200 Mr. W. Bateson. Heredity, Differentiation, 



distinction rests on fallacy 1 The reality of the problem as defined by 

 Professor Pearson depends on the assumption that there is an absolute 

 distinction between differentiation and variation among repeated parts, 

 and its solubility depends on the assumption that this distinction can 

 be perceived. The proviso that such a distinction is to be observed 

 stultifies the whole inquiry. So far are we from being able to perceive 

 this distinction, that we may even doubt whether variation among 

 repeated parts can take place except as differentiation. If the idea of 

 variation is to be extended to the case of differences between repeated 

 parts it must inevitably include differentiations among them. 



But, unless differentiation can always be detected or always reckoned 

 for, the average value of the homotyposis coefficient will have no more 

 natural significance than would the average variability of all organisms 

 measured by their " Standard Deviation " from their various means, or 

 the average size of living cells, or the average weight of all ponderable 

 bodies. 



I now proceed to a different point, in a sense the converse of the 

 former. Professor Pearson perceives that the correlation between 

 " undifferentiated like parts " has an analogy with the resemblance or 

 correlation between brothers. But does he recognise that variation 

 between brothers is comparable not merely with variation between 

 repeated parts, but also with differentiation, and with predominantly 

 orderly variation among such parts 1 The phenomena in a colonial or 

 social form will clearly illustrate this principle. Ova and spermatozoa 

 may be treated as "undifferentiated like parts " so long as their varia- 

 tions, judged by the resulting offspring, are sensibly irregular. Can we 

 recognise differentiation among them as distinct from variation ? 

 Certainly we sometimes can. In determining the correlation of con- 

 fraternities, the parentage enables us to distinguish the fraternal groups 

 correctly, and consequently a fraternal correlation may be truly deter- 

 mined. For to do so we are not compelled to distinguish differentiation from 

 variation. 



But I put it that the parallel Professor Pearson is seeking is im- 

 properly instituted in his paper. He compares the homotyposis of 

 " undift'erentiated like parts " with the correlation between brothers. 

 He ought to compare it with the correlation between undifferentiated like 

 brothers. As it is, he is trying to find for homotypes what he would be 

 trying to find if he set about a determination of the average value of 

 fraternal correlation for fraternal groups of like members taken from 

 families composed of differentiated members. Such an attempt would 

 immediately necessitate a distinction between differentiation and varia- 

 tion. Had his comparison been correctly instituted, Professor Pearson 

 could hardly have failed to discover the fallacy on which I submit his 

 reasoning is based. 



Let me state a case in illustration. In most species of Ants females 



