No. 595] THE MECHANISM OF CROSSING-OVER 



the frequency of crossing-over between factors is deter- 

 mined by their distance apart in the chromosome. It sup- 

 plements the other evidence for these conclusions that had 

 previously been found by Sturtevant in the linear manner 

 of linkage of the factors. 



3. It seems uncertain whether crossing-over occurs in 

 the strepsinema stage, as concluded by Janssens, or earlier 

 in synapsis. The cytological evidence at present at hand 

 would seem insufficient to settle this point. Possible tests 

 for various alternative mechanisms of crossing-over are 

 proposed. 



4. In order to study the nature of crossing-over by 

 means of " interference, " stocks were made up that dif- 

 fered in regard to many factors. Females heterozygous 

 for 22 pairs of factors were thus obtained, and a special 

 method was devised for testing their output. Other 

 special methods for obtaining multiple stocks, and for 

 eliminating discrepancies due to differential viability, 

 have also been presented. 



5. The results have been arranged in the form of a 

 curve showing the amount of interference for various dis- 

 tances. The results thus far obtained confirm those ob- 

 tained by less exact methods, and also give evidence that 

 interference decreases gradually with distance from a 

 point of crossing-over; this, taken together with certain 

 evidence from non-disjunction, lends some probability to 

 the view that crossing-over occurs at an early stage m 

 synapsis. 



6. A case of crossing-over in an embryonic cell of a 

 male is reported. 



7. Incidentally, the experiments have afforded an ex- 

 tensive test of Castle's assumption of contamination of 

 factors by their allelomorphs. Outcrossing in each gener- 

 ation for 75 generations has failed to change any of the 

 factors. 



The author is deeplv indebted to Professor Morgan, and 

 wishes also to convey his appreciation of the active co- 

 operation so often rendered him by E. B, Altenburg and 



