No. 592] THE MECHANISM OF CROSSING-OVER 205 



B rarely separate coincident ally, bnt usually remain to- 

 gether, and C, D, E, etc., all remain together also, separa- 

 ting in a body from A and B. In other words, the factors 

 are not interchanged singly, but stay together in sections, 

 according to their positions on the diagram, and whole 

 sections are exchanged at once. 



It may be objected that these conclusions are in many 

 cases based on linkage values obtained in different experi- 

 ments; that it is unwarranted to conclude, for instance, 

 that when C and D separate, E remains with D, simply 

 from the fact that in one experiment the frequency of 

 separation between C and D had the value m per cent., in 

 a second experiment DE was n, and in a third experiment 

 CE was m + n, for this conclusion would only be true on 

 the supposition that in all three experiments each factor 

 had the same frequency of separation with each of the 

 others as in the particular experiment where that fre- 

 quency was determined. The answer is that numerous 

 experiments have been performed in which three pairs of 

 factors (or more) could be followed at the same time, and 

 these experiments have given results precisely the same 

 in kind, although more accurate than the preceding. But 

 in experiments of the latter type, the coincidence of the 

 various separations and non-separations does not have to 

 be calculated out as in the case above, but is given directly 

 by the results. Thus in a hybrid which has received ABC 

 from one parent and the allelomorphs abc from the other, 

 gametes in which coincident separation between B and A 

 and between B and C has occurred will be distinguishable 

 by having either the composition aBc or the composition 

 AbC (see Fig. 3), and the number of such offspring can 

 thus be directly counted instead of it being necessary to 

 calculate them from the relations between separation 

 values for A and B, B and C, and A and C. And in the 

 experiment of the author's given in Section V, where the 

 inheritance of a large number of factors is followed simul- 

 taneously, the results show directly and graphically that 



