ABSTRACT
Campbell and Fiske introduced the multitrait-multimethod (MTMM) design at the end of the 1950s. The design involves the use of different methods to make repeated measurements of various traits. The covariances or correlations between the resulting measurements are then ordered in a matrix that is analyzed with a confirmatory factor model. The purpose of this analysis is to obtain estimates of measurement quality: validity, method effects, and reliability. De Wit and Billiet (1995) wrote an historical review of the development of the MTMM design and the conceptualization of validity, method effects, and reliability, which we summarize briefly. Campbell and Fiske (1959) spoke of “convergent validity” if measurements of the same traits, using different methods, correlated strongly. Although this convergence was considered to be a necessary part of the validation process, it was not considered, by itself, to be sufficient. Campbell and Fiske thus proposed a supplementary criterion, “discriminant validity,” which applied if measures of different traits, using the same method, did not correlate strongly. The overall analysis technique used by Campbell and Fiske was largely qualitative and provided a rough indication of the convergent and discriminant validity of the MTMM measures. A decade later, confirmatory factor analysis using structural equation models became the main method for analyzing MTMM data. CFA offered the possibility to test different model specifications for MTMM data, for example correlated uniqueness models versus models with latent method factors, models with correlated versus models with uncorrelated trait and method factors, single-indicor versus multiple-indicator models. Eid et al. (2003) gave a comprehensive overview of the main models and discussed their properties and limitations.
