Dean I. Radin, PhD Sources: off on

Mind-Matter Interaction via Random-Event Generators and Psychophysiological Correlates

For more than four decades, Dean Radin has investigated whether human intention and attention can produce statistically detectable anomalies in the output of truly random number generators (RNGs) and in psychophysiological measures such as electrodermal activity and EEG. His work spans individual micro-psychokinesis experiments, large-scale meta-analyses, and novel paradigms using quantum-optical systems, collectively forming one of the most extensive empirical programs on mind-matter interaction in contemporary parapsychology. A persistent methodological challenge in this literature is that typical effect sizes are very small (r ≈ 0.01–0.02), meaning that underpowered single studies can easily miss genuine effects or produce false positives, making cumulative meta-analytic approaches essential for evaluating the evidence.

Deeper dives — Radin:

Back to hub Automated Ganzfeld Double-Slit Consciousness Global Consciousness Project Presentiment and Electrodermal Intention-Mediated Effects Sequential Analysis Methods Meditation, Consciousness Anomalous Cognition 📖 Real Magic 📖 Entangled Minds 📖 Supernormal 📖 The Conscious Universe

Key findings

A 1989 meta-analysis of more than 800 RNG experiments found statistically significant non-chance effects in experimental conditions and chance-expected results in controls (though the magnitude and interpretation of this meta-analytic effect remain disputed), with effect sizes small but consistent across independent laboratories.¹
A 2003 updated meta-analysis spanning 1959–2000 confirmed the earlier findings, with the cumulative evidence resisting conventional publication-bias explanations under Radin and Nelson’s analysis of effect-size heterogeneity.²
A 1991 meta-analysis of dice-fall experiments (148 studies, over 2 million throws, 2,569 subjects) found evidence for a small but statistically significant consciousness-related bias after correcting for a physical loading artifact.³
Electrodermal presentiment experiments (N=109 participants, 3,709 trials) replicated an anomalous pre-stimulus EDA elevation before emotional photographs across three independent studies.⁴
A 19-year online forced-choice experiment (114 million trials, ~200,000 participants) found a null overall hit rate but a highly significant sequential pattern (z = 11.28, p = 1.7 × 10⁻²⁹) consistent with a pre-planned secondary hypothesis about psi structure.⁵
Smartphone-based micro-PK and precognition tests with thousands of participants found that psi performance was often in the direction opposite to conscious intention, and that gender and psi belief moderated outcomes.⁶

Overview

Radin’s engagement with random-event generator (REG/RNG) research began in the early 1980s with individual laboratory experiments testing whether mental intention could shift the statistical distribution of truly random bit sequences generated by hardware noise sources.⁷ He recognized early that single experiments were underpowered to detect the small effect sizes characteristic of this domain, and that the field required cumulative meta-analytic synthesis to assess the evidence fairly.¹ Alongside RNG work, Radin developed a parallel psychophysiological research program examining whether the body’s autonomic nervous system registers information about future emotional events before those events occur, a phenomenon he termed “presentiment.”⁸ Both lines of research share a common methodological logic: use objective, continuously recorded physical or physiological measures to detect anomalies that cannot be attributed to sensory information, and apply rigorous statistical methods to evaluate whether the anomalies exceed chance expectation.

Scope and Methodological Philosophy

Radin has argued that the mind-matter interaction literature suffers from a Type-II vulnerability: because true effect sizes are very small (typically r ≈ 0.01–0.02 in RNG studies), individual experiments are almost always underpowered, making null results uninformative and positive results fragile. His preferred solution has been meta-analysis combined with sequential analysis methods designed to detect cumulative signal across many trials.¹² He has also explored neural network approaches to identify person-specific “signatures” in RNG data, patterns unique to individual operators that might be masked in aggregate analyses.⁹¹⁰

RNG Meta-Analyses and the Foundational Evidence Base

The 1989 Radin and Nelson meta-analysis, published in Foundations of Physics, synthesized more than 800 experiments from the parapsychological literature and found that experimental conditions produced non-chance effects while control conditions conformed to chance expectation, a pattern the authors argued was inconsistent with a simple file-drawer or publication-bias account.¹ A subsequent 1991 meta-analysis of dice-fall experiments extended this conclusion to a different physical system, finding a small but statistically significant consciousness-related bias after correcting for a physical loading artifact that had inflated earlier estimates.³ These two meta-analyses established the empirical foundation that Radin and collaborators continued to build upon through the 2000s.

