t 



1907] DE VRIES^TWIN HYBRIDS 405 



L 



In the young plants, before the sending-up of the stems, it is of 

 course easy to count far larger numbers of plants, but the influence 

 of transgressive fluctuability is somewhat greater. I counted for 

 O. hiennisXbrevistylis 287, for O. biennis XLamarckiana 347, and for 

 O. biennis cruciataXbrevistylis 208 seedlings, and found 55 per cent., 

 60 per cent., and 64 per cent., or on an average 60 per cent* specimens 

 of O. velutina^ the remaining 40 per cent, being O. laeta. The figures, 

 although from the cause given not as exact as those given above, 

 evidently confirm the result. 



Tried in the second generation, from artificially and purely self- 

 pollinated seed, each of the twins yields a uniform progeny, with 

 exactly the same characters as its parent. Therefore they may be 

 considered as constant hybrid races. I made the majority of the 

 crosses of the following table in 1903; had the first generation, which 

 often consisted of only a few specimens on account of the dimin- 

 ished fertiUty, in 1905; and counted the forms in the second genera- 

 tion during the flowering-period of 1907. Most of the plants were 

 counted in full flower, some of them, however, in the condition of 

 rosettes of root leaves. Only for 0. muricataXbrevistylis the cross 

 was made in 1905 and the second generation grown in 1906- 



TABLE II 



Constancy of O. laeta and O. velutina in secoxd generation 



Second generation 

 Cross First generation laeta velutina 



O. biennis X Lamarckiana \ ^^^^^^^ '^ o 



O. biennis cruciata XLamarckiana •] i *: o q * 



O. muricata X Lamarckiana -I ^,^i^^i^^ q ^^ 



O. biennis X rubrinervis \ ^^^^^^ ^° t 



O. biennis cruciata X rubrinems j y^i^^[^g_ ^o 3 



O. muricataXbrevistylis. ............ | ^^^^^ ^ ^ 



Total 122 131 



I also tried this result by counting seedling plants as soon as they 

 clearly showed their differentiating marks. From O. laeta seed I got 



