In this paper, we have focused on two key sources of between-study heterogeneity, sampling and measurement, and elsewhere we consider the sources of geographical region, history and design characteristics (Curran & Hussong, 2009). Many other sources of between-study heterogeneity may also be present, and identifying the plausible primary sources is a preliminary step in hypothesis testing. To the extent that study membership is not completely confounded with these sources of between-study heterogeneity (e.g., differences in rates, though not in the presence, of maternal depression across studies), we can isolate the factors that underlie differences in our findings across studies (i.e., factors that explain failures to replicate). Such questions of methodological interest may allow insight into those of theoretical interest, with the ultimate goal of providing some synthesis regarding theoretical associations or even mechanisms. For those sources of between-study heterogeneity that are completely confounded with study, we obviously cannot identify what factors underlie observed study differences, though we can control for them.
