Miscellanea 
465 
liminary whether the proportions of different hair colours in the case of those who die early and 
in hospitals are the same as those of healthy children in the population at large. On the basis 
of our statistics this is improbable. A selective death rate in the matter of pigmentation, and 
peculiar selection of subjects, vitiates largely, we believe, returns based upon post-mortem room 
observations. All that has been done with Pfitzner's observations was to work out the Table for 
males from 0 to 75 jcAVfi of age. This is given below. 
TABLE V. 
Correlation of Afje and Hair Colour. Hales, Lower Elsass. 
Hair Colour. 
Blond 
Brann 
Schwarz 
Totals 
Age 0—15 
384 {83-8} 
74 {16-2} 
0 { 0} 
458 
15—30 
122 {35 -B} 
204 {59-8} 
15 { 4-4} 
341 
30—J,5 
119 {28 'O} 
261 {61-6} 
44 {10-4} 
424 
1,5—60 
120 {24-6} 
292 {60 -61 
75 {15-4} 
487 
60—75 
58 {28-7} 
105 {52-0} 
39 {19-3} 
202 
Totals 
803 
936 
173 
1912 
Using the method of mean squ.i,re contingency we find : 
Correlation between Age and Hair Colour = '451. 
Now if we were to trust this table we should have a very marked increase in black hair with 
increasing age, and this is the interpretation which Dr Pfitzner put upon results like this in his 
memoir. But a little consideration shows that this cannot be correct. Are we to say that there 
are no children with black hair in Lower Elsass, and that the 20 per cent, found among old 
persons is due to darkened pigmentation only ? Prussia is lighter than Elsass but it contains 
18 per 1000 of such children. We should have expected at least eight or nine black-haired 
children in Pfitzner's 458 under 15 j^ears of age. Baden close to Elsass shows 18 per cent, of 
conscripts with black hair, while Pfitzner gives only 4-4 with black hair between 15 and 30! 
Pfitzner's results, if attributed to age-effect, seem quite incompatible with what is known of 
the normal population for P^lsass. They are, however, quite comprehensible if there be a 
positive correlation between fairness and disease in childhood. Now this is exactly what our 
British school children show : t/iere is a correlation between health and darkness of hair colour. 
Hence if we do not follow up individuals, noting their pigmentation at different ages, but 
simply correlate age of different individuals with hair colour, we are liable to exaggerate the 
correlation between age and pigmentation, and this will be especially the case, if we use hospital 
returns. Hence, it is probable that our neglect of a selective death rate, based upon the known 
correlation between general health and pigmentation, really emphasises the values found for 
correlation between pigmentation and age. Further, while it is probable that if we take 
adult life into account we should find this correlation increased — perhaps even to -2 or -25 
— the value deduced from Pfitzner's observations of '45 may be safely considered to mark in 
the first place a selective death rate, i.e. a correlation between fitness //( childh<io<l and dark 
pigmentation. 
Biometrika in 59 
