The complementarity of behavior as moderated by the base rates of consequent behaviors, the differences attributable to friendly and hostile antecedent behaviors, and octant versus quadrant codes was examined. Complementarity was defined using 3 different but overlapping models: R. Carson (1969), D. Kiesler (1983), and J. Wiggins (1979). The Interpersonal Communication Rating Scale (ICRS; S. Strong & Hills, 1986) behavioral ratings on 80 female dyads provided by Strong, Hills, Kilmartin, et al. (1988) were analyzed using the randomization test of hypothesized order relations and correspondence analysis. Results demonstrated that (a) complementarity was supported for each of the 3 definitions (Carson's, Kiesler's, and Wiggins's), (b) complementarity fit the data better if account was taken of the base rate of behavior, (c) complementarity was more prominent if the antecedent behavior was friendly rather than hostile, and (d) complementarity of quadrant behavior was fairly strong. Results are discussed with respect to the measurement of interpersonal behaviors and the construct of complementarity.