The effect of sub-dividing a population during selection was studied by selecting downwards for sternopleural bristles in Drosophila melanogaster. The four structures used all involved the selection of 40 individuals out of 200 measured in each sex. The alternatives were:

(i) one large line with the same selection procedure in all generations,

(ii) eight small lines with the same procedure in all generations,

(iii) the ‘single cycle’ structure in which large lines derived from crosses between selected small lines at either the sixth or twelth generation, and,

(iv) the ‘repeated cycle’ structure in which the best five small lines were mixed at the sixth generation of selection to produce a new set of small lines on which the procedure was repeated for a total of three cycles.

Of the first two methods, the large line had a higher final response than did any of the small lines. The latter, selected with an intensity of 5 out of 25 in each sex, had lower average final response than had a similar group of lines selected from the same base population with the lower intensity of 10/25. The results showed no clear effect of the sub-dicision of the population, followed by selection between lines, though they were in general agreement with theortical expectations. Several second chromosome recessive lethals were found at high frequencies in different selected lines. One of these, found in eight of the lines, had an effect on bristles in the heterzygote of about 1·5 phenotypic standard deviations. The evidence strongly suggests a pleiotropic effect on bristle score.