Overview

Micro-psychokinesis, usually shortened to micro-PK, names a claimed effect in which focused intention nudges the output of a random process. The target is normally a hardware random number generator (RNG), also called a random event generator (REG): a device that turns electronic noise into a stream of unpredictable bits. The claim is narrow. A participant tries to make the device output more ones than zeros, and the bit stream drifts a tiny amount away from a fair coin.

The effect is statistical only. It is not visible to the eye and shows up only after many trials are pooled and compared against chance. This distinguishes micro-PK from macro-PK, which claims direct, visible movement of objects such as bending metal. It also differs from DMILS (direct mental interaction with living systems), where the target is another person’s body rather than non-living hardware. Micro-PK targets noise-seeded electronics.

Phenomenology: what is reported

In a typical session a person watches a screen and tries to push a running tally of random bits in an assigned direction. The reported signal is small. Across large pooled datasets the bias sits at the edge of noise, a fraction of a percent away from chance, yet investigators report it as consistent enough to reach statistical significance when many trials accumulate.^[1]

Researchers have described variations on the basic theme. Some report that the effect appears as a change in the variance of the output rather than a steady shift in the mean.^[2] Others have argued the effect behaves as if pulled by the intended goal rather than driven step by step, and may even read as running backward in time.^[3] A recurring report is that the effect weakens with repetition, a pattern often called the decline effect.

History of the label

The modern form of micro-PK began with Helmut Schmidt around 1969-1970, when he built electronic generators seeded by radioactive decay and asked people to bias their output. His design replaced the older method of throwing dice with a fast, automated, electronically logged device. The shift to electronic targets later supported a meta-analysis of dice studies that pooled more than two million throws and reported a weak but positive effect once methodological bias was controlled.^[4]

The label gained wider scientific visibility through the Princeton Engineering Anomalies Research (PEAR) program. Operators there tried to shift REG output above or below baseline, work later summarized by Roger D. Nelson and colleagues as the founding paradigm of the lab. The PEAR effort, associated with Brenda J. Dunne, accumulated very large databases over years of operation and became the most-cited body of micro-PK data.

How it is studied

The core method is simple to describe. A noise source generates random bits, a participant is assigned a direction, and the pooled output is compared to what chance alone would produce. Controls include calibration runs with no intention present and counterbalanced high/low conditions.

A neural-network analysis of PEAR data tested whether individual operators left a recognizable personal pattern, reporting that a trained network could link people to their own data and transfer that learning to new runs.^[5] Later groups extended the targets to quantum optics, asking whether attention perturbs a double-slit interference pattern^[6] or modulates entangled photon pairs.^[7] A separate field-style line of work, the Global Consciousness Project, runs networked generators continuously and looks for departures from randomness around moments of mass attention.^[8]

Evidence summary

The most influential positive synthesis pooled 832 REG experiments from 68 investigators and reported a small, statistically significant effect that the authors said survived variation in study quality and could not be wiped out by selective reporting.^[1] A broader review across decades argued that probabilistic mind-matter effects appear repeatedly across dice, RNG, and related tasks.^[9]

A meta-analysis of 832 random-event generator experiments from 68 investigators reported a small but statistically significant bias that the authors said persisted across differences in study quality.^[1]

More recent work has tried to specify when the effect should appear. One large preregistered study reported that observers’ fear-based personality traits predicted the direction of the bias.^[10] Another large online study reported that aligning outcomes with what participants implicitly expect can sustain the effect and blunt the decline.^[11] Replication attempts have been mixed: a careful blind replication of a correlation-matrix design found no significant effect in its main analysis.^[12]

Individual studies in the reference library

The studies listed below are individual micro-PK results and analyses currently held in the ESP-Nexus reference library beyond those the article already cites. What share of the published literature on micro-PK they represent has not been measured, so the table summarizes what the library holds rather than counting what has been published. The studies also report different kinds of number — z values, Bayes factors, hit rates and p values alone — and those cannot be added together into one bottom-line figure.

