AbstractThe advantages and disadvantages of evolutionary fitness (probability that a single mutant line will not become extinct) and genetic fitness (mean fecundity) are compared. For deterministic processes the two are equivalent, but for stochastic branching processes they may be totally unrelated except that an absolute genetic fitness of unity or less implies an evolutionary fitness of zero. To know the variance as well as the mean family size does not in general uniquely determine the evolutionary fitness.Except where genetic fitness is close to unity, the impact of selection is shown to be rapid for the binomial, Poisson, negative binomial, and truncated negative binomial distributions.Evolutionary fitness, though somewhat cumbersome, has greater relevance to evolution, genetic counseling, and voluntary population control; but genetic fitness which is much easier to handle is the more appropriate measure where a large number of mutants is involved.Some empirical data on the transmission of various types of characters from parent to child are analyzed to allow comparison of genetic fitness, Crow's index, and a Malthusian parameter, with evolutionary fitness. There is a fair, but far from perfect, agreement among them. Multiple correlation of evolutionary fitness with mean and variance of family size taken jointly suggests a much more satisfactory approximation. It thus appears that, at the least, the population geneticist cannot afford to ignore the variance (which is not adequately represented in Crow's index). These relationsips, based on two sets of data only may be accidental and should be invoked with caution. It seems more than likely that other aspects of the distribution of family size (eg, even higher moments) may contain relevant information in certain cases.