1989 Meta-Analysis: Effect Sizes and Publication-Bias Assessment

Radin and Nelson’s 1989 review covered more than 800 experiments reported in the parapsychological literature, finding a small but consistent positive effect in experimental conditions and chance-level results in controls. The authors addressed the file-drawer artifact (selective reporting of positive results) by estimating the number of unpublished null studies that would be required to reduce the overall effect to non-significance, a calculation that yielded an implausibly large number, suggesting publication bias alone could not account for the findings. The proposed mechanism, that human intention directly influences quantum-level random processes, was explicitly labeled as the researcher’s preferred interpretation; alternative interpretations including subtle methodological artifacts and optional stopping were acknowledged but argued to be insufficient to explain the full pattern of results.¹

Dice Meta-Analysis: Physical Bias Correction and Residual Effect

The 1991 Radin and Ferrari dice meta-analysis covered 73 English-language reports published from 1935 to 1987, comprising 148 studies by 52 investigators involving more than 2 million dice throws from 2,569 subjects. A critical methodological contribution was the identification and correction of a physical loading artifact: dice are not perfectly balanced, and heavier faces tend to land face-down, biasing outcomes toward lower-numbered faces. After correcting for this artifact, a residual effect consistent with a consciousness-related influence remained statistically significant. The multiple-comparisons artifact was partially addressed by treating each study as a single data point in the meta-analytic model rather than pooling individual trials, mitigating but not fully eliminating concerns about within-study optional stopping.³

In 2006, Radin, Nelson, Dobyns, and Houtkooper published a formal response in Psychological Bulletin to a competing meta-analysis by Bösch, Steinkamp, and Boller that had concluded the RNG evidence was attributable to selective reporting. Radin and colleagues argued that Bösch et al.’s analysis rested on an incorrect assumption, that effect size is entirely independent of sample size, and that this assumption mathematically guarantees the appearance of heterogeneity even when a genuine small effect is present.¹¹

The Bösch et al. Dispute: Effect-Size Heterogeneity and Selective Reporting

Bösch, Steinkamp, and Boller’s 2006 meta-analysis in Psychological Bulletin found that RNG studies showed statistically significant overall effects but with heterogeneous effect sizes, and concluded that selective reporting (file-drawer) was the most plausible explanation. Radin, Nelson, Dobyns, and Houtkooper’s published comment argued that the heterogeneity was an artifact of the “influence-per-bit” assumption, the incorrect premise that any mind-matter effect operates uniformly per random bit regardless of sample size or generation rate. Under this assumption, larger samples should show smaller per-bit effects, which is exactly what the data show; Radin et al. argued this pattern is consistent with a genuine small effect, not with selective reporting. The dispute remained unresolved in the primary literature, with both interpretations receiving continued support from different analysts.¹¹¹²

Modern Context

The methodological dispute over RNG meta-analyses sits within a broader mainstream debate about how to interpret heterogeneous effect sizes in small-effect literatures. The “influence-per-bit” assumption contested by Radin and colleagues parallels mainstream statistical discussions about whether effect sizes should be expected to be constant or to vary with sample characteristics, a question that has no settled answer in general statistics. Radin’s neural network approach to identifying person-specific signatures in RNG data⁹ anticipates later mainstream interest in individual-differences moderators of small effects, though the specific application to mind-matter interaction remains outside mainstream physics and psychology. The broader question of how consciousness relates to quantum measurement, invoked as a proposed mechanism in Radin’s work¹³, remains an open and contested problem in the foundations of physics, with no consensus interpretation of the quantum measurement problem available to adjudicate between conventional and parapsychological accounts.¹⁴

Psychophysiological Correlates: Presentiment and EDA

Beginning with a 1997 paper in the Journal of Scientific Exploration, Radin reported that electrodermal activity (EDA), a measure of sympathetic nervous system arousal, was anomalously elevated in the seconds before participants were shown randomly selected emotional photographs, compared to the seconds before calm photographs.⁸ He termed this pattern “presentiment” and subsequently conducted three replication experiments with 109 participants contributing 3,709 trials, again finding higher pre-stimulus EDA before emotional than before calm images.⁴ The proposed mechanism, that the autonomic nervous system registers information about future emotional states before those states occur, is the researcher’s preferred interpretation; alternative explanations including subtle response-bias artifacts and optional stopping were addressed but not fully eliminated.

