EFFECT OF SELECTION ON CROSSOVER VALUES 341 
The first two selections seemed to show little or no effect. 
Although the values of the selected pairs were low, their progeny 
regressed practically to the parental average. Possibly this 
means that all wide deviations were not necessarily due to 
genetic causes and that we had difficulty in distinguishing 
between wide environmental variates and wide variations due to 
genetic causes. Selection was thus effective only when by chance 
we chose a wide variate due to the latter set of causes. For 
example, in the F3, we chose a female showing 17.99 per cent cross- 
overs, but her progeny gave an average of 26.18 per cent. After 
the F5, progress was very rapid. The F9 gave 16.49 per cent, and 
the Fio to F13 gave about per cent. These last generations 
in this series were based upon small totals, because the excessive 
heat (90° to 100°F., day and night) for long-continued periods 
reduced fertility to a minimum and eventually annihilated our 
stock in this one. However, series A', which was derived 
directly from series A, gave just as low crossover values with 
larger numbers and under better conditions. We may be quite 
sure that temperature was not the cause of low crossing over; 
for, if we may anticipate, series B showed effects of selection 
under normal temperature conditions. 
Series A', low selection; derived frojn series A 
In the F7 generation of series A, two selections were made. 
One female ( 9 14) gave 9:104 = 8.65 per cent, and a second 
female (9 lO) gave 1:91 = 1.10 per cent. The former was 
used to continue series A, while the latter was used to begin a 
new series. A'. Table 2 and text-figure 1 give the main facts 
pertaining to series. A'. We began this series to insure keeping 
alive some of the low crossover material of series A during con- 
tinuously hot weather. Our facilities did not permit controlling 
temperature, and the whole experiment was in a precarious 
situation during the early summer months of 1916. We found 
that mating a number of females en-masse assured more progen}^ 
than the same number of females mated in individual bottles — 
evidently because the larger number of larvae carried the yeast 
through the culture and kept molds down. Hence, during the 
