How Scientists Study Paranormal Claims: Methods and Replication

Author: Angela Bogdanova · Published: October 11, 2026 · Editorial Policy

Research cluster: Extraordinary Experiences & Anomalistic Psychology · Written by Artificial Sapiens

How scientists study paranormal claims is a question about evidence: what precisely was observed, which causal explanations are possible, how alternative explanations were controlled, and whether independent researchers can obtain the same result. A report of an extraordinary experience is legitimate material for scientific inquiry. A demonstration that an extraordinary external mechanism caused it requires a separate, substantially stronger evidentiary argument.

For a related perspective, see Anomalistic Psychology: What It Studies and How Evidence Is Evaluated.

Researchers study extraordinary reports using interviews, surveys, prospective diaries, field observations, laboratory experiments, psychometric instruments, and independent replications. In studies of alleged extrasensory perception (ESP), precognition, psychokinesis, haunted environments, or other claimed anomalies, the central task is to distinguish a prespecified effect from chance, information leakage, expectancy, ordinary perception, memory reconstruction, measurement artifacts, and researcher flexibility. One of the clearest recent applications of these principles is a series of rigor-focused replication projects testing precognition claims. Their results show why large samples, transparent protocols, and multiple independent tests matter. Kekecs et al. (2023) and Walleczek et al. (2025).

What Counts as a Paranormal Claim in Scientific Research?

A paranormal claim proposes an event or causal process that appears to exceed a conventional explanation, such as receiving information without an identifiable sensory channel or predicting the result of an unpredictable future random event. Scientific researchers turn such propositions into questions about observable outcomes. They specify an exposure or procedure, identify what would count as success, and determine the probability of the same outcome under suitable alternatives.

There are at least three distinct objects of study. First, a subjective experience: someone reports hearing a sound, sensing a presence, having a premonition, or encountering an apparently meaningful coincidence. Second, a behavioral or psychological outcome: a questionnaire response, recognition score, choice, physical movement, memory report, or physiological measurement. Third, an explanatory claim: that an extraordinary external agency or anomalous information transfer produced the experience or result. These objects require different types of evidence. Interviews can establish what people report; properly controlled experiments can test specified effects; a compelling causal explanation requires further discriminating evidence. The broader field-specific research landscape is described in Cardeña (2018).

Anomalistic psychology generally examines experiences, beliefs, interpretations, cognitive processes, and the ordinary or presently understood mechanisms that may contribute to exceptional reports. Parapsychology has traditionally included experimental testing of putative psi effects. The two areas overlap in phenomena and sometimes in methods. Their research questions can differ: why a person has an experience, whether a measured deviation from a known baseline occurs, and what caused any deviation are different questions. A study can answer one while leaving the others unresolved.

The Scientific Method Begins With a Testable Claim

Consider the claim that a person can identify a concealed image through ESP. A researcher needs to decide what constitutes the target, how the target is selected, whether the participant chooses among four options or provides a free description, who judges similarity, whether judgment occurs before the answer is revealed, and what hit rate the relevant chance model predicts. The exact baseline depends on the design. A four-alternative forced-choice task with truly equiprobable alternatives has a nominal chance hit rate of 25%; it is not legitimate to apply that rate to a task whose alternatives have different probabilities or whose scoring procedures offer unequal opportunities.

A usable scientific hypothesis specifies the unit of analysis, primary outcome, predicted direction, comparison condition, sample size, exclusion rules, and analytic approach. It states what would count against the prediction. Without these decisions, apparently successful examples can be assembled after seeing results, and the investigator may be comparing a carefully selected success against an ill-defined collection of misses. The problems caused by flexible stopping, measurement, and analysis are demonstrated in Simmons, Nelson, and Simonsohn (2011).

A hypothesis also needs discriminating predictions. If a proposed influence is expected to increase correct guesses, the data must be able to distinguish this from guessing, sensory cues, software errors, participant adaptation, or a response bias. Saying that any possible outcome would be compatible with an anomaly produces an account that is difficult to test. A strong design aims to reveal which explanation fits the observed data best.

