Clairvoyance
Clairvoyance, from the French for “clear seeing,” is the claimed ability to perceive objects, events, or information directly, without any other mind acting as a sender. Researchers have tested it for nearly a century, from shuffled-card guessing to free-response remote viewing. The evidence is contested: positive results exist, but they are small, and skeptics point to flaws in how many of the studies were run.
Key findings
- A meta-analysis of 141 forced-choice ESP studies (1987-2022) found a tiny but statistically reliable effect, with no meaningful difference between clairvoyance, telepathy, and precognition.6
- A classic card-guessing clairvoyance experiment by Rhine and Pratt has been cited as significant evidence for the phenomenon.5
- Free-response remote viewing at Stanford Research Institute reported far larger effect sizes than card guessing.1
- Pre-selected participants outperformed unselected ones, and feedback during trials was linked to better scores.6
- Critics and proponents jointly agreed an overall effect exists in the ganzfeld database that selective reporting cannot explain, while disagreeing on whether it proves psi.13
Overview
Parapsychologists divide extrasensory perception into three kinds. Telepathy is mind-to-mind, where a sender knows the target. Precognition is knowing a future event. Clairvoyance is the odd one out: information about a thing or event that no living mind currently knows.
That last point is the whole reason clairvoyance matters as a separate category. If a target exists but nobody knows what it is, then a correct call cannot be blamed on telepathy. There is no sender to read. This is why researchers built “pure-clairvoyance” designs. They are partly a tool for ruling out one explanation in favor of another.
The phrase “information that seems to move between minds without any normal explanation” is sometimes used as a catch-all for ESP. Clairvoyance is the version where there is no other mind in the loop at all.
Where the Term Sits in the Literature
Many studies that look like clairvoyance cannot fully exclude precognition or telepathy, so researchers often relabel them “general extrasensory perception” (GESP) when a sender or a future component cannot be ruled out.3 One reported case is striking: the percipient Sean Harribance scored at his usual high level under GESP conditions but at chance under clairvoyance-only conditions, hinting that the distinction may matter for some people.2 A 1995 dream study used a clairvoyance design with no sender, but its authors noted precognition still could not be excluded.4
Why This Is Hard to Study
Clairvoyance is hard to study for three reasons stacked on top of each other. First, the reported effect is usually very small. Second, separating clairvoyance from its cousins is a genuine puzzle. Third, the early methods left room for ordinary mistakes that can mimic a real effect.
Start with size. In forced-choice card tests, the typical effect is tiny. The average score sits only a hair above chance. An effect this small means that any single study, even a large one, gives an ambiguous result. You need to combine many studies to see anything at all. That makes the whole field vulnerable to small biases that push results in one direction.6
Then there is the separation problem. To call something pure clairvoyance, you must guarantee that no living person knows the target when the call is made. You must also ensure that the result cannot be a lucky guess about the future. Both conditions are demanding. In practice, many designs leak one or the other.
Finally, the phenomenon may be state-dependent. Some research links psi scoring to relaxation, meditation, and creativity, and even to quiet periods in Earth’s magnetic field.8 If the effect only appears under specific conditions, a study run under the wrong conditions could find nothing even if the effect is real. That is why dismissing the topic on a single null result would be premature.
The Effect-Size Gap Between Paradigms
Forced-choice card guessing produces a mean effect size of about 0.02, meaning scores hover barely above chance.6 Effect size here is a standardized measure of how far results sit from chance. A value of 0.02 is near the edge of noise; it only becomes detectable when trial counts are enormous. The forced-choice effect is weak precisely because it is estimated from the number of trials rather than the number of participants.6 By contrast, the SRI remote-viewing program reported effect sizes above 0.6, a difference of more than an order of magnitude.1 Whether that larger number reflects a stronger genuine signal or looser judging is a central dispute.
How Clairvoyance Has Been Tested
The earliest systematic tests came from J. B. Rhine’s laboratory at Duke University in the 1930s. The basic tool was the deck of ESP cards, each printed with one of five simple symbols. A subject tried to name cards they could not see. With five symbols, pure chance gives one hit in five.
Rhine’s clever twist for clairvoyance was the “down-through” technique, or DT. The subject called the whole shuffled deck from top to bottom before anyone looked at a single card. Because no living person knew the order, a hit could not come from reading a sender’s mind. The target existed, but it was unknown.
This design tries to rule out one specific artifact: sensory leakage from a sender. The control is structural. If nobody has seen the cards, there is no facial cue, no whisper, no glance to leak. The ruling-out of telepathy is, in principle, complete under DT. What DT does not rule out is recording error, poor shuffling, and the possibility that the subject is somehow reading the future order rather than the present one.
