Organizational factors in information handling in task-oriented groups have been studied recently at M.I.T. Laboratory groups were given very simple information-collection tasks under rigidly controlled conditions and the communication recorded completely. Decisions were restricted to choices of message destination and the influence of various imposed communication nets on the adequacy of group performance analysed. A technique of group-action quantization was used, which made possible a well-defined measure of group efficiency and an easily understood group goal. Experimental results show that an equiprobable random assumption is a poor model even for initial behavior. Centralized nets are shown to learn to do as well as the net permits fairly rapidly but to be characterized by high minimum-action. A net designed to give the lowest average action for task completion with random message-destination choices is shown to be essentially incapable of achieving its limiting minimum. Another similar net with a higher random average and a favorable minimum is shown to learn its minimum readily. Differences among the results for different nets are found to be explained in terms of a concept of “locally rational behavior” and the combinatorial characteristics of the nets. Operations Research, ISSN 0030-364X, was published as Journal of the Operations Research Society of America from 1952 to 1955 under ISSN 0096-3984.