Observer influence on quantum interference: Testing the von Neumann–Wigner consciousness-collapse theory

Radin, D. (2025). Observer influence on quantum interference: Testing the von Neumann–Wigner consciousness-collapse theory. Physics Essays, 38(1), 64–80. https://doi.org/10.4006/0836-1398-38.1.64

AI Assessment

A preregistered test of the von Neumann–Wigner idea that conscious observation can act like a weak quantum measurement, using 47 attention-trained participants who each ran a custom optical interferometer at home. The honest headline is mixed: none of the three preregistered hypotheses was supported once the full dataset was combined. One preregistered subgroup contrast held (outward-focus “magicians” outperformed inward-focus meditators, p = 0.008), and an exploratory trend analysis produced a spectacular result (a decline in interference at the observed spot but not a simultaneously recorded unobserved spot, p = 5.9 × 10⁻¹⁴). But the striking number is exploratory, not confirmatory, and the author himself flags plausible artifacts, feedback-related electromagnetic effects, laser noise, and the unsettling fact that two sensors just 3.5 cm apart behaved very differently. Two dataset revisions were also made after preregistration. This audit reports what the study preregistered, found, and cautioned; it takes no position on whether consciousness collapses the wave function.

Provenance

DOI. 10.4006/0836-1398-38.1.64 · Physics Essays 2025, 38(1), 64–80. Received 18 October 2024, accepted 7 February 2025.

Study type. A preregistered experiment (preregistration at osf.io/5ngkm) with both confirmatory (preregistered) and exploratory analyses. Note that Physics Essays is a small, specialist venue rather than a mainstream physics journal.

Author. Dean Radin, Institute of Noetic Sciences. The study extends his long line of consciousness-and-interference experiments.

Data availability. The preregistration is public; the data and analysis scripts are described as available to qualified professionals on request, which is more restricted than fully open deposit.

Source basis. Every figure below is taken from the article’s own Abstract, Methods, and Results.

What the paper reports

The von Neumann–Wigner interpretation holds that a quantum system’s wave function collapses only when a conscious observer registers the outcome. The study operationalizes this by asking whether human observation of one region of an optical interference pattern changes the recorded illumination there, relative to a second region recorded simultaneously but never observed.¹ Forty-seven people experienced in focusing attention each ran a preassigned series of observed-versus-unobserved sessions on a custom apparatus, with environmental sensors and real-time encryption built in to protect data integrity.

With all data combined the results did not support three preregistered hypotheses, but for one of those hypotheses participants selected for experience in an outward versus an inward focus of attention achieved significantly better results in reducing interference (p = 0.008).

How it was run

• Apparatus. Each participant used a custom printed-circuit optical system (a laser, a 500-line-per-millimeter diffraction grating, and three 16-bit light sensors) to record an interference pattern; a feedback signal reflected the illumination in one observed region.
• Design. One region of the pattern (the observed spot) was watched by the participant while a second region recorded at the same time (the unobserved spot) served as a simultaneous internal control. Sessions alternated observation and no-observation periods, and separate control sessions were run with no observer present at all.
• Preregistered hypotheses. Three: a difference in the observed spot’s mean (H1) and variance (H2) between observed and unobserved conditions, and a talent effect (H3) whereby more attention-experienced participants show larger effects.
• Analysis. A preregistered 20-step pipeline including environmental-artifact detection, encryption-based integrity checks, Savitzky-Golay smoothing, permutation-based nonparametric z-scores (5000 permutations), Stouffer-Z combination across sessions, and False Discovery Rate correction within a 4-plus-or-minus-1-second lag window.
• Post-registration revisions. Two changes were made after preregistration: two participants’ sessions were replaced as outliers (suspected hardware misalignment), and the z-score computation was switched from parametric to nonparametric because the permutation distributions were non-normal.

Results, as reported

Values are reproduced from the article’s Abstract and Results. The key interpretive point is the split between the confirmatory and exploratory layers: the preregistered predictions largely failed, while the single most dramatic number in the paper (p = 5.9 × 10⁻¹⁴) comes from an exploratory trend analysis that the author treats cautiously.