The Pratt-Woodruff Series and Forced-Choice Scoring
The Pratt-Woodruff series (1939) at Duke was a notable controlled clairvoyance experiment built to tighten procedures against the criticisms aimed at earlier card work. Forced-choice designs have one clear advantage that free-response work lacks: the chance baseline is known exactly. With five symbols, the expected hit rate is exactly 20%, so any deviation can be tested with a clean statistical model. Card-guessing experiments were the dominant ESP method from the 1880s to the 1940s. The Rhine-Pratt clairvoyance experiment is cited as a classic non-ganzfeld source of significant clairvoyance evidence.5
The modern form of clairvoyance testing is remote viewing. Here the subject does not pick from a fixed menu. They describe a distant, hidden target in free-form words and sketches. The free-response programs at Stanford Research Institute, starting in the 1970s, are the best-known example.1
Free-response scoring works differently. Because the subject’s description is open-ended, a judge has to decide how well it matches the real target. To keep this fair, researchers use rank-order judging. A blind judge, who does not know which target was used, ranks the description against several possible targets. Blinding means the judge cannot let knowledge of the answer color their scoring. If the real target tends to land at the top of the rankings more often than chance allows, that counts as a hit.
SRI Remote-Viewing Protocols
SRI used psi-gifted participants such as Ingo Swann, Pat Price, and Hella Hammid, who were not randomly recruited but pre-selected for apparent ability. Reported protocols included outbound experimenters who traveled to randomly chosen sites, shielded rooms, double-blind handling, and rank-order judging.1 A separate training study with 36 trials by Army intelligence officers at Fort Meade reported 19 first-place matches where about 6 were expected by chance.1 These are reported figures from program participants; the program’s interpretation is contested and discussed below.
What the Experiments Found
The strongest quantitative claim for the forced-choice tradition comes from a meta-analysis. When you have many small studies, each ambiguous on its own, you combine them with a method called meta-analysis. Pooled together, a tiny effect can become statistically detectable even if no single study impresses.
A 2023 meta-analysis of 141 forced-choice ESP studies found a small but reliable effect above chance. Crucially for this page, it found no significant difference in effect size between clairvoyance, telepathy, and precognition.6 Selected participants did better than unselected ones, and studies with feedback did better than those without.
A meta-analysis of 141 forced-choice studies found a small effect above chance that selective reporting alone could not explain, with no significant difference between clairvoyance, telepathy, and precognition.6
The Forced-Choice Meta-Analysis in Detail
The 2023 analysis combined a revised 1987-2010 database (83 studies) with new 2011-2022 work (59 studies) to form its homogeneous combined database of 141, included only peer-reviewed studies using randomized targets and human participants, and used random-effects modeling.6 The combined mean effect size was 0.02 with a Stouffer Z of 8.52 and a p-value below 10 to the minus 16. That p-value means results this extreme would be astronomically unlikely if nothing were going on, but a p-value measures only whether an effect differs from zero, not whether it is large or important. About 21% of the studies were independently significant. The authors reported finding no experimenter bias and no publication bias on their tests.6 Note that publication bias, the tendency for positive results to get published while null results sit in file drawers, is exactly the kind of artifact a small pooled effect is most vulnerable to.
The remote-viewing tradition reported much larger numbers, but those figures rest on free-response judging rather than the exactly-known chance baseline of cards. That difference is the crux of the disagreement over what remote viewing shows.15
Null and downward reanalyses
Not every reanalysis points upward. A reanalysis of a dream-ESP database recalculated the effects using inverse-variance weighting, which gives larger studies more importance. This is a standard correction: a 200-person study should count for more than a 2-person study.
The Dream-ESP Reanalysis
Howard’s 2018 reanalysis of Storm et al.’s dream-ESP database found an overall effect of r = .07, roughly a third of the originally reported r = .20, and it became non-significant when a trim-and-fill procedure imputed likely missing studies.12 Effect size shrank as sample size grew: large studies produced near-zero effects while only small studies produced moderate ones. For the clairvoyance mode specifically, the effect was r = .19 and significant, but unstable; removing any one of five key studies would have flipped it to non-significant.12 The pattern of bigger effects in smaller studies is a classic warning sign of publication bias and small-study artifacts. This is a contested reanalysis of a pre-existing database, and the original authors dispute the weighting choices.
