166 ELMER ROBERTS 



b. Effect of 'crossing-in.' If selection had no effect in either 

 of the two series, then the differences existing must have some 

 other explanation. It will be of value to compare the generations 

 in the two series, which were produced during the same period of 

 time. Generations F22 to F34 of selected series A and F7 to F19 

 of 'crossed-in-and-selected' series C were produced at the same 

 time and were subject to the same conditions, except that the 

 latter series had been crossed to long-winged stock a few gen- 

 erations back in its history. 



Text figure 3 shows the curves of the means for the contempo- 

 rary generations of these series noted above. In only one case, 

 the twelfth generation, does the 'crossed-in-and-selected' series 

 fall as low^ as the simple selected series. By taking the average 

 of the means of the males and females in each of these series, we 

 have 19.07 ± 0.069 for selected series A (F22 to F34) and 25.62 ± 

 0.12 for 'crossed-in-and-selected' series C (F7 to F19). This is a 

 significant difference. As the average of all the generations in 

 each series, we have 19.27 ± 0.043 for selected series A and 



25.68 ± 0.073 for 'crossed-in-and-selected' series C. This is 

 also a significant difference when judged by their probable 

 errors.^ 



The males were more affected by 'crossuig-in' than were the 

 females. The average of the males for generations noted above 

 in the 'crossed-in-and-selected' series is 27.99 ± 0.19 and of the 

 females 23.40 ± 0.14. That this difference is due to having been 

 crossed with the long wing is proved by the fact that in the 

 generations of selected series A produced at the same time (F22 

 to F34) the average of the males is 19.43 ± 0.11, and of the females 



18.69 ± 0.078. For all the generations in 'crossed-in-and- 

 selected' series C, the males average 27.70 ± 0.12 and the fe- 

 males 23.73 ± 0.090, while selected series A gives 19.40 ± 0.068 



'* The probable error of an average of averages was calculated from the 

 formula: 



E = ^ Vri7e7+V2^"e7+T7777777nm2~e^ 



N 



in which n is the number of individuals in the generation, e the probable error, 

 and A'' the total number of individuals. 



