Improved Races After Selection Ceases. 125 



cockscoml^s, Celosia cristata, become much less flattened ; 

 but the doubling, branching and fasciation are never en- 

 tirely lost and if single individ- 

 uals should seem to lack them, 

 they will always reappear in 

 plants grown from their seed. 

 The same conditions obtain in 

 my experiments with Papaver 

 somnifcrum polycephaluin and 

 in Tri folium pratense quinque- 

 foliuin. On the cessation of se- 

 lection these plants lose the high 

 pitch of their improvement but 

 not the character itself. 



I have already referred (pp. 

 72-73, Figs. 17 and 18) to my 

 experiment with maize. Starting 

 in 1886 with an ordinary race 

 of maize whose average number 

 of rows varied between 12 and 

 14, I had succeeded by 1891 in 

 raising a race with a mean of 20 

 rows, a number which the orig- 

 inal hardly ever reached From 

 1892-6 the race was maintained 

 by selection at about the same 

 level. During the period 1897- 

 1899 however I selected the ears 

 with the smallest number of 

 rows. In 1897 I sowed the seeds 

 of a 16-rowed ear, but the mean 



of the harvest which they bore lay still at 20. The next 

 year, 1898, the mean lay at 18, and in 1899 at 14-16 



Fig 



26. Triticum turgidum 

 compositum, branched 

 Wonder or Smyrna 

 Wheat. 



