, Ltd.