Our results suggest a small negative effect between racial diversity and team performance (r = -0.13) and gender diversity and team performance (r = -0.09). These negative effects were observed only under field conditions. Our results bring some clarity to surface-level demographics historically associated with social categorization (i.e., age, gender, race). Results from previous meta-analyses have indicated either no relationship between these diversity variables and team performance (e.g., Horwitz and Hortwiz, 2007; Webber and Donahue, 2001) or little impact (Joshi and Roh, 2009) when context is relative. Our results indicate that the relationship between the diversity of surface-level variables and team performance varied depending on the specific variable. In particular, we found no confirmation of the relationship between age diversity and team performance, regardless of the conceptualization of age diversity. It is possible that generational differences (Smola and Sutton, 2002), rather than chronological age diversity, affect team performance. The negative relationship between racial diversity and team performance is interesting because diversity has been consistently operationalized as an indicator consistent with diversity (e.g., Blau, Teachman), most likely due to the categorical nature of the variable. However, racial diversity is typically hypothesized to be related to team performance through processes such as social categorization and intergroup bias, as well as the similarity and attraction hypothesis (e.g., Williams and O'Reilly, 1998), which is consistent with the conceptualization of separation (Harrison and Klein, 2007). Maximum diversity reflects a distribution in which each member in a unit is from a unique category of variable (e.g., race); that is, each member differs from the others in terms of race. The maximum degree of separation, however, reflects teams that have polarized subgroups. The Blau index has been widely used to operationalize racial and gender diversity. It is designed to reflect heterogeneity (or whether more categories are represented), and should be used when racial and gender diversity is expected to be related to team performance as more categories are represented. Researchers should consider how the operationalization of racial and gender diversity captures the mechanisms through which they expect diversity in these variables to be related to team performance. Pearsall, Ellis and Evan's (2008) recent laboratory study of gender dividing lines reflects the concept of gender separation. In particular, they compared homogeneous teams with teams equally divided in terms of men and women. In the case of actual organizational teams, researchers face the additional challenge of considering the range of race and gender categories represented in their dataset, and are likely to have fewer opportunities to manipulate the size and composition of teams. Researchers should consider whether a moderate level of diversity or another measure reflecting maximum separation (such as fault lines) better captures the form of differences suggested by their theoretical arguments. Further research using data simulations and other methods is needed to help understand these issues. The potential misuse of the Blau index to represent racial differences other than heterogeneity has been noted more widely in the social sciences (Rushton, 2008). Given the limited number of studies on race and gender diversity, we question whether most previous studies have used the most effective approach to examine the relationship between team performance and gender and racial diversity.Unfortunately, the effectiveness of mono-ethnic teams by race is not reported - this is important because it would indicate potential differences in effectiveness between races. It may be that the results presented are underestimated, example: a team of 8 whites and 2 blacks produces 5 cars per day - a diversity team has an efficiency of 0.5 cars per person. In monoethnic teams, a team of 10 whites produces 10 cars, and a team of 10 blacks produces 2 cars - monoethnic teams produce 0.6 car per person, which is more, but we completely failed to convey the essence of the problem. It's hard, it's cool that they published it at all.