Abstract: In this study, we used multitrait multimethod matrices and confirmatory factor analysis to examine the construct validity or structural models of a set of instruments making up an assessment center (AC). The AC consisted of three exercises tapping multiple performance dimensions. The three exercises included leaderless group discussions, in basket exercises, and personality questionnaires. One hundred and thirty six employees were assessed on four ability dimensions using these exercises. The convergent validity was found to be lower than the discriminant validity. These and other results suggest that AC ratings on various ability dimensions across multiple exerciese are situation specific rather than dimension specific. That is, the latent factors underlying AC ratings represent method but not trait dimensions.

Key words: assessment center, construct validity, structure modeling