Starting from the assumption that people direct more attention to the objective, non-evaluative, aspects of undesirable behaviours than of equally evaluatively-intense desirable ones and attach more weight to the defining descriptive features of undesirable traits than of desirable ones, we predicted and demonstrated, with samples of Dutch adjectives, that, on the average, undesirable trait adjectives, as compared to desirable ones, had a lower category breadth and a higher discriminative value (Study 1). Undesirable trait adjectives were also estimated as less similar in meaning and as less likely to co-occur in the same person, and were less strongly intercorrelated when used for memory-based self-ratings (Study 2). These positive-negative asymmetries tended to hold for all traits whether they belonged to the self-profitability dimension or to the other-profitability dimension, and did not depend on the evaluative intensity of the traits. Alternative interpretations of these asymmetries and the implications of these asymmetries for personality assessment and for research on implicit personality theory and on the factorial structure of rating correlations are discussed.