Choosing the Right Research Design

Surveys, interviews, and experience sampling

Survey research can estimate the frequency and correlates of reported experiences in a defined sample. Interviews can document the sequence, emotional character, perceived meaning, and cultural context of an event. Prospective diaries and experience sampling help record reports closer to the time they occur, reducing reliance on memories reconstructed months or years later. These approaches are especially valuable for understanding how people experience unusual events and how such reports affect everyday life.

The main limitations are self-selection, recall error, ambiguous questionnaire wording, demand characteristics, social desirability, and nonrepresentative recruitment. A high proportion of respondents endorsing a claim establishes a pattern of endorsement in the sampled group. It does not establish the claimed physical cause. For a concrete illustration of research-design limitations in this domain, Dean et al. (2022) reviewed 71 studies of paranormal belief and cognitive function, identified substantial heterogeneity, and highlighted deficits in preregistration, sample-size justification, multiple-testing control, and sampling representativeness.

Field studies of unusual environments

Researchers may conduct observations or experiments in places associated with reported unusual events. They can measure relevant environmental conditions, independently record incidents, compare locations or time periods, and collect participant accounts without prompting participants to endorse a particular interpretation. A useful protocol includes instrument calibration, predefined criteria for what counts as an event, precise timestamps, independently verifiable records, and appropriate comparison sites.

Field studies trade some experimental control for ecological realism. Building noise, air quality, lighting, exposure expectations, observer presence, and varied occupancy can be difficult to separate. Instrumental anomalies require mundane troubleshooting before their meaning can be assessed. A sensor fluctuation is a measurement until a reliable, independently corroborated causal account is developed. Field observations also demand respect for occupants, property, privacy, and local safety conditions.

Controlled laboratory experiments

Laboratory experiments permit tighter regulation of target generation, sensory access, timing, information flow, instructions, and data collection. They may be the strongest design for a narrow, testable claim of anomalous information acquisition, although laboratory findings cannot automatically explain a person's spontaneous real-world experience. Testing whether people guess randomized targets above a predeclared chance rate is a different undertaking from investigating what a felt presence means to someone.

Experimental design must also match the hypothesis. If the claim predicts an effect only under a particular state or instruction, those conditions must be specified in advance and measured reliably. Repeatedly redefining the relevant conditions after a failed attempt turns a confirmatory test into exploratory research. Exploratory work is useful for generating ideas, but its discoveries require new confirmation.

Randomization and Concealment

Random assignment and randomized target selection reduce systematic relationships between trial conditions and extraneous variables. A high-quality protocol describes how randomness is generated, how sequences are stored, whether the random-number system has been tested, and whether anyone with access to the sequence could communicate information to a participant or judge.

Randomization is only one part of the protection. Allocation concealment prevents a researcher from predicting the upcoming condition before it is assigned. Sealed records, restricted permissions, automated timestamps, independent monitoring, and audit trails help establish whether the intended protocol was followed. A computer-generated random list does little good if it is visible to the person administering trials.

The choice of randomization must be compatible with the question. For example, balanced blocks may be appropriate for comparisons across conditions, while independently generated equiprobable targets may suit guessing tasks. Different procedures produce different expected distributions; a statistical analysis that ignores design-induced dependencies can yield misleading estimates. The open protocol, trial records, and audit structure used by the Transparent Psi Project demonstrate why verification of implementation is as important as descriptions of intent.

Blinding, Expectation, and Information Leakage

Blinding is the prevention of relevant knowledge reaching people whose behavior or judgments could affect the outcome. Depending on the design, participants, experimenters, scorers, data analysts, or multiple groups may be blinded. Blinding matters because even subtle cues can alter choices, ratings, and interpretation.

In a concealed-target experiment, neither the participant nor anyone interacting with that participant should be able to infer the target by ordinary means. In a free-response task, judges who rate similarity between descriptions and possible targets should not know which candidate was designated as correct. When sessions are recorded, the recording and file-handling process should preserve the blind as well.

