No. 592] 



THE MECHANISM OF CROSSING-OVER 



215 



such variations would have gradually accumulated in the 

 course of evolution, until no correspondence remained 

 between the relative lengths of the chromosomes and the 

 total separation frequencies. But the parallel which does 

 exist between the observed chromosome lengths and the 

 usual (homozygous) total frequencies, would seem too 

 close to be meaningless, and so we should be led to be- 

 lieve that for some reason marked variations in the fre- 

 quencies of particular groups, even though they may be 

 possible, do not generally persist; in other words, the 

 frequencies seem usually to stay at least roughly propor- 

 tional to the actual chromosome lengths, and to furnish 

 another verification of the theory of crossing-over. Fur- 

 ther evidence of this will be met with when we consider 

 group IV. 



As the data whereby the positions of the factors and 

 the total frequency of separation have been determined 

 in group III, have not hitherto been published, it .may be 

 of interest to present some of them here. In order to 

 obtain data on as many combinations of factors as pos- 

 sible in the same cross, so that the linkage values between 

 the different factors would be comparable, I have en- 

 deavored to make up, by cross-breeding, stocks containing 

 six or seven mutant factors in group III at the same time. 



Since on account of the baffling linkage variation, the 

 order of these factors could not well be determined by 

 combining the results of separate experiments each of 

 which dealt with only two factors at a time, it required a 

 great many trial matings before such multiple stocks 

 could be made up, as of course the crosses have to be 

 made in a certain precise order, to secure a combination 

 of many linked factors. To obtain stock ABC, for ex- 

 ample, it would not suffice to make up AC and then mate 

 it to B, for it would then require two coincident recombi- 

 nations (which might never occur) to secure ABC. More- 

 over, as a first step it had been decided to get combina- 

 tions of ebony with various factors, and very much time 

 was lost in this attempt, as it was not then known that 



