Data Colada Flags Likely Fabrication in Ariely's Famous Procrastination Study

A forthcoming Psychological Science paper failed to replicate Study 2 of Ariely and Wertenbroch’s heavily cited 2002 article on deadlines and self-control — the one claiming people proofread far better under evenly spaced, externally imposed deadlines than under a single end-of-period deadline. That failure prompted the Data Colada team to reexamine the original data, which they obtained via replication author Kyle Hyndman, who had received the Excel files from Dan Ariely’s email address back in 2006. Their conclusion: the data in both Study 1 and Study 2 show signs of tampering.

The statistical case rests on several red flags. The reported effect is implausibly large — a Cohen’s d of about 2.5 for proofreading performance, meaning the evenly-spaced and last-day groups sat 2.5 standard deviations apart, a gap bigger than the effect of gender on height and larger than many manipulation checks that are true almost by definition. With only 20 participants per condition, the distributions barely overlap. A second flag is duplicate observations: multiple participants recorded not just identical error-correction totals but identical counts on each of the three separate documents, a pattern that does not occur naturally.

The episode plays out against Ariely’s broader credibility crisis and the Francesca Gino litigation (Gino sued Data Colada for $25 million). When the replication authors asked Ariely in 2024 to review an analysis of the Study 2 file, he refused and suggested the files might not be the real data, providing none himself. After being shown the new findings, both Ariely and Wertenbroch — who appears never to have had access to the data — asked the journal to retract the paper, a process now underway.