Vai-iatiim aud Seleotidii; ü Reply. 259 



thPiii tliat a minus variant should produce anytliin>>- l)ut similar minus 

 variants, if mated with its like. This particular case shows the ground- 

 lessness of their assumption. The iiiue youuy of the first pair of — 2 

 jiarents were actually distributed as follows: — 



Grade . . . —-74, —1, — 1''4, — I'Vi, —2 



P^reciuency . . 1, 2, .3. 1, 2. 



(See Castle and Phillips, Table 19.) 



lu this case is seen a marked regression in grade downward, toward (i. 

 Parents of this same grade { — 2) chosen from the minus series at the 

 present time (generation 15) produce more offspring of grades higher 

 than of those lower than grade — 2, that is regression now occurs in 

 the opposite direction, as compared with its original direction. Facts 

 such as these show that the phenomenon of regression as originally 

 described by Galton is real, not apparent merely, as Johannsen's 

 work seems to show, and that the direction of regression with reference 

 to a jmrticular condition can be altered by repeated selection. 



5. It is evident that for the array of grades just given an average 

 nuiy be computed which will express the direction and degree of the 

 regression. This average is — 1"36 showing that the regression from 

 the grade of the parents is back toward 0, and in amount 0-64. A 

 comparison with the grade of the offspring of — 2 parents in subsequent 

 generations is- instructive. For the first eight generations in which 

 pairs of — 2 parents occur, the average grades of their offspring were 

 as follows. The precise distribution of the offspring in grade is recorded 

 in Tables 19-26, Castle and Phillips. 



It will be observed that ujjon repeated selection away from 0, the 

 regression toward <• grows less and less until in generation 11 it 

 coincides with the direction of the selection. 



18* 



