Daryl Bem’s Feeling the Future: Can Experiments Actually Prove Precognition?

Imagine you’re sitting at a computer screen. An image is about to appear on one side of the screen. You don’t know which side. The computer doesn’t know which side. Nobody knows which side, because the side hasn’t been determined yet. You make your guess. The computer then randomly selects which side the image will appear on. Your guess and the random selection are completely independent, separated by the laws of probability.

Now imagine that across 1,000 participants and 9 experiments, people guess correctly more often than chance allows. Not by much, just slightly more than 50%. But consistently. Across every test. That is the claim made by Daryl J. Bem in 2011, in a paper that shook psychology to its foundations.

Bem did not claim to prove that psychic powers are real. He claimed something more unsettling: that under carefully controlled laboratory conditions, using standard psychological methods, human behavior showed statistically significant evidence of being influenced by future events. He called it precognition, and he published it in one of psychology’s most respected journals.

The paper did not change minds about precognition. It changed minds about psychology. It became the spark that lit the replication crisis, the reform movement that has reshaped how psychologists do and publish research. The question is not whether precognition is real. The question is: what happens when a respected scientist publishes an extraordinary claim using ordinary methods, and nobody can replicate it?

TL;DR

What Is Precognition?

Precognition is the claim that humans can become aware of events before they happen. Not through prediction, inference, or logical deduction, but through some direct cognitive process that bypasses normal information channels (according to Wikipedia).

The word comes from the Latin prae (before) and cognitio (acquiring knowledge). It is distinct from premonition, which is a vague feeling of impending disaster. Precognition, if it existed, would be a specific cognitive ability: the capacity to perceive future events with accuracy that exceeds chance (according to Wikipedia).

The concept violates one of the fundamental principles of physics: causality. In our current understanding of the universe, effects follow causes. An event cannot influence something that happened before it. Precognition, by definition, requires that future events somehow influence present cognition. This is why the scientific establishment considers precognition pseudoscience (according to Wikipedia).

But “considered pseudoscience” is not the same as “proven false.” And this is where the story gets interesting.

The term “precognition” first appeared in the 17th century, but serious experimental investigation did not begin until the 1930s, when J.B. Rhine at Duke University began using card-guessing experiments to test for extrasensory perception (according to Wikipedia). Rhine’s early results appeared positive, but his methods were later shown to be flawed, and subsequent researchers using more rigorous procedures were unable to reproduce his findings.

In 1927, British engineer J.W. Dunne published An Experiment with Time, in which he recorded and analyzed his own dreams, concluding that approximately 10% appeared to contain elements of future experience (according to Wikipedia). Dunne’s book was widely read and influenced writers including H.G. Wells, J.B. Priestley, and Vladimir Nabokov. But it was anecdotal and unsystematic.

For decades, precognition research remained in the margins. Then, in 2011, a social psychologist at Cornell University decided to test it using the same methods he had used for decades of mainstream research. The results would either prove precognition or prove something far more troubling about psychological science itself.

Who Is Daryl Bem?

Daryl J. Bem was born on June 10, 1938. He received a BA in physics from Reed College in 1960 and began graduate work in physics at MIT. But the civil rights movement was underway, and Bem became fascinated by how attitudes toward desegregation were changing. He switched fields, earning a PhD in social psychology from the University of Michigan in 1964 (according to Wikipedia).

Bem went on to teach at Carnegie Mellon, Stanford, Harvard, and Cornell. He retired from Cornell in 2007 as a professor emeritus. He is known for two major contributions to mainstream psychology: the self-perception theory of attitude change, proposed in 1972, and the exotic becomes erotic theory of sexual orientation (according to Wikipedia).

Self-perception theory challenged Leon Festinger’s cognitive dissonance theory by proposing that people infer their attitudes from their own behavior, rather than experiencing dissonance when their behavior conflicts with their attitudes. The two theories appeared contradictory but addressed different mechanisms: dissonance theory explains attitude change, while self-perception theory explains attitude formation (according to Wikipedia).

Bem was a serious scientist with a serious career. He was not a fringe figure, not a psychic promoter, not an outsider. He was a tenured professor at an Ivy League university, with decades of mainstream research. This is what made his 2011 paper so disruptive.

