This study focused on two methodological issues by testing for discriminant validity in a multitrait-multimethod matrix which contrasted the set of estimates obtained from two methods and the "regression weights" derived from each subject's estimates from the second method. In order to change the conceptions of social behavior held by persons from different cultural backgrounds, it is necessary to be able to measure the similarity of such conceptions, or attributions. A method for quantifying attributions requires that values be assigned to five weights incorporated in an equation. This raises the question of how to obtain good estimates of these weights. A direct method is to simply ask subjects to estimate the weight of each predictor in the equations. A second method of measurement is to have the subject estimate the perceived relationship among all possible pairs of variables in each equation. This is tantamount to asking the subjects to estimate directly the correlation coefficient between all pairs of variables in the equation. The question as to whether the subjective correlations estimate fails to mirror processes actually used by the subjects can only be answered empirically. If the two methods produce different types of estimates, they may not yield similar interpretations of the subject's conceptions. (Author/JM)