Results are presented of 130-145 generations of selection for low scutellar bristle number in four lines of D. melanogaster derived directly from an Oregon-RC wild-type stock and in one derived from an Oregon-RC line selected for low sternital bristle number. The most rapid initial response and the lowest mean scutellar bristle number ultimately reached, just below 2 bristles, occurred in a line in which the response was due to a new recessive gene located at approximately 17·4 on the X chromosome. Three of the other four lines reached a plateau just above a mean of 2 bristles after different patterns of response. These plateaux reflected a new canalization or threshold phenomenon at 2 bristles in these lines. The remaining line reached a mean of about 2· 5 bristles after some 50 generations and remained at that level or slightly higher thereafter, but had no indication of canalization at 2 bristles. Two relaxed lines were derived from each selection line at different times and showed variable patterns of regression towards the base population level.In the early generations of selection, as the mean bristle number slowly dropped below the canalized wild-type level of 4 bristles, flies with more than 4 bristles continued to appear, i.e. the probit width of the 4-bristle class tended to drop from its un selected level of 5·4-5·8". to quite low levels in the selection lines and to recover in the relaxed lines as their mean bristle number moved back towards the un selected level. This could not be interpreted in the same way as similar phenomena in highselection lines, i.e. selection of poor regulators of the scute locus. Other possible explanations for these results and for the new canalization observed at 2 bristles were considered without being able to accommodate them fully within a modified model of regulation of the scute locus. The most likely possibility seems to be that the low selection response is due to selection of less-efficient alleles of a gene which is the inducer of the scute locus. A large reduction in the degree of dominance of sc+ over scute (SCi) was also observed in these selection lines, which may be compatible with the above explanation.Correlated responses in sternital bristle number in the selection and relaxed lines, while generally positive, were not closely related to the pattern of change in scutellar bristles. The dominance of sc+ over SCi in respect of sternital bristles showed little change except in one line where it increased, while dominance in respect of scutellar bristle score decreased.