SummaryThis study is concerned with comparisons of potentials exhibited by the entire class of general combining ability methods which can be generated by one or two random-mating populations. By potential is meant the greatest value the population mean assumes with continued application of a given selection method initially applied to a population of specified genetic constitution. The argument is restricted to an arbitrary number of alleles at a single locus, and it is assumed that the populations are infinite in size.Considerable attention is devoted to the situation of overdominance. In this case, the potential of a mating system depends on whether or not a stable equilibrium is possible. The equilibrium theory for multiple alleles is used to demonstrate that a stable equilibrium state exists for a general combining ability test in which the popu· lation undergoing selection, itself, is used as tester. However, when the population undergoing selection is tested by some other population, the solution is more complicated. For example, with reciprocal selection, it is shown that the eqUilibrium (if it exists) in the hybrid population is unstable. This leads to fixation of alleles in the parent popUlations. However, if there are more than three alleles at the locus, more than one fixable region exists on the topography generated by the hybrid popu· lation mean. In this situation, reciprocal selection carries the hybrid population to one of several genotypic values which may. or may not, be the greatest value in the population.These equilibrium results are used in making comparisons among the various selection methods in terms of their potentials for the overdominant situation.Similarly. potentials of the mating systems are compared for the case of partial dominance.From these considerations of potentials, selection programmes are suggested for the situations in which heterosis is, or is not, used.