SummaryRegular cycles of intermittent mass selection have been compared with continuous mass selection at the same selection intensity for the quantitative trait of bristle number on one abdominal segment in a non-inbred strain of Drosophila melanogaster. All lines had 20 pairs of parents per generation. There were four replicates of each intermittent selection treatment, six of continuous selection at 20% selection intensity, and two at 10% selection intensity.In the treatment designations, the numbers of selected and non-selected generations per cycle are given first and second respectively, separated by a colon. The intermittent selection treatments, with number of generations of actual selection in parentheses, were 1:3 (5) and 1:1 (17) at 10% selection intensity, and 3:1 (24) and 6:2 (24) at 20% selection intensity. Rates of response of all treatments except 1:3 had declined by the time they were terminated, although none appeared plateaued. Sublines (continuous, 2:1, and 1:1 selection at an intensity of 20%) split off two replicates of treatment 1:1 at termination revealed little variation still available for utilization under selection in these two lines.All intermittent selection treatments gave similar amounts of response per generation of actual selection to those of the relevant continuous selection treatments.Measures of reproductive fitness, bristle number changes over five generations of relaxed selection under crowded mass culture, phenotypic variance for bristle number, and frequencies of lethal genes were obtained for almost all lines. As selection progressed, continuous selection lines were more prone to (1) exhibit increases in phenotypic variance for abdominal bristle number, and (2) have larger declines in abdominal bristle number on relaxation of selection.Although some replicates of almost all treatments showed little or no loss of reproductive fitness, fitness declined more under continuous than under intermittent selection.Failure to get increased response per generation of actual selection in the intermittent selection lines relative to their continuous selection controls may have been because responses obtained in both sets oflines were largely due to concentration of a few genes which were individually of large effect, and initially present at low frequencies. The greater decline in fitness under continuous selection suggests that linked complexes of genes deleterious to fitness and ones responsible for response in bristle number were important.