416 
Miscellanea 
The essentials of this note may be summarized as follows : 
In using Intra-class coefficients care must be taken to guard against spurious values arising 
through differentiation among the individuals of the class. 
Besides the orderly differentiation (due to age of individuals, position of organs on axis, etc.) 
for which Pearson has determined corrective fornuilae in terms of correlation coefficients, a 
disorderly differentiation for which such corrective formulae have not as yet been found some- 
times obtains. Illustrations of such cases are here given. 
Probably the empirical methods used here in correcting for this disorderly differentiation 
should be I'eplaced by formulae with a sounder theoretical foundation. This I have not as yet 
been able to do. 
The pm-pose of this note will have been served if it directs attention to a source of danger 
which may sometimes be encountered in the use of serviceable formulae, and indicates a method 
by which in the absence of more perfect methods jiractical results may be secured. 
Cold Spring Harbor, N.Y. 
February/ 3, 1914. 
II. On an Extension of the Method of Correlation by Grades 
or Ranks. 
In a memoir published in 1907* I have shown how, on the hypothesis of normal distribution, 
the true correlation of variates r may be ascertained from the correlation p of grades. If 
gi and g2 be the two grades, and V2 the corresponding ranks, x and _?/ the corresponding variates 
with means :v and ^, and standard-deviations o-i and 0-2, while 
* " On Further Methods of Determining Correlation," Drapers' Company Research Memoirs (Dulau 
and Co.), pp. 11, 12. 
By KARL PEARSON, F.R.S. 
