The composition of three-man military crews was varied experimentally by assigning members according to crew composition in terms of all possible combinations of levels of ability and motivation. The crews performed real military tasks in a military field setting, and unit commanders ranked the effectiveness of their performance at the end of 2 months of military activity (208 crews in all). It was found that both ability and motivation had an additive effect on crew performance. In addition, crew composition effects were found for ability but not for motivation. Most significant was the fact that the performance of uniformly high-ability crews far exceeded the levels expected on the basis of individual crew members' ability, whereas the performance of uniformly low-ability crews fell considerably below the expected level. It was concluded that when crews perform highly interdependent tasks, performance is also likely to be affected in a nonadditive manner by crew composition. The implications of the findings for crew formation are discussed.