The authors would like to thank Rod Kramer, Linda Johanson, and three anonymous ASO reviewers for their wonderful guidance and fabulous reviews. Thanks also to Donald Gibson, Barry Staw, John Turner, and Batia Wiesenfeld for their comments and help throughout the paper. Partial support for this article came from the Fred Frank Fund, Yale University. 803/ASQ, December 2000 be a useful explanatory construct in understanding workplace behavior (e.g., George, 1990George, , 1995.Research has shown trait positive and negative affect to be classic personality factors, congruent with extraversion and neuroticism (John, 1990: 86), a result repeatedly demonstrated in the literature (see Parkinson et al., 1996: 61; reviews by Larsen and Diener, 1992; Meyer and Shack, 1989). We chose to focus on trait affect, however, rather than other personality variables, as trait affect is a more narrowly affectively defined construct, which leads to specifically affective manifestations (Tellegen, 1985;Watson and Clark, 1992; Parkinson et al., 1996: 61). This is by contrast, for example, to extroversion, which in addition to affective components such as cold and warm includes many other, less purely affectively related components, such as degree of sociability, talkativeness, spontaneity, and being a joiner versus being a loner (Costa and McCrae, 1992). Trait positive affect appears to be the best candidate for an initial study of how affective diversity relates to the interaction and performance of top management teams. While there is not as much prior research supporting a negative affective diversity model, we feel it is too soon to rule out negative affect in this context and thus conduct exploratory trait negative affectivity tests for all of our hypotheses as well.
Group CompositionAnalyses of the effects of group composition have been used to explain a wide variety of group phenomena, such as turnover, interpersonal relations, innovation, and performance, in general work groups (for reviews, see Jackson, 1995; Williams and O'Reilly, 1999) and in top management teams (see Finkelstein and Hambrick, 1996). Here, we focus on the group's diversity in trait positive affect. As with other group composition variables, when a group is interacting, members should react to each other's trait positive affect. Although trait positive affect is not a demographic characteristic, it is still readily identifiable, perhaps more so than the underlying values demographic characteristics are meant to represent. Research by Ekman and colleagues (1982) has shown that internal emotional states are reliably observable and can "leak" even when people are trying to hide them (Ekman, 1992). Supporting this, strong correlations have been found between peers' ratings of trait positive affect and selfreport ratings of trait positive affect, as well as among the peer raters themselves (Barsade, 1995), indicating the observability and reliability of trait positive affect. Hypothesis 5a: Affectively homogeneous groups will have better group performance than will...