Individual studies in the reference library
The studies below are the individual clairvoyance results held in the reference library. How large a share of the published clairvoyance literature that represents has not been measured, so what follows is a summary of what the library holds rather than a count of what has been published. Two of the five report a standardized effect size and three report a hit rate, and the two kinds of number cannot be added together into one bottom-line figure.
The reported results differ by more than an order of magnitude. Rock and colleagues, working with subjects pre-selected for apparent ability, report a moderate effect whose confidence interval does not cross zero.16 Vasse and Vasse, using the classic Duke card protocol, report 264 hits where 195 were expected, a large departure from chance for a forced-choice series.19 Shen and colleagues report 5 complete five-character matches in 34 double-blind trials from a single participant.20 At the other end, Steinkamp, Milton and Morris found a mean effect size of 0.009 across their 22 matched study pairs, and an almost identical 0.010 for precognition, so their data give no sign that clairvoyance outperforms precognition.14 They caution that the small number of study pairs leaves the comparison underpowered, so a near-zero result there is not by itself evidence that no effect exists.14
One row carries a disclosure that governs how it can be read. The Delorme and colleagues 2016 study of mortality judgments from facial photographs was retracted by the journal after publication.18 It is listed here because it is in the library and its removal from the record is part of what the library holds, but a retracted paper supports no conclusion about clairvoyance in either direction. Independent replication of the exceptional-subject free-response result, by groups other than those that produced it, is not represented in the library.
Interpreting the Data
The data and the explanation for the data are two separate things. A small above-chance score in a clean study is a finding. What causes it is a question, and the proposed answers are plural and contested.
Some researchers read the results as anomalous information transfer that physics will eventually accommodate. Various quantum-flavored frameworks have been proposed, invoking a nonlocal informational substrate behind both mind and matter.7 These are interpretations, not measurements. No consensus mechanism has emerged, and the proposed quantum readings remain contested even among proponents.
A complication runs through the whole field: clairvoyance is genuinely hard to isolate from precognition. If a subject is reading the future feedback rather than the present target, then a “clairvoyance” hit is really precognition wearing a disguise. One theorist has argued that even in remote viewing, a backward-in-time influence from the people who later analyze the session could feed information to the viewer, blurring the line still further.11
Why Clairvoyance and Precognition Keep Collapsing Together
The meta-analytic finding of no significant difference between clairvoyance, telepathy, and precognition modes cuts two ways.6 A proponent might say one underlying ability shows up regardless of label. A skeptic might say the modes are not cleanly separable in practice, so the labels carry little information. A dedicated forced-choice meta-analysis comparing clairvoyance and precognition directly examined how trials were interspersed and which conditions were used, underscoring how much the result depends on design choices.14
What the Five Library Studies Do Not Settle
The five individual results listed above leave four specific questions open. First, Rock and colleagues report a moderate effect and Steinkamp, Milton and Morris a near-zero one, and the two papers cover different designs, different target materials and different ways of choosing participants, so neither reports a test that would reconcile the gap.1614 Second, Rock and colleagues selected subjects for apparent ability, so their figure describes those subjects and does not transfer to unselected participants without separate evidence.16 Third, Shen and colleagues report 34 trials from one person, which describes that person’s trials rather than any wider population.20 Fourth, Rock and colleagues pool clairvoyance and precognition trials together in a single free-response analysis, and Steinkamp, Milton and Morris found the two modes scoring almost identically, so neither dataset separates the two.1614
Skeptical Critiques and Discussion
Critique 1: Early card-guessing results can be explained by recording errors, sensory cues, and poor randomization.
Critics of the early Duke work raised recording errors, sensory leakage, inadequate shuffling, and optional stopping as routes to false positives.
The down-through technique was a direct structural response: by having subjects call the deck before anyone saw it, sensory leakage from a sender is eliminated, since there is no sender and no card on view. Later forced-choice work added randomized targets and peer review as inclusion criteria.6
Analysis. The DT design addresses sensory leakage specifically but not recording error or optional stopping, which is stopping data collection once results look favorable. The 2023 meta-analysis restricted itself to randomized, peer-reviewed studies and reported no detectable publication bias on its tests; whether those filters fully remove the older concerns is debated between the meta-analysts and their critics. Steinkamp, Milton and Morris coded procedural-safeguard variables directly, recording for each study how targets were randomized and whether a participant could reach the target during a trial, and reported that no safeguard variable correlated significantly with effect size across their 22 study pairs.14 That analysis covers forced-choice card work; the free-response designs used in the later exceptional-subject studies were not part of it.16
Critique 2: The Soal-Goldney series, long held up as strong ESP evidence, was fabricated.
The Soal-Goldney experiment became one of parapsychology’s most notorious cases, with the target series widely held to have been manipulated.
