48 PROCEEDINGS : BOSTON SOCIETY NATURAL HISTORY. 
features of the lower part of the face; on the other hand, the nose 
root breadth, though by no means inconspicuous, is, I think, 
obscured by the more prominent features near it. At any rate, 
these are the only two face dimensions remaining, which in any¬ 
wise lend themselves to this classification. Columns 13 and 14, 
incomplete through lack of data, show in turn complete correlation. 
There is, however, one head dimension which is eminently con¬ 
spicuous, the forehead height. This, in column 16, I have included 
with the dimensions of column 14; and in column 15, with the 
dimensions of column 13, I have included the breadth between the 
ears. I have chosen this last dimension because it is apparently 
the most variable of the remaining head dimensions, and, in conse¬ 
quence, least favorable to my case. Here, too, in comparing four 
conspicuous dimensions of the face and one of the head with four 
inconspicuous dimensions of the face and one of the head, I find 
in every instance that the more conspicuous dimensions are the 
more variable. 
It seems, at first sight, an easy matter to apply this last test to 
the data of Tables A, B, and D. But in the case of Table A the 
number of measurements is too few, and in Tables B and I) I have 
found it impossible to assign the data to the proper groups. 
Finally, in column 18, are given the mean coefficients of eight 
body measurements; these should evidently be less than the coeffi¬ 
cients of the more conspicuous face dimensions. So indeed they 
are, except in the case of the Chinese; and even here, as in every 
instance, the face including the nose is more variable than the body. 
Table D is like A and B, and Table 4 like 1 and 2, and need no 
comment; they give twelve good cases out of fourteen. 
I have now made 142 comparisons between the variability of 
various dimensions classified according to their importance for per¬ 
sonal appearance; and in 120 cases — more than 84% —the more 
conspicuous dimensions are the more variable. This per cent would 
doubtless be greater, if the coefficients could, in some cases, have 
been based on larger numbers of individuals. For in Tables A, B, 
C, and D, if I throw out five sets of coefficients based on fewer than 
twenty individuals, I get 89% of favorable cases. Furthermore, in 
all these comparisons, the number of dimensions is quite as impor¬ 
tant as the number of individuals; and it is noteworthv that Table 
B, with 26 dimensions and 129 individuals, gives 100% of good 
cases. Finally, these comparisons are based on 335 coefficients 
