300 
AntJiropometri/ of Scottish Insane 
flaxen, and golden yellow; medium included chestnut and all shades of brown 
except dark brown and black. The eye categories were light, medium and 
dark. Light included light grey, blue or bluish grey. Dark embraced simply 
hazel brown and dark brown, while medium covered a mixed class (including 
grey) which were neither light tior dark. Hair if turned grey was not recorded. 
The nose shapes recognised and recorded were straight (8), Roman (R), Jewish (J), 
concave (G). and wavy (W). 
It seems desirable at the outset to state the problems which, from the nature 
of the data, it appears necessary to deal with. 
(a) The fundamental problem clearly is : Does the insane population differ 
from the sane population ? and this necessitates a comparison between sane Scots 
and insane Scots. No general comparison can, however, be made between these 
two classes since samples of the normal population in the various districts from 
which the insane population is drawn have not yet been measured. Only two 
or three short series ai-e available for comparison. These will be dealt with 
under the districts to which they belong. Only pauper lunatics having been 
measured the population of each asylum is a local sample of the district served 
by that asylum. 
(b) Do the data differ in the form of their distribution from data already 
collected from other, presumably sane populations ? 
(c) Do different parts of Scotland differ sensibly from each other, assuming 
the insane population to be an anthropometric sample of each local population ? 
(d) Is there any reason for supposing greater homogeneity or heterogeneity 
iu one part of Scotland than in another ? 
(e) What general conclusions on other points may be drawn ? 
(2) Relation between the Nature of the Distributions for Sane and 
Insane Populations. Problem (6). 
In this section it is proposed to consider, not the absolute values of the type, 
variation and correlation of characters, but the general question of how closely 
the form of the frequency distribution is the same for these two classes of the 
general population. This may be done (1) by discussing the frequency curve for 
the distribution of a single character, or (2) by considering the nature of the 
regression curve for two characters. 
(i) Distributions. It has been shown by a number of writers (Fawcett*, Pearson 
and Leet, MacdonellJ, and Pearl§), that, with short series, frequency curves for 
anthropometric characters such as stature, head measurements, cranial measure- 
ments and indices follow closely, but not without sensible exceptions, the normal or 
Gaussian curve. It becomes therefore a problem of much interest to determine 
* Diometrika, Vol. i. p. 443. 
X Biometrika, Vol. iii. p. 227. 
t Biometrika, Vol. ii. pp. 361— 3C9. 
§ Biometrika, Vol. iv. p. 40. 
