To identify underlying patterns in the alliance literature, an empirical review of the many existing studies that relate alliance to outcome was conducted. After an exhaustive literature review, the data from 79 studies (58 published, 21 unpublished) were aggregated using meta-analytic procedures. The results of the meta-analysis indicate that the overall relation of therapeutic alliance with outcome is moderate, but consistent, regardless of many of the variables that have been posited to influence this relationship. For patient, therapist, and observer ratings, the various alliance scales have adequate reliability. Across most alliance scales, there seems to be no difference in the ability of raters to predict outcome. Moreover, the relation of alliance and outcome does not appear to be influenced by other moderator variables, such as the type of outcome measure used in the study, the type of outcome rater, the time of alliance assessment, the type of alliance rater, the type of treatment provided, or the publication status of the study.