The study of spontaneous abortion risk is complicated by the fact that a woman's successive pregnancies are not independent events. Non-independence may distort any analysis of spontaneous abortion risk that includes successive pregnancies. In order for an analysis of spontaneous abortion to take non-independence into account, a model must be developed. Previous attempts to describe non-independence have focused on the closely-related issue of whether the risk of abortion increases with successive pregnancies (i.e., with gravidity). Authors using various analytic strategies have reached conflicting conclusions about the gravidity effect. We review these attempts in detail, and in so doing, develop a synthesized viewpoint. We conclude that three factors powerfully affect the occurrence of spontaneous abortion: (1) a heterogeneous risk of abortion (constant for each woman but differing among women); (2) selective fertility (a tendency to control fertility on the basis of previous pregnancy outcomes); and (3) increased risk with maternal age. We demonstrate with a simulation model that these three factors alone can produce a spurious gravidity effect of a size similar to the effect found in actual data. Furthermore, we propose that these three factors offer a framework for a generalized analytic approach to spontaneous abortion, one that takes into account the non-independence of successive pregnancies.