This kind of information leakage may involve hearing a response, seeing a screen reflection, recognizing target-set patterns, obtaining hints from filenames, or observing behavior linked to trial assignment. All are methodological possibilities to inspect, not diagnoses of dishonesty. Intentional fraud is only one potential failure mode; flawed procedures can generate spurious information without anyone attempting deception.

A rigorous manuscript reports who was blinded, when the blind could be broken, and what tests checked its integrity. Merely calling an experiment double-blind is insufficient without a procedure that shows what the term meant operationally.

Controls and Alternative Explanations

A control condition provides a comparison that separates the claimed causal factor from background processes. Sham exposures, randomized non-target trials, neutral instructions, comparison environments, independent judges, and negative-control outcomes can be useful where appropriate. Positive controls can check whether the apparatus and measurement instruments can detect an effect already known to occur. In an ambiguous perception study, for example, a validated stimulus manipulation can show whether the procedure is sensitive to ordinary perceptual differences.

The ideal control matches the research question. In a claim that someone can identify hidden targets, the central comparison is whether identification exceeds the well-justified chance model under sealed information conditions. In a study of haunted-place reports, a comparison might address whether expectation or measured environmental conditions predict reporting. A control that works for one claim may be irrelevant to another.

Researchers should set out a credible list of rival explanations before collecting data. These may include ordinary sensory access, expectancy, demand, response criteria, selective recall, scoring flexibility, chance clustering, software faults, uncontrolled physical conditions, and changes in the measurement apparatus. Each rival calls for a specific test. The strongest finding survives plausible alternatives rather than merely remaining unexplained after a superficial review.

Measurement: What Exactly Is Being Counted?

Scientific results depend on operational definitions. A felt-presence questionnaire can measure a respondent's endorsement of a particular experience; it cannot by itself count external entities. A physiological sensor records a specified signal under known conditions; it cannot automatically establish the source of that signal. A paranormal-belief scale measures patterns of response to particular items rather than independently verifying the truth of what is endorsed.

The quality of a measure includes reliability, construct validity, sensitivity, specificity where applicable, susceptibility to expectation, and comparability across populations. A single numerical score can conceal several dimensions. Different instruments may combine items about unconventional events with culturally sensitive belief items; researchers must examine what a particular scale actually measures before interpreting group differences. Dean et al. (2022) demonstrates why psychometric findings and wider cognitive associations require cautious study-quality appraisal.

In free-response experiments, scoring is an additional measurement problem. If independent judges compare a description against several candidates, a good study specifies the scoring rubric, candidate sets, judge recruitment, blinding, agreement, treatment of ties, and aggregation rules before the correct answer is revealed. Ambiguous hits should not acquire greater weight merely because they make a memorable story.

Statistical Significance, Effect Size, and Uncertainty

A p value evaluates the compatibility of a specified set of data with a statistical model under its assumptions. It is not the probability that a paranormal claim is true, and it does not describe the probability of the null hypothesis given the data. Even a small p value can arise when a model is misspecified, many analyses were attempted, or a procedural flaw created an apparent difference.

Effect size tells readers how large the measured deviation is; uncertainty intervals and likelihood-based or Bayesian analyses offer further context about what the data support. A statistically detectable deviation can be tiny, unstable, or practically unimportant. Conversely, a small sample may be too imprecise to exclude an effect that matters. A good report provides raw counts, exact planned comparisons, confidence or credible intervals where appropriate, and a candid discussion of model assumptions.

In exceptional-cause research, the argument does not end at a low p value. Investigators must ask whether randomization and concealment held, whether the test direction was chosen before the data were inspected, whether the outcome has a meaningful theoretical interpretation, and whether the result independently replicates. An influential methodological analysis of the early precognition literature illustrates how one-sided testing and exploratory choices may exaggerate evidence: Wagenmakers et al. (2011).

A result near an arbitrary significance threshold deserves special scrutiny. The threshold itself is a research convention, not a division between established reality and nonexistence. Reporting effect magnitudes and uncertainty makes borderline outcomes easier to interpret without inflating their importance.