A 2010 computational reanalysis reported that the published target series can be retrieved from Chambers’ Tables of logarithms using Soal’s stated method, with Benford’s Law and entry-point distribution matching that source.9
Analysis. The historical fraud allegation is the foundational disclosure here, and it concerned documented manipulation of target records. Garton’s reanalysis reports statistical evidence that the targets trace to the claimed source, but it does not by itself establish that the original experiment was otherwise sound. Any claim resting on the Soal-Goldney data would require specific independent verification of the experimental conduct, not just the target sourcing. This evidence carries a documented historical-integrity flaw, so it yields no conclusion either way about clairvoyance.
Critique 3: The ganzfeld and related psi databases reflect selective reporting and flexible analysis rather than a real anomaly.
Ray Hyman’s long critique held that flaws in how studies were run and reported, including multiple analysis and selective reporting, could account for positive psi results.13
In a joint communiqué, Hyman and Honorton agreed that an overall significant effect in the database could not reasonably be explained by selective reporting or multiple analysis, while continuing to disagree on whether it constitutes evidence for psi.13
Response by Honorton: Honorton accepted the call for tighter standards on randomization, judging, feedback, and statistics, and both authors recommended that the verdict await future experiments by a broader range of investigators.13
Analysis. The communiqué is unusual in that a leading skeptic and a leading proponent published shared conclusions. They agreed an effect exists in the data and disagreed about its cause, explicitly deferring the question to better future studies. Bem and Honorton later argued the ganzfeld evidence met the standard of replicable anomalous information transfer, a claim that itself drew continued scrutiny.5
Critique 4: Forced-choice ESP effects are too trivial to matter even if real.
The forced-choice effect is acknowledged within the field itself to be very weak, estimated from huge trial counts rather than from participants.6
Proponents counter that the same database shows selected participants outperform unselected ones and that feedback improves scores, suggesting the effect is a real signal that can be enhanced rather than pure noise.6 Training programs that gave talented subjects immediate feedback reported sustained or improving performance.10
Analysis. Both sides agree the raw effect is small. The disagreement is whether the moderator patterns, like participant selection and feedback, mark a genuine modulable ability or whether they are post-hoc subgroup findings that small-study artifacts can also produce. The dream-ESP reanalysis showing effects shrinking with sample size is the kind of result that keeps this question open.12
Open Questions: What Would Settle This
The cleanest test of pure clairvoyance is one that no current study fully delivers: a target that no living mind ever knows, and feedback that cannot leak backward in time to the viewer.
A decisive study would be preregistered, meaning the hypotheses, sample size, and analysis plan are locked in publicly before data collection. It would use the down-through logic to exclude any sender. It would withhold feedback entirely, or generate the target only after the call is judged, to block the precognition and retroactive-analyzer routes that currently muddy the clairvoyance label.11 The randomization of targets would be auditable, and the judging blind and rank-ordered. Hyman and Honorton’s own recommendations on randomization, judging, documentation, and statistics describe much of what such a protocol would need.13
A second open question is whether the moderators are real. An adversarial study, designed jointly by a proponent and a skeptic, could preregister the prediction that pre-selected, feedback-trained participants beat unselected controls by a specified margin. If selection and feedback genuinely modulate clairvoyance, that effect should replicate under hostile scrutiny. If it does not, the subgroup findings would look more like artifacts.6
Until such studies are run and independently replicated, the honest summary is narrow. There is a small, persistent, much-debated above-chance signal in the forced-choice record, larger reported effects in free-response remote viewing whose judging methods remain contested, and no agreed mechanism for any of it.
