On the subjective character of the empirical base of Guilford's structure-of-intellect model.

Abstract

Using data from three studies that have been cited as indicating support for the structure-of-intellect (SI) theory, analyses demonstrated that Procrustean factoring methods can be used to provide results that appear to support randomly determined theories. Such results can be obtained even when (<z) variables are reliable and presumably interrelated in lawful ways, (b) factors are required to be orthogonal, (c) factor loadings used in interpretation are required to be at least .30, and (d) subject sample numbers are as large as 175, 205, and 240. These findings were interpreted as indicating that the factor-analytic support for SI theory is not compelling because it is not appreciably better than the support that can be provided for theories generated by random procedures. The structure-of-intellect (SI) model (Guilford, 1967; Guilford & Hoepfner, 1971) occupies a prominent place among current theories about human abilities. Evidence of this can be seen in a number of places. The theory has been discussed in leading textbooks and review articles as a major contribution to our under