Cornell University professor's p-hacking brag led to 18 retractions, $22M program lost

Brian Wansink was a professor at Cornell University and director of its Food and Brand Lab, a research group that studied how people make food choices. His work was widely cited and became the basis for the Smarter Lunchrooms Movement, a USDA-funded program that redesigns school cafeterias to nudge children toward healthier eating. The program was implemented in over 30,000 schools with $22 million in federal funding.

In January 2017, Wansink published a blog post titled 'The Grad Student Who Never Said No' in which he described asking a graduate student to salvage results from a study that had found nothing. The student ran the data through multiple analyses and produced five publishable papers. Critics recognized this as p-hacking — running enough statistical tests until one comes back significant. Researchers began scrutinizing his work and found problems across dozens of papers, including duplicated data, impossible values, and unsupported conclusions.

Cornell initially found errors but no misconduct in April 2017. By September 2018, a deeper investigation concluded Wansink had committed scientific misconduct: data falsification, failure to preserve data, inappropriate authorship attribution, and dual publication. At least 18 papers were retracted, including 6 by JAMA and 2 by the Annals of Internal Medicine. Wansink was removed from research and teaching and resigned from Cornell effective June 2019.

The blog post confessing to p-hacking triggered scrutiny that exposed falsified data, ending the Smarter Lunchrooms program and Wansink's career.

Wansink built a reputation on novel food psychology findings, but the pressure to keep producing results led him to let a grad student mine null data until something significant emerged.

The blog post that triggered the investigation was written as a celebration of persistence — Wansink did not realize he was publicly confessing to p-hacking, which is statistical misconduct.

Cornell's initial investigation was shallow and failed to catch the scale of the problem, giving Wansink another year to publish before the deeper investigation found fraud.

A researcher who treats null results as a data-mining opportunity will eventually be caught, and the institution that tolerated the practice pays the price in reputation and funding.