Presentiment EDA: Effect Sizes, Replication Record, and Artifact Mitigation

Radin’s 2004 replication series (N=109, 3,709 trials) reported that pre-stimulus EDA was significantly higher before emotional than before calm photographs (p = 0.002 in the original 1997 study; the 2004 replications again showed the differential effect). The specific artifact of response bias, the possibility that participants’ general arousal level differed systematically across trial blocks in a way that correlated with upcoming stimulus type, was partially addressed by the randomized, counterbalanced trial structure and by the use of a truly random stimulus-selection process, eliminating the possibility that participants could consciously anticipate upcoming stimuli. The optional-stopping artifact was partially mitigated by pre-specifying the number of trials per session, though stopping rules across the full series were not formally preregistered. A 2014 meta-analysis by Mossbridge, Tressoldi, Utts, Ives, Radin, and Jonas synthesized the broader physiological anticipatory activity literature (distinct from Bem’s 2011 behavioral paradigm) and found a consistent pre-stimulus effect across multiple independent laboratories.¹⁵

Radin also investigated EEG correlates of presentiment, examining whether electrocortical activity prior to unpredictable stimuli differed between experienced meditators and non-meditators.¹⁶ In a separate line of work, he explored whether EEG signals recorded in pairs of isolated individuals showed anomalous correlations when one member of the pair was stimulated, a paradigm testing for non-local physiological coupling.¹⁷

EEG Correlations Between Isolated Pairs: Design and Limitations

Radin’s 2004 EEG study recorded simultaneous EEGs from 13 pairs of volunteers (11 pairs of adult friends, 2 mother-daughter pairs). One member of each pair relaxed in a double steel-walled, electromagnetically and acoustically shielded room while the other, located 20 meters away in a dimly lit room, was stimulated at random times by the live video image of the first person. The key artifact addressed was electromagnetic cross-talk between recording systems: the shielded room was designed to eliminate direct electromagnetic coupling between the two EEG systems. The judging-contamination artifact was addressed by using automated event-related potential analysis rather than human judges. The primary limitation was the small sample (N=13 pairs), which provided insufficient power to detect the small effect sizes typical of this literature, making the results exploratory rather than confirmatory.¹⁷

Large-Scale Online Experiments and Sequential Analysis

Recognizing that laboratory experiments with small samples were chronically underpowered, Radin developed and deployed online psi experiments capable of collecting data from tens of thousands of participants over extended periods. A 19-year online forced-choice experiment running from August 2000 to December 2018 collected 114 million trials from an estimated 200,000 participants worldwide using a five-target protocol.⁵ The overall hit rate was consistent with chance, a null result on the primary hypothesis. However, a planned secondary analysis designed to detect a predicted sequential pattern in the data produced a small but statistically unambiguous outcome (z = 11.28, p = 1.7 × 10⁻²⁹), which Radin interpreted as evidence that psi effects manifest as structured sequential dependencies rather than simple mean shifts.

Sequential Pattern Analysis: The “Trickster” Hypothesis and Statistical Interpretation

The 19-year experiment combined data from two online psi tasks with a five-target forced-choice protocol. The primary hit rate across both experiments was p₀ = 0.32 (chance expected), consistent with a null effect. The planned secondary analysis tested a specific sequential hypothesis: that the probability of a hit on trial n+1 is predicted by the outcome of trial n in a particular pattern. The observed sequential statistic was p₁ = 0.320502 ± 0.000044, z = 11.28, p = 1.7 × 10⁻²⁹. Control tests showed no such pattern in pseudo-random sequences. The multiple-comparisons artifact is a concern here: the secondary analysis was one of several analyses applied to the dataset, and the degree to which it was truly pre-specified versus selected from a family of possible sequential statistics is not fully documented in the published report. Radin’s preferred interpretation is that psi effects manifest as sequential structure rather than mean shifts; an alternative non-psi interpretation is that subtle biases in participant response strategies could produce sequential dependencies that mimic the predicted pattern.⁵

A complementary smartphone-based study by Mossbridge and Radin deployed three iOS tasks related to micro-psychokinesis and precognition, collecting data from thousands of participants between 2017 and 2020.⁶ The large sample enabled examination of demographic and personality moderators of psi performance, finding that psi performance was often in the direction opposite to conscious intention, a pattern previously called “psi-missing“, and that gender and psi belief were related to performance direction and magnitude.

