Maximum rates of gene substitution are calculated, showing that Haldane's estimate (1957) of one completed gene substitution per 300 generations may be much too low. The approach differs conceptually from Haldane's in postulating that the mean fitness of the population is not affected by its genotype. The rate of death is assumed to be set by density-dependent factors, and rates of substitution and selective intensities consistent with this death rate are calculated. This formulation leads to no obvious upper limit for the rate of substitution. Rates of one completed substitution per generation or more are by no means excluded. The selective value consistent with a given death rate or set of fitness differentals is not independent of the rate of substitution but is approximately inversely proportional to it. The result is little affected by the degree of dominance or by the initial gene frequency.