Although the behaviors displayed by assessees are considered to be the currency of assessment centers (ACs), they have remained largely unexplored. This is surprising because a better understanding of assessees’ behaviors may provide the missing link between research on the determinants of assessee performance and research on the validity of performance ratings. On a practical level, a focus on assessees’ behaviors also informs dimension selection, exercise design, rating aids, and assessor training. Therefore, this study draws on behavioral personality science to scrutinize the behaviors that assessees express in interpersonal AC exercises. Our goals were to investigate (a) the structure of interpersonal behaviors, (b) the consistency of these behaviors across AC exercises, and (c) their effectiveness. We obtained videotaped performances of 203 assessees who took part in short interpersonal AC role-plays in a high-stakes context. Apart from assessors’ performance ratings, trained experts also independently coded assessees on over 40 specific behavioral cues in these role-plays (e.g., clear statements, upright posture, lively expressions, freezing). Results were threefold: First, the structure underlying behavioral differences in interpersonal AC exercises was represented by four broad behavioral constructs: agency, communion, interpersonal calmness, and intellectual competence. Second, assessees’ behaviors showed more consistency across exercises than performance ratings did. Third, the behaviors were related to role-play performance and predicted future interpersonal performance. We discuss the theoretical and practical implications of this study’s granular, behavior-driven perspective.