Abstract: If two series of psychological tests X and Y involve p and q subtestsseparately, we can take ρ_(XY)=1/r sum from i=1 to r λ_i as the validity coefficient betweenX and Y, where λ_i is the ith nonzero characteristic root of matrixM=V_(YY)~+V_(YX)V_(XX)~+V_(XY). When p=q=1 or p=1, q>1, the generalized correlationcoefficient ρ_(XY) is just the ordinary validity coefficient of psychologicaltests. This ρ_(TT) may be a measurement of the reliability of series test T.A practical example is in the paper.

Key words: NULL