CH. II] THE PODOSTEMACEAE 11 



pound, from pinnate to palmate, from dorsiventral (facing 

 upwards, with different anatomy on the two sides) to isobilateral 

 (facing sideways, hke Gladiolus leaves, with the same anatomy 

 on both sides), from parallel veined to net veined, or in other ways 

 that could be mentioned. Now variations in length and breadth 

 are rarely of much importance for distinction of species, unless so 

 great that there is a wide difference between the averages in the 

 two cases, while the other characters that have just been men- 

 tioned will be seen at once to be such as are of great importance 

 in distinction between one species and another. This is another 

 fatal objection to the use of this kind of variation as part of the 

 mechanism of evolution. Some kind of variation was required 

 that was not only inherited and irreversible, but also differen- 

 tiating and not merely linear, or up-and-down. 



Another serious difficulty was the fact that species were very 

 rarely distinguished from one another by a single character only. 

 Usually there were from two to six characters marking them off 

 from one another, some of them more variable than the rest, and 

 more liable to overlap from one species to the other, so that one 

 had to examine a great number of specimens of each of the 

 species to be sure that their overlap was not due simply to lack 

 of real difference. Thus in Cornus, to take the first genus that 

 comes to hand, C. kousa and C. capitata are closely allied species. 

 The key division is that the former has the calyx truncate, the 

 latter 4-lobed, and the involucral bracts more or less ovate as 

 against obovate. But there are so many minor points of difference 

 that the description of either takes up nearly twenty lines (49). 

 None of the characters afford any opening for natural selection 

 to work upon, so far as can be seen, but supposing that it had 

 worked upon one, were all the rest simply correlations? One could 

 hardly imagine it working upon one at a time, for what would 

 ensure that a should be followed by b, which was unconnected 

 with it (as are the two distinguishing characters above quoted)? 

 Nor could one imagine it picking out a variation that included a 

 little of each of a, b, c, d, etc., when these were unconnected, 

 unless they were in some way correlated.^ But if correlation were 

 to be invoked to this extent, it must be the principal, though 

 perhaps only passive, factor in evolution as sho^\Ti by the 

 characters that distinguish its finished product. Nearly all the 



1 Cf. Origin of Species, chap, vii, first few pages, for remarks upon this 

 subject. Incidentally Darwin there suggests the "somewhere" which has 

 proved such a useful refuge to the defender of natural selection. 



