The numbers of sternopleural chaetae of 18 inbred lines derived from the Texas population of
Drosophila melanogaster
were ascertained for flies raised at each of two temperatures, 18 and 25°C. Two characters were then defined:
M
, the average of the chaeta numbers at the two temperatures; and
S
, half of the difference between the average chaeta numbers at the two temperatures. The inbred lines differed among themselves in both characters, so revealing genetic variation in the Texas population for them both. There was no correlation between the values of
M
and
S
among the lines. The mean of the inbred lines did not differ significantly from that of the progenies of 22 single-pair matings among flies taken from the population in either character, so suggesting that the 18 inbred lines provided a fair sample of the genes in the population. The differences between the variances among the inbred lines, on the one hand, and the biparental progenies, on the other, suggested that dominance was present for both characters and that dominant alleles were, in general, present at higher frequencies than their recessive fellows. This was confirmed by a half-diallel experiment with 11 of the 18 inbred lines as parents. The half-diallel further showed that dominance was ambidirectional for character
S
, but gave no conclusive evidence in this respect for
M
. The theoretical consequences of ambidirectional dominance combined with higher frequencies of dominant alleles were considered for a simple model involving four loci. This showed, in particular, that when one homozygous line (the recurrent parent) is crossed to a number of other lines (the non-recurrent parents), the amount by which the phenotype of the F
1
exceeds the mean of its two parents shows a regression of negative slope on the phenotype of the non-recurrent parent. The regression line is expected to cut the axis of zero excess at a point that is, on average, as far away from the phenotype of maximum dominance as is the expression of the character in the recurrent parent, but in the opposite direction. The point mid-way between the recurrent parent and the intercept thus provides an estimate of the phenotype of maximum dominance. Estimates of the phenotypes of maximum dominance were obtained in this way for both characters,
M
and
S
, by means of data from the half-diallel and from a further experiment carried out for the purpose. For character
M
, the phenotype of maximum dominance does not depart significantly from the mean of the inbreds or from that of the Texas population itself; but, for
S
, though close to the means, it is significantly higher than them. It is concluded that this type of genetical architecture is to be expected with characters under stabilizing selection and that the phenotype of maximum dominance is the optimal phenotype towards which the selection has been acting.
Modifying the sternopleural hair pattern in Drosophila by selection
