This article traces the history of the Susceptible, Infected, Recovered (SIR) model in early 20th-century epidemiology (1904–27). The aim is to test the hypothesis that the active stance taken by Ross, Hudson, McKendrick, and Kermack represents a turn in the history of modern epidemiology, shifting the classical method of statistical epidemiology from a data-based model to a theory-based model. The article shows that epidemiological modeling is based on a mathematical simplification of epidemics at the time of the microbiological complexification of epidemiology. What were the implications of the mathematical theory of epidemics for the development of modern epidemiology? How can epidemic dynamics be expressed mathematically without compromising the biological representation? This is exactly what is at stake here: from the work of Ross to the standardization of the SIR model by McKendrick and Kermack, modeling has been confronted with the problem of relating its equations to the facts.