The forced-choice and a questionnaire format of the Thorndike Dimensions of Temperament were administered under regular and desirability directions to college students. The forced-choice format produced low scale intercorrelations under the regular directions, but under the desirability directions a common factor appeared. The same factor appeared in the questionnaire format under both the regular and desirability directions. The forced-choice format under the regular directions also produced mean scores that were generally below the neutral point of the scales, while under the other three conditions of administration the mean scores were generally above. These data are interpreted as showing that the forced-choice format cannot prevent subjects from presenting a more favorable image of themselves if they choose to do so, but that this problem is normally less in the forced-choice than in the questionnaire format. In the questionnaire format there is a stronger effect of a social desirability factor which increases the mean scores on the scales and the variance from individual differences in the tendency to respond in a self-regarding manner. The increased variance from a social desirability factor produced correlations between items from the different scales which obscures the differential measurement of the traits. (Author)