In search of the hypothesis.

Epidemiologic studies of racial differences sorely lack sound and explicit hypotheses. Race is a social convention, not a biological concept. Its careless use in epidemiology demonstrates a failure to generate appropriate hypotheses to study its role in health. Studies of hypertension in blacks illustrate the point. Two underlying pitfalls plague hypothesis generation: directionality involving the null and alternative hypotheses and circularity, where efforts to understand social factors have the effect of emphasizing racial differences. The proper prescription is to identify explicitly the hypotheses of interest, including their origins and implication.