Both directions are well represented in the table. The positive results include the PEAR program review, whose composite deviation across more than 1,000 series carries p = 3.5 × 10⁻¹³ with a high–low separation of z = 3.81,^[21] the remote-operation database, in which high-intention efforts reached z = 3.185 while low efforts were indistinguishable from chance,^[20] Schmidt’s report that some subjects can affect the random generator, an effect he states was confirmed by a large number of different experimenters,^[19] and the opening results of the Munich series: strong Bayesian evidence for an effect among smokers (BF₁₀ = 66.06)^[25] and in a positive-priming condition at 4,092 participants (BF₁₀ = 13.35).^[28] The null and critical results include the 380-study meta-analysis, which found a significant but very small overall effect and closed by extending the older verdict of “not proven” to human intentionality on RNGs,^[22] the three-laboratory replication of the PEAR benchmark, which came in at a nonsignificant overall z = 0.6,^[23] Grote’s 720,000-bit correlation study, in which none of the pre-defined analyses was significant (combined p = 0.315),^[24] the 12,571-participant online experiment that reported strong evidence for the null (BF₀₁ = 10.07),^[26] and the preregistered correlational replication that ended with the discovery of a data error in the original dataset.^[29]

Several of the papers qualify their own results directly. Bösch, Steinkamp, and Boller accept that the pooled database shows a significant overall effect while attributing the pattern of larger effects in smaller studies to publication bias, a reading contested in the published response by Radin, Nelson, Dobyns, and Houtkooper.^[13] Varvoglis and Bancel attribute the failed consortium replication to the assumption that the PEAR data were homogeneous across participants, when two extreme outlier operators contributed nearly a quarter of the total data, and conclude that micro-PK “is not widely distributed but is exceptional.”^[23] Dechamps, Maier, Pflitsch, and Duggan report that the data of their four studies align with standard quantum mechanics while adding that the data “do not entirely falsify” the observer models under test.^[28] Dechamps and Maier report a peak Bayes factor of 421.2 at participant 134 followed by a decline to an overall Bayes factor below 1, and tested a damped-oscillation description of that trajectory against 10,000 simulated datasets, with the preregistered confirmatory analyses yielding no definite results.^[27]

Skeptical critiques

The sharpest published challenge came from a 2006 meta-analysis in Psychological Bulletin by Bösch, Steinkamp, and Boller, which concluded that psychokinesis on RNGs was not proven. The authors pointed to a striking pattern: the smaller studies showed larger effects, which they read as a fingerprint of publication bias rather than a real signal. Defenders replied that this reading rests on assuming effect size should be independent of sample size, and argued that correcting that assumption leaves the cumulative data persuasive.^[13] A companion paper restated the case that publication bias and quality concerns do not plausibly account for the pooled result.^[14]

A second line of critique is internal to the field. Even sympathetic analysts disagree about what the effect is. One exchange framed the dispute as whether large-network results reflect a field-like influence of group attention or instead goal-oriented psychokinesis by the experimenters themselves.^[15] A Bayesian critique of one micro-PK dataset argued the evidence pointed against the effect; the original team replied that controlling for cumulative-score artifacts restored a marginal signal.^[16] These debates show that even within parapsychology the interpretation is contested.

A third line of critique comes from the micro-PK laboratories’ own preregistered continuations. Maier, Dechamps, and Pflitsch ran an online experiment to a preset Bayesian stopping criterion, reached 12,571 participants, and reported strong evidence for the null hypothesis (BF₀₁ = 10.07), with the mean score for positive stimuli at 50.02% against 50% chance; the paper’s own title states that the analysis reveals evidence against micro-psychokinesis.^[26] In the same group’s four-study series, strong initial evidence in Study 1 was followed by three preregistered replications that showed no deviations from the Born rule,^[28] and a later preregistered replication of a correlational effect found that the original finding rested on a data error — the authors describe it as a failed replication of an erroneous finding.^[29]

Current status

Micro-PK remains an active but unsettled research area. The reported effect, when present, is small and inconsistent, and the decline effect continues to complicate replication. Recent effort has moved in two directions: toward quantum-optical targets such as double-slit and single-photon systems,^[17] and toward theory-driven designs that predict the size and direction of any bias from observer psychology before data collection.

Field-style work also continues, with networked generators monitored around collective events and group settings.^[18] Whether these correlations reflect a genuine mind-matter interaction or unresolved statistical and selection artifacts is the central open question. The two camps agree on the raw observation that pooled RNG output sometimes departs from chance; they disagree on whether intention is the cause.