References
- Targ, R. (2019). What Do We Know About Psi? The First Decade of Remote Viewing Research and Operations at Stanford Research Institute. Journal of Scientific Exploration, 33(4). https://journalofscientificexploration.org/index.php/jse/article/view/1669 R001 [Targ 2019] ↩︎
- Morris, R., Dalton, K., Delanoy, D., & Watt, C. (1995). Comparison of the sender/no sender condition in the ganzfeld. Proceedings of the 38th Annual Convention of the Parapsychological Association, 244–259. R002 [Morris 1995] ↩︎
- Dalton, K., & Sherwood, S. (1999). A dream GESP experiment using dynamic targets and consensus vote. Journal of the American Society for Psychical Research, 93, 145–166. R003 [Dalton 1999] ↩︎
- Sherwood, S., Dalton, K., & Watt, C. (1995). Dream GESP study II using dynamic video-clips: Investigation of consensus voting judging procedures and target emotionality. Proceedings of the 38th Annual Convention of the Parapsychological Association. R004 [Sherwood 1995] ↩︎
- Bem, D., & Honorton, C. (1994). Does psi exist? Replicable evidence for an anomalous process of information transfer. Psychological Bulletin, 115(1), 4–18. https://doi.org/10.1037/0033-2909.115.1.4 R005 [Bem 1994] ↩︎
- Storm, L., & Tressoldi, P. (2023). Assessing 36 Years of the Forced Choice Design in Extra Sensory Perception Research: A Meta-Analysis, 1987 to 2022. Journal of Scientific Exploration, 37(3), 517–535. https://journalofscientificexploration.org/index.php/jse/article/view/2967 R006 [Storm 2023] ↩︎
- Williams, G. (2016). What Can Consciousness Anomalies Tell Us about Quantum Mechanics? Journal of Scientific Exploration, 30(3). https://journalofscientificexploration.org/index.php/jse/article/view/1001 R007 [Williams 2016] ↩︎
- Radin, D. (1994). Geomagnetism and psi in the ganzfeld. Journal of the Society for Psychical Research, 59, 352–362. https://www.deanradin.com/publications R008 [Radin 1994] ↩︎
- Garton, R. (2010). Retrieving “fraudulent” Soal-Goldney target series from the source originally reported: Confirmation by Benford’s Law and empirical expectation. Journal of Scientific Exploration, 24(3). https://journalofscientificexploration.org/index.php/jse/article/view/34 R009 [Garton 2010] ↩︎
- Radin, D. (2023). Charles Tart’s Contributions to Parapsychology: A Scientific Career Sparked by the Cabbage Patch…. Journal of Anomalous Experience and Cognition, 3(2), pp. 248–266. https://journals.lub.lu.se/jaex/article/view/25229 R010 [Radin 2023] ↩︎
- Dahmen, T. (2023). Why a Retroactive Analyzer Influence Should be Considered in Remote Viewing Research. Journal of Scientific Exploration, 37(3), pp. 560–566. https://journalofscientificexploration.org/index.php/jse/article/view/2973 R011 [Dahmen 2023] ↩︎
- Howard, M. (2018). A Meta-Reanalysis of Dream-ESP Studies: Comment on Storm et al. (2017). International Journal of Dream Research, pp. 224–229. https://journals.ub.uni-heidelberg.de/index.php/IJoDR/article/view/52040 R012 [Howard 2018] ↩︎
- Hyman, R., & Honorton, C. (2019). A Joint Communiqué: The Psi Ganzfeld Controversy. Journal of Parapsychology, 82(3), 108–117. https://doi.org/10.30891/jopar.2018s.01.09 R013 [Hyman 2019] ↩︎
- Steinkamp, F., Milton, J., & Morris, R. L. (1998). A meta-analysis of forced-choice experiments comparing clairvoyance and precognition. The Journal of Parapsychology. R014 [Steinkamp 1998] ↩︎
- Utts, J. (1996). An Assessment of the Evidence for Psychic Functioning. Journal of Scientific Exploration, 10(1), 3–30. R015 [Utts 1996] ↩︎
- Rock, A. J., Houran, J., Tressoldi, P. E., & Laythe, B. (2023). Is Biological Death Final? Recomputing the Drake-S Equation for Postmortem Survival of Consciousness. International Journal of Transpersonal Studies, 42(1), 9–26. https://doi.org/10.24972/ijts.2023.42.1.9 R016 [Rock 2023] ↩︎
- Delorme, A., Pierce, A., Michel, L., & Radin, D. (2016). RETRACTED: Prediction of Mortality Based on Facial Characteristics. Frontiers in Human Neuroscience, 10. https://doi.org/10.3389/fnhum.2016.00173 R017 [Delorme 2016] ↩︎
- Frontiers Editorial Office. (2016). Retraction: Prediction of Mortality Based on Facial Characteristics. Frontiers in Human Neuroscience, 10. https://doi.org/10.3389/fnhum.2016.00515 R018 [Frontiers Editorial Office 2016] ↩︎
- Vasse, P., & Vasse, C. (1955). PK and ESP experiments with Martie Vasse. Parapsychology Foundation. R019 [Vasse 1955] ↩︎
- Shen, D., Shen, R., & Li, G. (2021). Laboratory Investigations of Extrasensory Identification of Concealed 5-Character Codes by a Presumably Gifted Teenager. Journal of Scientific Exploration, 35(3), 570–589. https://journalofscientificexploration.org/index.php/jse/article/view/1827 R020 [Shen 2021] ↩︎