Smartphone Micro-PK and Precognition: Moderator Analysis

The Mossbridge and Radin smartphone study used three iOS tasks: one testing micro-psychokinesis (mental influence on an RNG-driven display) and two testing precognition (prediction of future random outcomes). The large convenience sample drawn from app-store users enabled moderator analyses that would be impossible in small laboratory studies. Key findings included that psi-missing (performance opposite to intention) was common, and that gender and psi belief moderated performance direction. The sampling-bias artifact is a significant concern: participants self-selected by downloading a psi-testing app, creating a sample biased toward psi believers and potentially toward individuals with prior psi-testing experience. The demand-characteristics artifact, participants performing in the direction they believed was expected, was partially addressed by the psi-missing finding (performance opposite to intention is inconsistent with simple demand compliance), but not fully eliminated. The pseudo-RNG used in smartphone apps, rather than true hardware RNG, is a methodological limitation relative to laboratory REG studies.⁶

Radin also explored whether group mental coherence, induced through binaural beat entrainment during multi-day workshops, produced anomalous deviations in RNGs located in the same building, finding exploratory evidence for correlations between entrained group states and RNG output across 14 workshops compared to 8 control periods.¹⁸ The proposed mechanism, that collective mental coherence influences nearby random physical systems, is the researcher’s preferred interpretation; the alternative explanation of subtle environmental confounds (temperature, vibration, electromagnetic fields correlated with group activity) was partially but not fully addressed by the shielded RNG design.

Group Coherence and RNG Anomalies: Workshop Study Design

The Radin and Atwater workshop study used two RNGs based on electronic noise and one based on radioactive decay latencies, continuously recording data during 14 workshops (experimental) and 8 control periods (same locations and times, no workshop activity). Coherence was entrained by having groups listen to prescribed binaural beat rhythms. The environmental-confound artifact, the possibility that group activity (body heat, movement, electromagnetic emissions from equipment) directly affected RNG hardware, was partially addressed by the use of radioactive-decay-based RNG (less susceptible to electromagnetic interference) alongside electronic-noise RNGs. The order-effects artifact was partially addressed by comparing workshop periods to matched control periods at the same times and locations. The study was explicitly labeled exploratory, and no formal stopping rule or preregistration was reported.¹⁸

Skeptical Critiques and Discussion

Critique 1: Conventional anticipatory mechanisms explain presentiment EDA findings without invoking precognition

Skeptic source: May, Paulinyi, and Vassy argued in a 2005 paper that anomalous anticipatory skin conductance responses to acoustic stimuli, the class of findings that includes Radin’s presentiment work, can be explained by Decision Augmentation Theory (DAT) and conventional anticipatory mechanisms rather than by genuine precognition. Under DAT, participants unconsciously use existing psi-like information-gathering to select into experimental conditions in ways that mimic pre-stimulus effects, without requiring that the body actually registers future information. May et al. also proposed that conventional physiological anticipation processes, operating below conscious awareness, could produce pre-stimulus EDA elevations that are mistakenly attributed to anomalous perception of future events.¹⁹

Response: Radin published a direct response to May et al. in the same journal, arguing that DAT does not eliminate the anomaly but merely relocates it, DAT itself requires a psi-like process (unconscious information access) to operate, so it does not constitute a conventional explanation. He further argued that the specific temporal structure of the pre-stimulus EDA effect, peaking in the 2–5 seconds before stimulus onset and then declining, is inconsistent with conventional anticipatory arousal, which would be expected to increase monotonically as the stimulus approaches. The 2014 Mossbridge et al. meta-analysis of predictive anticipatory activity across multiple independent laboratories, using physiological measures including EDA, EEG, and fMRI, found a consistent pre-stimulus effect that the authors argued could not be attributed to conventional anticipation alone.²⁰¹⁵

Analysis. The Mossbridge et al. 2014 meta-analysis provides moderate support for the rebuttal across independent laboratories, but the DAT alternative has not been definitively ruled out. Neither the conventional anticipation account nor the psi account has published a specific mechanism for the pre-stimulus temporal profile.