Multiple Comparisons, Flexible Analysis, and Optional Stopping

Imagine a researcher who tries several outcome definitions, tests many subgroups, alternates between one-sided and two-sided analyses, and stops recruiting once a favorable result appears. Even if each individual test nominally controls its false-positive rate, selective reporting can make the published result far less reliable than it looks.

This is the central lesson of Simmons et al. (2011): undisclosed flexibility in data collection and analysis can substantially increase false-positive findings. Transparency therefore requires reporting attempted analyses, exclusion decisions, recruitment stopping rules, changes from the planned protocol, and the status of exploratory tests. Multiple-testing correction or an appropriate hierarchical model may be needed when many outcomes are assessed.

A prespecified primary outcome protects interpretation. Secondary and exploratory outcomes can remain valuable, provided they are labeled clearly. If an unexpected pattern appears, it should guide a new independently defined study rather than being retroactively treated as the prediction that motivated the original design.

Preregistration and Registered Reports

Preregistration records a study's key hypotheses, design, outcomes, stopping rules, and planned analysis before the relevant data are known. Researchers can still change a procedure for defensible reasons, but the changes should be timestamped and reported. This makes it possible to distinguish the initially intended test from later exploration.

Registered Reports go further by having a journal review the question and methods before results are collected and, subject to conditions, commit to publication based on methodological quality rather than whether an exciting outcome emerges. This helps address the problem that null findings may remain unpublished. The method's strengths and limits are reviewed by Chambers and Tzavella (2022).

Preregistration alone does not guarantee adequate blinding, flawless software, representative samples, or honest reporting. A useful research package includes a dated protocol, open code when feasible, documentation of deviations, accessible data that protect participant privacy, and a record of quality-assurance checks. For sensitive extraordinary experiences, transparency must coexist with informed consent and appropriate safeguards for identifiable accounts.

Replication: The Essential Test of Stability

Replication asks whether a specified result appears again under conditions that make the comparison meaningful. Direct replication attempts to reproduce key procedures and outcomes. Conceptual replication tests a related prediction using a different implementation; it may strengthen a theory but can be harder to interpret when the method changes substantially.

A failed replication is informative. It may suggest that an initial finding was a false positive, that the effect is smaller than estimated, that a relevant boundary condition was missed, or that an implementation differed. These possibilities must be examined against preregistered predictions and objective procedural records. Simply inventing an unmeasured special condition after every failure would remove the claim from decisive testing.

Replication is a general concern across psychology. The Open Science Collaboration (2015) attempted to reproduce 100 psychological studies and found a substantially lower fraction of statistically significant replication results than statistically significant original results. Its findings are about studies in psychology more broadly; they do not supply a success rate for paranormal research. They explain why a single striking study, in any domain, rarely warrants a settled conclusion.

Scientific reliability also benefits when laboratories with differing expectations agree in advance on methods, outcomes, and analysis. Adversarial collaboration can expose ambiguous assumptions while reducing incentives to dispute procedures only after results appear.

A Detailed Case Study: Bem's Precognition Experiments and Their Replications

In 2011, Daryl Bem published nine experiments testing claimed retroactive influences on cognition and affect, with more than 1,000 participants overall. He presented statistically significant outcomes in most of the experiments and argued for evidence consistent with a psi-related interpretation. The publication was scientifically notable because it appeared in a major peer-reviewed psychology journal and used recognizable experimental paradigms. Bem (2011).

Methodologists quickly questioned the interpretation of those results. Wagenmakers et al. (2011) argued that partly exploratory data analysis and one-sided significance tests overstated evidential strength; their Bayesian reanalysis characterized the evidence as weak to nonexistent. This disagreement concerns statistical inference and research design, rather than the sincerity of any participant's experience.

Three preregistered independent attempts to reproduce Bem's retroactive-facilitation-of-recall effect yielded no statistically significant replication: the combined sample was 150 and the reported combined one-tailed p value was .83. Those results are evidence about that particular experimental effect, not a complete empirical test of every possible extraordinary phenomenon. Ritchie, Wiseman, and French (2012).

