72 WILLIS : GEOGRAPHICAL DISTRIBUTION 



connections above surmised will account for the presence of the 

 species in Malaya, and interest centres in the Ceylon forms. 

 One of them, A. unifloru.m, is fairly widespread, while the other 

 six are all rare, though they all occur near to localities of A. 

 miiflorum, which is "extremely variable, almost every district 

 producing a local form " (Trimen). With regard to one of 

 them, A. dissectum, Trimen remarks : " The plant at Hiniduma 

 was growing along with A. 'miiflorum, var. sylvaticum, and 

 many of the specimens are believed by Thwaites to be hybrids 

 between the two species, the various forms of the leaves making 

 a complete connecting chain." No one who has examined the 

 flowers of this species, and has had experience of the really 

 extraordinary lack of insect visitors to small plants in the 

 tropics, will suppose that hybridization has really occurred 

 here ; it is far more likely that various mutations are still living 

 side by side , illustrating the transition from one species to the 

 other. I am at present conducting some experiments with a 

 form of A. uniflorum, endeavouring to persuade it to mutate. 



Another species of interest is A. lyratum, which is charac- 

 terized by very long peduncles. This occurs solely on the 

 summit of Nillowekanda , an isolated precipitous rock in the 

 Hinidum pattu , and , as I have pointed out in my previous paper, 

 it is impossible to suppose that the long peduncles can have 

 been evolved on Nillowekanda by infinitesimal variations. 



The easiest way by far to explain the distribution of Acro- 

 trema is to suppose that A. uniflorum, which we know to be 

 very variable, travelled about the south-west of the Island, and 

 sJied the other species by mutation in different places, while it, 

 or one of its mutants, got as far as Travancore, where it gave 

 rise to A. Arnottianum. 



The next genus, Schumacheria, diiiers from Acrotrema 

 mainly in being shrubby, and is confined to Ceylon, where it 

 may be supposed to have arisen from Acrotrema by mutation. 



The fifth genus is Wormia, of which there are seven species. 

 W. triquetra is confined to Ceylon, while there are six species 

 further east, and again the distribution is most easily explained 

 by imagining a land connection to Malaya. W. triquetra 

 belongs to the subgenus Euwormia, to which also belong W. 

 pulchella in Perak, Malacca, and Sumatra, W. mellosmcejolia 