Bem also had a longstanding interest in parapsychology. He defended the ganzfeld experiment as evidence for psi phenomena, co-authoring a review with Charles Honorton in 1994 (according to Wikipedia). But the ganzfeld was a small corner of his career. His 2011 paper on precognition was something different: a direct, large-scale, methodologically rigorous attempt to demonstrate psi using the tools of mainstream social psychology.

He was also, as it happens, an amateur magician. Bem had studied stage magic since childhood, and he used his knowledge of illusion and deception to design experiments that he believed were immune to both. He understood how people could be fooled, and he designed his studies to prevent it (according to Wikipedia).

“Feeling the Future”: The Nine Experiments

In 2011, the Journal of Personality and Social Psychology (JPSP) published Bem’s paper, “Feeling the Future: Experimental Evidence for Anomalous Retroactive Influences on Cognition and Affect.” JPSP is one of the most prestigious journals in social psychology, published by the American Psychological Association. It does not publish fringe science. Or at least, it had not before (according to APA).

The paper reported nine experiments with a total of over 1,000 participants. All used standard psychological methods: random assignment, double-blinding, statistical testing. The only unusual element was the direction of the hypothesis. Bem was not testing whether past experience influences present behavior. He was testing whether future events influence present behavior.

The mean effect size across all nine experiments was d = 0.22. Eight of the nine experiments yielded statistically significant results. The probability of getting 9 significant results out of 10 tests, assuming no real effect, was approximately 1 in 53 billion (according to Replicability-Index).

Here is what the nine experiments actually tested:

Experiment 1: Precognitive Detection of Erotic Stimuli. Participants sat in front of a computer screen. Two curtains appeared on the screen. Behind one curtain was an erotic image; behind the other was a neutral or negative image. Participants were asked to guess which curtain hid the erotic image. The computer randomly selected which curtain would contain the erotic image after the participant had already guessed. Across 100 participants, participants selected the erotic curtain at a rate of 53.1%, significantly above the 50% chance level (p = .01). For non-erotic stimuli, the hit rate was at chance.

Experiment 2: Precognitive Avoidance of Negative Stimuli. The same setup as Experiment 1, but with negative images (spiders, snakes) instead of erotic images. Participants avoided the negative images at a rate of 52.2%, above chance (p = .03).

Experiment 3: Retroactive Priming. Participants were shown a prime word (positive or negative) followed by a target word. They had to categorize the target word as positive or negative as quickly as possible. The prime was randomly selected by the computer after the participant had already responded. Participants responded faster when the prime matched the valence of the target word, as if the future prime had influenced their reaction time. Effect size: d = 0.27, p = .01.

Experiment 4: Retroactive Habituation. Participants were shown a series of images and rated their preference for each. The computer then randomly selected which images would be shown again. Participants showed reduced preference for images that would later be repeated, as if the future repetition had already habituated them. Effect size: d = 0.20, p = .02.

Experiment 5: Diamante Retroactive Habituation. A replication of Experiment 4 with erotic images. Effect size: d = 0.28, p = .003.

Experiment 6: Retroactive Facilitation for Recall. Participants studied a list of words and then took a surprise recall test. The computer then randomly selected a subset of words that would be used in a practice session. Participants recalled the words that would later be practiced at a higher rate than the words that would not be practiced. Effect size: d = 0.16, p = .04.

Experiment 7: Retroactive Facilitation for Recall II. A replication of Experiment 6 with a different word list. Non-significant (d = 0.09, p = .19). This was the one experiment that did not reach significance.

Experiment 8: Retroactive Induction of Boredom. Participants viewed a series of images and rated how boring they found them. The computer then randomly selected which images would be shown again in a “practice” session. Participants rated images that would later be repeated as more boring, as if the future repetition had already bored them. Effect size: d = 0.16, p = .05.

Experiment 9: Retroactive Facilitation of Recall III. Another recall experiment, this time using a practice session that involved rehearsing some words and not others. Participants recalled more of the words that would later be rehearsed. Effect size: d = 0.21, p = .01.

