No. 594] THE MECHANISM OE CROSSING-OVER 355 



present in heterozygous condition (and the effect of the 

 one or two characters wherein the father may not have 

 been dominant). Thus error due to differential viability 

 may be held within safe bounds. 



It may be objected, however, that we have, as it were, 

 killed the patient in curing the disease— that there is no 

 use in overcoming the discrepancies in the count due to 

 differential viability, if we thereby eliminate the possi- 

 bility of making any count at all, by making all the off- 

 spring appear alike! It is true that, in such an experi- 

 ment, it is impossible to tell by inspection of any offspring, 

 what maternal factors were present in the ova from which 

 they sprang, since these factors are made invisible, so to 

 speak, by the dominant factors brought in by the sperm. 

 But the factorial composition of each of these offspring 

 (which we will for convenience call "F 2 ") can be deter- 

 mined by breeding tests. The plan which was followed 

 was to mate the F 2 flies, each in a separate bottle, to indi- 

 viduals containing the recessive factors. Thus whatever 

 recessive factors were present in the eggs of the original 

 heterozygous female ("F^'), whose output it was desired 

 to test, would become visible the generation after (in 

 "F 3 "). Whereas, in an ordinary linkage determination, 

 each bottle produces a large number of flies, which need 

 merely be classified according to their api>earance, and 

 counted— in this case, each of the offspring themselves re- 

 quires to be mated and given a whole bottle to itself, and 

 its progeny in turn ("F 3 ") must be examined. In other 

 words, in ordinary cases, there is only one bottle necessary 

 for a count of many flies, but in this case one bottle repre- 

 sents one fly of the count. The numerical relations exist- 

 ing between the flies ("F 3 ") hatching in one of these final 

 testing-out bottles need not be determined, however ; that 

 is, these flies need not be counted ; all that is necessary is a 

 "qualitative" determination of what recessive characters 

 appear among them, in order to judge of the composition 

 of their parent (F 2 ), which is the fly recorded in the count. 

 Thus far 1008 of these test bottles have been recorded. 



In preparation for this experiment the main task was to 



