Abstract
The evolution of gene frequencies, at a locus at which selection is not so strong as to make it essentially deterministic, can be modelled by an appropriate stochastic differential equation describing a diffusion process. Unless the number of possible alleles is rather small, it is difficult to extract useful information from such models except under assumptions of statistical equilibrium. This paper suggests some different approaches to the non-equilibrium problem, with particular reference to the case of a neutral locus with a very large number of potential mutant alleles.