EFFECT OF SELECTION ON CROSSOVER VALUES 337 
were in position to know that pair 15 would be selected, prac- 
tically all of its Fo offspring had emerged. Therefore it seemed 
expedient to mate en masse the F2 offspring (i.e., red long females 
w m 
W M' 
miniature white males w m) from each of several 
promising pairs, to perpetuate the promising lines. Hence, in 
table 1, the F2 sibs came from the selected Fi pair, no, 15, and 
were mated en masse, giving 25.46 per cent crossovers. The F3 
offspring were then mated in pairs, and selection was again 
exercised. This means that an odd-numbered generation (Fi, 
Fs, etc.) in table 1, for example, represents the mating of pairs, 
while an even-numbered generation represents the mating en 
masse of sibs from the selected pair. It will be clear that inbreed- 
ing was very intense throughout all series, for the pair gave sibs, 
and the sibs from the selected pair mated en masse gave a popu- 
lation in which the most remote relationship could be double 
cousins, but it might be as close as sibs again. Thus we had 
alternate generations of double cousins (or nearer relatives) 
mated in pairs of which we selected the offspring (a sibship) of 
the most promising pair to mate en masse. Selection therefore 
really took place in alternate generations. While we recognized 
that this procedure was not ideal theoretically, at least from the 
point of view of a strict selection experiment, the advantages out- 
w^eighed the disadvantages, inasmuch as it made the whole selec- 
tion experiment possible in a practical sense and yet maintained 
inbreeding. The chief disadvantage lies in the fact that this 
method precludes calculation of the parent-offspring correlation 
and regression coefficient for any two successive generations. 
In all these selection experiments, after the Pi generation, all 
w m 
of the matings \vere of the type ,„ ^ X w m; i.e., red 
long females heterozygous in white miniature mated to white 
miniature males, except where special tests were made for the 
sake of genetic analysis. This type of mating gave crossovers 
among the offspring of both sexes and thus a more effective 
criterion for selection, since numbers were doubled. It also 
gave the doubly heterozygous females and the ultimate recessive 
