446 WILLIS : MORPHOLOGY OF THE PODOSTEMACBÆ 
thalli, and so on. Most of the present genera and tribes 
would be united with others, or perhaps the tribes might 
become genera, the genera species, though in nearly all 
their essential features, i.e., in those related to the environ- 
ment and to the mode of life, the plants would be as different 
as ever. This insecurity must remain until the study of 
experimental morphology with variation and correlation 
shall provide us with a more reliable quantitative test of 
generic and other differences. On the whole, perhaps, our 
specific and generic grouping will be found ultimately 
supported by the facts of evolution so far as they may be 
discovered, but it is well at times to remember the insecurity 
of our position. It is partly in view of this insecurity of 
the larger grouping, that though in the classification of the 
Indian Podostemaceæ in the preceding paper I have used 
the large Linnean species, I have used proportionately small 
genera, in which vegetative characters have been possible of 
use in addition to the floral. 
The next question that may be raised is whether many of 
our groups, such as the Linnean species, and still more 
the genera, tribes, sub-orders, orders, and larger groups are 
or are not polyphyletic. The question is already being 
raised by the fact that so many of the largest groups, e.g.^ the 
Fungi, the Gymnosperms, the Sympetalæ are proving to be 
polyphyletic, so far as the evidence of palæobotany and 
morphology enables us to judge, but the considerations 
above given raise it from a different standpoint, and enable 
us to see one way in which the phenomenon may have 
arisen. We can easily imagine a case like the following. 
Let A be a species of Podostemaceæ, Avith a marked but not 
extreme dorsiventrality of the vegetative organs, and let it 
give rise by gradual evolution to three species, K, L, M, 
of which K and M are more, L less, dorsiventral than itself in 
the vegetative organs, while in all other respects K is more 
nearly allied to L than to M. Now let these go on to evolve 
further species, X, Y, Z, respectively, the dorsiventrality 
still increasing in the case of the first and last, but X and Z 
