One purpose for combining research studies is to estimate a population treatment effect. The internal validity of a model for how effect size estimates should be computed and combined will hinge upon the homogeneity of the effect size variation. Effect size variation may be assessed in the form of a summary fit statistic, and a direct consideration of the extent of individual effect variation from the population estimate. This paper presents some diagnostic techniques that facilitate the analysis of effect size variation. Bivariate plots of effect size residuals can aid in detecting sources of variation inconsistent with the model. Particularly, plotting the standardized residual of each study against the homogeneity of the sample if that study were removed is of interest for assessing the extent of heterogeneity contributed by individual studies. It is emphasized that the use of diagnostic techniques is useful for revealing why a lack of fit occurred, and is not advocated for the ad hoc purpose of finding a best-fitting subset of studies. (BW)