Intergroup Conflict and Cooperation: The Robbers Cave Experiment
In the chapters to follow the main points of a large-scale experiment on intergroup relations are
reported. It was carried out as a part of the research program of the Intergroup Relations
Project at the University of Oklahoma. In this first presentation, sufficient time and facilities were
not available to make use of data contained in recorded tapes and half a dozen short moving
picture reels. Nor was it found feasible to include introductory chapters surveying major
theories on intergroup relations and elaborating on theoretical outlines of the present approach,
which determined the formulation of the hypotheses advanced and the design of the study in
successive stages. These are presented more fully in our Groups in Harmony and Tension
(Harper, 1953), which constituted the initial work unit in the present intergroup relations project.
Therefore, a brief statement of the cardinal considerations that shaped the conception of this
approach to the study of intergroup relations is in order. It is not unfair to say that the major
existing theories fall within two broad categories in terms of the emphasis placed in formulation
of the problem and methods involved.
In one broad category of theories, the problems are expressed in terms of actualities of events
in group relations as they exist in everyday life. On the whole, theories advanced by many
social scientists fall in this broad category. In this concern over actualities the problem is
frequently not stated and discussion not developed in a way that can be tested rigorously. In
the second broad category of theories, problems are stated and analysis carried out in terms of
more rigorous-appearing concepts and units of analysis. Theories coming from psychologists
and social scientists heavily influenced by them fall within this broad category. In this line of
approach, theories are advanced without due regard to actualities, and consequently they are
plagued with serious questions of validity.
The present approach starts with a serious concern over the rise and functioning of actual small
groups in social life. The hypotheses advanced are formulated on the basis of recurrent events
reported in sociological accounts of small groups. Testing these hypotheses under conditions
that appear natural to the subjects has been a theoretical and methodological consideration of
prime importance. Therefore, a great point was made of carrying on observations without the
awareness of subjects that they were being observed and of giving priority to the uninterrupted
and uncluttered flow of interaction under experimentally introduced stimulus conditions. The
techniques of data collection were adapted to the flow of interaction, rather than cluttering or
chopping off interaction for the convenience of the experimenter. This imposed the task of
securing an experimental site which is isolated from outside influences so that results could not
be accounted for primarily in terms of influences other than the experimentally introduced ones
and the interaction on that basis.
In such a natural, life-like interaction situation, there are so many items that can be observed at
a given time that it becomes impossible to observe and report all behavioral events. Therefore,
there is the possibility of being selective in the choice of events to be observed. In testing vital
hypotheses related to intergroup relations, restricting the number of subjects to just a few is not
the proper remedy. Circumscribing the number of reactions of the subjects is no remedy.
Asking the subjects to remain within optimal distance of a microphone and asking them please
to speak one at a time will destroy the very properties of the interaction process in which we are
interested. The dining hall adjacent to the kitchen is not the place conducive to getting the
subjects to cooperate in preparing a meal of their own accord. By trying to eliminate selectivity
through such resorts we would have eliminated at the same time the essential properties of the
very things we set out to study.
(1) One remedy lies in unmistakable recurrences of behavioral trends so that the observer
cannot help observing them even if he tried to ignore them. If these trends are independently
reported by the observers of two different groups, then they serve as a check against each
other. We have secured such checks time and again in this study.
(2) The danger of selectivity can be avoided (without disrupting the flow of interaction) by
having outside observers in crucial problem situations and by having them make, for example,
their own independent status ratings in terms of effective initiative in getting things started and