80 
Variation and Correlation in Brain-Weight 
between 0 and + 1. There is no evidence from this source as to whether the 
correlation is positive or negative. Suppose we take still another case, stating the 
problem this time in this way : Assuming that the Cambridge undergraduates 
have brain-weight correlated with skull length to roughly the same degree as is 
shown by Bohemian males, what will be the limits of the correlation between 
braiti-weight and intelligence? The necessary coefficients are: 
= '0861 correlation coefficient between skull length and intelligence. 
?'i3 = "5482 „ „ „ „ and brain-weight. 
The limits of ?'23 or the correlation between brain- weight and intelligence, are -f '7 
and — '6, or practically we reach much the same result as before. Now the 
Bohemian coefficient, as we have seen, probably is in error in the direction of 
being too small, or the probability is that the limits of n-i are still wider than the 
figures indicate. As to the reliability of these limits this may be said : in order to 
fix correctly the sign of the correlation between braiu-weiglit and intelligence, both 
the coefficients between intelligence and physical characters and brain-weight and 
the same physical characters would have to be greater than V'S or '707. Now 
I think it is reasonably evident from what we already know, that for civilized man 
at least there is no likelihood whatever of ever finding values as high as "707 for 
the coefficients concerned. Or in other words the degree of the correlation 
between brain-weight and intelligence is indeterminate, with the probability that 
it is sensibly equal to zero. That is, brain-weight and intelligence in the sense of 
mental capacity are probably not sensibly correlated. All the inferential evidence 
when subjected to careful scrutiny leads to the same conclusion, I think. 
In closing I may perhaps be permitted to point out what seem to me to be the 
most promising lines for future investigation of the problems regarding the weight 
of the brain. In the first instance it seems to me highly desirable to transfer the 
problem for a time at least from man, where in the nature of the case all the 
fundamental records must be vitiated from the fact that we cannot weigh the 
brains of any considerable number of normal men instantly killed. As a conse- 
quence of this impairment of the data resulting from complex ante-mortem con- 
ditions the absolute values of certain of the biometric constants concerned cannot 
in the case of man be proven to be sensibly identical with what we should consider 
necessarily normal values. To be sure, such a sensible identity cannot, on the 
other hand, be disproven, but a state in which one can neither prove nor disprove 
is scientifically not altogether satisfactory. By working with animals it is possible 
to weigh any number of brains of instantly killed normal individuals. From 
records so obtained the values of the important biometric constants like the 
coefficients of correlation can be satisfactorily determined. In the hands of 
Prof Donaldson this line of work is yielding results of the greatest significance 
and value. For continuing the work on the human side the primary need is for 
more and larger collections of brain-weights in which close attention is paid to the 
racial homogeneity of the material and to the recording of other subsidiary facts 
