Note on Standardisation of Method of Using the Coefficient of Racial Likeness

Abstract
Highly correlated characters and those with large errors of measurement should be avoided. A coefficient <3 probably does not mark a real racial difference. A correction for size of population is proposed.