Critique 2: Effect-size heterogeneity in RNG meta-analyses is best explained by selective reporting rather than genuine mind-matter interaction

Skeptic source: The Bösch, Steinkamp, and Boller 2006 meta-analysis in Psychological Bulletin, addressed directly by Radin and colleagues in the same issue, concluded that while RNG studies show statistically significant overall effects, the heterogeneous distribution of effect sizes across studies is most parsimoniously explained by selective reporting (file-drawer effect): laboratories with positive results publish, while those with null results do not. Under this account, the apparent cumulative evidence for mind-matter interaction is an artifact of publication bias rather than a genuine physical phenomenon. This critique applies funnel-plot asymmetry reasoning: if a true effect exists, effect sizes should be independent of sample size; the observed negative correlation between effect size and sample size is the signature of publication bias.¹¹

Response: Radin, Nelson, Dobyns, and Houtkooper’s published comment in Psychological Bulletin argued that the negative correlation between effect size and sample size is not the signature of publication bias in this literature but rather a mathematical consequence of the incorrect “influence-per-bit” assumption. If a mind-matter effect operates at the level of individual experimental sessions rather than individual bits, then larger samples (more bits per session) will show smaller per-bit effect sizes even when the true session-level effect is constant. This argument implies that funnel-plot asymmetry tests, which assume effect-size independence from sample size, are inapplicable to RNG data as typically analyzed. The Parapsychological Association‘s policy against selective reporting of negative results, adopted in 1975, was also cited as a structural mitigation of the file-drawer problem, though its effectiveness in practice is difficult to verify.¹¹²¹

Analysis. Both the selective-reporting interpretation and the influence-per-bit rebuttal have been formally published in a mainstream peer-reviewed journal. The dispute turns on a statistical modeling assumption that has not been resolved by subsequent independent analysis. Neither side has produced a definitive empirical test that distinguishes between the two accounts, leaving the interpretation of RNG meta-analytic heterogeneity genuinely open.

