An Example of Convergent and Discriminant Validation of Personality Questionnaires

Abstract

Within the last 5 years efforts at validating personality questionnaires via the multitrait-multimethod strategy (MTMM) originally proposed by Campbell and Fiske (1959; Jackson & Paunonon, 1980) have been redoubled. This interest is of course spurred on by the realization that what is actually being measured in any personality instrument is not solely the trait itself but rather the trait confounded with the method used. Rezmovic and Rezmovic (1981) succinctly summarize this state of affairs by saying that “each measure is a trait-method unit in which observed variance is a combined function of variance due to the construct being measured and the method used to measure that construct” (Rezmovic & Rezmovic, 1981, p. 61). In any attempt to introduce new personality questionnaires it has become standard practice to include some form of multitrait-multimethod analysis. While this practice is straightforward in itself, many different statistical approaches have been suggested for this purpose (Jackson, 1975; Hubert & Baker, 1978). In fact, the variety of different analyses has justified a general review and critique of approaches (Schmitt, Coyle, & Saari, 1977). Of the many proposals suggested for MTMM, the approach via confirmatory factor analysis seems to be the most straightforward and promising in this context (Schmitt, 1978; Rezmovic & Rezmovic, 1980, 1981).