Psychophysical interactions with a double-slit interference pattern

Radin, D., Michel, L., Johnston, J., & Delorme, A. (2013). Psychophysical interactions with a double-slit interference pattern. Physics Essays, 26(4), 553–566. https://doi.org/10.4006/0836-1398-26.4.553

AI Assessment

Three experiments testing whether focused human attention toward a double-slit optical system reduces the interference pattern, as a possible probe of the quantum measurement problem. The reported pattern is internally consistent in direction: a strong supervised effect (50 sessions, effect size 0.73, p = 2.4 × 10⁻⁷), a much smaller but well-powered online effect (685 people, effect size 0.09, p = 2.6 × 10⁻⁶), and a small third replication (effect size 0.62, p = 0.006), with matched no-observer control sessions null throughout and a combined Stouffer z of 6.81. But the participants were hand-selected for prior performance and high absorption rather than sampled at random, the large modeled effect sizes come from a subset of 12 sessions chosen because they were already significant, the psychological-modulator table runs many uncorrected comparisons, and the paper carries four internal numerical inconsistencies (documented below). This audit reports what the paper preregistered, ran, and found, and surfaces its own internal conflicts; it takes no position on whether consciousness affects the wave function.

Provenance

DOI. 10.4006/0836-1398-26.4.553 · Physics Essays 2013, 26(4), 553–566. Received 2 April 2013, accepted 19 September 2013, published online 30 December 2013.

Study type. A set of three laboratory and online experiments with confirmatory (pre-specified lag and threshold) and exploratory (modeling, subset, and psychological-modulator) analyses. Note that Physics Essays is a small, specialist venue rather than a mainstream physics journal.

Authors. Dean Radin and Leena Michel (Institute of Noetic Sciences), James Johnston (Biofeedback Health), and Arnaud Delorme (University of California, San Diego). The paper extends the authors’ earlier interferometer and six-experiment double-slit work.

Funding. Grants from the Federico and Elvia Faggin Foundation, Inc. (stated in the Acknowledgments).

Data availability. Statistics are reported in-text and in Table I; the paper describes storing full camera-image data on lab and cloud servers, but does not state that the raw data were publicly posted.

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

What the paper reports

In a double-slit optical system, obtaining which-path information makes photons behave more like particles and reduces the interference pattern. The authors ask whether focused attention alone, without any physical detector, can produce a similar decline in interference, and treat that question as a possible empirical handle on the long-unresolved quantum measurement problem.¹ Across three experiments, participants concentrated their attention toward or away from a shielded double-slit apparatus while the spectral magnitude and phase of the interference pattern were compared between the two conditions.

Three new experiments were conducted to further investigate this phenomenon. The first study consisted of 50 half-hour test sessions where participants concentrated their attention-toward or -away from a double-slit system located 3 m away… the combined results provided evidence for an interaction (effect size = 0.73 ± 0.14, p = 2.4 × 10⁻⁷).

How it was run

• Apparatus. A HeNe laser (Experiments 1 and 2) or a temperature-stabilized solid-state laser (Experiment 3) passed through two 10 µm slits separated by 200 µm; the interference pattern was recorded by a CCD line camera (14.0 cm from the slits in Experiment 1, about 16 cm in Experiment 2, 18 cm in Experiment 3). A footnote corrects an earlier publication that had listed the Experiment 1 distance as 10.4 cm.
• Shielding. Experiments 1 and 3 ran inside a double steel-walled, electromagnetically shielded chamber with line-power filtering; the apparatus sat in a matte-black sealed aluminum housing. Participants sat about 3 m away and were asked never to approach the optical system.
• Task. Counterbalanced “concentrate” and “relax” epochs (roughly 30 s each, with a random 0–9 s addition to prevent accommodation), announced by a recorded voice, with a real-time auditory feedback tone tied to the interference measure.
• Participants. Experiment 1 used 21 people selected for prior performance or an active attention discipline (meditators, musicians, healers), who scored well above the population norm on the Tellegen Absorption Scale (z = 3.5, p = 0.0003). Experiment 3 used 10 such people. Experiment 2 was open to anyone on the Internet.
• Analysis. Interference patterns were FFT-decomposed into spectral magnitude (log-transformed) and phase (unwrapped), detrended, and compared per session with Mann–Whitney U-tests, combined into a composite z and then a Stouffer z across sessions, checked against a 10,000-iteration permutation bootstrap. A 3 s lag was pre-specified from earlier studies, with a Bonferroni threshold of z = 2.4.
• Controls. No-observer control sessions used the identical equipment, protocol, and analysis. In Experiment 2 the web server automatically ran a control session every even hour, streaming to a Java client that simulated a human, with no one observing.

Results, as reported

Values are reproduced from the article’s Abstract, Results, and Discussion. The direction is consistent across all three experiments (a decline in interference during observation, with null controls), but the magnitude is very different: the well-powered online experiment yielded an effect size of 0.09, roughly an eighth of the supervised Experiment 1’s 0.73.

Eleven-dimension audit

Pre-registration

There is no formal registered protocol. The paper does pre-specify two things from its earlier studies, the 3 s analysis lag and the Bonferroni threshold z = 2.4, and Experiment 2 declared a stopping rule in advance (a minimum of 2000 experimental plus 2000 control sessions). But the modeling analysis, the distance analysis, and the twelve-way psychological-modulator table in Table I are exploratory, and the credit here is limited to the pre-committed lag and threshold rather than a full registered plan.

Randomization

Epoch order was counterbalanced and epoch lengths were randomized (30 s plus a uniform 0–9 s) to keep participants from accommodating to a fixed rhythm. Participant selection, however, was deliberately non-random in Experiments 1 and 3: people were chosen for superior prior performance or an active attention discipline and scored far above the norm on absorption. That is a reasonable design choice for a talent-sensitive effect, but it means those two samples are not representative.

