Approval of a generic alternative to an existing formulation carries the implication that the two formulations are "equivalent". However, there are many notions of equivalent. Current practice in bioequivalence studies defines equivalence solely in terms of average (or median) measures of bioavailability. Current methods for this average bioequivalence approach are commonly based on the two one-sided tests principle. Average bioequivalence is the special case of population bioequivalence, where the entire distribution of bioavailabilities is considered. Statistical approaches for population bioequivalence are suggested. Population bioequivalence is an improvement over average bioequivalence, because average bioequivalence does not consider the variabilities of the formulations. However, population bioequivalence is still not sufficient to ensure that an individual will respond similarly to the two formulations; that requires individual bioequivalence. Again, statistical approaches are suggested, one of which, a tolerance interval approach, appears to directly address the clinical question of switchability of formulations. We conclude that population bioequivalence is sufficient for marketing approval, but that individual bioequivalence is necessary for switchability.