Feeling the Future

The design of these experiments is their whole rhetorical point. Bem took procedures every social psychologist recognises, took nothing else from parapsychology, and reversed the time order. In one, participants studied a list of words, were tested on recall, and only afterwards did a computer select a random subset of those words for them to practise; the reported finding was that they had recalled more of the words they would later practise. In another, a variant on a standard priming task, the prime appeared after the response. In another, participants guessed which of two curtains on a screen concealed an image, with the image's position selected by the computer only after the guess. Across nine experiments and more than a thousand participants, eight returned statistically significant results in the direction predicted.

The paper appeared in the Journal of Personality and Social Psychology in 2011, one of the most selective outlets in the discipline, after normal peer review. Bem's framing was that if these effects are real they matter enormously, and that if the methods are sound the results must be reported regardless of prior plausibility. The journal's editor later defended the decision to publish on the ground that the paper met the field's standards.

The replication record is unambiguous and is the most important fact on this page. Stuart Ritchie, Richard Wiseman and Christopher French ran three direct replications of the retroactive recall experiment in 2012, found nothing, and had considerable difficulty publishing the null result, an experience that itself became evidence in the wider argument about what journals accept. Galak, LeBoeuf, Nelson and Simmons reported seven further failed replications together with a meta-analysis showing no effect.

Alongside the original paper, Wagenmakers and colleagues argued that the statistical approach was inappropriate for a claim of this kind and that a Bayesian reanalysis of Bem's own data gave no evidential support. Bem, with Tressoldi, Rabeyron and Duggan, answered in 2015 with a meta-analysis of ninety experiments reporting a small positive overall effect; critics read that as an accumulation of the same biases rather than as evidence.

What followed was larger than the dispute. Because the methods were standard, the paper became the discipline's worked example of how ordinary practice could yield a false positive, and it is cited throughout the literature that produced pre-registration, larger samples, open data and organised replication efforts. The claim did not survive. The reforms did.