Parapsychology

Abstract

Parapsychology is the scientific study of ostensibly paranormal phenomena, chiefly extrasensory perception and psychokinesis, that appear to lie outside known physical and biological mechanisms. Defined by its methods rather than its subject, it applies forced-choice card tests, free-response protocols such as the ganzfeld, and random-number-generator experiments, evaluated with standard inferential statistics. Its central unresolved problem is replication: meta-analyses report small, cumulatively significant departures from chance, while pre-registered replications and Bayesian reanalyses often fail to confirm them, so the same database reads as weak evidence for an anomaly to proponents and as selective-reporting artefact to critics. This has made parapsychology an influential test case in the wider debate over evidential standards and the replication crisis. Three interactive demonstrations let the reader test a hit rate against chance, weigh the file-drawer problem, and calibrate what a significant p-value is worth.

Keywords: parapsychology, extrasensory perception, psychokinesis, ganzfeld, replication

Parapsychology denotes the empirical investigation of anomalous processes of information or energy transfer that are not explicable by known physical or biological principles, a domain that MeSH classifies among the behavioral sciences (Cardeña, 2018). Its subject matter divides into extrasensory perception, the apparent acquisition of information without the recognised senses, and psychokinesis, the apparent influence of mind on matter without a physical intermediary. What makes the field a scientific enterprise rather than a catalogue of anecdotes is its commitment to controlled experiment: a claim counts only if it survives randomisation, sensory shielding, blind judging, and statistical analysis (Utts, 1991). The field matters to cognitive psychology less for the reality of the phenomena it studies, which remains firmly rejected by the scientific mainstream (Reber & Alcock, 2020), than for the methodological mirror it holds up: the disputes over its evidence anticipated, and helped precipitate, the discipline-wide reckoning with underpowered studies, flexible analysis, and publication bias (Wagenmakers et al., 2011).

What Parapsychology Is

Parapsychology is the study of putative psi phenomena, a neutral term coined to name the hypothesised anomalous process without committing to any mechanism. Psi is conventionally partitioned into two families. Extrasensory perception (ESP) covers cases in which information seems to reach a person without any known sensory channel, and is subdivided by the source of that information: telepathy (mind to mind), clairvoyance (from a physical state of affairs), and precognition (from a future event). Psychokinesis (PK) covers the converse direction, in which intention appears to affect a physical system directly, from the fall of dice in early experiments to the output of electronic random-number generators in later ones (Bösch, Steinkamp, & Boller, 2006). The definitions are deliberately negative: psi is what is left when every conventional explanation, from sensory leakage to fraud to statistical artefact, has been excluded.

This negative definition is the source of both the field's rigour and its difficulty. Because a positive theory of the mechanism is absent, an ostensible psi effect can only be established as a reliable departure from chance under conditions that rule out ordinary causes, never as the confirmation of a predicted process (Reber & Alcock, 2020). Parapsychology is therefore best understood as a methodological discipline attached to a set of claims: it shares the experimental and inferential apparatus of mainstream psychology, including the misjudgements of probability studied under heuristics and the pattern-finding of ordinary perception, and differs only in the phenomena it hopes those methods will reveal. The scientific consensus is that they reveal none, but the demonstration of that null result has been unusually instructive.

Types of Parapsychology

The Medical Subject Headings vocabulary places Parapsychology within the behavioral sciences and gives it a single narrower descriptor, shown in Table 1. Because MeSH is an indexing classification rather than a theory of the paranormal, this hierarchy records how the biomedical literature is catalogued, not a claim that the field divides cleanly at exactly these joints; the subtype below is not mutually exclusive with the broader construct but one heavily studied region of it, and the field's own working taxonomy (ESP versus psychokinesis, forced-choice versus free-response) cross-cuts the MeSH scheme.

Table 1. Direct subtypes of Parapsychology in the MeSH classification (tree F02.550).

The Experimental Paradigms

Three paradigms carry most of the modern evidential weight. The oldest is the forced-choice test, in which a participant guesses which of a fixed set of targets, classically the five Zener card symbols, has been or will be selected. With five equiprobable targets the chance hit rate is exactly 20%, and the whole method reduces to testing whether an observed hit rate exceeds that baseline across many trials. Rhine's programme at Duke industrialised this design and gave parapsychology its statistical character, but forced-choice tasks are dull for participants and vulnerable to sensory cues and recording errors, and their aggregate effects proved small and fragile (Utts, 1991).

