372 Origin of Each Species Considered Separately. 



remained a year in the soil. Such plants also occurred 

 in the main culture referred to on page 224, although 

 they are not referred to there. I selfed 38 plants, in 

 bags, with their own pollen. They were all mutants from 

 Lamarckiana, some with three, some with two genera- 

 tions of tall ancestors. I saved and sowed their seed 

 separately ; and recorded the seedlings at the stage shown 

 at Fig. 78 B. Doubtful ones were allowed to grow a 

 little further. The average number of seedlings from 

 each of the 20 seed-parents was 500, the maximum was 

 860, and only in three cases was it less than 100. 



The total number of seedlings was 18649; they were 

 without exception dwarfs. Three of them were also ob- 

 long a and one also elliptica. 



Thus the seed-parents of the second experiment proved 

 themselves to be, like the 20 of the first, perfectly con- 

 stant. 



It seemed to me important to test the constancy of 

 nanellas from other sources. I chose for this purpose 

 two plants from a scintillans-isimily. This family arose 

 from the lateral branch of the Lamarckiana group (p. 262) 

 and indeed from the only individual referred to there. 

 This was biennial and flowered in 1896. I sowed some 

 of the self-fertilized seed of this in 1898 and selfed the 

 scinfillans plants again with their own pollen. From the 

 seed thus produced I obtained nine examples of nanella, 

 which I transplanted and selfed. Only two of them 

 however set seed. They had two generations of scin- 

 tillans behind them, and behind these two generations 

 of Lamarckiana. 



The plants had become very weak; and the harvest 

 was a meagre one. Only 64 seeds germinated ; but they 

 were all nanella. This shows that the dwarfs even when 



