Analyses of spontaneous abortion rates often include special provisions to account for the presumed distorting effects that arise because of the presence of habitual aborters. In this paper, I establish a framework for investigating the effects of the presence in study groups of such women. I then examine the strengths and weaknesses of some of the analysis strategies in current use, both in situations where the study group is homogeneous and where there are high-risk women present. I show that the concept of habitual aborter is not useful and can be misleading, and I suggest that one take a broader view of heterogeneity.