A simulation study was undertaken of methods of subdividing populations into several small sublines and utilizing the variances generated between lines by selecting among them. Crosses of chosen lines were made, and either selection was continued in a single large population (single cycle) or the population was subdivided again (repeated cycles). As a control for the efficiency of these schemes, a single large population was maintained and selected at the same intensity from the outset. Simple models were used of additive or completely dominant genes, usually of equal effect and equally spaced on a single chromosome.

The single and repeated cycle structures give similar results, but the repeated cycle structure is more extreme.

With additive models intense selection between lines gives short-term advances, but causes a reduction in the limit when compared with a single population. The effect on the limit is greatest with free recombination, very small with complete linkage. If no selection is practised between lines the limit is unaffected, but takes longer to attain.

With complete dominance, and the recessive allele initially at low frequency, greater responses from selection are obtained within sublines than in the large population, large gains are made from selection between sublines, and a higher limit can be reached. If the recessive allele is at high initial frequency the subdivision is not beneficial.

Some simple theory is developed to explain these results. It is concluded that subdivision and crossing schemes are unlikely to be very useful except for elimination of deleterious recessive genes.