The pattern was consistent: participants behaved as if future events had already influenced them. The effects were small but persistent. The methods were standard. The statistics were conventional. And the claim was extraordinary.

The Journal’s Unusual Decision

JPSP’s decision to publish the paper was itself controversial. The editors, Charles Judd and Bertram Gawronski, published an accompanying editorial explaining their reasoning. They wrote that they found the paper methodologically sound and that the results, while surprising, met the journal’s standards for publication. They emphasized that they expected and hoped for replication attempts from other researchers (according to Replicability-Index).

The editorial also noted that the paper had been reviewed by four independent reviewers, two of whom recommended publication and two of whom did not. The editors decided to publish despite the split decision, in part because they believed the paper would stimulate productive scientific debate (according to Wikipedia).

They were right about the productive debate. They may have been wrong about what it would prove.

The Meta-Analysis: 90 Experiments

In 2015, Bem joined with Patrizio Tressoldi, Thomas Rabeyron, and Michael Duggan to publish a meta-analysis of 90 experiments on precognition. The meta-analysis, published in F1000Research, included Bem’s own studies plus 81 experiments conducted by other researchers between 1978 and 2010 (according to PubMed).

The meta-analysis reported a mean effect size of d = 0.20, 95% CI = [0.13, 0.29], with a combined z-score of 5.30 and a p-value of 5.7 x 10^-8. The authors calculated that 87 unpublished null studies would be needed to reduce the effect to trivial levels, suggesting that publication bias alone could not explain the result (according to F1000Research).

The meta-analysis also found that the effect was larger for erotic stimuli (d = 0.28) than for non-erotic stimuli (d = 0.17), and that the effect was not moderated by the number of sessions, sample size, or type of analysis (according to PMC).

This meta-analysis is the strongest statistical evidence for precognition. But it is also the most criticized. Critics argue that the 90 experiments include many conducted by Bem himself or by researchers who share his theoretical commitments, that the inclusion criteria were too permissive, and that the statistical methods used to assess publication bias are themselves subject to debate (according to Replicability-Index).

The meta-analysis is also vulnerable to a more fundamental critique: if the individual studies are biased by questionable research practices, a meta-analysis of biased studies will produce a biased result. Averaging biased data does not unbias it.

The Replication Failures

The first independent replication attempts came quickly. In 2012, Stuart Ritchie, Richard Wiseman, and Chris French conducted three studies with a total of 150 participants, following Bem’s protocol. They found no significant effect (according to Columbia University).

The sample size was small, though, and Bem’s supporters argued that 150 participants were insufficient to detect a small effect. The real blow came from Jeff Galak, Robyn LeBoeuf, Leif Nelson, and Joseph Simmons, who conducted seven studies with a combined sample of 3,289 participants. Their average effect size was d = 0.04, not significantly different from zero. They found that the only significant moderator was whether the experiment was conducted by Bem or not (according to Replicability-Index).

This was devastating. An effect size of d = 0.04 is invisible to the naked eye. It is the statistical equivalent of noise. And with 3,289 participants, the study had far more power than Bem’s original experiments to detect a real effect if one existed.

Other replication attempts followed. In 2012, Jeff Galak and Leif Nelson conducted three failed replications of Bem’s retroactive facilitation of recall experiment, published in PLoS ONE. Their combined effect size was d = 0.01, with a non-significant p-value (according to Columbia University).

In 2018, psychologist Ulrich Schimmack obtained Bem’s original data and discovered that neither of two datasets Bem provided reproduced the results reported in the 2011 paper. The data Bem shared did not match the data he published. Schimmack wrote a formal letter to JPSP’s editor calling for retraction (according to Replicability-Index).

In 2020, JPSP editor Shinobu Kitayama responded to Schimmack’s letter. The paper was not retracted (according to Replicability-Index).

The Transparent Psi Project: The Definitive Replication

In 2023, the most rigorous replication attempt yet was published. The Transparent Psi Project, led by Zoltan Kekecs of Lund University and an international team of 30 co-authors, aimed to replicate Bem’s Experiment 1 (the erotic image detection test) under the strictest possible methodology (according to Royal Society Open Science).

