In the summer of 1948, Muzafer Sherif brought twenty-two boys to Robbers Cave State Park in Oklahoma, divided them into two groups, staged competitive games to raise tension, then introduced cooperative tasks requiring both groups to work together.
Sherif's hypothesis — that hostile groups unify around a superordinate goal neither can achieve alone — became foundational to conflict resolution theory taught in every business school and policy institute. But the entire architecture rests on something he never established. He never determined whether the boys actually formed genuine in-groups with real psychological weight or merely performed the competitiveness the environment demanded.
This distinction matters because the two are not the same thing. In 1948, the world needed Sherif's framework to work. Europe fractured, the Soviet Union consolidating control, policymakers desperate for a scalable formula that said hostile groups aren't locked in conflict by history or irreconcilable interests but by a failure of coordination.
When Sherif engineered the competition, he may not have been observing real group formation at all. He may have been training the boys to perform competitiveness because the environment demanded it. The scarcity was artificial, the tasks designed, the goals imposed by an adult with power. Subjects complying because that's what you do in a controlled setting with an experimenter in authority.
For seventy-five years, no one interrogated whether the groups formed at all or whether Sherif had simply created a behavioral script and watched it run. Policy recommendations and organizational interventions cascaded from this single unexamined assumption, meaning every conflict-resolution program that assumed hostility dissolves with shared goals might have been treating a symptom of artificial setup rather than a feature of human group psychology. The difference between manufacturing cooperation and observing it is the difference between having a useful theory and having theater. And for decades, institutions couldn't tell the two apart because they didn't want to.
When you design an experiment to produce an outcome, you need to be ruthlessly precise about what outcome you've actually produced.
”