The ganzfeld was introduced to remedy the psychological poverty of card guessing. A receiver reclines in a state of homogeneous sensory input, halved ping-pong balls over the eyes under red light and white noise in the ears, while a distant sender concentrates on a randomly chosen target image. Afterwards the receiver ranks four candidate images, one the target and three decoys, so the chance hit rate is 25%. The autoganzfeld studies of the late 1980s automated target selection and randomisation to close the sensory-leakage and recording loopholes that had marred earlier free-response work, and the resulting database became the strongest single case for ESP (Bem & Honorton, 1994). It also became the sharpest point of dispute, because an independent set of studies then failed to reproduce the effect (Milton & Wiseman, 1999).

The third paradigm addresses psychokinesis. In random-number-generator experiments a participant attempts to bias the output of an electronic device driven by a quantum or thermal noise source toward more ones than zeros. The design is attractive because the physical baseline is known exactly and enormous numbers of trials can be accumulated automatically, and the largest meta-analysis of it examined more than 350 experiments; the reported deviation from chance was extremely small and, the authors argued, best explained by a small-study bias rather than a genuine mind-matter interaction (Bösch et al., 2006). A related design, the feeling-of-being-stared-at experiment, tests whether a person can detect unseen observation at above-chance rates (Schmidt, Schneider, Utts, & Walach, 2004).

Meta-Analysis and the File-Drawer Problem

Because any single psi experiment yields at best a weak signal, the field's case rests on meta-analysis, the statistical pooling of many studies into a combined estimate. A meta-analysis of free-response ESP studies from 1992 to 2008 reported an overall hit rate reliably above the 25% baseline, which proponents read as evidence that the ganzfeld effect is real and replicable when studies are aggregated (Storm, Tressoldi, & Di Risio, 2010). Critics counter that a small pooled effect near chance is exactly what selective publication would manufacture, and that the ganzfeld literature's apparent success and failure track the analytic choices and inclusion criteria of whoever assembles it (Milton & Wiseman, 1999). The psychologist Ray Hyman, the field's most persistent methodological critic, replied directly to that meta-analysis that pooling heterogeneous studies conceals more than it reveals: a combined estimate averages over incompatible protocols and quality levels, so a significant summary statistic can reflect a few flawed high-yield studies rather than a coherent effect (Hyman, 2010).

The core threat is the file-drawer problem: studies with null results are less likely to be written up and published, so the visible literature is a biased sample of all the studies actually run (Rosenthal, 1979). Rosenthal's fail-safe N quantifies the vulnerability by asking how many unpublished null studies would have to be sitting in file drawers to reduce a significant pooled result to non-significance. When that number is implausibly large, the effect is robust to publication bias; when it is small, a modest file drawer suffices to erase it. In parapsychology the fail-safe numbers are frequently large enough to be rhetorically impressive yet, critics note, still compatible with the low base rate of publication and the high flexibility of analysis that characterise the field (Reber & Alcock, 2020). The tool cuts both ways, and the Worked Example below shows how sensitive it is to the assumptions fed into it.

Feeling the Future: Psi and the Replication Crisis

In 2011 the social psychologist Daryl Bem published nine experiments in a leading journal that appeared to show precognition: standard cognitive effects, such as the facilitation of recall by practice, seemed to operate backwards in time, with events after a test influencing performance on it (Bem, 2011). The paper used conventional methods and conventional statistics, and that was precisely the point of the controversy it ignited. If orthodox procedures could yield strong support for an impossible conclusion, then the procedures, not the conclusion, were suspect.

The response reshaped methodological psychology. A Bayesian reanalysis argued that Bem's data, correctly weighed, provided little or no evidence for precognition, and that the significant p-values reflected the well-known tendency of null-hypothesis testing to overstate evidence against a true null, compounded by undisclosed analytic flexibility (Wagenmakers et al., 2011). Independent laboratories then ran direct replications of Bem's retroactive-recall experiment and found nothing (Ritchie, Wiseman, & French, 2012), a null result echoed by a larger multi-study replication effort (Galak, LeBoeuf, Nelson, & Simmons, 2012). The episode is now cited as a proximate trigger of the replication crisis: it made vivid that flexible analysis and selective reporting could produce publishable evidence for anything, and it accelerated the adoption of pre-registration and Bayesian evidence assessment across psychology. The demonstration below makes the statistical heart of the critique concrete, showing how weak the evidence in a just-significant result actually is.

