EFFECT OF SELECTION ON CROSSOVER VALUES 353 
of cases. While we cannot subscribe fully to this theory for 
cogent reasons advanced by Sturtevant ('17), Bridges ('17), and 
Jennings ('18), nevertheless Goldschmidt's proposed theory 
would not appear entirely supererogatory, for a crossover value 
is apparently a variable and the variation is due to or controlled 
by multiple hereditary factors. A cross between low crossover 
stocks and the original population, and testing out a large num- 
ber of Fa segregates should throw the desired light on this ques- 
tion. Unpublished data indicate that segregation in crossover 
values does take place as one would expect on the basis of the 
multiple-factor explanation. 
5. May we suppose that we have been taking advantage of 
small mutations in the nature of modifying factors arising during 
the course of selection? While this is possible we are inclined to 
doubt it, for favorable mutations evidently do not take place 
in the direction of selection as readily as this view would imply 
(cf. Muller and Altenburg, '19). 
The following conclusions may be drawn from the data pf this 
paper : 
1. Crossover values are very variable and part of this vari- 
ability is due to genetic causes. 
2. Low selection has been effective in two entirely independ- 
ent series, A and B. 
3. The low crossover stock bred true to about 0.6 per cent 
(almost zero) for nine generations in series A' (derived from 
series A) . 
4. The low crossover stock bred true to about 6 per cent for 
twenty- two generations in series B. 
5. High selection probably induces double crossing over, as 
shown by series C. 
6. Crossing over in the various regions of the sex chromosome 
(and other chromosomes?) is probably controlled by multiple 
incompletely^ dominant factors. 
