196 "Journal of Comparative Neurology and Psychology. 



The division for head circumference was taken as nearly as 

 possible at the mean.^ From this table I find the following values 

 for the constants: 



h =.03620, kr:. 08456, 



whence .000635 6>--. 001409 03_ooi53i 6^+ '9-. 103222-0 

 Accordingly r = . 103^.034. 



Or, the coefficient of correlation is almost exactly three times its 

 probable error. The chance, then, of such a system of frequen- 

 cies as that of Table III arising if intelligence and head circum- 

 ference were not correlated in the population from which the sam- 

 ple is drawn is approximately i in 1000. We may reasonably 

 conclude then that there is a sensible correlation between these 

 characters. 



There is one further point needing discussion. It will be noted 

 that the coefficient of contingency deduced from Table II is some- 



'The constants for the head circumference distribution as given in Table II were found to be as 

 follows: 



Mean =56.499 ± .031 cm. 

 Standard Deviation = 1.413 zb -022 cm. 

 Coefficient of Variation = 2.501 ± .040 per cent. 

 These values are in close accord with those obtained from well-known homogeneous series, and point 

 clearly to the very substantial physical homogeneity of the present sample. Indeed this fact is so strik- 

 ing that I cannot refrain from reproducing for comparison the following tables of coefficients of varia- 

 tion for horizontal circumference from Macdonell's memoir on the skull (Biometrika, Vol. 3, p. 

 223). I insert in the table the coefficient deduced from the present paper. 



Actually the present data show the least variability of any of the male series in spite of the fact 

 that the measurements were made with a tape on the living subjects. 



