This paper is a classic in discussion of causal affect in epidemiology. Causation is discussed looking at dichotomous variables with the annulling assumption that no random error attributable to sampling variability exists. This paper present the concept of causal theory. The limitations on randomized studies are reasons why methods for causal inference from observational data are needed. Contrast this proviso to that we see in prospective interventional trials in which using the same dichotomous variables gives the ability to conclude risk and therefore causation