CH. XII] C. TAXONOMIC 139 



lepidaceae, and Restionaceae. Each family makes its first divi- 

 sion (in some classifications) into Haplanthereae and Diplan- 

 thereae, or groups with monothecous and with dithecous anthers, 

 a well divergent and clearly marked division. In the Erio- 

 caulaceae the Diplanthereae contain half the genera of the 

 family, including Eriocaulon and PaejJalanthns, which are by far 

 the largest genera, while the Haplanthereae include only very 

 small genera, whose species make only about 2 J per cent of those 

 in the other group. And whilst the Diplanthereae cover the 

 warmer parts of the world, the Haplanthereae are found only in 

 warm America. In the Centrolepidaceae and Restionaceae, on 

 the other hand, the larger group is the Haplanthereae. In the 

 former, they include five genera and thirty-five species, against 

 one and two in the Diplanthereae ; and in each case the distri- 

 bution is much more extensive. 



There is no conceivable reason why dithecous anthers should 

 suit America better, and monothecous the Old World, and yet 

 the former are more common in the one, the latter in the other. 

 It is clear that we must be dealing here with a divergent muta- 

 tion, and that one family began with dithecous anthers, the other 

 two with monothecous, and that probably each one subsequently 

 split off the other division. The Eriocaulaceae, for example, 

 beginning dithecous, spread over the world, but split ofP the 

 monothecous group in America. Perhaps the splitting off was too 

 late for the plants to cross to the Old World in any case, or it may 

 have been that as we have elsewhere explained the early growth 

 and dispersal of the new forms was too slow for them to be in time 

 to cross. 



Cases of the same kind, showing exact parallelism, are very 

 numerous indeed. To take a few examples, the Marantaceae 

 divide into a group with 3-locular ovary, and a group with 

 1-locular, and each of these divides into a group with two lateral 

 staminodes, and a group with one. In Amaryllidaceae both the 

 groups Amaryllideae and Narcisseae divide into groups with 

 many ovules and with few, whilst this is the first division in the 

 related Haemodoraceae. In Araceae, most of the principal groups 

 divide into groups with endosperm and without. In the Palma- 

 ceae, several widely separated groups have fan leaves, others 

 feathers. And so on, in hundreds of cases. 



This phenomenon has always been a great difficulty to explain 

 upon the theory of selection, for it makes it obvious that none 

 of these characters — for example, those of climbing plants, else- 



