ON THE CALCULATION OF INTRA-CLASS AND INTER- 
CLASS COEFFICIENTS OF CORRELATION FROM CLASS 
MOMENTS WHEN THE NUMBER OF POSSIBLE COM- 
BINATIONS IS LARGE. 
By J. ARTHUR HARRIS, Ph.D., Carnegie Institution of Washington, U.S.A. 
I. Introductory Remarks. 
If one designates as classes the groups of measurements or evaluations 
segregated out of the whole series under consideration, then those correlations 
which measure the similarity in character of the individuals of the same class 
may be designated as intra-class coefficients of correlation, while those which 
measure the resemblance between the individuals of different but (in some 
manner) related classes may be designated as inter-class coefficients of corre- 
lation. 
Intra-class and inter-class should not be confused with intra-racial and 
inter-racial. These latter terms seem to have been first used in a memoir on 
Naquada Crania by C. D. Fawcett* and discussed for the local races of the 
lesser celandinef and in greater detail with anthropometric illustrations by 
Jacob, Lee, and Pearson \. Examples of the practical use of inter-racial 
correlation coefficients are also given by Tschepourkowsky §. By intra-racial 
correlation as the term is used in these discussions, one understands those 
coefficients expressing the interdependence between the measures on a series 
of individuals of the same race. Under inter-racial are included the correlations 
between the types (means, modes) of a series of races. 
In intra-class and inter-class correlation, on the other hand, the unit is always 
the individual (in the statistical sense), not the type of a series of individuals. 
The purpose of the correlation is to determine the resemblance for any series of 
characters of the individuals of sub-classes, or of an associated pair of sub-classes, 
* Biomeirika, Vol. i. pp. 460—461, 1902. 
t Biometrika, Vol. n. pp. 152—153, 1903. 
X Biometrika, Vol. n. pp. 347—356, 1903. 
§ Biometrika, Vol. iv. pp. 161—168, 286—312, 1905. 
