( 28 ) 



pattern and outline, shown in a reduction of the undulations- 

 of the outline, the comparative obscurity of the marginal and 

 submarginal pattern, and in the loss or faint development of 

 minute details in the pattern of both surfaces, especially the 

 under. This last effect is far more evident in frohenia. In 

 one feature, however, the upper sui-face of comorarum departs 

 fiu-ther from mayottensis than frohenia from dumetorum, viz, 

 in the loss of the mark on the inner margin which, in other 

 species of the saclava group, prolongs the band of the hind- 

 on to the fore-wing. This is an isolated feature ; for in the 

 retention of minute white points in the cell of the fore-wing 

 upper surface and in the whole pattern of the under surface, 

 comoraruvi is immensely nearer the saclava group than 

 frohenia or even dumetorum. 



The pronounced simplification which has occurred in the 

 two most isolated species has been accompanied by a remai-k- 

 able persistence in the larger and more conspicuous elements 

 of the pattern, so that during flight or at rest with wings 

 expanded, yroftema, at a little distance, would closely resemble 

 dumetorum, and comorarum, under similar circumstances, 

 viayottensis. 



Such are the changes which have taken place in the two 

 most isolated species of this group of Kejytis, and it is im- 

 portant to separate sharply the remarkable and interesting 



[xlii 

 facts from any attempt — in this case peculiai"ly difficult — to 

 explain the causes. 



The minutifE of an aposematic pattern such as that of 

 JVeptis may well be kept up by selection on an area where 

 two or more species exist together, and where e.g. an elabora- 

 tion of the marginal markings is common to all. Under these 

 circumstances, too, elaboration would probably be gained by a 

 species with simpler pattern brought by migration into the 

 area in question. When, however, an aposematic species 

 with elaborate pattern becomes isolated, we should expect, on 

 the very same principles that are believed to account for the 

 growth of Miillerian resemblances generally, that enemies 

 would continue to test with especial severity large departures 

 from the average specific pattern. But in this case the 



