and the Mcndclian hypothesis. 529 



dimorphism which appeals to the human eye — e. g. of 

 coloration, as in our Xanthorhoe, is necessarily correlated 

 to the true gametic differentiation. This is, at least, a 

 possible interpretation of Doncaster's words {loc. cit.) that 

 in some cases, such as that of the melanism of Aplecta 

 nebulosa, " the inheritance is not Mendelian." 



The only in-bred broods of which I reared any imagines 

 deserve mention here, in spite of their small numbers. 

 (1) From a wild purple $, which produced 21 black, 

 15 purple, and one intermediate in her progeny, a pairing 

 of two of the purple examples was obtained. From this 

 pairing, 10 purple and 2 black were reared, none of which 

 were successfully paired together. (2) From another wild 

 purple $, which produced 9 black and 4 purple, a pairing 

 of two of the black was obtained. From this there resulted 

 only 4 imagines, all of which were black. Again the 

 strain failed at this point. (3) From a third wild purple ^, 

 which produced 5 black and 8 purple, a pairing of purple 

 ^ with black $ was obtained. This pairing resulted in 

 the brood of 7 black and 10 purple, which has already 

 been alluded to as "black x purple, No. 5." Yet again 

 the strain failed at this point. 



Neither of these results seems to suggest with any 

 clearness that either form was a "recessive"; one looks in 

 vain for any approximate 3 : 1 ratio, such as might have 

 been expected, by the ordinary laws of chance, even when 

 allowance had been made for the large percentage of loss 

 in rearing. Rather do they suggest some kind of biometric 

 interpretation, and with the potency of each colour ap- 

 proximately equal (or black somewhat the stronger) and 

 equally direct in its effect. In the case numbered (1), a 

 brood in which yV were purple, gave, in the next genera- 

 tion and by the aid of purple selection, ^f purple, or just 

 double the percentage. In that numbered (2), a brood in 

 which -^-g (69 per cent.) were black, gave one, by black 

 selection, in which one hundred per cent, were black. In 

 that numbered (3), where there was a cross-pairing of the 

 two colours, the percentages in the two generations were 

 very little disturbed, working out (roughly) at 38'o per 

 cent, and 41*2 per cent, of black specimens respectively. 



It remains to consider whether the larger experiments 

 detailed earlier in this paper throw any further light on 

 the statistical aspects of the question. It is manifest from 

 the figures I have given, that there has been no behaviour 



