426 Henry Edward Crampton. 



the following spring, when many individuals proved to be unable 

 to form perfect moths, and thus offered a comparison with the 

 pupiE that metamorphosed successfully and perfectly. The 

 material possessed a peculiar interest for the reason that the pupa 

 does not "use" many of its structures at all during its long period 

 of quiescence but remains practically inactive until the time for 

 emergence approaches. To state the results of this investigation 

 in briefest form, it was found in many instances that the pupae 

 which died before metamorphosis were of somewhat different 

 types and were more variable than the surviving individuals; and 

 the same relation between elimination at the time of metamor- 

 phosis and structural variation appeared on the basis of a statis- 

 tical examination of the two groups of pupae distinguished at that 

 time.^ 



Since 1900, the same species has been much more extensively 

 examined to see if similar phenomena were exhibited in other 

 years, and other forms of related moths have also been investi- 

 gated — /. €., cecropia, promethea, ruber, joruUa, etc. The results 

 have been confirmatory, m that in every series where successful 

 pupae have been compared with unsuccessful pupae some of the 

 characters exhibit the stated relation between variation and 

 elimination. 



Nevertheless, the conclusion that this relation was a general 

 and a final one did not appear to me to be justified, for there 

 were many cases where the reverse was true, where, that is, the 

 survivors were of the same type as the others but were more 

 variable, or they were of the same type and variability. In addi- 

 tion, the characters that exhibited ''selection," where it was indi- 

 cated, were of such a nature that they could hardly have served 

 the pupa either advantageously or disadvantageously. The pupa 

 does not "use" its antenna, and yet the pupae that succeeded in 

 producing perfect moths often possessed antennae that were cer- 

 tainly selectively different from those of the contrasted unsuc- 

 cessful group. It was inferred, therefore, that selection was 

 "indirect," and that the real basis for selection was not to be 

 sought in the series of individual characters themselves but in 

 the condition of correlation between and among these char- 

 acters. The conclusion of the preliminary study reads as follows: 



'See Biometrika, vol. iii. 



