Raymond Pearl 
29 
question of the homogeneity of this material will be treated fully in connection 
with the discussion of the variability in brain-weight. 
The interesting questions brought up by this table of means, regarding sexual 
differences, association of brain-weight with other characters, etc., will be discussed 
in later sections of the paper. 
5. Va7-iation in Brain-weight. Homogeneity of Material. 
Variability in the weight of the brain has been subjected to exact investigation 
by only one worker, Pearson {loc. cit.), so far as is known to the writer. He deduced 
from his material (of p. 15 supra) coefficients of variation ranging in value from 
7'93% to 10"64<%, the higher values being from admittedly heterogeneous series. 
The numerical values for the variation constants and their probable errors found 
in the present work are given in Table I. 
The question of homogeneity of material should be first discussed. For reasons 
which have been set forth above it would be hopeless to look for any high degree 
of homogeneity in any collection of human brain-weighings at present available. 
The best we can hope for is a fair degree of homogeneity, and reasonably the same 
degree in different series which are to be compared. Unfortunately there is not 
available here, as in the case of craniometrical investigations, series in which fair 
homogeneity can be inferred with high probability, so that it is not possible to 
make a direct estimate by comparing variabilities with such a " known base." In- 
stead resort must be had to indirect methods. The best of such indirect methods 
is based on the fact that if a random sample be taken from a homogeneous collec- 
tion of material the vai'iation constants for the sample and the whole collection 
will not significantly differ. On the contrary if the material is non-homogeneous 
such a sampling will give different values for the constants. If the sample be 
selected, i.e. not random, the variation constants for the character selected will 
of course be lowered. Now in the material as treated here we have in the 
" young " series for each racial group a selection from the " total " series, but a 
selection based on age, not on brain-weight directly. If there were no correla- 
tion between brain-weight and age such a selection on an age basis would be, of 
course, a random sample so far as brain-weight is concerned. Unfortunately for 
the argument, there is, as will be shown later, a sensible though in general low 
correlation between age and brain-weight. This being the case it would be 
expected that such a selection as has been made in separating out the " young " 
from the " total " series would result in a lowering of the standard deviation and 
coefficient of variation in brain-weight. Now, as a matter of fact, as the following 
table shows, the correlation between brain-weight and age is so low that in the 
relatively small series under discussion, the "young" series forms practically a 
random sample of the " total " series, within the limits of error, with respect to 
brain-weight. 
