ABSTRACT
Campbell and Fiske (1959) argued that every measurement we obtain in psychology is a trait-method composite—a measure purportedly of a particular trait construct obtained using a given method of measurement. Campbell and Fiske introduced the multitrait–multimethod (MTMM) matrix as a tool for evaluating systematically the correlations among a set of measures obtained using multiple methods. The primary utility of the MTMM matrix approach is the opportunity such a study affords to determine the preponderance of trait-related and method-related variance in measures in a battery. To aid in this evaluation, Campbell and Fiske argued that researchers should measure each of t traits (e.g., Extraversion, Neuroticism, Fluid Intelligence) using each of m methods (e.g., self-report, objective tests, observer ratings), so that each trait is measured using each method. By arranging trait measures in the same order within methods, the MTMM matrix should exhibit clear patterns to satisfy the dictates of convergent and discriminant validation. Convergent validation is satisfied if the researcher finds high correlations among measures of putatively the same construct using different methods of measurement, and discriminant validation is satisfied if low correlations are found among measures of presumably different constructs. Campbell and Fiske described several rules of thumb for evaluating patterns of correlations in the MTMM matrix. Specifically, (a) correlations between measures of the same construct obtained using different methods of measurement should be large; (b) correlations between measures of the same construct obtained using different methods of measurement should be larger than correlations of those measures with measures of different constructs obtained using the same or different methods; and (c) the same pattern of trait correlations should hold for all combinations of methods.
