178 CARL DOWNEY LA RUE 



continued to increase to a marked degree. Obviously this continued 

 increase was not due to selection; on the contrary one would expect a 

 lessened divergence between the high and low series following the cessation 

 of selection. The increased divergence during the non-selection period 

 was almost entirely due to increase in size and spine number in the high 

 series. That this series later produced some very large forms, and that 

 great difficulty was experienced in securing offspring from this group are 

 facts that may be indicative of abnormality in the group. As in HEGNER'S 

 other experiments, relatively few individuals were studied. 



The extensive and painstaking studies of JENNINGS (1916) on Difflugia 

 present what appears to be the least questionable evidence of the effective- 

 ness of selection yet published. In one experiment he made selections for 

 seven periods, each of which, except the first, included one generation. 

 Selection was made for number of spines, individuals with from 1 to 3 spines 

 being chosen as parents for a "low" set, and those with from 5 to 8 spines 

 being retained as parents of a "high" set. 



The selection was apparently effective and is so interpreted by JEN- 

 NINGS. Two lines were formed by the selection which differ in mean spine 

 length. However, a closer examination shows that the divergence of the 

 two lines varied from generation to generation. In the second period the 

 divergence is equal to nearly a whole spine (.95 spines), but in the third 

 period the divergence falls to .03 spines, which means the lines are really 

 identical statistically. From the third period on the divergence continu- 

 ally rises and falls but never again reaches so high a point as in the second 

 period. If the divergence is due to selection why is the effect so great in 

 the second period, and why do further selections result in a decreased 

 divergence? 



The individual measurements are not published, but if we compute the 

 error of the difference between the means of the two groups, using the 

 generation means, we find it to be . 103. The difference between the means 

 of the two groups is .22 .103 spines. The divergence is barely more than 

 twice its probable error and so is not statistically significant. 



During the selection periods the two groups showed fluctuations in 

 regard to diameter of shell and length of spine which are similar to those 

 already noted for number of spines. Table 11 shows the divergence of the 

 two groups for each of the three characters. 



It may be seen at a glance that the differences do not steadily increase 

 with an increased number of selections but rise and fall as if by chance. 

 The difference in spine number is greatest in the second period, that of 

 shell diameter is greatest in the fourth period; and that for spine length 



