Cohesion is the most popular of the relationship constructs in the group therapy literature. This article reviews common definitions of cohesion, the most frequently studied measures, and a measure that may clarify group relations using two latent factors (quality and structure) to explain common variance among frequently used group relationship instruments. We present the results of a meta-analysis examining the relation between group cohesion and treatment outcome in 55 studies. Results indicate that the weighted aggregate correlation between cohesion and treatment outcome was statistically significant, r = .26, z = 6.54 (p < .01), reflecting a moderate effect size (d = .56). Heterogeneity of effect sizes was significant (Q = 260.84, df = 54, p < .001) and high (I2 = 79.3%), supporting moderator analyses. Six moderator variables were found to significantly predict the magnitude of the cohesion–outcome association (type of outcome measure, leader interventions to increase cohesion, theoretical orientation, type of group, emphasis on group interaction, and dose or number of group sessions). Patient contributions, diversity considerations, and evidence-based therapeutic practices are highlighted.