Overview

In everyday physics, causes come before effects. Retrocausation flips that: a future event influences the present.^[7] This idea has become useful in parapsychology because some findings are hard to explain any other way. Suppose a participant correctly guesses an image that a computer picks after the guess is recorded. No normal signal could have carried that information forward in time. Retrocausation offers a different explanation: the future event leaves a backward-in-time trace in the person’s mind or body right now.

This is not purely a fringe idea. The American Association for the Advancement of Science has held symposia asking whether precognition and retrocausation fit with quantum theory.^[2] Some theoretical physicists have proposed time-symmetric versions of quantum mechanics that formally allow backward-in-time influence. Researcher Dean Radin has argued that quantum entanglement, which is not limited by ordinary space or time, may offer a path to understanding how future events connect to present responses.^[8]

Not everyone in parapsychology agrees on the explanation, though. Philosopher Stephen Braude has drawn a distinction between a “passive” retrocausal view, where the future simply imprints on the present, and an “active” view, where ordinary-direction ESP or psychokinesis does the work. What most researchers do agree on is that the experimental record needs an explanation. Whether that explanation is retrocausation, some other psi process, or flaws in how the studies were run is still debated.

Key research programs

Bem’s “Feeling the Future” paradigm

Daryl J. Bem, emeritus professor of psychology at Cornell University, built his research strategy around a simple idea: take a well-known psychology effect and run it backward in time.^[9] In a normal psychology experiment, you show someone a stimulus and measure their response. Bem flipped this. He measured the response first, then let a computer randomly pick the stimulus afterward. He then asked: did the future stimulus shape the earlier response?

His 2011 paper in the Journal of Personality and Social Psychology reported nine such experiments.^[1] They covered four time-reversed effects: approaching erotic images before seeing them, avoiding negative images before seeing them, being primed by words shown afterward, and recalling words better when they were practiced after the recall test. Eight of the nine experiments returned statistically significant results. The mean effect size across all nine was d = 0.22. To put that in plain terms: the average score was about one-fifth of a standard deviation above what chance would predict. That is a small effect, but it showed up consistently.^[1]

The paper’s publication in a top journal sparked intense debate, both about the findings and about how peer review works.^[10] Bem encouraged others to replicate his work and shared all his materials on request.^[2] A 2015-16 meta-analysis he co-authored with Tressoldi, Rabeyron, and Duggan pulled together 90 experiments from around the world. Even when Bem’s own nine studies were left out, the 69 independent replications from 33 labs in 14 countries still showed a significant combined effect. The combined z-score was 4.16, with p = 1.1 x 10 to the power of -5. That p-value means: if there were no real effect, results this strong would occur by chance only about 1 in 90,000 times. The Bayes Factor for those independent replications was 3,853, far above the conventional threshold for what statisticians call “decisive evidence.”^[2]

Presentiment research

A separate line of retrocausation research skips conscious decisions entirely. It watches the body instead. In a presentiment experiment, a participant sits quietly while sensors record their skin conductance, heart rate, or pupil size. A computer then randomly selects and displays either an emotionally arousing image or a calm, neutral one. The key question is: does the body start reacting differently before the computer has made its random choice?^[11]

Dean Radin pioneered this work in the mid-1990s. His results were later replicated by psychologist Dick Bierman at the University of Amsterdam and by other labs using heart rate and pupil dilation as measures.^[12] A 2012 meta-analysis by Mossbridge, Tressoldi, and Utts combined results from seven independent laboratories. It found a presentiment effect size of about 0.21. That is almost identical to the effect sizes seen in Bem’s behavioral experiments and in the wider meta-analysis of those studies.^[2] Supporters point to this match as meaningful. Two very different methods, one behavioral and one physiological, are producing nearly the same size effect. That convergence is hard to explain as a shared accident of method.

