1 68 H. E. EWING. 



+ 22/14, + 10/13, + 16/13- + 2 13, + ii/io, o, - - 6/3, + 6/2, 

 + 16/1, + 5/5. + 9 '6. + 1,6, + 8/8, 4/9, + U.'iS, + 42/28, 

 + 22/31. 



We may also express this series in the form of decimals, which 

 will be as follows: 



1.57, .76, 1.23, .15, 1. 10, o, - 2.00, 3.00, 16.00, i. oo, 1.50, .16, 

 i. oo, - - .44, .72, 1.50, .70. 



These fractions added together and divided by their number 

 should give us the average amount of regression. If the regres- 

 sion is according to Galton's law the decimal should be 0.333; if 

 according to Johannsen's predictions, that is if the regression is 

 complete, it should be i.oo. The figure which we actually obtain 

 by adding these fractions and dividing by their number is 1.64. 

 In other words, the regression is more than complete, or beyond 

 the mean of the strain. However, it should be noted that the 

 number of individuals included in this computation is too small 

 to permit the results to be conclusive. Yet the results show 

 that regression in a pure line of a parthenogenetic form does not 

 follow Galton's law, also that there appears to be some justifica- 

 tion in the contention made in a previous paper of mine, that 

 regression under these conditions is somewhat pendulum-like, 

 swinging beyond the mean of the strain, or line. 



