126 Relatioi>sliip of liitcllvicnce to Size and Shape of Head 
the correlation with other physical and mental characters. My school measure- 
ments enable results to be obtained for the following series : 
Physical Characters Mental Characters 
Athletic Power Temper 
Health Popularity 
Head Length Self-Consciousness 
Head Breadth Shyness 
Head Height Conscientiousness 
Cephalic Index Quiet Habits 
Hair Colour and the psyclio-physical character 
Eye Colour Handwriting 
Curliness of Hair 
Age. 
But the method by which the relationship between intelligence and these 
characters can be best obtained must be varied with the completeness of classification 
which it is possible to apply. Thus where one quantity, as in the case of the head 
measurements and age, is quantitative the correlation ratio rj has been deter- 
mined. Where no quantitative measurement is available but a fairly numerous 
system of classes as in the case of the relation between intelligence and hand- 
writing, health and hair colour, the method of mean square contingency has been 
adopted*. Even when one of the characters has only a threefold division, as in 
the cases of temper, curliness of hair and eye colour, the contingency table gave an 
18-fold grouping. In the remaining cases with only two alternatives for one 
character, we were perforce thrown back on the fourfold division table. But even 
here many tests were made by dividing the intelligence grouping at more than 
one point. The chief (juestion is whether the slow intelligent shall in the fourfold 
division be classed with the intelligent or dull groups. In the case of the boys 
all the fourfold tables were worked out both ways, and the mean taken of the 
results, but the labour proved excessive and was abandoned in the case of the 
girls, the division being taken as nearly as possible through the median — which 
gives the least probable error — i.c. the (juick intelligent and intelligent were taken 
as a single group. 
It will be of value to compare some of the results obtained by different 
methods. 
First, we may take as a comparison of correlation ratio and correlation coefficient 
found by a fourfold table method : 
, T, ,,, „ TT 1 1 T i 11- (■ Correlation Ratio = -100+ -014. 
Boys Lreadth ol Head and Intelligence ^ „ , „ „, . 
1 Correlation Coefficient = -084 ±-024. 
Next as a comparison of contingency and fourfold method : 
r, , TT 1 -i- 1 T , 1,- f Mean Square Contingency = '283. 
Boys Handwriting and Intelh-ience , ,. ^, ^ . , „ 
o o I Correlation Coefficient = -31 2. 
A more complete comparison may be taken in the case of temper and intelli- 
gence in girls. Here four fourfold tables were worked out ; the good-natured 
* Drapers'' Compunij Research Memoirs. Biometric Series I. Dubiu and Co. ' 
