Social science has a replication problem — a new massive study found that only half of published findings hold up when researchers try to repeat them and many that made it into textbooks

May 28, 2026 - 10:11

It didn't start with a paper. It started with a classroom. I was teaching a unit on classic social psychology -- the foundational studies that most of us in the field absorbed as canonical truth. Milgram's obedience experiments. Zimbardo's Stanford Prison Experiment. Baumeister's work on ego depletion, the idea that willpower is a depletable resource.

Then I read the new study. And I had to rethink my syllabus.

A massive replication project, the largest of its kind, has confirmed what many researchers have quietly suspected for years: social science has a serious credibility problem. The project attempted to repeat 21 prominent studies published in top journals. The results were sobering. Only about half of the original findings held up under rigorous retesting. Some of the studies that failed had been cited thousands of times. A few had made their way into standard textbooks.

The problem is not fraud in most cases. It is more subtle. Original studies often rely on small sample sizes, which produce dramatic results that later fail when tested on larger groups. Researchers also face pressure to publish exciting findings. Journals want surprising results, not null ones. This creates a system where weak evidence can look like a breakthrough.

The replication crisis is not new. Psychologists have been wrestling with it for over a decade. But this latest project is notable because it targeted studies that many considered settled science. If these findings cannot survive a careful retest, then what else in the field is shaky?

For teachers, the implications are uncomfortable. We have been telling students that ego depletion is a proven fact. That a simple act of self-control drains a limited resource. The replication data suggests the effect is much smaller than originally claimed, if it exists at all.

The field is not broken beyond repair. Many labs are now preregistering their studies and sharing data openly. But the lesson is clear: science is a process, not a collection of facts. And sometimes, the facts we thought we knew turn out to be wrong.