EFFECT OF SELECTION ON CROSSOVER VALUES 343 
Series A', like series A, was eventually lost in the latter part 
of the summer of 1916 because of an unavoidable succession of 
events. We regretted the loss of this stock because we had 
hoped to make a genetic analysis of the last generations in an 
attempt to learn what was taking place during selection. How- 
ever, the data as they stand indicate that crossing over is not a 
very stable phenomenon and that it can be rather easily modi- 
fied. We surely cannot concur in Morgan's ('19) view that 
crossing over ''gives numerical results of extraordinary con- 
stancy." 
We immediately began a new selection experiment, hoping 
that we could duplicate the results of series A and A'. 
Series B; low selection 
Series B, like the preceding series A and A', began with the 
mating of a single white miniature female and a wild red long 
male. In fact, as a prelude to series B, we made eighty such 
paired matings, for we had found some non-disjunction in our 
original stocks and in series A and A'. Since non-disjunction 
theoretically lowers the percentage of crossing over (Bridges, '16), 
we wished to assure ourselves, if possible, that this cause might not 
be operative in producing low crossover values in our selection 
experiment. Of the eighty white miniature females tested we 
found eleven giving either matriclinous daughters or patri- 
clinous sons or both. This must mean secondary non-disjunc- 
tion in the white miniature stock, for the exceptions were -too 
numerous to be considered primary. We chose white miniature 
9 53 mated to a wild male as the foundation pair for our experi- 
ment, because this pair gave fifty-two wild-type daughters and 
seventy-eight miniature white sons. While they showed no 
exceptions, it does not prove that 9 53 may not have been non- 
disjunctional (XX Y), for a ratio of 0: (52 + 78) might well 
occur as a chance ratio where an average of 4.3 per cent of excep- 
tions is expected from XXY females (Bridges, '16). However, 
in the present paper we are concerned only with the question 
whether selection based on variable crossover ratios can be effec- 
THE JOURNAL OF EXPERIMENTAL ZOOLOGY, VOL. 32, NO. 2 
