Studies of the therapeutic alliance typically use a one-with-many (OWM) design in which each therapist (the one) treats multiple clients (the many). This study used Kenny, Kashy, and Cook's (2006) OWM method to examine the composition of the therapeutic alliance and to analyze the association between alliance and outcome in a sample of 398 adolescents treated for substance abuse by 14 therapists. Both the client and therapist alliance ratings yielded large relationship variances, with limited consensus among clients treated by the same therapist about the quality of the alliance. If a client reported an especially strong alliance with his or her therapist, the therapist was likely to also report an especially strong alliance with that client (dyadic reciprocity). The association between the components of the alliance and treatment outcome was complicated, with different levels of measurement and different components of the alliance (perceiver, partner, or relationship) derived from different informants (therapist or client) relating to different outcomes. (Contains 2 tables and 5 footnotes.)