The purpose of this study was to test the impact of two predictor variables, one representing extratherapeutic factors and one representing relationship factors, on outcome at a university counseling training clinic. A naturalistic design was used to collect session-by-session outcomes on 135 clients seen by 88 counselors. Hierarchical linear modeling was used to test hypotheses about the effects of clients' pretreatment social support (extratherapeutic factor) and working alliance (relationship factor) at session 3 on change over time. Results showed that higher rated alliances predicted greater change over the first 8 sessions. When clients reported poorer social supports, the therapeutic alliance was even more important in predicting a positive outcome. Limitations of the design and implications for further outcome studies are discussed. © 2011 Wiley Periodicals, Inc. J Clin Psychol 67:1–11, 2011.