The formula in question gives the correlation between two variables for a number n1 and nz of individuals when the ordinary parameters, including correlations, are available for the separate groups of n1 and n2 individuals; and it is extended to more than two groups. It is possible, even probable, that the formula is known, but the writer has not found any direct references to it; in any case its practical value is such that there can be no harm in drawing attention to it.