The study was published in Royal Society Open Science. It used pre-registration: the researchers committed to their analysis plan before collecting data. It used Registered Reports: the journal agreed to publish the results regardless of outcome, eliminating publication bias. It used a large sample. And it used both believers in psi and skeptics as experimenters, to rule out the possibility that experimenter expectancy effects explained Bem’s results (according to The Skeptic).

The result was unambiguous: no evidence for precognition. The hit rate was at chance. The effect size was indistinguishable from zero. The paper concluded: “The failure to replicate previous positive findings with this strict methodology indicates that it is likely that the overall positive effect in the literature might be the result of recognized methodological biases rather than ESP” (according to Royal Society Open Science).

This was not a small study run by hostile skeptics. It was a large, international collaboration that included both believers and skeptics, published in a top journal, with pre-registered analysis plans and Registered Reports. If precognition existed, this study should have found it. It did not.

The Questionable Research Practices Problem

The replication failures raised a question that went beyond precognition: how did Bem get such strong results in the first place? The answer, or at least the most plausible explanation, involves questionable research practices.

Questionable research practices (QRPs) are methods that inflate the probability of finding statistically significant results without being outright fraud. They are common in psychology and were common before the replication crisis made them visible. In 2012, researchers John, Loewenstein, and Prelec surveyed over 2,000 psychologists and found that the majority admitted to at least one QRP (according to Replicability-Index).

Ulrich Schimmack’s analysis of Bem’s data identified several specific issues:

Stopping data collection early. Bem’s studies aimed for a minimum number of participants but did not pre-register a stopping rule. If the data looked significant at an interim check, Bem could have stopped collecting data. If it did not look significant, he could have continued collecting data until it did. This practice inflates false positive rates substantially.

Selective reporting. Bem reported nine experiments that worked and one that did not. But how many experiments did he run that he did not report at all? If Bem ran 20 experiments and published only the 9 that reached significance, the publication rate would be 45%, which is consistent with what QRPs can produce when there is no real effect (according to Replicability-Index).

p-hacking. When there are multiple ways to analyze data, researchers can try each and report only the one that gives a significant result. Bem’s studies had some analytical flexibility, though the degree to which Bem exploited this is debated (according to Replicability-Index).

None of these practices constitute fraud. They are, or were, normal parts of how psychologists conducted research. But they produce false positives at a rate far higher than the 5% that the statistical threshold assumes. A field where QRPs are common will produce a steady stream of significant findings for effects that do not exist.

Bem’s paper exposed this problem by applying QRPs to a phenomenon that virtually no psychologist believed was real. If QRPs could produce significant results for precognition, they could produce significant results for anything. And if they had been producing significant results for everything else in psychology, then a lot of published psychology was wrong.

Bem’s Own Words

In 2017, journalist Daniel Engber interviewed Bem for Slate. The interview produced a remarkable quote:

I’m all for rigor, but I prefer other people do it. I see its importance, it’s fun for some people, but I don’t have the patience for it. If you looked at all my past experiments, they were always rhetorical devices. I gathered data to show how my point would be made. I used data as a point of persuasion, and I never really worried about, “Will this replicate or will this not?” (Daryl J. Bem, in Engber, 2017, reported by Replicability-Index)

This is the paper’s author, admitting in plain language that he used data as persuasion rather than as evidence. He did not say the data was fabricated. He did not say the experiments were fraudulent. He said that he did not worry about whether his results could be replicated, because replication was not the point. The point was persuasion.

This admission does not prove that precognition is not real. It proves that Bem’s paper, as published, cannot be treated as reliable evidence for precognition. The author himself does not stand behind it in the way that scientific evidence requires.

The Replication Crisis: What Bem Really Started

The legacy of “Feeling the Future” has nothing to do with precognition. It has to do with the reliability of psychological science.

Before Bem’s paper, the replication crisis was brewing but had not yet erupted. Researchers knew that QRPs were common. They knew that publication bias favored positive results. They knew that many published findings were fragile. But they had not yet confronted the full implications.

Bem’s paper forced the confrontation. Here was a study that met every existing standard of psychological research: large sample, standard methods, peer-reviewed journal, statistically significant results. And the claim was absurd. Precognition violates the laws of physics. If this study was valid, then everything we know about the physical universe was wrong. If this study was not valid, then the methods that produced it could not be trusted to produce valid results for anything.

