Pearl, Correlation in Brain-Weight. 471 



ians for the females, the differences between the coefficients of 

 variabiHty with their probable errors are respectively .526± 

 .217 and .958 ± .337. In these cases the difference is less than 

 three times its probable error and cannot be considered as cer- 

 tainly significant. The lists show merely the general trend of 

 the results. It is possible that with much larger series of brain- 

 weight statistics and consequently reduced probable errors some 

 such relation as that just stated might be definitely proven. 



An analytical study of the frequency distributions leads to 

 the conclusion that variation in the weight of the brain may, 

 iox practical ^nx^os&'s,, be considered to follow the "normal" law 

 of the distribution of errors. Some of the series deviate from 

 strict normality, but for the present we cannot do better than 

 use the normal curve. If the mean and the mode do not ex- 

 actly coincide the mean will be slightly greater than the mode 

 in brain-weight frequency distributions. The various analytic- 

 al constants for the different series are in close agreement. This 

 fact is strong evidence for the general trustworthiness of the 

 data. 



4. The Correlation of Brain-weight with other Characters. 

 Brain-weight statistics usually return data for but {q.sn other 

 characters ; in the majority of cases only sex, age and stature. 

 In two of the series used in this work, viz., the Hessian and 

 the Swedish, this was the case. For these two groups all pos- 

 sible correlations for both sexes and two age groupings ("young" 

 and "total") were determined. This gave as the pairs of char- 

 acters (a) brain-weight and stiture, (b) brain weight and age, 

 (c) stature and age. Bischoff's Bavarian material furnished, 

 in addition to the above, data on body-weight, so that it was 

 possible to correlate this character with brain-weight and with 

 stature. In the case of the Bohemian material data were fur- 

 nished from which the correlation between brain-weight and 

 skull length and skull breadth could be determined, in addition 

 to the characters age and stature. The degree of correlation 

 was measured by the well known coefficient, 



S {xy) 

 r = , 



