Power Posing: The Rise and Fall of a Nonverbal Miracle

In 2010 Dana Carney, Amy Cuddy and Andy Yap published a study reporting that holding an expansive, open posture for two minutes produced measurable physiological change: testosterone up, cortisol down, increased tolerance for risk in a gambling task, and increased self-reported feelings of power. Forty-two participants.

Cuddy's TED talk about it has been viewed tens of millions of times. Power posing entered interview preparation, executive coaching, and a large number of school assemblies.

In 2015 Eva Ranehill and colleagues ran a replication with two hundred participants — nearly five times the original sample — with a standardized video instruction rather than an experimenter present, and measured the same outcomes. The self-reported feeling of power replicated. The testosterone effect did not. The cortisol effect did not. The behavioral risk-taking effect did not.

In 2016 Dana Carney, the first author of the original paper, posted a statement on her own website saying that she did not believe the effects were real, that she no longer studied the topic, that she discouraged others from doing so, and setting out in detail the analytic decisions in the original work that she had come to view as problematic: a small sample, dropped participants, flexible outcome analysis, and hypothesis-driven choices made after seeing the data.

This is one of the most direct public recantations by a lead author in the history of the field, and it is why the chapter is worth reading even if you never intend to stand with your hands on your hips.

The subsequent literature has not been simply negative. Multi-lab work, including adversarial collaborations involving Cuddy, has found reasonably consistent evidence for one thing: expansive posture produces a modest increase in self-reported felt power. What has not survived is the hormonal mechanism, and with it the causal story that made the finding remarkable — that a posture could reach into the endocrine system and change what a person is, rather than merely how they feel for a few minutes.

So the defensible residue is: posture affects your mood a little, and affects how others read you, which was already known from the dominance-signaling literature. That is a real finding. It is not a miracle, and it is not what was sold.

Two general lessons come out of this episode and both are worth more than the specific result.

The first concerns effect sizes and sample sizes. A striking effect found in forty-two people, published because it was striking, is exactly the profile of a result that will shrink or vanish under replication. This is not a scandal about individuals; it is the predictable output of a publication system that rewarded novelty and did not require replication, and it produced a great many findings in this literature that are now uncertain.

The second is about the pipeline from a laboratory to a stage. The gap between what the study measured and what the talk claimed was large at every step, and each retelling made the claim rounder and more actionable. By the time it reaches a corporate training slide there is no uncertainty left in it at all — which is a good general reason to treat any behavioral claim that arrives fully confident as having been through that pipeline.