In many mammalian species, males are selected to kill unrelated infants and/or fetuses in order to cause lactating and pregnant females to begin cycling sooner than they otherwise would. As a result, females have evolved numerous counterstrategies to prevent infanticide and feticide. One such proposed counterstrategy is the Bruce effect, an apparently costly strategy in which inseminated or pregnant females cease reproductive investment in a developing embryo or fetus following exposure to nonsire males. Here I present a quantitative model that seeks to explain under what conditions females will be selected to exhibit the Bruce effect (i.e. to block or terminate pregnancy) rather than risking future infanticide or feticide. I first present an analytical model of the costs of the Bruce effect relative to the costs of potential feticide or infanticide. I then test the resulting predictions using an individual-based model operating under ecologically relevant conditions. The individual-based model predicts that moderate and high, but not low, levels of infanticide can produce selection for the Bruce effect. In contrast, feticide risk alone is unlikely to lead to selection for the Bruce effect, although feticide risk coupled with a substantial risk of female mortality following feticidal attack can. The model correctly predicts the evolution of the Bruce effect in geladas, Theropithecus gelada, and correctly predicts the absence of a Bruce effect in chacma baboons, Papio ursinus, and yellow baboons, Papio cynocephalus. Finally, I present a framework by which researchers can predict whether they expect to find infanticide, feticide and/or the Bruce effect in their study species.