Karl Pearson 
105 
Conclusions. The illustrations will have indicated that the new process of 
determining correlation can be applied to a great variety of problems. In none of 
the cases dealt with was the correlation very high, but this is purely a result of 
the material selected, which was chosen, not from any knowledge of the existence 
of correlation, but to indicate the type of problem to which the new method can be 
applied. Hitherto such problems could only be treated by the fourfold table 
method. The examples given show how much more expeditious is the new process, 
and further how it frees us from all doubt that exists in the old method as to the 
suitable position for the division of the measured variate. That its probable error 
will be less than that of the fourfold table method will I have little doubt be 
demonstrated, when its value has been worked out. The method is, however, so 
convenient and so frequently of service that I have not delayed its publication 
until the leisure came to determine the probable error. 
Biometrika vii 
14 
