When a reasonable basis exists for postulating the distribution of risks (or of survival times) in a population, it may also provide a basis for choosing a desirable combination of sample size (n) and duration of follow-up (t). In studies of conception risk, if a Beta distribution of risks is assumed, the asymptotic precision and bias of maximum likelihood estimators of the parameters are improved by a moderate prolongation of t. Not surprisingly, the desirable value of t increases as the mean risk decreases, i.e. as the probability of "survival" beyond a fixed t increases.