Since its origin in the 19th century, epidemiology has faced an internal tension between an approach oriented toward biology and the study of mechanisms, and an approach oriented toward populations and their interactions with the environment. Initially, this tension took the form of an opposition between microbiology and statistics. We describe the early roots of the quantitative approach to health and disease and several historical examples of the above tension. The search for the causes of pellagra exemplifies our thesis. In Italy, where pellagra was endemic, contrasting opinions coexisted between the hypothesis of contaminated maize, supported by Cesare Lombroso, and the hypothesis of a prevailing role of poverty and poor nutrition. In the United States, Joseph Goldberger found no evidence for the hypothesis of contaminated maize or for a microbiological agent, but recognized the central role of nutrition. The "cure" Goldberger proposed was land reform, but he continued studying the disease from a mechanistic point of view; shortly after his death, niacin deficiency was identified as the cause of pellagra. The tension between mechanistic and population-based studies is still present within epidemiology and is in fact essential for the success of the discipline.