Method Variance, Inadequate Constructs, Or Things That Go Bump In The Night?

Jackson's (1975) revised method for multimethod factor analysis and his critique of Golding and Seidman's (1974) similar "two step principal components" procedure are critically examined. While Jackson's revised method is more empirically defensible, the conceptual assumptions upon which it is based may not be warranted, and his critique of the Golding-Seidman method may be misleading. It is argued that "method variance" is properly viewed as a hidden facet of nomologicals underlying a construct, and that multimethod investigations that fail to specify the nomological, assumptions surrounding particular trait-method units are conceptually and empirically unsound.