Multitrait-Multimethod Analysis (MTMM)

An approach to construct validation that examines the degree to which methods and constructs are distinguishable. The approach was first formalized by Campbell and Fiske (1959). In particular, an MTMM study provides evidence of the convergent and discriminant validity of a test that is designed to measure a particular construct of interest to a researcher. An MTMM study examines at least two different traits measured by at least two different methods. Measures of the same trait should converge (have strong positive correlations with one another). Measures of different traits should be “discriminant” (have low or zero correlations with one another or an inverse relation to one another if it is theoretically meaningful). Examining different methods provides evidence of the robustness of a construct to different ways of measuring it. If correlations between measures of the same construct are moderate or highly positive, even though they are based on different measurement methods (e.g., a Likert scale and an interview), then the validity of the construct is supported. However, if there are stronger correlations between different constructs measured by the same measurement methods (two different constructs measured by Likert scales) than between measures of the same construct based on different methods, then there are problems with the validity of the construct because of measurement artifacts. More recent versions of MTMM studies have used structural equation models to test the fit of theoretical models designed to examine the relationships that would be expected under the conditions of convergent and discriminant validity. For further reading, see Campbell and Fiske (1959).