One area that researchers agree on is the disruptive nature of especially strong fault lines: deep differences between team members that allow subgroups to emerge. If a minority of a group are white, young, female, and of a technical background, and the remainder share none of those attributes, the team are likely to be fragmented and performance will suffer . But before now other questions have remained unanswered: is it worse to have more or fewer subgroups? And is it better if they are balanced, of a similar size?
The study looked at 326 teams from a food service and processing company, all participants in an internal tournament to demonstrate excellence through actions like developing better customer service. Carton and Cummings developed an algorithm that examined several possible fault lines in each team to determine whether it contains subgroups, and if so, how many and on what basis. These subgroup features were then related to how the team performed in the tournament.
The researchers predicted that number and balance of subgroups have different effects depending on the fault lines that define them. Identity-based fault lines, including age and gender, are strong lines that can encourage in-group/out-group thinking. As expected, groups that contained two, evenly matched identity-based subgroups performed worse than any other combination, as this 'us and them' situation can increase territoriality, where both sides feel threatened by the other.
Knowledge-based fault lines occur when a subgroup shares different sources of information, such as when they have a different reporting channel in the business. These lack overt in-group cues, and so they’re less disruptive. In fact they offer multiple perspectives – handy for solving problems. As predicted, the more of these, the better, and more so when the groups were balanced in number, as this reduces the likelihood that voices are discounted.
The categorisation of subgroups by algorithm, without any validation, seems a limitation of the study. We don't actually know if team 23 actually had two subgroups based on gender, only that the algorithm was satisfied there would be. I would like to see follow-up work that verifies the finding is genuinely due to grouping factors, not the mix of members in a more general sense. Nevertheless, it takes us a step closer to understanding team performance, and how decisions about team composition can have emergent effects upon performance.
Cohen, S. G., & Bailey, D. E. (1997). What makes teams work: Group effectiveness research from the shop floor to the executive suite. Journal of Management, 23, 239 –290. DOI:10.1177/014920639702300303