The principles of “interpersonal complementarity” posit that one person’s behavior tends to evoke reactions from others that are similar with regard to affiliation, but dissimilar with regard to status (Carson, 1969). Empirical support for these assumptions has been mixed, especially with regard to the status dimension. The present study investigated influences of level of observation (behaviors vs. traits), personality, and instructional set on complementarity. Previously unacquainted participants (N = 182) were randomly assigned to dyads who engaged in videotaped discussions with either a cooperative or a competitive framing. Their behaviors were rated every 10 s in terms of affiliation and status. We used a multilevel modeling approach that enables powerful omnibus tests of core tenets of interpersonal theory, including previously overlooked influences of person and situation, as well as time-lags. Interpersonal complementarity was found for both affiliation and status. It was highest for simultaneous behaviors, (i.e., occurring within the same 10-s interval), but for status we also found lagged effects being anticomplementary. Complementarity was also moderately predicted by some personality traits. The situational context predicted the mean levels of interpersonal behaviors but not complementarity. Concerning consequences of complementarity, the participants’ individual complementarity coefficients predicted their being liked by the interaction partner, observer judgments of their social competence, and their overall discussion performance. We clearly recommend this type of comprehensive multilevel modeling for future research into these and related issues.