496 Proceedings of Royal Society of Edinburgh. [sess. 
tribution whose constants are those required by type IV. The fit 
of both these curves is good. At the beginning and end the 
divergence of the actual figures from the theoretical distribution is 
most marked, but it is at these points that the probable error of 
the statistics is very large. As an equal number was subtracted 
Diagram XII. 
from the whole series, and as at the extremities this number, in 
place of being only a moderate percentage of the whole cases, 
becomes equal to many times the residual number of cases, it will 
be seen that the accuracy of the residual numbers cannot in- 
dividually be very great. The curve of zymotic diarrhoea deaths, 
London,* for the years 1853-1903 (diagram XV., Table A, No. 
Report to Registrar-General for England, 1903. 