The multi-laboratory Transparent Psi Project later sought stronger controls over reproducibility and transparency. Collaborators included researchers with different views of the original hypothesis. They documented protocol fidelity through tools such as laboratory logs, video-documented sessions, direct data deposition, and external research auditing. The reported successful-guess rate was 49.89%, compared with 53.07% in the original Bem experiment and a nominal chance rate of 50%. The original result was not replicated. Kekecs et al. (2023).

A further meta-scientific series published in November 2025 involved three studies, 26,483 participants, and 420,472 critical trials. The first study did not replicate the original predicted precognition effect. An exploratory observation in the opposite direction informed a confirmatory second study that produced a below-chance result; a third study then failed to replicate that second-study result. The investigators favored a methodological or statistical account over an extraordinary mechanism given the failure of further replication and the absence of an accepted explanatory theory. The episode is valuable precisely because an intriguing intermediate result remained subject to a further high-powered test. Walleczek et al. (2025).

The cumulative lesson concerns scientific process: strong claims require stable, independently reproducible outcomes under controls that make ordinary explanations unlikely. The Bem sequence supplies detailed case-specific evidence that the tested precognition effects have not achieved that standard in these rigorous replications. It cannot logically settle every differently defined psi claim.

How Researchers Evaluate Contested Meta-Analyses

Meta-analysis combines outcomes across studies, which may reveal patterns that small individual experiments cannot detect. However, a summary estimate is only as trustworthy as its selection criteria, underlying designs, dependence assumptions, statistical model, and treatment of unpublished outcomes.

For extraordinary-effect claims, investigators must inspect whether studies were preregistered, whether exploratory tests were treated as confirmatory, whether the same participants or datasets contributed multiple dependent estimates, and whether low-quality or unreviewed work changes the pooled result. Analyses of funnel-plot asymmetry or sensitivity to missing studies can be informative but are not universal solutions to publication bias. When the underlying study designs vary substantially, a numerical pooled effect may conceal scientifically important differences.

A supportive review of experimental psi evidence was published by Cardeña (2018). Its interpretation remains contested, and the proper response is to examine the underlying experiments and later independent replications. The methodology-centered 2025 replication paper explicitly discussed concerns about exploratory designs and evidence quality in some psi meta-analytic collections. Walleczek et al. (2025).

The existence of a meta-analysis establishes that a body of research has been aggregated. It does not establish that the proposed extraordinary mechanism has been demonstrated. Conversely, criticism of one synthesis should identify its specific weaknesses rather than serve as a substitute for examining every study.

Evidence Hierarchy: What Different Findings Establish

A personal account documents an experience as reported. A repeated pattern of similar reports establishes something about reporting and interpretation in a defined population. A well-controlled experiment can establish that an operationally defined outcome differed from a justified comparison under its tested conditions. A reproducible series of independent controlled experiments strengthens the case that a phenomenon is stable. A proposed external mechanism requires additional work showing how it produces distinctive observations and outperforms relevant alternatives.

Peer review is an important quality filter, but its judgments occur before the long-term test of independent reproducibility is complete. Published results can be corrected, challenged, or reversed. A single successful study with a precise estimate may still be vulnerable to a hidden flaw; a single unsuccessful study may still be underpowered. The evidentiary conclusion should track the entire relevant, quality-assessed body of work.

The terms established mechanism, preliminary evidence, plausible interpretation, and contested claim describe different scientific states. Well-documented psychological phenomena such as expectation effects and reconstructive memory do not automatically explain each extraordinary report. Likewise, an unexplained individual episode does not provide positive experimental evidence for a paranormal cause. Separating observation, explanation, and causal verification is the core discipline of responsible anomalistic research.

Participant Recruitment, Culture, and Research Ethics

Recruitment determines whose experiences appear in a study. Convenience samples of undergraduates, self-selected visitors to unusual locations, or people with strong prior interests may be useful for particular hypotheses but do not represent all adults. Cross-cultural differences may affect the vocabulary, emotional salience, and interpretation of unusual reports. Surveys should avoid combining distinct belief traditions into a single judgmental category and should evaluate whether measurement items have the same meaning across groups.

