To be responsive to clients' evaluations of the unfolding therapy process, therapists must first accurately "read" client behavior, a particularly challenging task in conjoint family therapy. In this study, the authors compared client behavior in 28 sessions that one family member and the therapist concurred, on the Session Evaluation Questionnaire (W. B. Stiles & J. S. Snow, 1984), were relatively better or worse than their other sessions. Client behavior was rated from videotapes using the System for Observing Family Therapy Alliances (SOFTA-o; M. L. Friedlander et al., 2006). In contrast to the worse sessions, the better sessions were characterized by significantly greater client Engagement in the Therapeutic Process and Safety within the Therapeutic System. Notably, whereas only the worse sessions had exceptionally poor within-family collaboration, 40% of the better sessions were characterized by mild family conflict. Implications are discussed for building theory on therapist responsiveness and for future research that may benefit practice, training, and supervision. (Contains 2 tables and 3 footnotes.)