438 Tricotyloiis Races. 



were planted out in the summer uf 1893 and eight of 

 them yielded an ample quantity of seed. In each lot, of 

 from 200 to 1300 seedlings, I counted the percentage of 

 tricotvls and found most of them to be distributed be- 

 tween 2 and 15%, the highest numbers being 24 and 

 32^0 (these latter amongst 1300 and 1060 seedlings re- 

 spectively). The mean w^as 11%. 



Therefore the original plant obviously belonged to a 

 tricotylous intermediate race, a fact which, however, was 

 only pro\'ed by the behavior of its descendants. 



In the spring of 1894 I only planted out the tricotyl- 

 ous ofl'spring of a parent with 32%, and obtained 22 

 ])Iants which set seed. From these I obtained in the 

 following year 22 hereditary values which were distrib- 

 uted between 26 and 55% and reached a mean of 37%. 

 There were three individuals with 5^1 — 55 and 55%. We 

 see that the mean was higher than the corresponding 

 \alue of the parent plant, and from this we conclude that 

 regression did not take place in the direction of 0, but 

 towards the other side ; and this is exactly what should 

 happen in the isolation of new races from their mix- 

 tures. 



In the summer of 1895 I planted out tricotylous off- 

 spring from the two parents with 55% ; I kept them in 

 two groups and determined their values from the har- 

 vests of 31 individuals. There was no essential differ- 

 ence between the two groups ; the numbers were dis- 

 tributed between 26 and 73 and their mean was 53%. 

 The mean had, therefore, reached the value of the parent, 

 and the race could be regarded as an intermediate race, 

 isolated but not yet subjected to selection. 



For the next generation I selected the tricotylous 

 offspring of three plants with 66 — 68 and 73%. Of 



