LINKAGE VARIATION IN DROSOPHILA / 



is to be investigated. In the 'repulsion' experiment (table 6) 

 each of the four females showed a marked decrease, and the totals 

 show a decrease of nearly five units. 



The most convincing experiment is the following, in which 

 three second chromosome loci (namely, black, purple, and 

 curved) are run together in the same back-cross experiment. 

 Such an experiment is much more satisfactory than three separ- 

 ate experiments would be in studying linear arrangement. In 

 spite of the labor of getting the triple recessive used, and the 

 time of classifying the flies in a more complex manner, there 

 are many advantages. Since the same data furnish three 

 linkage values, only a third as many individuals need ultimately 

 be raised, or three times the amount of data may be obtained 

 in the same time. The resulting values are more exactly and 

 safely comparable, since they were produced under the same 

 conditions. The order of gens is most strikingly shown by 

 means of the smallness of the two contrary classes which are the 

 result of double crossing-over. The amount of double crossing- 

 over can be directly observed instead of being calculated from 

 linkage values obtained in three separate experiments and 

 perhaps not strictly comparable. A comparison of the observed 

 amount of double crossing-over with the expected amount 

 gives a measure of the amount of interference. Finally, if there 

 is any disturbance of the linkage, it is important that the differ- 

 ent values be derived from the same experiment, so that the 

 disturbance can be shown to be localized or to affect the whole 

 chromosome alike, as in the present case (table 7). 



In the totals of the broods the change for each value is sig- 

 nificantly a fall. The thirty-five cases of changed ratios are dis- 

 tributed as shown in table 8. In this table the percentage of 

 decrease are calculated from the brood totals of table 7. It is 

 evident that the fall has affected each section of the chromosome 

 by approximately the same amount. 



Of the twelve families of table 7, eight showed a fall and two 

 a rise in both component values (black purple, and purple 

 curved). It might therefore seem that there is some correlation, 

 such that when black purple changes greatly purple curved 



