Smith et al. (2007a) examined the nature of the distinctions among the four constructs using both comparative factor analysis and multitrait, multimethod matrix (MTMM) approaches. In a sample of 1,886 undergraduates, they compared several, alternative factor structures of measures of the four putative traits. In one model, they tested the hypothesis that indicators of the four traits fall together on a single, common trait (the “impulsivity” model). That hypothesized structure fit the data poorly: Comparative Fit Index (CFI) = .58; Tucker-Lewis Fit Index (TLI) = .48, root mean square error of approximation (RMSEA) = .26; standard root mean square residual (SRMR) = .19. A model in which measures of the four traits were represented as separate but related factors fit much better: CFI = .99, TLI = .98, RMSEA = .05, SRMR = .03. Ultimately, they endorsed a third model, in which lack of planning and lack of perseverance were facets of an overall “deficits in conscientiousness” factor, and (negative) urgency and sensation seeking were separate factors. That model, which imposed the additional constraint of the higher order factor, produces precisely the same fit indices as the model specifying four separate traits.