The answer, as the replication failures demonstrated, was that the methods were not trustworthy. QRPs, publication bias, and analytical flexibility had been producing false positives in psychology for decades. Bem’s paper was just the most visible example (according to Royal Society Open Science).

The response was transformative. Psychologists began adopting pre-registration, where researchers commit to their analysis plan before collecting data. Registered Reports, where journals agree to publish studies regardless of outcome, became more common. Open data, where researchers share their raw data for independent verification, became a standard expectation. Replication studies, once considered unglamorous, became valued contributions (according to Royal Society Open Science).

Bem did not intend to start a revolution. He intended to prove precognition. But the revolution happened anyway, and it happened because his paper made the cost of QRPs visible in a way that no methodological critique ever could.

The Case Against

Intellectual honesty requires presenting the strongest objections. Here is what the skeptical literature says about Bem’s precognition research, and where the critique is strongest.

1. The replication failures are definitive. Galak et al. (2012) tested 3,289 participants and found d = 0.04, not significantly different from zero. The Transparent Psi Project (2023) used the strictest methodology available and found nothing. Ritchie, Wiseman, and French (2012) found nothing with 150 participants. Galak and Nelson (2012) found nothing with three studies. When independent researchers cannot reproduce a finding, the finding does not stand (according to Replicability-Index).

2. Bem’s own data does not support his claims. Ulrich Schimmack obtained Bem’s original data in 2015 and again in 2018. Neither dataset reproduced the results reported in the 2011 paper. The data Bem shared did not match the data he published. This is not a matter of interpretation or statistical method. The numbers do not add up (according to Replicability-Index).

3. Bem admitted to using data as persuasion. In his 2017 interview with Slate, Bem said he “used data as a point of persuasion” and “never really worried about” replication. This is not the statement of a researcher who stands behind his findings. It is the statement of a researcher who used data to make an argument, not to test a hypothesis (according to Replicability-Index).

4. Precognition has no proposed mechanism. No physical theory explains how future events could influence present cognition. Quantum mechanics does not provide one. Classical physics does not provide one. Relativity does not provide one. The absence of a mechanism does not prove an effect is impossible, but it does mean that the evidence for the effect must be overwhelming. Bem’s evidence is not overwhelming. It is, at best, suggestive, and it has not survived replication (according to Wikipedia).

5. The meta-analysis is contaminated. Bem’s 2015 meta-analysis of 90 experiments includes many studies conducted by Bem himself or by researchers with theoretical commitments to psi. If the individual studies are biased by QRPs, the meta-analysis is biased too. Garbage in, garbage out. The meta-analysis also uses statistical methods for assessing publication bias that are themselves debated among statisticians (according to Replicability-Index).

6. The erotic stimulus effect is suspicious. Bem’s strongest results were for erotic images. Why would precognition be specifically tuned to erotic content? The most parsimonious explanation is that erotic images are more engaging and that Bem’s stopping rules or analytical choices were more likely to produce significant results when the stimuli were more engaging, not that the future was more detectable when it involved sex.

7. The paper’s legacy proves its weakness. “Feeling the Future” is cited far more often as an example of what is wrong with psychological science than as evidence for precognition. The paper did not advance our understanding of precognition. It advanced our understanding of research methodology. When a study’s primary contribution is demonstrating the unreliability of the methods that produced it, the study has failed on its own terms.

The Deep Question

Bem’s paper raises a question that extends far beyond parapsychology: how do we know when scientific methods are working?

Before Bem, the implicit answer in psychology was: if a study is published in a peer-reviewed journal, uses standard methods, and reports statistically significant results, it is probably correct. After Bem, the answer became: maybe not. Peer review does not catch QRPs. Standard methods do not prevent false positives. Statistical significance does not mean what we thought it meant.

This is the real lesson of “Feeling the Future.” Not that precognition is real or not real. Not that Bem was right or wrong. But that the tools psychologists used for decades to separate truth from fiction were themselves unreliable. The tools could produce evidence for anything, including things that were almost definitely not true.

