Three retractions in one day, three different failures

Three papers pulled, three different failures

Today's ledger carries three retractions. The only thing they share is the word.

Psychological Science retracted Dan Ariely and Klaus Wertenbroch's 2002 paper on deadlines and procrastination, cited nearly 1,000 times, after Data Colada found duplicated values and altered grades in both experiments. A replication published in the same journal in July had already failed to reproduce the result. It is Ariely's second retraction.

Nature retracted a June 2025 paper that attributed roughly 40% of the worsening in global drought severity to rising atmospheric evaporative demand, after the Oxford-led authors found a timestepping coding error in their own work. The paper had been cited 337 times. The authors say the corrected number will be slightly lower and the conclusion unchanged.

Remote Sensing retracted the 2022 synthetic aperture radar paper by Filippo Biondi and Corrado Malanga that claimed 20 hidden corridors, rooms and ramps inside the Great Pyramid of Khufu, citing serious methodological flaws and statistical errors. The authors disagree with the retraction, and the journal made no finding of misconduct.

One is a data integrity finding with a failed replication behind it. One is a bug the authors found and reported in a result they still expect to hold. One is a contested judgment about method, with nobody accused of anything. A reader who counts three retractions in a day and concludes the literature got three times worse has learned nothing about any of the three.

The interval, not the count

The useful number here is not how many papers were pulled but how long each stood. The drought paper was corrected roughly fifteen months after publication, by the people who wrote it. The deadlines paper stood for 24 years and nearly 1,000 citations, and came down only after outside forensic work and a failed replication. The pyramid paper stood four years and is still disputed by its authors.

Those are three different systems failing at three different speeds, and only the first one is the machine working as advertised.

The papers that say what they cannot say

Set against that, most of today's live results arrive carrying their own limits, stated by the authors rather than extracted from them.

• Insilico Medicine, with Harvard Medical School and Peking University, applied six independently built proteomic aging clocks to serum from 42 patients in its phase 2a rentosertib trial and found roughly 3 to 4 years of predicted biological age reversal at week 4 on 30 mg twice daily. With 42 patients whose lung fibrosis was also being treated, the study cannot separate slowed aging from treated disease, and the authors say so.
• Cedars-Sinai reported overall survival of up to 22 months in an early-phase trial adding L-glutamine to first-line chemotherapy for stage 4 pancreatic cancer, against the 8 to 13 months typical of chemotherapy alone. The investigators say it needs a larger trial before it means anything for practice.
• Kavli IPMU identified four candidate pairs of little red dots in JWST imaging from 12.5 to 12.8 billion years ago. They are photometric candidates from pixel-level colour selection, not confirmed black hole binaries.
• NICER and the Einstein Probe resolved a 374 second X-ray source into a pair of white dwarfs at an estimated chirp mass of 0.43 solar masses. The distance is unknown, so the amplitude a detector like LISA would see cannot yet be calculated.
• Penn State and the University of North Carolina at Chapel Hill linked marine heatwave exposure across 9,800 coastal sites in 36 low- and middle-income countries to a 5.4% increase in the odds of child mortality per standard deviation of heatwaves over two years. The design is observational and establishes association, not cause.

None of those hedges are hidden in a limitations paragraph nobody reads. They are the claim.

A negative result is not a failure of the record

The same day, NSABP B-59/GeparDouze missed its primary endpoint: adding atezolizumab to neoadjuvant chemotherapy in 1,550 patients with stage II-III triple-negative breast cancer produced an event-free survival hazard ratio of 0.80 at P = 0.083, a 3.3% difference at four years, with immune-related adverse events in 27.6% of the atezolizumab arm against 11.4% on placebo.

That trial answered its question. It is the healthiest item on today's list, and it will never be retracted, because it never claimed more than it measured.