No. 594] THE MECHANISM OF CROSSING-OVER 351 



other, or to actual differences in the behavior of the chro- 

 mosomes in the two experiments. As the last two influ- 

 ences seemed by no means negligible, and as the experi- 

 ment involving many points at once gave a more direct 

 and graphic picture of the results, it was decided to use 

 this method of attack in preference. Moreover such an 

 experiment incidentally afforded an opportunity of attack- 

 ing certain other questions, such as the effect, on crossing- 

 over, of having the two chromosomes different in regard 

 to many factors. Meanwhile, the indirect method of attack 

 would be followed by other workers, and the two sets of 

 results could finally be used as checks upon each other. 



The many factor method is in itself for several reasons 

 very laborious, but this is compensated for by the fact that 

 when the results are obtained they are the equivalent of 

 an entire series of different experiments involving in turn 

 the linkage of each factor with every other one, and indeed, 

 the results are much more than the equivalent of these, 

 for in the latter cases the linkages are obtained in differ- 

 ent experiments, so that there is much more chance for 

 error in determining the relation of one linkage to another. 



It was evident from the outset, however, that there was 

 one very important obstacle to be overcome in any study 

 of linkage exact enough to give useful information regard- 

 ing coincidence, and that the difficulty was especially great 

 in the type of experiment contemplated. The difficulty re- 

 ferred to is " differential viability," for it is found that 

 in nearly all experiments not involving the characteristics 

 of seeds or other structures dependent upon the maternal 

 organism for support, the individuals belonging to differ- 

 ent genetic classes may be very differently equipped in 

 respect to their ability to meet the struggle for existence. 

 Thus, since the count generally takes note only of the indi- 

 viduals which survive, the ratios obtained may be very 

 different from the ratios of the different classes of gam- 

 etes. These discrepancies apply especially to forms like 

 flies, the larval life of which can not be well controlled, and 

 they are, of course, particularly great in crosses involving 

 many factors at once. 



