242 



MISSOURI BOTANICAL GARDEN 



of many melastomataceous genera in tropical America first 

 called my attention to the problem of the geographical 

 distribution of these floral types. 



It would appear that in general the most satisfactory 

 statistical comparison of the floras of two or more regions 

 is to be made not on the basis of genera but of species. 



While this proposition would seem to hold for general 

 comparisons of floras, we have in the present problem a 

 special case in which the morphological differentiation of 

 certain organs of the species entering into the composition 

 of the floras rather than the taxonomic differentiation of 

 the floras as a whole is to be considered. The comparison 

 in this case should be based upon 

 than taxonomic units, but it becomes at once clear that 

 it is quite impossible to separate the two conceptions. 



I think that on the whole the special data which we seek 

 are furnished by the characteristics of genera rather than by 

 the peculiarities of species. In the limitations of the former, 

 the floral organization has been assigned a more impor- 

 tant place than in the separation of the latter where the 

 vegetative features are usually more minutely studied 

 and described. But even if all species of a genus have an 



mor 



mus 



, not be forgotten that spe- 

 cific differentiation may permit of adaptation to more varied 

 environmental conditions and so increase the relative 

 importance of the genus as a constituent of the flora where 

 the area under consideration is so large as to present a 



multiplicity of conditions. 



In a comparison which embraces ecological considerations, 

 as the present one does, there is one strong objection to 

 taking the number of genera as data. 



In ecological corn- 



much 



In 



floral ecology where the anthophilous fauna is the environ- 

 ment, one species may have characteristics which enable it 

 to secure more visits than a dozen others. It is obviously 

 incorrect to present merely a numerical statement of the 



