A method is provided for estimating the nonspurious correlation of a part of a test with the total test in which it is included by considering a hypothetical "subtest", parallel to the actual subtest but separate from the total test, and determining its correlation with the total test. Two cases are considered: one in which the actual subtest is parallel to the total test, and the other, in which the actual subtest is not parallel to the total test. Data are given to illustrate the applicability of the formulas.