The open science movement that grew from this realization has been one of the most significant reforms in the history of psychology. Pre-registration, Registered Reports, open data, and replication studies have become standard practice. The field is, by most accounts, more rigorous now than it was in 2011. Whether this reform would have happened without Bem’s paper is debatable. That Bem’s paper accelerated it is not.

Precognition remains unproven. The evidence that existed was not replicated, the original data does not match the published results, and the most rigorous replication attempt found nothing. But the question of whether humans can perceive future events is less important than the question of whether our methods for answering questions about human behavior are trustworthy. On that question, Bem’s paper provided an answer, just not the one he intended.

Video: Daryl Bem Presents “Feeling the Future”

Video: Is ESP Real According to Science?

Video: Feeling the Future – Precognition and Daryl Bem

Key Researchers

Name	Affiliation	Role	Key Contribution
Daryl J. Bem	Cornell University (Emeritus)	Author, “Feeling the Future”	Published 9 experiments claiming evidence for precognition in JPSP (2011)
Ulrich Schimmack	University of Toronto Mississauga	Independent analyst	Analyzed Bem’s original data, found it did not match published results; called for retraction
Stuart Ritchie, Richard Wiseman, Chris French	Goldsmiths, University of London	First independent replicators	Three failed replication attempts with 150 participants (2012)
Jeff Galak, Robyn LeBoeuf, Leif Nelson, Joseph Simmons	Carnegie Mellon, UC Berkeley, UPenn	Large-scale replicators	7 studies, N=3,289, found d=0.04 (not significant); found Bem was the only significant moderator
Zoltan Kekecs et al.	Lund University	Transparent Psi Project	Pre-registered, Registered Reports replication of Bem Experiment 1; found no evidence (2023)
Patrizio Tressoldi, Thomas Rabeyron, Michael Duggan	Various	Meta-analysis co-authors	Co-authored 90-experiment meta-analysis with Bem (2015)

Sources

1. Bem, D.J. (2011). “Feeling the Future: Experimental Evidence for Anomalous Retroactive Influences on Cognition and Affect.” Journal of Personality and Social Psychology, 100(3), 407-425. APA (PDF).
   Funding disclosure: Cornell University. COI: Bem has a longstanding interest in psi research and has defended the ganzfeld experiment. Published in a top-tier APA journal but accompanied by an editorial acknowledging controversy. Evidence tier: 🟠
2. Bem, D.J., Tressoldi, P., Rabeyron, T., and Duggan, M. (2015). “Feeling the Future: A Meta-Analysis of 90 Experiments on the Anomalous Anticipation of Random Future Events.” F1000Research, 4, 1188. PubMed | F1000Research.
   Funding disclosure: Not specified. COI: Bem is the first author and the original study author. The meta-analysis includes his own studies. F1000Research is an open-access journal with post-publication peer review (not traditional pre-publication review). Evidence tier: 🟠
3. Galak, J., LeBoeuf, R.A., Nelson, L.D., and Simmons, J.P. (2012). “Correcting the Past: Failures to Replicate Psi.” Journal of Personality and Social Psychology, 103(6), 933-948. Replicability-Index.
   Funding disclosure: Carnegie Mellon, University of Florida, UC Berkeley, University of Pennsylvania. COI: Independent replication, no psi affiliation. Published in the same journal as Bem’s original paper. Evidence tier: 🟢
4. Ritchie, S.J., Wiseman, R., and French, C.C. (2012). “Failing the Future: Three Unsuccessful Attempts to Replicate Bem’s Retroactive Facilitation of Recall Effect.” PLoS ONE, 7(3), e33423. Columbia University.
   Funding disclosure: Goldsmiths, University of London. COI: Richard Wiseman is a well-known skeptic of psi research. Independent replication. Evidence tier: 🟡
5. Kekecs, Z. et al. (2023). “Raising the Value of Research Studies in Psychological Science by Increasing the Credibility of Research Reports: The Transparent Psi Project.” Royal Society Open Science, 10(2), 191375. Royal Society Open Science.
   Funding disclosure: Lund University, multiple international institutions. COI: 30 co-authors including both psi believers and skeptics. Pre-registered with Registered Reports. Evidence tier: 🟢
6. Schimmack, U. (2018). “Why the Journal of Personality and Social Psychology Should Retract Article DOI: 10.1037/a0021524.” Replicability-Index.
   Funding disclosure: University of Toronto Mississauga. COI: Author is a known critic of Bem’s work but provides raw data analysis. Evidence tier: 🟡
7. Engber, D. (2017). “Daryl Bem Proved ESP Is Real.” Slate. Referenced at Replicability-Index.
   Funding disclosure: Slate (media company). COI: Journalistic interview, not peer-reviewed. Bem’s quotes are direct. Evidence tier: 🟡
8. Judd, C.M. and Gawronski, B. (2011). Editorial accompanying Bem (2011) in JPSP. Referenced in Replicability-Index.
   Funding disclosure: JPSP (APA). COI: Editors of the journal that published Bem. Evidence tier: 🟡
9. Bem, D.J. and Honorton, C. (1994). “Does Psi Exist? Replicable Evidence for an Anomalous Process of Information Transfer.” Psychological Bulletin, 115(1), 4-18. Referenced at Wikipedia.
   Funding disclosure: Not specified. COI: Bem and Honorton are both proponents of psi. Publication venue: Psychological Bulletin (top-tier journal). Evidence tier: 🟡
10. Fox, J. (2025). “The Possibility of Precognition.” Psychology Today. Psychology Today.
    Funding disclosure: Psychology Today (media). COI: Author is a psi proponent. Publication venue: Psychology Today (popular science magazine, not peer-reviewed). Evidence tier: 🟠

