The purposes of this study were twofold: (a) to examine the relative performance of general CFA models, correlated uniqueness model and composite direct product models to the analysis of MTMM data in educational and psychological research, and (b) to investigate the concordance of evidence associated with convergent and discriminant validity provided by these models. Based on goodness of fit and ,number of improper parameter estimates, performance of the correlated uniqueness model was the most successful. Although three alternative models were used, the performance of CFA models was the worst. As for the concordance of convergent and discriminant validity, the agreement for discriminant validity was stronger than that for convergent validity. Since the correlated uniqueness model fits most of the data, the effect of violating zero method correlation assumption on parameter estimates should be examined in future research.