R. C. Lewontin (1970) proposed three criteria as jointly sufficient for a unit of evolutionary change and immediately went on to argue that selection could operate simultaneously at a variety of levels of organization. For natural selection to operate on a unit, there must be variation in the objects in question: natural selection “chooses” between alternatives, so there must be different alternatives. This variation must affect fitness: the variation available must create different probabilities of reproductive success. Third, this variation in fitness must be heritable: there must be a correlation between parents and offspring for the traits yielding variation in fitness. In brief, for there to be evolution by natural selection, there must be heritable variation in fitness.Among the concerns that might be raised over Lewontin's formula, one is particularly important. It was clearly brought out by Williams (1966) and is emphasized by Elliott Sober (1981).