Pedhazur and Tetenbaum speculated that factor structures from self-ratings of the Bem Sex-Role Inventory (BSRI) personality traits would be different from factor structures from desirability ratings of the same traits. To explore this hypothesis, both desirability ratings of BSRI traits (both "for a man" and "for a woman") and self-ratings were obtained from the same sample and factor analyzed. Factor analyses performed on the three sets of ratings of the 40 BSRI traits (self-ratings, desirability ratings "for a man", and desirability ratings "for a woman") confirmed that the factors across ratings were diverse. Thus, the underlying constructs must be studied independently. Predictive discriminant analyses replicated the finding that two traits alone ("Masculine" and "Feminine") provided nearly all of the discrimination of males and females in the sample when self-ratings were employed. Also, predictive discriminant analyses revealed that the classification of participants into gender groups was very accurate using self-ratings but deteriorated remarkably when using the desirability ratings. (Contains 5 tables.)