Tricotylous Races Do Not Arise by Selection. 403 



siderably larger number of individuals, I could yet im- 

 prove the prospects of success. For this purpose 1 se- 

 lected the two plants of 1895 which seemed to be the 

 best, and for which I had found the highest values in 

 the spring of 1896, viz., 5.3% and 5.5%. In order 

 to have a large crop to select from, I sowed 15 culjic 

 centimeters of seed, and raised from 15,000 to 20,000 

 seedlings, of which the strongest tricotyls and hemi- 

 tricotyls were planted out singly in pots, and later put 

 out in the beds. Three sowings were made, at the end 

 of March, at the end of April and in the middle of May, 

 in the hope of possibly increasing thereby the varialjility 

 and the prospect of a mutation. Furthermore, within the 

 three groups, widely different positions, different degrees 

 of remoteness of the individual plants, and different 

 treatment in the matter of pruning, obtained. Many 

 plants gave more than 30 cubic centimeters of seed each, 

 but more than ten cubic centimeters was never saved. 



Altogether I saved the seeds of about 450 plants, 

 and sowed them separately. For each seed-parent 300 

 seedlings were recorded in 1898; and the proportion of 

 tricotyls was calculated from these data. The result was, 

 however, that a very great difference was seen to exist 

 between the two grandparents of 1895; the one with the 

 \'alue of 5.3% proved to be a bad stock plant. Amongst 

 its offspring, of which there were 30, the ratio was 

 greater than 3% in ten cases only, and on the average 

 it was 0.3 — 0.5%.' And this in spite of the treatment, 

 which, though varied, was the best that could be given, 

 and in spite of the complete exclusion of atavists. 



The second grandparent, with the value of 5.5%, 

 proved as fortunate in its progeny as the former had 

 been unfortunate. Its offspring had all been sown to- 



