of the Amazon Valley. 219 



out significance, therefore, that several families of Butterflies or 

 Rhopalocera have only four perfect legs, the first pair being more or 

 less aborted or rudimentary. This character is not inconstant or 

 intermittent ; it is absolutely universal throughout an immense 

 diversity of generic forms. The degree of abortion of the fore legs, 

 moreover, is different in the sexes, the male taking the lead in the 

 atrophy of the organs ; the nature of the sexual difference also being 

 constant and characteristic of large groups of genera constituting, as 

 I believe, natural families. Thus, in this division of the order we 

 have, first, two families in which the fore legs are perfect in both 

 sexes , then, two in which they are imperfect in the male, but per- 

 fect in the female — in the one being slightly, in the other greatly 

 aborted in the former sex ; and, lastly, one in which they are im- 

 perfect in both sexes. The Papilionidae and Pieridae of authors 

 possess six perfect legs ; they are, with the Hesperidae, the only 

 groups of the division which are in this condition, and they should, 

 I think, on this account occupy the lowest places in the series of 

 families ; whilst that group in which the atrophy of the fore legs is 

 most complete should be placed at the head, as being the farthest 

 removed from the Heterocera, and therefore the extreme develop- 

 ment of the Rhopalocerous type. 



A natural classification of the Rhopalocera, then, according to this 

 view, wovdd commence, in an ascending series, with the Hesperidae. 

 In this family all the legs are perfect, and the hind tibiae, with only 

 a few exceptions, have two pairs of spurs, as in nearly the whole of 

 the Heterocera. Next to them would follow the Papilionidae (and 

 Pieridae), which, although quite unconnected with the Hesperidae, no 

 connecting links between the two families being known, have like 

 them six perfect legs. Then would succeed the groups with imper- 

 fect fore legs. The characters thus derived from the structure of the 

 legs harmonize well with those furnished by the metamorphoses, 

 and partly with those derivable from the wing-neuration, as will be 

 seen in the following table, wherein I have attempted to establish 

 five families, subordinating to them as subfamilies the greater part 

 of those proposed by authors. 



Family 1. Hesperidce. Six perfect legs in J $; hind tibiae, with 

 few exceptions, having two pairs of spurs. Larva in- 

 habiting a rolled-up leaf; pupa secured by many threads, 

 or enclosed in a slight cocoon. 



Family 2. Papilionidce. Six perfect legs in d $ . Wing-cells (at 

 least of the hind wings) closed by perfect tubular ner- 



