104 Mr. G. A. K. Marshall on Diaiwsematism, with reference 



the mind of the enemy unpleasant experiences in con- 

 nection with B. On these grounds it is argued that both 

 varieties, A' and B', are advantageous, and that therefore 

 the two species can mimetically approach one another at 

 the same time. But a little consideration will soon show 

 that the above argument does not deal with all the factors 

 in the case. In the first place, the enemies are divided 

 into only two categories (X and Z) ; but there must 

 obviously be a third (Y) which will derive its knowledge 

 through a mixed experience of A and B together, and the 

 effect of this has been quite left out of consideration. 

 Now it is evident that reciprocal mimicry could only take 

 place where A and B are approximately equal. When 

 this is the case, the law of probabilities shows that the 

 numbers of Y will be very large, those of X and Z very 

 small; therefore the net result of the Miillerian factor will 

 depend upon the effect produced by Y, and this is entirely 

 covered by the general argument set forth above, which 

 thus remains unaffected. 



There is yet another more important objection to this 

 criticism. It is contended that X will discriminate 

 between B and B', and Z between A and A', to the 

 advantage of B' and A' respectively; but the relation of 

 X to A and A', and Z to B and B' is not taken into 

 account at all. But if A' is sufficiently different from A 

 that Z will discriminate between them, it must be admitted 

 that X will do so likewise. Such discrimination means that 

 A' will be subjected to special tasting experiments by X, 

 as apart from A. But ex hypotJiesi the numbers of A' will 

 be very much smaller than those of A, and therefore these 

 experiments will involve a much higher percentage of loss 

 for A' than for A, so that the former will be at a decided 

 disadvantage in relation to the attacks of X. The same 

 applies to B' with regard to the attacks of Z. Thus A' 

 will have an advantage over A during the attacks of Z, 

 and a disadvantage during those of X, and the net result 

 will depend upon the relative numbers of X and Z. But 

 the relative numbers of X and Z are directly dependent 

 on the relative numbers'of A and B respectively ; for where 

 A is abundant the members of X will be large and vice 

 versa. Therefore when A and B are equal, X and Z will 

 be equal, and the advantage which A' derives from Z will 

 be balanced by the disadvantage due to X ; and similarly 

 for B'. In these circumstances there will be a condition of 



