1908] The Polymorphism of Ants 55 



as we may regard the macrergates, macrogynes, and macraners 

 as due to overfeeding. These are, of course, cases of nanism and 

 giantism, variations in stature, not in form. Simiharly, all cases 

 in which, as in certain species of Formica, Camponotus, Pheidole, 

 etc., the workers or desmergates vary in size, must be regarded 

 as the result of \'ariable quantitative feeding in the larval stage. 

 Here we are confronted with the same conditions as Weismann 

 observed in prematurely pupating blow-fiies and entomologists 

 have noticed in many other insects. Such variations are of the 

 fluctuating type and are therefore attributable to the direct effects 

 of the environment. The soldier and worker, however, dift'er 

 from the queen in the absence of certain characters, like the 

 wings, wing-muscles, spermatheca, some of the ovarian tubules, 

 etc., and the presence of other characters, like the peculiar shape 

 of the head and mandibles. In these respects the sterile castes 

 may be regarded as mutants, and Weismann's contention that 

 such characters cannot be produced by external conditions, such 

 as feeding, is in full accord with De Vries's hypothesis. His fur- 

 ther contention, however, that they must therefore be produced 

 by natural selection need not detain us, since it is daily becoming 

 more and more 'evident that this is not a creative but an elimina- 

 tive principle. It is certain that very plastic insects, like the ants, 

 have developed a type of ontogeny which enables them not only 

 to pupate at an extremely early period of lar^■al life, but also to 

 hatch and survive as useful though highly specialized members of 

 the colony. It is conceivable that this precocious pupation 

 may be directly responsible for the complete suppression of cer- 

 tain organs that require for their formation more substance than 

 the underfed lar\-a is able to accumulate. At the same time it 

 must be admitted that a direct causal connection between under- 

 feeding on the one hand and the ontogenetic loss or development 

 of characters on the other, has not been satisfactorily established. 

 The conditions in the termites which are often cited as furnishing 

 proof of this connection, are even more complicated and obscure 

 than those of the social Hymenoptera. While Grassi and San- 

 dias (1893) and Silvestri (lyoi) agree with Spencer in regarding 

 feeding as the direct cause of the production of the various castes, 

 Herbst (1901) who has reviewed the work of the former authors, 

 shows that their observations are by no means conclusive; and 

 Heath (1902) makes the following statement in regard to his ex- 

 periments on Californian termites: "For months I have fed a 



