Power Posing

Forty-two participants split into two groups of ~21 is far too small to detect hormonal changes from a two-minute manipulation — salivary hormones vary widely between people and across the day. When underpowered studies do produce significant results, effect sizes are almost certainly inflated: only exaggerated estimates clear the significance bar.

Participants either held expansive or contractive poses; no one sat normally. The study cannot determine whether high-power poses raise testosterone or low-power poses lower it, yet the paper frames results as if high-power poses actively boost hormones. Three outliers (7% of the sample) were removed, and in a study this small each exclusion materially shifts results. No sensitivity analysis shows what happens if they're included. Risk-taking was measured by a single choice — keep two dollars or gamble on a coin flip for four — not a validated measure of risk tolerance. The paper reports four separate tests with no correction for multiple comparisons, and several p-values sit just below 0.05. Large-scale replications have failed to find the hormone effects (Ranehill 2015).

No power analysis justified the sample size. The reported degrees of freedom are inconsistent with the described model: the paper says it controlled for baseline hormone levels, sex, and the other hormone, but the degrees of freedom match a simple two-group comparison with no covariates. Either the covariates weren't included or the reporting contains errors.

Three of the four primary p-values cluster just below 0.05. With four uncorrected tests, the probability of at least one false positive by chance alone is ~19%. A standard correction for multiple comparisons would render most results non-significant. The reported effect sizes — medium to large — are implausibly big for a two-minute posture change on tightly regulated hormone systems. The confidence intervals (unreported but calculable) stretch from negligible to very large, meaning the data are consistent with almost any true effect size. The authors' own follow-up of 49 participants dropped the hormone measures entirely — the expected pattern if those results didn't hold even in their own lab.

The paper treats small hormone fluctuations over 17 minutes in ~20 people per group as evidence that poses produce "power." Hormone levels fluctuate naturally; a wobble after holding an unusual position doesn't mean the body has entered a dominance state. The shorthand "the dominance hormone testosterone" and "the stress hormone cortisol" obscures how much interpretation these labels carry.

Risk tolerance was measured with a single coin flip for two dollars. The expected payoff of gambling equals keeping the money, so the "risky" option isn't risky in any economic sense. A single binary choice cannot measure a psychological disposition.

The central inference — that posing produces "the same physiological and behavioral profile" as actual power — compares a two-minute exercise to the effects of leading organizations and controlling resources over years. Directionally similar results on three measures in 21 people per group do not establish equivalence. The study cannot distinguish bodily feedback from a simpler explanation: participants placed in obviously dominant poses may have thought about dominance and responded accordingly.

The paper's narrative does more work than its data. The opening — peacocks, cats, chimpanzees, boardroom executives — frames body posture as a deep evolutionary channel for power before any evidence appears, making the hypothesis feel inevitable rather than speculative.

The abstract claims posing "caused advantaged and adaptive psychological, physiological, and behavioral changes." "Adaptive" implies real-world benefit, but the study measures hormone fluctuations over minutes and a two-dollar coin flip. The closing paragraphs leap to claims about job interviews, public speaking, health, and well-being — none of which were tested or approximated. The most obvious alternative explanation — that participants arranged into expansive poses simply guessed what responses were expected — is never discussed. The paper defines power as testosterone, cortisol, risk-taking, and feelings of power, measures exactly those things, and concludes it has "produced power" — confirming its framework rather than testing it.

Bad even by 2010 standards: no pre-registration, no shared data, no analysis code, no power analysis, no ethics statement mentioning IRB approval (despite physical touch, saliva collection, and deception), and no funding disclosure. There is no limitations section at all. The discussion moves directly from results to sweeping practical applications without acknowledging the small sample, the outlier exclusions, or the possibility that demand characteristics explain the findings.

The citations are real but consistently stretched beyond what they support. Studies of chronic cortisol differences in baboon hierarchies (Sapolsky 1997) and testosterone responses to winning real competitions (Booth 1989) are cited as if they predict that sitting with feet on a desk for a minute will shift hormones. The embodied cognition citations — studies showing that smiling increases enjoyment or fist-clenching increases self-rated power — are presented as strong precedent, but none demonstrated hormonal changes. A key claim — high testosterone plus low cortisol predicting "leadership abilities" — rests on an unpublished manuscript (Mehta and Josephs 2010, "in preparation"). The literature on demand characteristics — directly relevant to a study where an experimenter physically poses participants and then asks how powerful they feel — is entirely absent.

The paper discloses no financial conflicts and has no industry funding. The pressure is structural: academic publishing rewards novel, clean, positive findings, and this study's design — tiny sample, no pre-registration, flexible analysis, sweeping conclusions — is what that incentive structure produces. The conclusions extend far beyond what 42 participants can support, claiming "real-world, actionable implications" for job interviews, health, and chronically powerless populations. The authors' divergence is instructive: Carney publicly disavowed the findings in 2016, while Cuddy, who had built a prominent brand around power posing (including one of the most-viewed TED talks), continued defending a narrower version.