CHAP. XV.] 



tarsus of Blatta. 



419 



Five-jointed tarsus. 



Four-jointed tarsus. 



It is thus seen that the percentage variation of the ratios of the 

 several joints to the total length is very little greater in the case of the 

 abnormal than it is in the normal tarsus. 



As regards the longer joints these results are probably a trustworthy 

 indication of the amount of Variation, but in the case of the shorter 

 joints the errors of observation must no doubt be so great in proportion 

 to the smallness of the lengths to be measured that no reliance should 

 be placed on results obtained from them. 



As evidence that in spite of the small number of instances examined 

 the general result is satisfactory it may be mentioned that the mean 



Q 3 + Qi 



obtained as the value of — ^r— agrees fairly well in each case both 



-j 



with the value of M, the middlemost value, and also with the arith- 

 metic mean given above. It may therefore be taken that the curve is 

 regular and the series nearly uniform. 



The correlations between the lengths of the joints and that of the 

 whole tarsus have also been examined by Mr Brindley using the 

 method proposed by Galton I.e., the results closely agreeing with those 

 obtained by the ordinary method here described 1 . 



If the 4-jointed tarsus be a congenital variation the sig- 

 nificance of the fact that the abnormality is in its constancy to 

 its normal hardly less true than the type-form must be apparent 



1 It is hoped that a fuller account of this subject will be given separately. I am 

 indebted to Mr F. Galton for advice kindly given when this investigation was 

 begun, and Mr Alfred Harker has most obligingly given much help in connexion 

 with it. 



27—2 