Cognitive Accounts of Paranormal Belief

Alongside the experimental programme runs a second, more productive line of research that takes paranormal belief, rather than paranormal phenomena, as its object. This work asks why psi experiences feel so compelling even if no anomalous process exists, and locates the answer in ordinary cognition. Believers in the paranormal show, on average, a greater susceptibility to the misperception of randomness: they underestimate how often coincidences arise by chance and detect meaningful structure in noise, a bias in probabilistic reasoning that generates the sense of an uncanny connection where only pattern recognition run ahead of the data is at work (Blackmore & Trościanko, 1985). The relationship is not simple, and its strength is moderated by other traits such as schizotypy (Dagnall, Denovan, Drinkwater, Parker, & Clough, 2016). The same programme extends to specific anomalous experiences: near-death experiences, for instance, are analysed as products of a stressed or dying brain and of reconstructive memory rather than as glimpses of an afterlife (French, 2001).

A complementary account explains why belief in one's own psychic ability persists in the face of failure. On the misattribution hypothesis, people who believe themselves psychic are not more accurate but more prone to notice and remember the hits and to discount the misses, reconstructing a chance run of guesses as evidence of a gift (Wiseman & Watt, 2006). These mechanisms connect parapsychology to the mainstream study of heuristics and decision making: the same shortcuts that make everyday judgement efficient also make chance look like agency, so paranormal belief becomes a predictable output of a normally functioning mind rather than a failure of one. This is the part of parapsychology that has yielded durable, replicable findings, and it belongs squarely to cognitive psychology.

Worked Example

Consider a stylised ganzfeld dataset with the free-response chance baseline of p₀ = 0.25 (one target among four). Suppose N = 300 sessions yield 96 direct hits, an observed hit rate of 96 / 300 = 0.32. The standard error of a proportion under the null is √(p₀(1 − p₀) / N) = √(0.25 × 0.75 / 300) = √0.000625 = 0.025. The test statistic is z = (0.32 − 0.25) / 0.025 = 2.80, giving a one-tailed p of 0.0026. By the ordinary conventions of the field this is a clear success: the hit rate exceeds chance at better than one in three hundred.

Two cautions follow immediately, and they are the whole lesson. First, the effect is tiny. The excess of seven percentage points corresponds to a Cohen's h of about 0.16, a small effect by any standard, so significance here is bought with sample size, not with a large signal. Second, that significance is fragile under publication bias. Suppose a meta-analysis combines k = 28 such studies into a Stouffer combined score of Z = 5.7, an emphatic result. The sum of the individual z-scores is 5.7 × √28 ≈ 30.16. To dilute the combined score below the significance threshold of 1.645 requires √(28 + x) > 30.16 / 1.645 = 18.33, hence 28 + x > 336, so x ≈ 308 unpublished null studies would suffice to erase it. Whether 308 is reassuringly large or uncomfortably plausible depends entirely on one's estimate of how many ganzfeld studies were run and never reported, an estimate no one can supply with confidence (Rosenthal, 1979). The arithmetic is exact; the inference it supports is not.

Discussion

More than eight decades of laboratory parapsychology have produced a stable stalemate. The proponents' strongest claim is not that any single experiment is decisive but that meta-analysis reveals a small, persistent, above-chance signal across paradigms that ought to average to zero if nothing were there (Cardeña, 2018). The critics' reply is that a small signal near chance is the exact signature of a literature shaped by selective publication, optional stopping, and flexible analysis, and that no proposed psi effect has met the field's own standard of a protocol that transfers intact to a sceptical laboratory (Reber & Alcock, 2020). The failure of the autoganzfeld to survive independent replication (Milton & Wiseman, 1999) and the collapse of Bem's precognition effect under direct replication (Ritchie et al., 2012) are, on this reading, not anomalies to be explained away but the expected outcome once the degrees of freedom are removed.

The lasting contribution of the dispute lies outside its nominal subject. Parapsychology functioned as an extreme test bed for inferential method: because its central hypothesis is regarded as antecedently near-impossible, the field forced a clarity about evidence that friendlier hypotheses never demanded. It was in arguing about psi that psychologists confronted, in an especially clean form, the gulf between a significant p-value and a convincing weight of evidence, the corrosive power of publication bias, and the necessity of pre-registration (Wagenmakers et al., 2011). The productive residue of the enterprise is twofold: a sharpened methodology that the whole discipline has adopted, and a genuinely cognitive account of why paranormal belief is so natural a product of an ordinary mind (Wiseman & Watt, 2006).

