EFFECT OF SELECTION ON CROSSOVER VALUES 351 
3. In series A and A' we found much evidence of non-disjunc- 
tion. Bridges ('16) stated that XXY females should logically 
show a decrease in crossing over, because heterosynapsis takes 
place in about 16.5 per cent of the cases and precludes crossing 
over in these cases. However, Bridges also showed that the 
experimental results disagree with such an expectation, for cross- 
ing over was not decreased among the regular sons of XXY 
females, but as far as the evidence goes it was slightly increased. 
For some time we labored under the impression that much, if 
not all, of our decreased crossing over was associated with the 
presence of non-disjunction (Detlefsen and Roberts, '20). We 
are now rather inclined to believe we were in error. It should 
not be a difficult matter to free our low crossover stock in series 
B from non-disjunction and thus dissociate this possible cause 
from the others. We could in this way demonstrate that non- 
disjunction was only accidentally present in our experimental 
material and that our results are quite independent of non-dis- 
junction. 
4. Have we reduced the frequency of the usual single 'chromo- 
some twist' between white and miniature to a minimum? Wein- 
stein's ('18) results indicate that crossing over takes place in the 
sex chromosome with about forty-six units as the modal distance 
between successive crossovers. Similarly, Gowen ('19) found 
twenty-five to thirty units in the case of the third chromosome. 
We began with two genes which were about thirty-three units 
apart, and which should therefore show a single crossover as the 
characteristic or modal number. This would mean that in 
series A, A', and B we have practically eliminated the usual 
single crossover in this region, while in series C we were on the 
way to increasing it to two crossovers (i.e., a double crossover), 
which would give us no crossing over as far as these two genes 
were concerned. Does this mean that we can decrease or 
increase the amount of 'twisting' which members of an homo- 
logous pair of chromosomes may show, and which is supposed 
to be responsible for crossing over according to the chiasmatype 
theory? If selection can accomplish this, then we may reason- 
ably suppose that numerous hereditary modifying factors are 
