ABSTRACT
Campbell and Fiske’s (1959) landmark paper on the multitrait-multimethod (MTMM) matrix is one of the most highly cited articles in all of psychology (Fiske & Campbell, 1992). As of February 2007, it had been cited 4,338 times in the Web of Science database alone, in diverse fields such as political science (Funke, 2005), marketing (Kim & Lee, 1997), leisure studies (Glancy & Little, 1995), medicine (Bernard, Cohen, McClellan, & MacLaren, 2004), sociology (Bollen & Paxton, 1998), biology (Ittenbach, Buison, Stallings, & Zemel, 2006), law (Rogers, Sewell, Ustad, Reinhardt, & Edwards, 1995), education (Krehan, 2001), and sports sciences (Cresswell & Eklund, 2006), as well as in various subdisciplines of psychology (e.g., social, personality, industrial-organizational). One of the reasons for the widespread adoption of the MTMM methodology is that the establishment of convergent and discriminant validity by its use has been seen as one of the cornerstones for the documentation of a measure’s construct validity (Benson, 1998; Messick, 1995), which is itself a unifying construct for the organization of validation evidence (Cronbach & Meehl, 1955; Landy, 1986). Thus, the history of the MTMM methodology is one of widespread adoption and endorsement as an important tool in construct validation efforts. However, through its use at least one “urban legend” has arisen that we address in this chapter. This legend is that if one crosses two measurement facets, one of which constitutes the substantive constructs of interest (i.e., the Traits), then the other measurement facet constitutes, de facto, the Method
measurement facet. As we explain later, this default assumption has had some very unfortunate consequences in at least two bodies of literature in industrial and organizational psychology.
