A large number of previous studies have used self-construal to predict communication outcomes. Recent evidence, however, suggests that validity problems may exist in self-construal measurement. The current study conducted a multimethod multitrait (Campbell & Fiske, 1959) validation study of self-construal measures with data (total N = 578) collected in Korea (N = 200), Japan (N = 212), and the U.S. (N = 166). The data showed that the Singelis (1994) Self-Construal Scale, the Cross, Bacon, and Morris (2000) Relational Interdependent Self-Construal Scale (RISC), and the Kuhn and McPartland (1954) Twenty Statements Test (TST) lacked convergent and discriminant validity, both pan-culturally and within each of the three countries included in the study. Scores on the TST were not significantly related to scores on the self-construal scales, and the various self-construal measures correlated more highly with measures of communication directness than with alternative measures of the same type of self-construal. Substantial method effects were also observed. The results were tested for both 2- and 3-dimensional models of self-construal and for refined scales and scales with all items retained. The results of all analyses were inconsistent with the claim that self-construal measures are construct valid.