Current Directions

The field's most consequential recent turn is procedural rather than substantive. Registered reports, in which a study's hypotheses and analysis plan are peer-reviewed and accepted before any data are collected, were pioneered in part within parapsychology precisely because it is the domain where analytic flexibility is least tolerable; the earliest examples predate the format's general adoption and are now studied as a template for the wider reform (Wiseman, Watt, & Kornbrot, 2019). Under this constraint the burden shifts from post-hoc meta-analysis to prospective, pre-committed prediction, and several pre-registered psychokinesis experiments have returned results that favour the null on Bayesian criteria (Maier, Dechamps, & Pflitsch, 2020).

A parallel current is methodological reflection on why the field's findings replicate so poorly. Recent analyses frame parapsychology's difficulties as an especially acute instance of the general replicability crisis, arguing that the same statistical pathologies that inflate false positives across the social sciences are simply more visible here, where the prior probability of a true effect is vanishingly small (Rabeyron, 2020). Bayesian reviews continue to debate whether the accumulated database, taken at face value, constitutes meaningful evidence at all, with proponents and sceptics agreeing on the numbers while disagreeing on the priors that turn them into a conclusion (Tressoldi, 2011; Cardeña, 2018).

Common Misconceptions

Parapsychology is the same as belief in ghosts and the occult.

Parapsychology is an experimental discipline defined by controlled tests and statistical analysis, not by any belief; much of its most robust output is a cognitive account of why paranormal beliefs arise in normal minds, which is the opposite of endorsing them (Wiseman & Watt, 2006).

A statistically significant psi result proves that ESP is real.

A small p-value near the conventional threshold carries far weaker evidence than it appears to; a just-significant result can correspond to a Bayes factor of only about two or three to one, which is why the significant psi findings did not persuade and why Bayesian assessment became central to the debate (Wagenmakers et al., 2011).

Meta-analysis settles the question in favour of psi.

A significant pooled effect near chance is exactly what selective publication and flexible analysis would produce, and the same ganzfeld database yields success or failure depending on inclusion criteria, so meta-analysis constrains the dispute without resolving it (Milton & Wiseman, 1999).

Glossary

Clairvoyance.

A form of extrasensory perception in which information is apparently acquired directly from a physical state of affairs rather than from another mind.

Decline effect.

The tendency of a reported psi effect to shrink as experiments are repeated and controls tightened, often toward the chance baseline.

Extrasensory perception (ESP).

The ostensible acquisition of information without the use of the recognised senses, encompassing telepathy, clairvoyance, and precognition.

Fail-safe N.

The number of unpublished null studies that would have to exist to reduce a significant meta-analytic result to non-significance; a gauge of robustness against the file-drawer problem.

File-drawer problem.

The bias that arises when null results go unpublished, leaving the visible literature an unrepresentative and inflated sample of all studies conducted.

Forced-choice test.

An ESP task in which the participant selects from a fixed set of possible targets, so that the chance hit rate is known exactly, as in Zener-card guessing.

Ganzfeld.

A free-response protocol in which a receiver in uniform sensory input ranks four candidate images against a distant sender's target, giving a 25% chance baseline.

Meta-analysis.

The statistical combination of results across many studies into a single pooled estimate, central to the parapsychological case and to the critique of it.

Misattribution hypothesis.

The account on which belief in one's own psychic ability reflects a selective memory for hits and neglect of misses rather than genuine accuracy.

Precognition.

A form of extrasensory perception in which information is apparently acquired about a future event that could not be inferred from present knowledge.

Psi.

A neutral cover term for the hypothesised anomalous processes studied in parapsychology, embracing both extrasensory perception and psychokinesis without implying a mechanism.

Psychokinesis (PK).

The apparent influence of mind on a physical system without any known physical intermediary, tested in dice-throwing and random-number-generator experiments.

Registered report.

A publication format in which hypotheses and analysis plans are peer-reviewed and accepted before data collection, removing analytic flexibility after the fact.

Sensory leakage.

The unintended transmission of target information through a normal sensory channel, a classic confound that tighter protocols such as the autoganzfeld were designed to eliminate.

Telepathy.

The ostensible transfer of information directly from one mind to another without a recognised sensory channel; the most tested form of extrasensory perception.

Zener cards.

A deck of five symbols (circle, cross, waves, square, star) used in early forced-choice ESP testing, giving a 20% chance hit rate.