Retrocausation has also been proposed as a unifying framework for other psi phenomena. Some researchers argue it could cover telepathy, clairvoyance, and even micro-PK, since it is logically impossible to fully close the “precognition loophole” in any psi experiment.^[13] Some theorists have also interpreted synchronicity and aspects of deja vu through a retrocausal lens.

Skeptical critiques

Critique 1: The statistics do not hold up under closer scrutiny

Wagenmakers, Wetzels, Borsboom, and van der Maas reanalysed Bem’s data using Bayesian statistics. They concluded the evidence was not strong enough to reject the null hypothesis (the null hypothesis being that there is no real effect). They also argued that the standard p-value method Bem used tends to find positive results too easily.^[5]

Proponents respond that the 2015-16 meta-analysis used multiple statistical methods, including Bayesian ones, and still found a significant effect. The Bayes Factor of 3,853 for independent replications alone is well above any standard threshold for strong evidence.^[2]

Evidence strength: The statistical debate is genuine. Both sides use legitimate tools. The Bayesian reanalysis raised real concerns, but the meta-analytic Bayes Factor is a direct response using the same framework.

Critique 2: Independent replications have failed

Galak, LeBoeuf, Nelson, and Simmons published seven replication attempts in the same journal that ran Bem’s original paper. They found no evidence for precognition.^[14] Ritchie, French, and Wiseman also failed to replicate Bem’s retroactive recall effect. Their struggle to get a null result published drew attention to publication bias in psychology more broadly.^[5]

The 2015-16 meta-analysis addressed publication bias directly. It estimated that 544 unreported null-result studies would be needed to wipe out the overall effect. That number is considered implausibly large. Seven of eight statistical tests applied to the database found no significant sign of p-hacking or selective reporting.^[2]

Evidence strength: Failed replications are a serious concern. The meta-analytic file-drawer analysis does not eliminate the problem, but it sets a high bar for how many hidden null results would need to exist.

Critique 3: Retrocausation is conceptually problematic even before looking at data

Some philosophers of science argue that no data strictly require retrocausation as an explanation. Every claimed example of retrocausation outside parapsychology is either a thought experiment or a contested reading of quantum equations. The parapsychological data could, in principle, be explained by ordinary forward-causal psi mechanisms. Critics also raise concerns about researcher degrees of freedom and the possibility that significant results come from questionable research practices rather than real effects.

Proponents note that the AAAS has taken the theoretical questions seriously enough to hold dedicated physics-and-psi symposia.^[2] The shift in parapsychological theory toward time-symmetric, biologically grounded models gives retrocausation a more naturalistic footing than it once had.^[6]

Evidence strength: The conceptual objection is legitimate. Retrocausation is a stronger claim than most psi explanations. The theoretical work is real but not yet settled.

Responses and evaluation

The cumulative statistical record is substantial. The 2015-16 meta-analysis ran eight different tests for bias and p-hacking. Seven of the eight found no significant problem. A P-curve analysis, which estimates the true effect size by looking at the distribution of p-values, put the real effect for independent replications at 0.20 to 0.24.^[2] That range matches the presentiment effect size of about 0.21 from the physiological studies. Two entirely different methods pointing to the same number is considered by supporters to be unlikely if both results are just noise.

The broader shift in how parapsychologists think about psi also matters here. The old view treated psi as a mysterious non-physical signal passing between minds. The newer view treats psi as a correlation between future and present states, possibly grounded in biology.^[6] That framing fits retrocausation naturally. It also fits better with the kind of physics that the AAAS symposia have been exploring.^[2]

As of 2026, the field is genuinely split. The statistical record is hard to dismiss on effect-size grounds alone. But conceptual objections and replication failures keep skepticism alive. Both critics and supporters now recommend pre-registering studies before running them. Pre-registration means researchers publicly commit to their methods and analysis plan before collecting data. This removes the ability to adjust the analysis after seeing results. Future pre-registered, multi-site programs are expected to be the clearest test yet of whether these effects hold up under the strictest conditions.