The recommendation of Popper's philosophy of science should be adopted by epidemiologists is disputed. Reference is made to other authors who have shown that the most constructive elements in Popper's ideas have been advocated by earlier philosophers and have been used in epidemiology without abandoning inductive reasoning. It is argued that Popper's denigration of inductive methods is particularly harmful to epidemiology. Inductive reasoning and statistical inference play a key role in the science; it is suggested that unfamiliarity with these ideas contributes to widespread misunderstanding of the function of epidemiology. Attention is drawn to a common fallacy involving correlations between three random variables. The prevalence of the fallacy may be related to confusion between deductive and inductive logic.