Summary Several characteristics of Fiedler's (8, 9) least preferred co-worker (LPC) scales were investigated. It was found that although the LPC scales reflect two evaluative dimensions, there is a general negative evaluative component common to both dimensions. When asked to evaluate someone they had known in rating their least preferred co-worker, approximately two-thirds of the respondents rated a real person, while the other third rated a stereotyped stimulus object. LPC and its subfactors were found to be substantially unrelated to a large number of personality measures. Theoretical and methodological problems inherent in a bidimensional interpretation of a univariate score were discussed, and an alternative unidimensional approach was suggested.