Frequently Asked Questions

Did Daryl Bem prove that precognition is real?

No. Bem published nine experiments that reported statistically significant results for precognition, but multiple independent replication attempts failed to reproduce the findings. The most rigorous replication, the 2023 Transparent Psi Project, found no evidence for the effect under pre-registered methodology. Bem’s own shared data does not reproduce the published results (according to Replicability-Index).

Why was the paper published in a top journal if the results were wrong?

The paper met the journal’s existing methodological standards. The editors acknowledged that the results were controversial but found the methods sound. The replication crisis revealed that the standards themselves were insufficient: QRPs could produce significant results for non-existent effects, and peer review did not catch them (according to Wikipedia).

What are questionable research practices?

Questionable research practices (QRPs) are methods that inflate the probability of finding statistically significant results without being outright fraud. They include stopping data collection when results look good, selectively reporting studies that worked, trying multiple analyses and reporting only the significant one, and excluding data points that reduce significance. A 2012 survey found that the majority of psychologists admitted to at least one QRP (according to Replicability-Index).

What is the Transparent Psi Project?

The Transparent Psi Project is a 2023 international collaboration of 30 researchers led by Zoltan Kekecs of Lund University. It attempted to replicate Bem’s Experiment 1 using pre-registered methodology and Registered Reports (where the journal agrees to publish regardless of outcome). It found no evidence for precognition (according to Royal Society Open Science).

How did Bem’s paper start the replication crisis?

Bem’s paper demonstrated that standard psychological methods could produce significant results for a phenomenon that virtually no psychologist believed was real. This showed that the same methods could produce significant results for other phenomena that might also not be real. The realization that QRPs and publication bias were contaminating the psychological literature led to reforms including pre-registration, open data, and Registered Reports (according to Royal Society Open Science).

Was Bem’s data fabricated?

There is no direct evidence that Bem fabricated data. However, Ulrich Schimmack’s analysis of Bem’s original data found that it did not reproduce the published results. Bem provided two different datasets on two occasions, and neither matched the published statistics. This discrepancy has not been explained (according to Replicability-Index).

Could precognition be real but too small to detect reliably?

This is the “file drawer” argument: perhaps precognition exists but is so weak that only very large studies with optimized conditions can detect it. Bem’s meta-analysis of 90 experiments claims d = 0.20, which is small but theoretically detectable. The problem is that the meta-analysis is contaminated by the same QRPs that affected the individual studies. If the underlying effect is zero and the methods are biased, a meta-analysis of biased studies will still show a non-zero effect.