134 Biometric Constants of English Brain-imights 
Now, although this table is very incomplete, and impresses upon one the need 
for the further collection of material, we can safely, we think, draw one or 
two conclusions : 
(i) The order of racial average brain-weights is very far from the order of 
average racial intelligence. Nor is the order bettered if we allow in any manner 
for stature. We think that even allowing for differences in racial stature, it is 
quite impossible to compare the brain-weights of men of different races, and any 
consideration of intelligence and brain-weight which proceeds by "pooling" the 
brain-weights of distinguished men of a great variety of races is fundamentally 
fallacious*. 
(ii) The variability, as measured by the coefficient of variation, of brain-weight 
is for the three races for which we possess fairly reliable data greater than that of 
skull capacity. Roughly, the average brain- weight variability is 9, the average 
skull capacity 8. On the whole the female shows very slightly greater variability 
in both cases. 
(iii) The ratio of brain-weight to skull capacity has an average value of about 
'9, and this is almost exactly tlie value of the ratio of the variation in brain-weight 
to that in skull capacity. It would appear that the cranial chamber is more closely 
packed with bi-ain in the case of the female than in that of the male, but the 
evidence, which is based upon heterogeneous material, is not convincing"]". 
(iv) The standard deviation of brain-weight in grammes is fairly close to the 
standard deviation of capacity in cubic centimetres. 
(v) Gladstone's results are, as far as the means are concerned, in close agreement 
with the results obtained by Pearson from mixing the data of a variety of observers. 
The increased variability, however, of the English Series I shows its real hetero- 
geneity, and not too much stress ought accordingly to be laid on the sensible 
identity of the mean values. The means of the component groups, indeed, vary 
widely. 
(G) On the Correlations of the various Characters measured. 
Table XIII gives the values of the correlation coefficients and their probable 
eiTors. They are tabled to four decimal places, not because they are worth more 
than one or two, but because these values must be used if the regression coefficients 
are to be tested. We propose to consider the separate entries in this table, and 
test them against such scant data as are available at present from other sources. 
As there are nine observed characters, each sex would provide 36 coefficients of 
correlation, but only 17 have been calculated for each case, partly because of the 
* Especially when the next stage is to compare this heterogeneous group with the fairly homo- 
geneous results for a local general hospital population ! 
t Data are being collected on more homogeneous material, and it is further hoped to test more 
completely than has hitherto been done the relative densities of different parts of the male and female 
brain. 
