Karl Pearson 
253 
agreeing for all practical purposes with the value found for r) above, and indicating 
that although the arrays as shown in the fourth column are heteroscedastic, it is 
practically sufficient to give y<Txl<^x its mean value. Examining Column 1, we see 
that the mean eye colour of the arrays shifts continuously towards the light as we 
pass from good to bad handwriters. The association is small, but can hardly be 
without significance. 
(4) Alcoholism is a character which may be reasonably treated as one of 
graduated intensity, and one accordingly to which our metliod may be reasonably 
applied. I take as an illustration some statistics of crime most kindly provided by 
Dr Charles Goring. 
(D) Alcoholism and Type of Grime. In the following table the crime 
categories have been arranged in the order of relative intelligence as ascertained 
from Dr Goring's correlation investigations on crime and intelligence. 
Class of Crime. 
Arson 
Rape 
Violence 
Stealing 
Coining 
Fraud 
Totals 
Alcoholic 
50 
88 
155 
379 
18 
63 
753 
Non-alcoholic ... 
43 
62 
110 
300 
14 
144 
673 
Totals 
93 
1.50 
265 
679 
32 
207 
1426 
We find : xja^ = + '0704, and for the arrays*: 
x.ja, = + -0,946, x^ja. = + 2190, x./a, = + -214.5, 
xja, = + -1463, ;r,/o-g = + -1573, x./a, = - -5119. 
Using formula (iii), we determine, rj = -234. There is thus quite a sensible 
correlation between the extent of alcoholism in the criminal and the type of crime 
he will commit -f*. Starting with fraud, there is increased alcoholism as we pass to 
stealing, coining, violence and rape. An exception, however, occurs in the case of 
arson, the class of crime whose perpetrators are least intelligent among criminals. 
Those who commit arson appear to be less alcoholic than coiners and thieves — 
criminals who require considerable sobriety for success in their callings. Probably 
the coiners would stand above thieves in the scale, did not the returns include the 
offenders who pass false coin as well as the producers ; further, highly skilled but 
alcoholic workmen, who cannot retain on that account their berths, may in the 
case of certain trades tend to gravitate into this class of criminal. While rape 
and crimes of violence arise from a low grade of intelligence combined with 
alcoholism, it would seem that arson has a low grade of intelligence for its source, 
but probably some character such as revenge rather than alcoholism ultimately at 
its source. 
* Eeduced to a common unit: 2i/o-^= + '0920, J2/o-x= + -2129, 53/0-^= + -2085, Jj/o-^^ + -1422, 
^6K=+"1529, xel<T^=--i9n. 
t The reader must remember that this has no bearing on the problem of whether alcoholism leads 
to crime. 
Biometrika vii 33 
