Within the context of the general superiority of meta-analytic reviews over qualitative reviews, we emphasize the fragility of meta-analytic outcomes, especially when the effects of moderator variables remain unexamined. We consider the contribution to increased theoretical understanding made by moderator analyses both of methodological and substantive variables in psychological, health, and medical research. The potential contribution of meta-analysis to understanding the process underlying established relationships and to assessment of the convergent and discriminative construct validity of previously used concepts is discussed.