59 



opment. The species are all leaf-miners and are very small, the pupae 

 of the largest species examined being 7 mm. in length. Very few of 

 the generalized families have been available for study, so that it is 

 exceedingly difficult to trace the relationships existing between the 

 more specialized families without first having carefully studied a num- 

 ber of more generalized forms. There is included in this group the 

 Nepticulidae, in many respects the most generalized pupae studied, 

 next to the Eriocraniidae, and certainly resembling the latter more 

 than any of the other generalized forms examined. It is just at this 

 point in our investigation that more material is needed to clear up the 

 relationships of the groups which have apparently branched off here 

 and have had a common ancestor with the Nepticulidae. From all the 

 evidence at hand it seems probable that development has proceeded 

 along two well-defined lines, the first, represented by the superfamily 

 Gracilarioidea, having early lost the maxillary palpi while still retain- 

 ing the covering of spines on the dorsum of the abdominal segments, 

 and having developed the triangular type of prothorax ; the second 

 having retained the maxillary palpi for a much longer time, but having 

 lost the covering of spines, while 'developing the same type of pro- 

 thorax. 



Of the second type no material has been examined which would 

 show any intermediate stages between the families Nepticulidae and 

 Epermeniidae. The latter family has apparently continued the line 

 of development begun in the Gracilarioidea as it still retains the 

 seventh abdominal segment free in the male though it is fixed in the 

 female. The presence of the maxillary palpi precludes its derivation 

 from the Gracilarioidea and would lead us back to some point below 

 the Heliozelidae because this family also has lost them. As we have 

 only the Nepticulidae for comparison, it has been assumed that this 

 branch has arisen coordinately with them. 



In the superfamily Gracilarioidea, with the exception of the family 

 Lyonetiidae, all the pupae have free appendages, the cuticle is very 

 slightly chitinized, and the dorsum of the abdomen is covered, in part 

 at least, with fine spines. There is a tendency in some genera, as 

 Lithocolletis and Ornix, towards the development of a single row of 

 spines, so that there is often one or more rows of larger spines at 

 either the cephalic or caudal margin, or at both margins, of the seg- 

 ment. This seems to indicate the way in which the rows of spines were 

 developed in the Tineoidea and Tortricoidea. The characters which 

 are common to all the members of the superfamily are the long vertex, 

 which is always longer than the prothorax at the median line, scarcely 

 ever less than twice its length and often much longer, and the long 



