Psychophysical interactions with a single-photon double-slit optical system

Radin, D., Michel, L., Pierce, A., & Delorme, A. (2015). Psychophysical interactions with a single-photon double-slit optical system. Quantum Biosystems, 6(1), 82–98.

AI Assessment

Six experiments using a single-photon (rather than bright-beam) double-slit apparatus, testing whether directed attention changes the number of photons arriving at an interference minimum. The honest headline is mixed and the authors present it that way: combining the six experiments for a mean shift in photon counts gave nothing (Stouffer z = 0.58, p = 0.56), and the single largest effect (Experiment I, z = 4.50) went in the direction opposite to the paper’s primary prediction. The significant result the paper rests on comes instead from a second, variance-based combination (chi-square = 30.1, p = 3.77 × 10⁻⁵), driven by two experiments that reached significance in opposite directions, and the authors reinterpret this post hoc as a goal-oriented rather than a passive observation effect. The single-photon runs also report no observer-free control sessions (unlike the authors’ earlier beam-laser work), and two of the six experiments used just two hand-selected participants. This audit reports what the paper predicted, ran, and found; it takes no position on whether consciousness influences quantum systems.

Provenance

Citation. Quantum Biosystems 2015, 6(1), 82–98 (Special Issue). No DOI is printed on the article; it is self-archived by the corresponding author. Quantum Biosystems is a small, specialist open-access journal rather than a mainstream physics venue.

Study type. A series of six exploratory experiments with a pre-stated directional prediction, plus post-hoc reinterpretation and secondary (variance and trend) analyses.

Authors. Dean Radin, Leena Michel, and Alan Pierce (Institute of Noetic Sciences), and Arnaud Delorme (University of California, San Diego). Note on corpus metadata: the local record stored only the lead author; the audit uses the paper’s own four-author byline.

Funding. The Bial Foundation, the Federico and Elvia Faggin Foundation, and members of the Institute of Noetic Sciences (stated in the Acknowledgments).

Data availability. Per-experiment Stouffer z, sample counts, and effect sizes are reported in the text and summarized in Table 1; the paper does not state that raw data were publicly deposited.

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

What the paper reports

Earlier experiments by the group used bright-beam lasers carrying trillions of photons, so a decline in interference could in principle be described by statistical rather than quantum mechanics. To probe individual quanta, this study used a commercial single-photon double-slit apparatus (TeachSpin) in which, roughly 99.7% of the time, no photon is in the device.¹ Participants directed attention toward or away from the sealed slits while a photomultiplier counted photons arriving at an interference minimum. The paper sets up two competing predictions: a consciousness-collapse hypothesis (CCH), under which any observation should reduce interference and therefore raise the count at a minimum, and a consciousness-influence hypothesis (CIH), under which the count shifts toward whatever the feedback defines as the goal.

When the results of all six experiments were combined via Stouffer z, overall there was no evidence for a systematic mean-shift in the photon counts (z = 0.58, p = 0.56). However, when these results were combined in the form of a chi-square statistic to assess a shift in variance, the outcome was highly significant.

How it was run

• Apparatus. A commercial TeachSpin single-photon double-slit system: a dim incandescent source, a collimating slit, the double slit, and a movable slit in front of a photomultiplier tube counting photons at a chosen point in the pattern (about 1,000 counts per second at a maximum).
• Task. A participant-initiated (volitional) protocol of “get ready / now concentrate / now relax” epochs, 30 trials per session, with a droning “om” feedback tone whose volume tracked the count rate over a 3 s sliding window.
• Environmental controls. From Experiment II onward, apparatus and participant were inside an electromagnetically shielded chamber, the participant sat about 2 m away, and humidity and temperature were logged once per second (both rose with occupancy, but photon counts did not track them). A vibration-isolation table was tested and did not reduce count variance.
• Analysis. Photon counts were linearly detrended per session; a mean difference between concentrate and relax samples was compared to a distribution of 1,000 randomly scrambled mean differences to yield a per-session z, then combined across sessions as a Stouffer z. All statistics are two-tailed.
• Six experiments. I Initial test; II EEG; III Replication without EEG; IV Neurofeedback (two selected participants); V Hypnosis (two participants); VI Enhanced “illuminated Buddha” feedback.

Results, as reported

Values are reproduced from the article’s Abstract, Results, and Table 1. The interpretive crux is that the pre-stated mean-shift prediction was not supported, while the significant outcome is a variance effect assembled from two experiments that reached significance in opposite directions (Experiment I negative, Experiment VI positive).

