A recent post on the DataColada blog argues that Study 2 of Ariely and Wertenbroch’s 2002 article “Procrastination, Deadlines, and Performance: Self‑Control by Precommitment” contains data that were likely tampered with. The original study, published in Psychological Science, reported that participants who faced evenly spaced weekly deadlines corrected far more errors in a proofreading task than those who set their own deadlines or faced a single final deadline. The effect was large: the evenly spaced‑deadline group averaged 136.1 corrections, while the final‑deadline group averaged 71.1 corrections, yielding a Cohen’s d of 2.5 and a correlation of r =.79. The authors of the post note that such an effect size exceeds those of obvious real‑world differences, such as the gender‑height gap (d ≈ 1.8), and is unusually large for a noisy behavioral measure like error detection. They also highlight a distribution with almost no overlap: no participant in the final‑deadline condition made more than 100 corrections, whereas about 90 % of the evenly spaced‑deadline group did. A second red flag concerns duplicate observations. In the data file, many participants recorded identical total correction counts, and further inspection showed that these identical totals were driven by identical counts for each of the three separate proofreading tasks. A screenshot provided in the post shows that 18 of the 20 participants in the “Last Day Deadline” condition exhibited this pattern. The post explains how the data came into the authors’ possession: in 2006 Kyle Hyndman received three Excel files (pilot, Study 1, Study 2) from an address ending in @danariely.com, with file properties indicating they were last saved by Dan Ariely. Hyndman later forwarded those files to the DataColada authors in August 2023, after a lawsuit involving Francesca Gino. Analyses of the files allowed the researchers to reproduce the means and standard errors reported in the original paper. Based on the observed irregularities, the authors conclude that the data for Studies 1 and 2 were tampered with. They have shared their full analysis in two posts, provided the data and code via a ResearchBox repository, and have asked Psychological Science to retract the original article. As of the writing of the post, the retraction request is still under review. The post emphasizes that the allegations are based solely on the presented analyses and invites readers to examine the evidence themselves. It also notes that Klaus Wertenbroch, the other co‑author, never had access to the data files, according to the authors’ understanding.
