Discovery Lab Answers

RNG / Micro-PK Research on ESP-Nexus

A note on coverage first: the ESP-Nexus library’s share of the published literature on RNG micro-PK has not been measured. What follows summarizes the 39 studies and 43 result rows currently in the library — treat this as a snapshot of what the library holds, not a settled account of the full field.

What RNGs are in this context

Random Number Generators (RNGs), also called Random Event Generators (REGs) in parapsychology, are electronic or computer-based devices that produce statistically unpredictable binary outputs. In psi research, they serve as the target system: a participant attempts to mentally bias the output in a pre-specified direction, and the deviation from expected randomness is measured. The paradigm is called micro-PK (microscopic psychokinesis) because any detected effect is vanishingly small — a shift in the probability of a binary “1” on the order of fractions of a percent.

What the library’s evidence shows

The 43 result rows in the library span studies from 1989 through 2025 and report a mix of outcomes — positive, null, and below-chance — across several non-comparable metrics (standardized effect sizes, raw hit rates, z-scores, p-values, and correlations). These metrics cannot be averaged or pooled into a single bottom-line figure.

Here is a structured view of selected results across the evidence base:

Unsettled signals — these must be taken seriously

Direction disagreement is the central unsettled feature of this evidence base. Fourteen of the 43 result rows report a null or below-chance result. The Bösch (2006) meta-analysis — the largest aggregate in the library at k = 380 intentional studies — returned a below-chance overall effect (z = −3.67), not a positive one. Williams (2021) pooled a more recent and smaller set (k = 42 retro-PK studies) and found a strongly positive aggregate. These two pooled estimates point in opposite directions and are not reconcilable by the figures alone.

Heterogeneity is also a live issue: effect sizes vary widely across experiments, and Maier (2022)’s preregistered primary result — a correlation of r = 0.01 across 2,052 participants — stands in sharp contrast to the positive results reported by Collesso (2021) and Dechamps (2025) under similar micro-PK framing.

Decline effects have been noted in the broader literature. A review by Williams notes a significant trend (p < .03) across z-scores of 264 RNG-PK experiments from 1959 to 1987, where scores first rose and then fell over time — a pattern sometimes called a “parabolic” decline.

Small effects and file-drawer risk: even positive results in this area typically report effect sizes on the order of a fraction of a percent shift in hit rate. The practical consequence is that very large trial counts are required to detect them, and the literature’s positive results are sensitive to publication bias.

Theoretical context

Two contrasting frameworks are active in the library. The conventional micro-PK interpretation holds that intention directly perturbs a physical random process. Edwin C. May’s Decision Augmentation Theory (DAT) reframes the same data as goal-directed information acquisition — participants unconsciously select when to engage with an already-favorable RNG output rather than altering the output itself. The DAT and RNG research page covers this in depth.

Roger D. Nelson and colleagues at Princeton’s PEAR laboratory ran one of the longest systematic programs testing whether operator intention could correlate with detectable REG deviations; Alexander (2019) — a study covering PEAR operators — returned a null combined result (ES = 0.024, p = 0.3) across 127,000 trials. Nelson’s full research context is on his ESP-Nexus profile.

Skeptical critiques

What critics argue. The standard methodological objection to RNG micro-PK research is that the effects, when they appear, are so small that they fall within the range explainable by publication bias, optional stopping, or subtle hardware non-randomness — not genuine mental influence. Bösch, Steinkamp, and Boller (2006) made this case quantitatively: their meta-analysis of k = 380 studies found the overall effect was below the chance baseline rather than above it, and they argued that the positive results in the literature were an artifact of small-study bias (smaller studies showed larger effects, which is a signature of publication bias).

What the experimental data show. Williams (2021) pooled a subset of retro-PK RNG studies and reported a strongly positive aggregate (z = 6.82, p = 4.57 × 10⁻¹²). Mossbridge (2021) reported p < 0.001 in a preregistered confirmatory micro-PK run. Against this, Maier (2022)’s preregistered primary test — expressly designed to avoid optional stopping — returned r = 0.01. Pisoni (2024)’s reanalysis of a three-way interaction returned p = 0.21.

Analysis. The disagreement between Bösch (2006)’s below-chance pooled result and Williams (2021)’s positive pooled result has not been resolved in the retrieved sources — the two analyses draw on overlapping but not identical study sets and use different inclusion criteria. Preregistered replications with sufficient power (Maier 2022, Pisoni 2024) have so far returned null primary results, while some preregistered confirmatory analyses (Mossbridge 2021) have returned positive ones. Independent large-scale preregistered replication outside any single laboratory group remains an open item in the literature.

For a fuller quantitative digest — all 39 studies with figures traceable to their sources — see the Micro-PK (RNG/REG) evidence page, and for methodological background on how RNG experiments are designed and analyzed, see the RNG/Micro-PK Research methods page.