Imagine the establishment publishing the paranormal — not in a fringe newsletter, but in the most prestigious journal its discipline has. That is what happened in 2011, when Daryl J. Bem, an eminent social psychologist at Cornell, published 'Feeling the Future: Experimental Evidence for Anomalous Retroactive Influences on Cognition and Affect' in the Journal of Personality and Social Psychology. The inverted twist is the one most coverage misses: the scandal was never really about whether psychic precognition is real. It was about what Bem's paper accidentally revealed regarding how normal, respected, rule-following science was actually being done.

What Bem did was clever. He took well-established psychology experiments and ran them backwards in time. In one, students were shown a list of words, given a recall test, and then — after the test — were shown a random subset of the words to 'practice.' Ordinarily practicing words before a test helps you recall them. Bem reported that practicing them after the test still improved earlier recall, as if the future practice reached back. In another, erotic images were placed behind one of two curtains on a screen, their position randomized after the participant guessed; participants, he reported, guessed the erotic location at above-chance rates, as if anticipating a randomization that had not yet happened. Across nine experiments and more than a thousand participants, eight reached conventional statistical significance.

The evidence, on its face, looked like textbook science: large samples, standard methods, peer review at a top journal, p-values below .05. And that is precisely the point that made it dangerous. If the exact same methods that produce 'serious' psychology produce statistical proof of seeing the future, then either the future is visible — or the methods are broken. The skeptical community took the second horn immediately and hard. Critics led by Eric-Jan Wagenmakers re-analyzed Bem's data using Bayesian statistics and showed that the evidence was far weaker than the p-values implied, and that with reasonable priors it barely moved the needle at all.

Then came the part that actually mattered. In 2012 Stuart Ritchie, Chris French, and Richard Wiseman ran direct replications of Bem's retroactive-recall experiment — and found nothing. No precognition. Tellingly, the journal that published Bem's positive result initially declined to even consider the failed replications, on the grounds that it did not publish replications. That editorial stance — positive novel findings welcome, attempts to check them unwelcome — was the smoking gun. It meant the published literature was a filter that let through false positives and blocked the corrections. The Bem affair became the catalyzing example for what is now called the replication crisis: the realization that 'questionable research practices' like flexible analysis, optional stopping, and selective reporting could manufacture publishable proof of literally impossible effects.

The skeptical-but-fair reading is not that Bem cheated. By all accounts he ran his studies sincerely and reported them transparently — which is exactly why the case is so damning. He played by the field's rules and arrived at the impossible, demonstrating that the rules themselves were the problem. The reforms that followed — pre-registration, open data, registered reports, the founding of large replication consortia, Bayesian re-analysis — owe a real debt to a precognition paper. The man who tried to prove psi instead helped prove that mainstream psychology's quality control was broken, and the field is more rigorous today because of it.

The unresolved question is the uncomfortable residue. Later meta-analyses, including one Bem co-authored over ninety experiments, still report a small overall positive effect that critics attribute to the same publication and analysis biases the crisis exposed — a tidy loop where the anomaly and its likeliest mundane explanation point at the same data. Almost no working scientist believes the curtains predicted the future. But 'Feeling the Future' left a genuinely open problem behind it, and it is not about ESP. It is this: a discipline only discovered the depth of its own unreliability because one of its own respected members used its accepted methods to prove something that could not be true. How many ordinary, believed, never-questioned findings sit in the same journals, produced the same way, that simply never tripped the alarm by being absurd enough to check?