Results are presented of 135 generations of selection for high scutellar bristle number in two lines M and M3 derived from the same original mating of one female with 5 bristles by one male with 4 bristles, the latter being the wild-type canalised phenotype. Results are also given of two relaxed lines per line and of a reselection line M2 derived from the first relaxed line of line M which had regressed almost to base population level. The effect of introducing the sc(1) allele into the M and M3 selected backgrounds was studied at generations 39-44. At the end of selection the effect of an extra dose of sc(+) was also studied in males of all selected backgrounds. The correlated responses in abdominal bristles were followed in all lines.-Considering their common origin, the selection lines differed markedly in pattern of scutellar response and in most other aspects observed, namely correlated responses in abdominals and p.c. scutellars, sex differences, and behaviour on relaxation. Selection limits for scutellar bristles in lines M and M2 were equal to or greater than the most extreme reported in the literature.-The probit span of the canalised 4 bristle class decreased in each selection line as the mean scutellar bristle number increased, and increased again in the relaxed lines as the mean bristle number decreased. In the context of an hypothesis that canalisation at 4 bristle is due to regulation of the scute locus, this result is now interpreted as being due mainly to selection for poor regulators of sc(+), in contrast to a previous interpretation that only the minor gene background was altered by selection, the canalisation (regulation) genotype not being affected.-Introducing the sc(1) allele into the selected backgrounds M and M3 showed a reduced effect on sc(1) flies compared with sc(+) flies, and an interaction of sc(1) and sc(+) with selected background. sc(1) flies had about the same number of bristles in both backgrounds though the mean of sc(+) flies in line M was about 3sigma higher than in line M3. Dominance of sc(+) to sc(1) was reduced slightly in M3. However, the effect of an extra dose of sc(+) at the end of selection was about the same as in unselected in all lines, so the first or dominance level of regulation of the scute locus was not significantly affected by selection, though the second or canalisation level of regulation was.-A large positive correlated response in abdominal bristles occurred in all lines. The response in line M was about twice that in M2 and M3 and was in fact as large as can be obtained from direct selection on abdominals. In line M some genes may have been selected with a proportionately greater effect on abdominals than on scutellars. This is supported by the further observation in line M that the abdominal scores of flies with particular scutellar bristles scores increased as the scutellar mean increased. An attempt was made to apply to these results Rendel's (1962) model of competition between scutellars and abdominals for common bristle-making resources. This could not be done satisfactorily mainly because the assumptions in the model about the similarity of effects in scute and wild-type flies were not met in the present material.