434 EDWIN CARLETON MacDOWELL 



Averages of r/Ej. 



From direct observation of figure 1 it appears that three groups 

 of generations may be roughly distinguished : group 1 , genera- 

 tions 2 to 10; group 2, generations 11 to 34; group 3, generations 

 85 to 49. In the first group the correlation coefficients are gen- 

 erally significant and positive; in the second group the coefficients 

 fluctuate, their values are smaller and, in some generations, sig- 

 nificantly negative; in the third group, the only significant 

 coefficients are negative. 



Since the second group of generations involves the only un- 

 certain results, it should be especially examined. Positive cor- 

 relation is surely present, but so is significant negative corre- 

 lation. From figure 2 it will be observed that especially low- 

 grade parents were among those used in generations 16 to 23. 

 These formed the return-selected line already mentioned. The 

 means of this line are generally parallel with the means of the 

 high-selected race during the corresponding period (Mac- 

 Dowell, '17, p. 125 and fig. 7), yet there is a tendency for the 

 means of the return-selected line to fall a little lower. As 

 already indicated, this difference in the means is probably due 

 to the unconscious selection of a weak race. 



A few unpublished data on the relation of the size of the fly 

 to the bristle numbers may well be presented at this point. 

 Sixty-four small flies from an old, dried-up bottle were turned 

 into a new bottle whose conditions proved to be especially fav- 

 orable. When the bottle was next examined the new generation 

 had begun to hatch. The offspring were so much larger and 

 brighter colored than the parents that there was not the slightest 

 doubt as to which generation each fly belonged. Fifty-one flies 

 had hatched in the next generation; their average weight was 

 0.000943 gram; the average weight of the parents at this time 

 was 0.000310. The bristle counts of the two groups were as 

 follows: 



