J. F. Tocher 
311 
Diagram VII. Stature. — 3915 Females. 
1360 
1280 
<200 
1120 
1040 
960 
880 
800 
720 
G40 
560 
480 
400 
1 
j \ 
f- 
\ 
I 
\ 
—1 
li 
li 
li 
— u — 
\ 
// 
f- 
;/ 
— ^ 
'/ 
\ 
j — 
— /-I 
i 
* 
1 — 
;/ 
— 
-\ — 
I 
'J 
+ 
i 
1 
— t 
1 
\ — 
A 
w 
i 
" 
Skew Curve 
Normal Curve 
removed for the reasons given above, it cannot be said whether others should 
or should not be removed by a similar process of reasoning. Heterogeneity, 
however, may exist. If, for instance, local groups at each individual asylum are 
taken and their degrees of goodness of fit to the normal curve tested, it is 
found that, with a 5 mm. grouping, this curve gives very good fits in 125 cases 
out of 176 (see Table IV.). This, indeed, shows no more than that, for small 
samples of the iusane, the noruial curve describes the distribution within the 
probable errors of the constants, exactly as Favvcett, Macdonell, and others have 
shown for small cranial series. The divergence from normality in both the " entire 
insane " and the " general insane " populations of Scotland is therefore either 
(a) real because the greater numbers allow of more accurate determinations of the 
kurtosis and asymmetry constants, or (h) spurious and due to the introduction of 
local heterogeneity. The evidence for and against heterogeneity will presently be 
considered. 
(ii) Correlations. The correlations and the nature of the regression curves 
for two characters will now be briefly considered. The first main point to be 
noted is that the values of the coefficients for the " entire insane " population are 
uniformly greater than those of the " general insane." This is chiefly due to the 
