58n 



THE AMERICAN NATURALIST 



[Vol. XLV 



becomes a question how many of these races are to be 

 included in each ' ' species, ' ' and it is often a purely arbi- 

 trary matter whether the line between two of the species 

 shall be drawn so as to include certain races in one or 

 the other. 



As an instance of this I may cite the case of 0. biennis 

 L. and 0. muricata L. I have cultivated many wild races 

 of both, including races of 0. muricata from such locali- 

 ties as Woods Hole, Mass., Gay Head, Mass., Seal 

 Harbor, Me., Middleton, N. S., St. John, N. B., Winni- 

 peg, Man., and St. Paul, Minn. Certain of the races, for 

 instance from St. John and Winnipeg, appeared to be 

 identical, but in most cases they produced uniform races 

 differing constantly in such features as width and color 

 of leaves (varying in different races from very narrow 

 to very broadly lanceolate) and average height of the 

 plant, as well as other characters. The only feature in 

 which all agreed was in certain flower characters, all hav- 

 ing a smaller flower than 0. biennis. In most cases the 

 leaves are also narrower than in 0. biennis, but as certain 

 races having smaller flowers than 0. biennis also have 

 broad leaves, it becomes an arbitrary distinction whether 

 these races be included in 0. biennis on the basis of their 

 broad leaves or in 0. muricata on the basis of their 

 smaller flowers. The latter course has been followed 

 and they are accordingly classed as 0. muricata races. 

 On this basis, the only distinction between 0. biennis 

 forms and O. muricata forms is in the size of the flow- 

 ers, and even this distinction is an arbitrary one. I will 

 not discuss in this connection certain interesting ques- 

 tions relating to the geographical distribution of these 

 races. 



My cultures have similarly isolated a number of races 

 of 0. biennis, some of which differ from each other in 

 most unexpected characters. There is a wide range of 

 variation in flower-size in the different races of 0. bi- 

 ennis, as there is also to a less extent in 0. muricata. 

 The general distinction is usually drawn, that self-polli- 



