The degree to which individual subjects tend to give socially desirable responses in self-description on each of thirteen MMPI scales was determined by computing, for each subject, a point–biserial correlation between his true/false responses and the social desirability (SD) scale values of the items. The results indicated that, on the average, from 10% to 20% of the variance in individual self-descriptions was associated with SD scale values. In comparison with these results, individual differences in the tendency to give SD responses, as measured by (a) these point–biserial correlations and (b) Edwards's SD scale, account for much more of the variance in raw scores on many of the MMPI scales. The differences between these methods of appraising SD aspects of test performance and implications for constructing SD-free scales were considered.