Compensation refers to the fact that a group perceived as higher than another on one of the fundamental dimensions of social judgment (competence and warmth) is also perceived as lower than the other group on the other dimension. Relying on a full-crossed design, the present work tested compensation in a real-life situation using existing groups involved in an ongoing relation. As predicted, compensation emerged when (a) the difference between the groups, and thus the perceived legitimacy of the status difference, was large as opposed to small, and (b) the relation between the groups was asymmetrical. In contrast, the smaller the difference (the lesser the legitimacy), the more ingroup bias emerged.