Eleven-dimension audit

Pre-registration

The study is preregistered (osf.io/5ngkm), which is a real strength, but two things temper it. Two dataset revisions were made after preregistration (replacing two participants’ sessions and switching from parametric to nonparametric z-scores), and the paper’s most eye-catching result comes from an exploratory analysis outside the registered plan. The registered predictions themselves were not confirmed, so the preregistration mainly earns credit for making that failure visible.

Randomization

Sessions were preassigned rather than randomized in the treatment sense; the design’s cleverness is spatial rather than random, using a second, simultaneously recorded but unobserved region of the same pattern as a within-session control, plus fully observer-free control sessions.

Sensory leakage

Here the relevant leakage is instrumental and environmental rather than sensory, and it is the study’s central vulnerability. The author builds in environmental sensors and real-time encryption, and runs no-observer controls, but he also candidly flags feedback-related electromagnetic artifacts and laser noise as possible causes, and reports that two sensors only 3.5 cm apart behaved very differently, which is exactly the kind of local instrumental effect that could masquerade as an observation effect.

Blinding

The simultaneous unobserved region and the no-observer control sessions function as the blinding-equivalent here: the comparison is between a watched and an unwatched signal recorded under otherwise identical conditions. There is no human rater to blind; the measure is instrumental.

Optional stopping

Sessions were preassigned, so optional stopping in the classic sense is not the concern. The analytic flexibility lies instead in the post-registration dataset revisions and in the exploratory trend analysis that produced the headline result.

Outcome measure

Multiple measures of the interference pattern (differential mean, variance, skew, and a slope-trend analysis) rather than a single pre-committed statistic. The primary preregistered measures (mean, variance) did not show the predicted effect; the effects that appeared were in a subgroup contrast and in the exploratory trend measure.

Effect size

Divergent. The confirmatory effects are absent or marginal; the subgroup contrast is moderate (p = 0.008); and the exploratory trend statistic is enormous (t = 9.24). A very large t from an exploratory pipeline on high-dimensional time-series data is precisely the kind of result that demands independent replication before it can be weighed against the null confirmatory outcome.

Multiple comparisons

The analysis spans several metrics, subgroups, and lags. The author applies False Discovery Rate correction within the lag window, which is appropriate, but the exploratory EH3 result is nonetheless one analysis selected from a large space, and it should be read as hypothesis-generating rather than confirmatory.

Internal replication

The internal controls are the study’s best feature: the effect appears during observation but not during no-observation periods (p = 0.77), and no-observer control sessions are uniformly nonsignificant. That internal contrast is genuinely suggestive, though it does not rule out an artifact that is itself tied to the act of running an observed session.

External replication

This is a novel apparatus and design, not a replication of an established protocol, and it has not yet been independently replicated. Given the extraordinary exploratory statistic and the acknowledged artifact possibilities, independent replication is the decisive next step.

Transparency

Mixed but leaning strong. The registered plan is public, the confirmatory failures are reported rather than buried, and the author devotes real space to alternative (artifact) explanations, including the awkward 3.5 cm sensor discrepancy. The weaker points are that the data are available only to qualified professionals on request rather than openly deposited, and that the two post-registration revisions add analytic latitude.

The adversarial record

• Confirmatory failure. The plainest reading is that the study did not confirm its preregistered predictions. On the registered tests, this is a null result; the positive findings are a preregistered subgroup contrast and an exploratory analysis.
• The headline is exploratory, and possibly instrumental. A p of 5.9 × 10⁻¹⁴ from an exploratory slope analysis is not evidence of the strength that number suggests, especially when the author himself lists electromagnetic feedback, laser noise, and a 3.5 cm sensor discrepancy as candidate causes. A skeptic can reasonably attribute the trend to an instrumental artifact tied to observed sessions.
• Post-registration latitude. Replacing two participants’ sessions and switching the z-score method after preregistration, however defensible individually, widens the researcher degrees of freedom and complicates the “preregistered” label.
• What the author does right. He reports the null confirmatory results, foregrounds the artifact possibilities rather than hiding them, builds in simultaneous unobserved and no-observer controls, and calls for replication. That candor is what lets a reader assess the claim honestly, and it is the appropriate posture for an extraordinary exploratory finding.