4 1 8 Studies in the Theory of Descent. 



consequently no longer variable. The transferred 

 characters may thus have acquired a greater regu- 

 larity, i. e. a less degree of variability, than they 

 possessed at their first origination. Extensive 

 investigations in this special direction must be 

 made if the precise laws, in accordance with 

 which the backward transference of new characters 

 takes place, are to be discovered. By such re- 

 searches only should we arrive with certainty at 

 the causes which determine the lesser variability of 

 the young larval stages. 



It may also occur that the early stages are 

 variable, whilst the later stages are constant, 

 although this case appears to happen less fre- 

 quently. Thus, the caterpillars of Gastropacha 

 Quercifolia vary considerably in the second stage 

 but are constant at a later period, and the same is 

 the case with Spilosoma Urtica, which in the 

 second stage may be almost considered to be di- 

 morphic, but which subsequently becomes constant. 



Cases in which the first stage is variable appear 

 to be of the least frequent occurrence. I know of 

 only one such instance, viz., Anceryx Pinastri, of 

 which the newly hatched larvae (PI. VI., Fig. 53) 

 show considerable differences in the brownish-black 

 crescentic spots. The second (Fig. 54), third, 

 and fourth stages are then tolerably constant, while 

 the fifth stage again is very variable. 



An instance of this kind can be easily explained 

 by two waves of variation, the first of which now 