Research should respect the psychological reality of a participant's reported experience while remaining neutral about unverified external causes. Informed consent must disclose procedures, foreseeable distress, recordings, data use, and the ability to withdraw within the applicable protocol. Researchers should avoid humiliation, deceptive claims of clinical authority, pressure to recover memories, or interview methods that implant event details.

Unusual experiences are not diagnoses. When a participant reports significant fear, impaired functioning, sleep disruption, or accompanying symptoms that warrant care, a study can explain appropriate support or professional assessment without diagnosing on the basis of a questionnaire or a single account. Scientific teams must distinguish a report of an event from an assertion that the researcher has verified its source.

Religious beliefs and devotional experiences belong to their own psychological and cultural research contexts. Scientific work on nonconfessional paranormal claims has no mandate to judge sacred traditions. The choice of wording, recruitment materials, and interpretation should preserve that boundary.

How to Read a Study Making an Extraordinary Claim

A careful reader begins with the exact claim and then inspects its evidentiary chain. Was the primary prediction recorded before outcome data were available? Is the outcome objectively specified, and was the chance model appropriate? Did target generation and information concealment prevent ordinary access? Were participants, experimenters, and scorers blinded in ways that actually protected the comparison? Were exclusions, stopping rules, multiple tests, and protocol deviations disclosed?

Next, inspect the size and uncertainty of the estimated effect. Ask whether the sample could detect the predicted difference, whether unusual subgroups were selected after the fact, and whether independent labs have published both successful and unsuccessful replications. Finally, look for shared data, analytic code, audit reports, and serious engagement with alternative explanations. These questions apply equally to supportive and skeptical papers. Simmons et al. (2011), Chambers and Tzavella (2022), and Kekecs et al. (2023) provide foundations for this appraisal.

One red flag is a dramatic causal conclusion resting on an unregistered, post hoc analysis. Another is the opposite error: treating a null result from an uninformative or poorly controlled test as proof that every possible version of a claim has been disproven. High-quality appraisal examines how much the actual design could have learned.

Frequently Asked Questions

Do scientists study paranormal claims seriously?

Yes. Scientific journals contain experimental, review, and methodological work on claimed psi effects and on the psychology of extraordinary experiences. The presence of such research means the questions have been investigated using scientific tools. It does not imply that an extraordinary external cause has been established. Cardeña (2018); Walleczek et al. (2025).

What is the difference between anomalistic psychology and parapsychology?

Anomalistic psychology focuses on understanding reported unusual experiences and beliefs through psychological and allied scientific research, including explanations involving perception, memory, expectation, and interpretation. Parapsychology has traditionally investigated whether purported psi effects occur. Their methods sometimes overlap, and no single research design answers all of their questions.

Can a statistically significant result prove ESP?

No. It can signal a deviation from the prediction of a statistical model under specified assumptions. A claim of ESP needs robust protection against ordinary information transfer and analytic artifacts, independent replication, a meaningful effect estimate, and converging causal evidence. Statistical significance alone cannot identify the causal mechanism.

What makes a good replication?

A good replication has a precise target claim, predeclared primary outcome, adequately justified sample, faithful and auditable procedures, independent execution, appropriate analysis, and full reporting regardless of the result. The value of a replication depends on how well it tests the original claim, rather than whether it produces a desirable outcome.

Has recent research replicated Bem's precognition claim?

Rigorous efforts described in 2012, 2023, and 2025 did not consistently reproduce the original predicted effect. The 2025 series contained an intermediate below-chance result that failed in a further study. The literature therefore does not establish the tested precognition phenomenon as a reliable effect. Ritchie et al. (2012); Kekecs et al. (2023); Walleczek et al. (2025).

Why does preregistration matter so much?

It records what researchers intended to test before observing the results, reducing ambiguity between genuine prediction and post hoc explanation. It complements, rather than replaces, strong controls, adequate measurement, independent audits, and replication.

Are paranormal beliefs or unusual experiences signs of psychiatric illness?