Eleven-dimension audit

Pre-registration

There is no formal registration. The paper does state its two competing hypotheses in advance and uses two-tailed statistics throughout, which is appropriate for a non-directional prediction. The weakness is that the pre-stated combined mean-shift test was null, and the result the paper builds on is a variance-based combination introduced in the analysis rather than committed to beforehand.

Randomization

The statistical null distribution was built by randomly scrambling each session’s samples 1,000 times, which is a sound nonparametric approach. Participant selection was not random: participants were recruited volunteers, and Experiments IV and V used just two hand-selected individuals each. Condition timing was participant-initiated rather than randomized.

Sensory leakage

Reasonably well controlled. From Experiment II onward the apparatus and participant sat inside an electromagnetically shielded chamber about 2 m apart, temperature and humidity were logged (they rose with occupancy but the photon counts did not follow), and a vibration-isolation table was tested and found not to change count variance. The single-photon apparatus itself is sealed.

Blinding

The measure is instrumental, so there is no human rater to blind, and the comparison baseline is the within-session resampling distribution. A notable gap relative to the group’s earlier beam-laser papers is that this study reports no observer-free control sessions for the single-photon runs, so the “no observer” baseline that anchored the earlier work is absent here.

Optional stopping

Each experiment states a planned size (for example 20 participants for Experiment II), which limits classic optional stopping. The countervailing concern is that the samples are small and uneven, and two experiments rest on only two participants, so individual sessions carry heavy weight.

Outcome measure

The primary measure is the mean difference in photon counts between concentrate and relax epochs, combined as a Stouffer z. The paper then adds a secondary variance measure (a chi-square on the sum of squared z-scores) and, within some experiments, chronological trend correlations. The headline claim rests on the secondary variance measure rather than the primary mean measure.

Effect size

The effects are large but inconsistent in sign: Experiment I at 0.90 (negative), Experiment VI at 0.57 (positive), and the others near zero. The combined mean-shift effect is essentially zero. A set of large effects pointing in opposite directions is exactly what a variance statistic will flag as significant, so the significant variance result should be read together with the null mean result, not instead of it.

Multiple comparisons

The analysis space is wide: six experiments, two different ways of combining them (mean and variance), plus per-experiment trend correlations and lead/lag analyses. No family-wide correction is reported. The significant variance outcome is one selected result from that space and is best read as hypothesis-generating.

Internal replication

This is the study’s weakest dimension by its own account. Experiment III, the designated replication of Experiment II, was null (z = 0.02), and the six experiments did not reproduce a consistent direction. The paper is explicit that the direction of the effect “was not constant,” which is the opposite of a clean internal replication.

External replication

The study deliberately uses a commercially available apparatus (TeachSpin) to make independent replication easier, and it builds on the group’s earlier beam-laser experiments, but it is not itself an independent replication and has not been independently replicated here.

Transparency

Mixed, leaning honest on the reporting and weak on the inference. To the authors’ credit, the null combined mean shift, the opposite-direction results, and the post-hoc nature of the goal-alignment interpretation are all stated plainly, and funding is disclosed. Points a reader should weigh: the significant claim rests on a post-hoc variance analysis; no observer-free controls are reported for the single-photon runs; and Experiment IV’s session count is reported inconsistently, with the Methods describing seven-plus-six sessions and Figure 12 showing 13, while the Results text and Table 1 state 25.

The adversarial record

• The primary prediction failed. The pre-stated combined mean-shift test was null (z = 0.58, p = 0.56). On the measure the paper set up in advance, this is a negative result.
• The headline is a post-hoc variance effect. The significant chi-square (p = 3.77 × 10⁻⁵) is driven by two experiments significant in opposite directions and by a combination method that rewards precisely that pattern. It is reinterpreted after the fact as goal-oriented (CIH), which the authors themselves flag as speculative.
• Direction reversal on the strongest effect. The single largest effect, Experiment I at z = 4.50, went opposite to the CCH prediction that counts at a minimum should rise. A skeptic can read the whole pattern as noise plus flexible interpretation rather than a stable effect.
• Thin samples and missing controls. Two experiments used only two selected participants, and, unlike the group’s earlier work, the single-photon experiments report no observer-free control sessions, weakening the artifact defense.
• What the authors do right. They report the null and the opposite-direction results openly, label the goal-alignment account as post-hoc, shield and environmentally monitor the apparatus, choose a commercial device to aid replication, and disclose funding. That candor is what lets a reader weigh the mixed result honestly.