The Laevifolia-Family. 277 



rubrinervis. But as I had not intended to go on with 

 more than one of these two races I discontinued the ex- 

 periment with this branch of the family after I had 

 recorded the 1893 crop. 



The object of the other two experiments was to 

 breed a race with long fruits and one with short ones 

 in order to find out the way in which this character re- 

 sponded to cumulative selection. (See Part III.) The 

 long-fruited race gave rise to no mutants from 1892 to 

 1894 inclusive; the short-fruited race was, on the other 

 hand, considering the relatively small extent of the ex- 

 periment, extending, as it did, over from 4 to 6 square 

 meters each year, fairly rich in mutants (see page 273). 

 It gave rise, in 1893, to a beautiful rosette of eUiptica 

 and two dwarfs (an annual one and a biennial one); 

 and, in 1894, to two annual latas which however had not 

 succeeded in flowering by the middle of September ; two 

 nibrinerz'is,^ a rosette and a weakly stem which did not 

 flower; and, lastly, a rosette of elliptica which I did not 

 succeed in wintering. 



In conclusion I will remark that the tricotylous and 

 long-fruited races referred to are excluded from the table 

 on page 273, whereas the short-fruited race (1892-1894) 

 forms the direct line of descent of the laevifolia- family . 



In this family, therefore, there have appeared 41 

 mutations altogether, 13 in the main line of descent, 21 

 from Lamar ckianas and 7 from examples of rnhrincrvis. 

 This evening-primrose experiment occupied a relatively 

 wide area in 1889 so that the characters of the young 

 plants had more chance of developing than in previous 

 years; this circumstance may, in part, account for the 

 greater number of mutants that appeared that year. This 



* See the table on page 273. 



