DRDING to recent field and laboratory studies, cohesiveiiess or "attraction-to-group" is a potent determiner of the strength of uniformity pressures (1), channels of face-to-face communication (3), the quantity of production (8), the expression of hostility (7), and other social behavior. There is, however, relatively little dependable information bearing on the factors that cause changes in cohesiveness. The present field experiment is concerned with this problem, namely, the effects upon cohesiveness of "threat" and "frustration."Everyday observations of how these variables operate on cohesiveness are highly inconsistent. It is often apparent, for example, that groups under stress "pull together" and "close ranks" more than under normal conditions. On the other hand, there are circumstances such as battles where groups completely "fly apart." Beyond the fact that such observations are usually uncontrolled, they are difficult to interpret because of the multiple and sometunes dubious meanings attached at least implicitly to the concept of cohesiveness. Thus, changes in physical propinquity, uniformity of opinion, productivity, etc., which appear to result from threat and frustration do not necessarily mean a change in underlying cohesiveness-when the latter is strictly denned as attraction to the group. What adds to the difficulty is that the conditions of "threat" and "frustration" are defined in a variety of ways, and in many instances, appear to be regarded as conceptually and operationally equivalent. Finally, there seems to be an absence of even the most rudimentary theory with respect to this problem from which empirical predictions can be made.Conceptually, "threat" is taken for present 1 This study was made possible by a grant from the Albert M. Greenfield Center for Human Relations of the University of Pennsylvania. We also gratefully acknowledge the help of Mrs. Helen Leighty of the United Neighbors Association who made available the subjects and other facilities at Camp Linden, Northfield, Pa., where the study was carried out. Finally, the contributions of George Moed are appreciated.