I will describe our recent research pursuing the topic of etiologic heterogeneity. The concept of etiologic heterogeneity is important since investigation of etiologic heterogeneity has the potential to improve our ability to discover new risk factors and thus to improve our ability to predict the risk of cancer in the population. To explore this phenomenon one needs a quantitative definition of etiologic heterogeneity. Risk heterogeneity in the population can be defined by the coefficient of variation in risk among members of the population. Greater risk heterogeneity implies greater predictable of the disease. Etiologic heterogeneity of two sub-types is defined by the alignment of the risks of the sub-types among members of the population, i.e. by the coefficient of covariance of the risks of the sub-types. Lower covariance implies greater heterogeneity. It can be shown that the incremental explainable risk variation on the basis of classifying disease into two sub-types is inversely related to the etiologic heterogeneity of the sub-types as defined above. Thus one can explore different sub-typing options to find the sub-types that optimize etiologic heterogeneity and consequently optimize the explainable risk variation. This strategy is easily generalized to multiple sub-types and it will be illustrated using data from two breast cancer case-control studies where the cases can be classified into various alternative sub-types on the basis of hormonal tumor markers.