No. 594] THE MECHANISM OF CBOSSIXd-oVEi; 



359 



ample, in the last step shown in the preceding diagram, 

 where factors ywAb and vmsrf are to be put together, it 

 was found that females of both of these kinds were ex- 

 tremely difficult to keep alive. It was, therefore, decided 

 to mate a vmsrf male by a female which contained ywAb 

 in one chromosome and normal factors in the other. Such 

 a female would be easy to breed from, as the normal fac- 

 tors dominate. About half the daughters (let us call them 



F,) would be of composition " -- ymsrf (representing 

 the mutant factors in the maternally derived chromosome 

 on the upper line, those from the father on the lower line). 

 All the daughters (FJ would, however, appear normal, 

 but if these F\ females were bred in separate bottles, those 

 of the desired composition ^^~^^f woul( i De distin- 

 guishable from the others by their offspring (F 2 ). All 

 bottles in which the parents (F\) had not been of the de- 

 sired composition could then be discarded. Next, among 

 the offspring (F 2 ) of those females which proved to be of 

 composition ^ Wf ^\, msr f? it- was necessary to select the 

 ones which, by reason of crossing-over between b and v, 

 contained all nine factors in the same chromosome (?*. e., 

 ywAbvmsrf ) . But such individuals, if homozygous, never 

 live long enough to mate, so great is the lowering of via- 

 bility produced by all these mutant factors at once. Con- 

 sequently, some method must be used of obtaining in this 

 cross heterozygous individuals (F 2 ) which received this 

 cross-over "nontuple" chromosome from their mother, 

 and of distinguishing these from other individuals pro- 

 duced by the cross. The natural suggestion would then 

 be that the F, females should be mated by normal males, 

 and the F 2 which receive this cross-over chromosome could 

 then be distinguished by breeding tests as their mother 

 had been. The crossing-over desired, however, does not 

 occur in more than one eighth of the flies, and so breeding 

 tests designed to be certain of securing at least one indi- 

 vidual of the required composition would have to be 

 rather extensive. In this case, however, the desired F 2 