References

Radin, D. I., & Nelson, R. D. (1989). Evidence for consciousness-related anomalies in random physical systems. Foundations of Physics, 19(12), 1499–1514. https://doi.org/10.1007/bf00732509 R001 [Radin & Nelson 1989] ↩︎
Radin, D. I., & Nelson, R. D. (2003). A meta-analysis of mind-matter interaction experiments from 1959 to 2000. Churchill Livingstone. R002 [Radin 2003] ↩︎
Radin, D. I., & Ferrari, D. C. (1991). Effects of consciousness on the fall of dice: A meta-analysis. Journal of Scientific Exploration, 5(1), 61–83. https://web.archive.org/web/20230331215150/https://www.scientificexploration.org/docs/5/jse_05_1_radin.pdf R003 [Radin 1991] ↩︎
Radin, D. I. (2004). Electrodermal Presentiments of Future Emotions. Journal of Scientific Exploration, 18(2), 253–274. https://www.semanticscholar.org/paper/Electrodermal-Presentiments-of-Future-Emotions-Radin/e00ddef190ee9c5134e60ced96a691d75c97fdc8 R004 [Radin 2004] ↩︎
Radin, D. I. (2019). Tricking the Trickster: Evidence for Predicted Sequential Structure in a 19-Year Online Psi Experiment. Journal of Scientific Exploration, 33(4), 549–568. https://journalofscientificexploration.org/index.php/jse/article/view/1429 R005 [Radin 2019] ↩︎
Mossbridge, J., & Radin, D. I. (2021). Psi Performance as a Function of Demographic and Personality Factors in Smartphone-Based Tests. Journal of Anomalous Experience and Cognition, 1(1-2), 78–113. https://journals.lub.lu.se/jaex/article/view/23419 R006 [Mossbridge 2021] ↩︎
Radin, D. I. (1984). Mental influence on machine-generated random events: Six experiments. Research in Parapsychology 1983 (R. A. White, Ed.). https://www.fourmilab.ch/rpkp/pk-rad.html R007 [Radin 1984] ↩︎
Radin, D. I. (1997). Unconscious perception of future emotions: An experiment in presentiment. Journal of Scientific Exploration, 11(2), 163–180. R008 [Radin 1997] ↩︎
Wallisch, Pascal, Dean I. Radin, Lusignan, Michael, Benayoun, Marc, Baker, Tanya I., Dickey, Adam S., Hatsopoulos, Nicholas G. (2009). Neural Network Part II. Matlab for Neuroscientists, 319–337. https://doi.org/10.1016/b978-0-12-374551-4.00029-4 R009 [Wallisch 2009] ↩︎
Radin, D. I. (1993). Neural Network Analyses of Consciousness-Related Patterns in Random Sequences. Journal of Scientific Exploration, 7(4), 355–374. R010 [Radin 1993] ↩︎
Radin, D. I., Nelson, R., Dobyns, Y., & Houtkooper, J. M. (2006). Reexamining psychokinesis: Comment on Bösch, Steinkamp, and Boller (2006). Psychological Bulletin, 132(4), 529–532. https://doi.org/10.1037/0033-2909.132.4.529 R011 [Radin 2006] ↩︎
Radin, D. I. (2006). Experiments Testing Models of Mind-Matter Interaction. Journal of Scientific Exploration, 20(3), 361–374. https://web.archive.org/web/20240712074323/https://www.scientificexploration.org/docs/20/jse_20_3_radin_1.pdf R012 [Radin 2006] ↩︎
Kauffman, Stuart A., Dean I. Radin (2023). Quantum aspects of the brain–mind relationship: A hypothesis with supporting evidence. Biosystems, 223, 104820. https://doi.org/10.1016/j.biosystems.2022.104820 R013 [Kauffman 2023] ↩︎
Radin, D. I. (2019). Don’t Look at My Hand: A Response to “Quantum Misuse in Psychic Literature”. Journal of Near-Death Studies, 37(3), 171–173. https://digital.library.unt.edu/ark:/67531/metadc1752583/ R014 [Radin 2019] ↩︎
Mossbridge, J. A., Tressoldi, P., Utts, J., Ives, J. A., Radin, D., & Jonas, W. B. (2014). Predicting the unpredictable: Critical analysis and practical implications of predictive anticipatory activity. Frontiers in Human Neuroscience, 8, 146. https://doi.org/10.3389/fnhum.2014.00146 R015 [Mossbridge 2014] ↩︎
Radin, D. I., Vieten, C., Michel, L., & Delorme, A. (2011). Electrocortical Activity Prior to Unpredictable Stimuli in Meditators and Nonmeditators. EXPLORE, 7(5), 286–299. https://doi.org/10.1016/j.explore.2011.06.004 R016 [Radin 2011] ↩︎
Radin, D. I. (2004). Event-Related Electroencephalographic Correlations Between Isolated Human Subjects. The Journal of Alternative and Complementary Medicine, 10(2), 315–323. https://doi.org/10.1089/107555304323062301 R017 [Radin 2004] ↩︎
Radin, D. I., & Atwater, F. H. (2009). Exploratory Evidence for Correlations Between Entrained Mental Coherence and Random Physical Systems. Journal of Scientific Exploration, 23(3), 263–272. https://journalofscientificexploration.org/index.php/jse/article/view/93 R018 [Radin 2009] ↩︎
May, E. C., Paulinyi, T., & Vassy, Z. (2005). Anomalous Anticipatory Skin Conductance Response to Acoustic Stimuli: Experimental Results and Speculation About a Mechanism. The Journal of Alternative and Complementary Medicine, 11(4), 695–702. https://doi.org/10.1089/acm.2005.11.695 R019 [May 2005] ↩︎
Radin, D. I. (2005). May et al.’s “Anomalous Anticipatory Skin Conductance Response to Acoustic Stimuli”. The Journal of Alternative and Complementary Medicine, 11(4), 587–588. https://doi.org/10.1089/acm.2005.11.587 R020 [Radin 2005] ↩︎
Radin, D. I. (2007). Finding Or Imagining Flawed Research? The Humanistic Psychologist, 35(3), 297–299. https://doi.org/10.1080/08873260701578384 R021 [Radin 2007] ↩︎

Deeper dives — Radin: