76 DIVERGENCE OF VARIATION [ch. ix 



these families, one finds the same kind of divergences, more and 

 more marked on the whole in approaching the top of the list (the 

 first divisions in the keys), and least marked in the characters 

 that distinguish one species from another. This fact of increasing 

 divergence as one gets nearer to the top of the list has always 

 been a great difficulty in the path of the supporters of natural 

 selection, and has been left discreetly unmentioned by them. 



Opening a volume of Engler, the family displayed is the Cyclan- 

 thaceae, composed of six genera only. The first and most obvious 

 division, into Carludoviceae with male flowers in fours, and 

 Cyclantheae with male and female flowers in alternating rings or 

 spirals, picks out the two most important genera, one in each of 

 the groups, though the first group, with five genera and forty-five 

 species, is much larger than the second, with one and four. 

 IncidentaUy, how did selection, or gradual adaptation, produce 

 these two very distinct types of inflorescence, and what was 

 intermediate between them? Taking first the Carludoviceae, the 

 genus Ludovia, with two species in Guiana and Amazonas, is 

 first cut off*, having only a rudimentary perianth in the male 

 flower (why, on the theory of selection, did it spoil its attractive- 

 ness to insects?). The four genera left are divided into Carlu- 

 dovica, which has forty species covering the whole range of the 

 family in tropical America (north and south) and the West 

 Indies, and which has a short perianth in the female flower, 

 against a long one in the other three (again, attractiveness 

 apparently spoiled), and an inferior ovary against a superior. 

 Carludovica is by far the largest genus in the family, far out- 

 numbering all the rest put together, and has a distribution 

 covering that of the whole family, just as we have seen to be a 

 general rule (p. 64). Does it owe its "success" to its inferior 

 ovary, and if so, wherein does the advantage lie, for the flowers 

 are so crowded that one cannot tell from the outside that the 

 ovary is inferior? And if there is an advantage there, what about 

 the reduced perianth? 



Evodianthus, next, is distinguished from the two other genera 

 by having the stamens inserted in the tube of the perianth, while 

 the others have them on the disk ; and finally Stelestylis has the 

 stalk of the male perianth flat and hollow, and a pyramidal style, 

 while Sarcinanthus has the male perianth forming six-sided 

 pyramids, and no style. All three are small and little dispersed 

 genera with two species in Costa Rica and the West Indies, one 

 in eastern Brazil, and one in Costa Rica, respectively. 