Sensory leakage

This is a genuine strength of the design. Experiments 1 and 3 used a double-walled electromagnetically shielded chamber with the apparatus in a sealed matte-black housing, and Experiment 2 was run over the Internet expressly to isolate the optical system from any heat, vibration, or electromagnetic effects of a nearby human body. The near-identical direction of the online result is the paper’s main argument that the effect is not a proximity artifact.

Blinding

There is no human rater to blind; the outcome is instrumental. The relevant safeguard is that preprocessing transforms were applied uniformly, without regard to the attention condition, so the objective (optical) and subjective (instruction) components were, as the authors put it, strictly independent. No-observer control sessions provide the comparison baseline.

Optional stopping

Experiment 2’s pre-declared 2000+2000 session target guards against optional stopping for the main test. The authors themselves flag a residual concern in a Table I footnote: the “high motivation” subgroup (participants who contributed 10 or more sessions) outperformed low-motivation participants, and they note this outcome “might also be due to optional-stopping behavior” by motivated participants continuing until they did well. Disclosing that is to their credit.

Outcome measure

The primary measure is a composite z combining double-slit spectral magnitude and phase, evaluated at the pre-specified lag. Beyond that, the paper reports modeled optical parameters (the A/B field ratio, total field A+B, center point x0, and y-max) and, in Experiment 2, a dozen psychological and environmental subgroup contrasts. The core interference measure is pre-committed; the modeled and subgroup measures are secondary.

Effect size

The effect sizes span an order of magnitude: 0.73 (supervised Experiment 1), 0.09 (online Experiment 2), and 0.62 (Experiment 3). The much larger modeled effect sizes (for example the A/B ratio at 1.26 ± 0.29) are computed only on the 12 of 50 Experiment 1 sessions that were selected because they were already individually significant, so they describe that hand-picked subset and cannot be read as the study’s effect size. The most trustworthy single number is the small online effect of 0.09, precisely because that sample was large and unselected.

Multiple comparisons

The Bonferroni correction is applied to the choice of lag (z = 2.4). It is not applied across the wider analysis space: Table I reports twelve data subsets (internet speed, motivation, meditation, ESP belief, levitation belief, geomagnetic quiet or stormy) each tested at p < 0.05, and the modeling examined four parameters. Those exploratory comparisons are not corrected as a family, so the subgroup findings should be read as hypothesis-generating.

Internal replication

The three experiments internally replicate the same direction, and Experiment 3 did so with a newly designed apparatus, a more stable laser, and a simpler analysis, which the authors present as evidence that the Experiment 1 and 2 hardware and protocols were not themselves producing the effect. The consistency of direction, with null controls each time, is the paper’s strongest internal feature.

External replication

These experiments build on the authors’ own earlier interferometer and six-experiment double-slit work, but they are not independent replications by another laboratory. The Discussion is candid on this point, stating that the experiments are “by themselves, insufficient to definitively demonstrate” the effect and that confidence would require “scores of successful replications conducted in different laboratories.” Independent replication is the decisive outstanding step.

Transparency

Mixed. On the positive side, the authors self-correct a prior paper’s stated camera distance (14.0 cm, not 10.4 cm), disclose the optional-stopping caveat, and report unexpected null findings (belief in ESP and meditation experience did not predict success, contrary to their prior work). On the negative side, the paper carries four internal numerical inconsistencies that a reader should be aware of:

• Distance range. The Abstract states participant distance ranged from “1 km to 18,000 km,” while the Results section (III.B.3) states “4 km to 18,000 km.”
• Experiment 2 intercept. The in-text value for the significantly negative session-score intercept is z = 3.08 (p = 0.002), but Table I’s “All data” row reports the intercept as z = 3.39.
• Experiment 3 result. The English Abstract and Results report effect size 0.62 ± 0.22, p = 0.006, but the French résumé reports 0.52 ± 0.22, p = 0.02 for the same experiment.
• Combined statistic. The Discussion computes the three-experiment Stouffer z from component values 5.16, 4.31, and 2.33, yielding 6.81, but Experiment 3’s own Results section reported its composite z as 2.76, not 2.33.

None of these individually overturns the reported pattern, but together they mean the exact combined figure and the online distance and intercept values should be treated as needing the underlying data to reconcile.

The adversarial record

• Selected participants, collapsing effect. Experiments 1 and 3 used people chosen for prior performance and high absorption, and produced the large effect sizes; the one unselected, well-powered sample (the 685 online participants) produced an effect size of just 0.09. A skeptic can read that gradient as selection and expectancy rather than a stable physical effect.
• The big numbers come from selected subsets. The dramatic modeled effect sizes (A/B ratio 1.26) are computed only on the 12 sessions pre-filtered for significance, and the striking subgroup contrasts in Table I are uncorrected multiple comparisons. Neither should be weighed as if it were the study’s primary confirmatory result.
• Internal inconsistencies. Four numerical conflicts between the abstract, text, table, and French résumé (documented under Transparency) mean the precise combined statistic and several online figures cannot be taken at face value from the paper alone.
• Venue. Physics Essays is a small specialist journal, not a mainstream physics venue, which bears on how much independent scrutiny the claims have received.
• What the authors do right. They isolate the apparatus (especially the online design), run null controls throughout, apply preprocessing uniformly across conditions, pre-specify the analysis lag, disclose the optional-stopping risk, self-correct a prior error, report unexpected nulls, and explicitly call for independent multi-laboratory replication. That candor is what lets a reader weigh the claim honestly.