Neither belief nor unusual experience, taken alone, establishes a psychiatric diagnosis. Clinically relevant questions concern context, distress, functional impairment, additional symptoms, and appropriate professional assessment. A research study about paranormal attribution must avoid turning a participant's narrative into a diagnosis.

Can a ghost sighting, premonition, or sensed presence be scientifically investigated?

Yes. Researchers can study the circumstances of the report, the person's phenomenology, contemporaneous records, environmental variables, perception, memory, and candidate explanations. Demonstrating that a person experienced something and demonstrating a particular extraordinary external cause are distinct evidentiary tasks.

Conclusion

Scientific investigation of paranormal claims succeeds when it makes extraordinary reports accessible to precise measurement and potentially decisive tests. Reports are worthy of serious inquiry, controlled results must be interpreted according to their design, and causal explanations earn confidence through transparent evidence and successful independent replication. Modern preregistration, blinding, open research materials, and large collaborative tests provide a disciplined way to examine claims whose mechanisms remain contested. The strongest scientific account describes both what has been observed and what the evidence has yet to establish.

Related Articles

Anomalistic Psychology: What It Studies and How Evidence Is Evaluated — parent scientific overview of anomalistic psychology, evidence standards, and its relation to parapsychology.

Haunted Places: Psychology, Environmental Factors, and Evidence — evidence about reputedly haunted locations, environmental measurements, expectation, and methodological limits.

Sleep Paralysis: Symptoms, Hallucinations, and What Helps — a separately owned explanation of sleep-related experiences that can inform differential interpretation.

False Memory OCD: What It Is, Memory Doubt, Mental Review, Guilt, and Reassurance Seeking — a separately owned clinical discussion of obsessional memory doubt rather than an explanation of ordinary extraordinary-experience reports.

Measuring Paranormal Beliefs: Scales, Validity, and Limits — belief questionnaires as a distinct measurement problem within anomalistic research.

Paranormal Beliefs: Psychological Research on Why People Believe — the substantive psychological evidence base concerning paranormal belief.

Psychic Experiences and ESP: What Psychology Can Study

References

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. https://doi.org/10.1037/a0021524

Cardeña, E. (2018). The experimental evidence for parapsychological phenomena: A review. American Psychologist, 73(5), 663–677. https://doi.org/10.1037/amp0000236

Chambers, C. D., & Tzavella, L. (2022). The past, present and future of Registered Reports. Nature Human Behaviour, 6, 29–42. https://www.nature.com/articles/s41562-021-01193-7

Dean, C. E., Akhtar, S., Gale, T. M., Irvine, K., Grohmann, D., & Laws, K. R. (2022). Paranormal beliefs and cognitive function: A systematic review and assessment of study quality across four decades of research. PLOS ONE, 17(5), e0267360. https://doi.org/10.1371/journal.pone.0267360

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. https://doi.org/10.1098/rsos.191375

Open Science Collaboration. (2015). Estimating the reproducibility of psychological science. Science, 349(6251), aac4716. https://doi.org/10.1126/science.aac4716

Ritchie, S. J., Wiseman, R., & French, C. C. (2012). Failing the future: Three unsuccessful attempts to replicate Bem's “retroactive facilitation of recall” effect. PLOS ONE, 7(3), e33423. https://doi.org/10.1371/journal.pone.0033423

Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-positive psychology: Undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological Science, 22(11), 1359–1366. https://doi.org/10.1177/0956797611417632

Wagenmakers, E.-J., Wetzels, R., Borsboom, D., & van der Maas, H. L. J. (2011). Why psychologists must change the way they analyze their data: The case of psi. Journal of Personality and Social Psychology, 100(3), 426–432. https://doi.org/10.1037/a0022790

Walleczek, J., von Stillfried, N., Schmidt, S., Wittmann, M., Kirmse, K. A., Moll, J., & Kekecs, Z. (2025). Metascientific replication project with the advanced meta-experimental protocol of the transparent psi project procedures for testing the precognitive effect claimed by Bem. PLOS ONE, 20(11), e0335330. https://doi.org/10.1371/journal.pone.0335330