Key Researchers

Daryl J. Bem (b. 1938). Emeritus Professor of Psychology at Cornell University; his 2011 experiments claiming evidence for precognition, and the failures to replicate them, became a catalyst for the replication crisis in psychology. Faculty Page - Wikipedia

Susan Blackmore (b. 1951). Visiting Professor at the University of Plymouth; a former parapsychologist who became a prominent sceptic after a decade of her own experiments failed to find psi, and who studies the cognitive roots of paranormal belief. ORCID - Faculty Page - Google Scholar

Etzel Cardeña (b. 1957). Thorsen Chair in Psychology at Lund University; author of a widely cited 2018 review arguing that the experimental evidence for psi deserves serious scientific consideration. ORCID - Faculty Page - Google Scholar

Christopher C. French (b. 1956). Emeritus Professor at Goldsmiths, University of London; head of its Anomalistic Psychology Research Unit, which studies the psychology of paranormal belief and experience. Faculty Page - Google Scholar - Wikipedia

Charles Honorton (1946-1992). Parapsychologist who designed the autoganzfeld protocol to close the sensory-leakage and randomisation loopholes of earlier work, and co-authored the influential 1994 review with Daryl Bem. Wikipedia - Wikidata

Ray Hyman (b. 1928). Emeritus Professor of Psychology at the University of Oregon; a leading methodological critic of parapsychology and co-author of the joint communiqué that set standards for ganzfeld research. Faculty Page - Wikipedia

J. B. Rhine (1895-1980). Founder of the Duke University Parapsychology Laboratory; introduced extrasensory perception as an experimental term and established the forced-choice, statistically evaluated method that defined the field. Wikipedia - Wikidata

Robert Rosenthal (1933-2024). Psychologist at Harvard and the University of California, Riverside; his file-drawer analysis and fail-safe N are the standard tools for gauging how publication bias could inflate a meta-analytic result. Wikipedia - Wikidata

Eric-Jan Wagenmakers (b. 1972). Professor of mathematical psychology at the University of Amsterdam; his Bayesian reanalysis of Bem's precognition data became a landmark critique of null-hypothesis significance testing. ORCID - Faculty Page - Google Scholar

Caroline Watt (b. 1962). Koestler Chair of Parapsychology at the University of Edinburgh; an advocate of registered reports and open-science practice in the study of anomalous experience. Faculty Page - Google Scholar - Wikipedia

Richard Wiseman (b. 1966). Professor of the Public Understanding of Psychology at the University of Hertfordshire; known for rigorous replications of psi claims and for research on the psychology of luck and deception. ORCID - Faculty Page - Wikipedia

Frequently Asked Questions

What is parapsychology?
Parapsychology is the scientific study of ostensibly paranormal phenomena, chiefly extrasensory perception and psychokinesis, using controlled experiments and standard statistical analysis to test for reliable departures from chance (Cardeña, 2018).

Is parapsychology a real science?
It uses genuine experimental and statistical methods, but the scientific mainstream rejects its central claims because the reported effects fail to replicate under independent, pre-registered conditions (Reber & Alcock, 2020).

What is the difference between telepathy, clairvoyance, and precognition?
All three are forms of extrasensory perception distinguished by the source of the information: telepathy from another mind, clairvoyance from a physical state of affairs, and precognition from a future event (Utts, 1991).

What is the ganzfeld experiment?
The ganzfeld is a free-response test in which a receiver in uniform sensory input ranks four images against a distant sender's randomly chosen target, giving a 25% chance hit rate; its automated version was the strongest single case for ESP (Bem & Honorton, 1994).

Why did Daryl Bem's precognition study cause a crisis in psychology?
Because it used orthodox methods to reach an impossible conclusion, it implied that those methods were unreliable; a Bayesian reanalysis and failed replications then helped trigger the wider reckoning with flexible analysis and publication bias (Wagenmakers et al., 2011).

Does meta-analysis prove that psi is real?
No; the small pooled effects reported are near chance and are exactly what selective publication and flexible analysis would produce, and the same database yields different verdicts depending on which studies are included (Milton & Wiseman, 1999).

Why do so many people believe in the paranormal?
Cognitive research links paranormal belief to the ordinary misperception of randomness and to a selective memory for coincidences, so such beliefs are a predictable output of normal cognition rather than evidence of real phenomena (Blackmore & Trościanko, 1985).

What is the file-drawer problem in parapsychology?
It is the bias created when null studies go unpublished, inflating the visible literature; the fail-safe N estimates how many such hidden studies would be needed to erase a significant result (Rosenthal, 1979).

References