Social epidemiologist Davey Smith (2011) argues that epidemiologists should accept a gloomy prospect: considerable randomness at the individual level means that they should keep their focus on modifiable causes of disease at the population level. The difficulty epidemiology has had in moving from significant population-level risk factors to improved prediction of cases at an individual level is analogous to the lack of success in the search for systematic aspects of the non-shared environmental influences that human quantitative genetics claims overshadow common environmental influences (e.g., the family’s socioeconomic status which siblings have in common). This article responds to the argument and analogy, aiming to draw three audiences—social epidemiologists, human quantitative geneticists, and philosophers of science—into a shared discussion that centers not on randomness, but on heterogeneity in various forms. The first half undercuts the analogy, providing a critical account of human quantitative genetics that explains why its estimates are unreliable and typical interpretations, such as that of non-shared environmental influences, unjustified. In the process, attention is drawn to the possibility of underlying heterogeneity—when similar responses of different types in a species or of different groups of people are observed, it need not be assumed that similar conjunctions of genetic and environmental factors (or, in epidemiology, risk and protective factors) have been involved in producing those responses. The second half introduces several additional forms of heterogeneity and captures their potential significance for epidemiology in four conceptual themes. The themes complicate Davey Smith’s advice for epidemiologists not to seek improved prediction of cases at an individual level and his view of the limited prospects for personalized medicine.