INTEKFERENCE 127 



The data also show that the series may break at two 

 points, and that when this happens the three blocks of one 

 set always correspond to the three blocks of the other 

 series of genes. Thus interchange at two levels gives : 



abcdEFGHijkl 

 ABCDefghlJKL 



The same relation holds in principle for three or more 

 breaks in the series. 



If in such a system the blocks have no commonest 

 length, the break in the series at one level should not 

 bear any relation to the place at which another break 

 takes place. For example, if it is true that when a break 

 occurred between D and E it had no influence on a break at 

 any other point of the series, the blocks resulting from two 

 breaks would not tend to be more of one length than of any 

 other length. But if the evidence shows that when a 

 break occurs between D and E the chance of another break 

 occurring in that vicinity is decreased, or increased, the 

 results would be expected to follow some definite law or 

 principle, rather than be simply the result of chance. This 

 is in fact the case. An illustration may make this clear. 



Suppose when crossing over takes place within the 

 blocks A B C D, smd E F G H, and I J K L it can be 

 recorded. If we know how often, when the break occurs 

 only once in the series, it takes place in the first, in the 

 second, or in the third block, we can then determine in 

 those cases where breaking occurs in the first block, 

 whether it is as likely to tal^e place in the second block as 

 when no break occurs in the first, etc. Such tests have 

 been made (MuUer, Sturtevant, Bridges, Weinstein, 

 Growen) with Drosophila, and the same kind of results con- 

 sistently obtained. It has been found, for example, that 

 when a crossing over takes place between G and H, a sec- 

 ond one is less likely to take place on either side, i.e., 

 between F and G or between H and I than when no cross- 



