Performance at a Precognitive Remote Viewing Task, with and without Ganzfeld Stimulation: Three Experiments

Roe, C. A., Cooper, C. E., Hickinbotham, L., Hodrien, A., Kirkwood, L., & Martin, H. (2020). Performance at a Precognitive Remote Viewing Task, with and without Ganzfeld Stimulation: Three Experiments. Journal of Parapsychology, 84(1), 38–65. https://doi.org/10.30891/jopar.2020.01.06

AI Assessment

A strong within-condition result, but the comparison the design was built to make falls just short. Across three counterbalanced experiments, novice participants identified randomly selected target locations well above chance when placed under ganzfeld stimulation, yet barely above chance in an ordinary waking state. The result is unusual in this field for holding across all three runs, but the ganzfeld-versus-waking contrast the study was designed to test only approached significance.

Provenance

DOI. 10.30891/jopar.2020.01.06 · Open access, CC BY.

Study type. Precognitive remote viewing with ganzfeld induction. Not classical sender–receiver telepathy.

Funding. Perrott-Warrick Fund; SPR Research Grants Committee; Parapsychological Association Research Endowment Fund.

Data availability. Psi Open Data repository (SPR): open-data.spr.ac.uk.

Source basis. Figures confirmed against the primary article (University of Northampton repository copy, record 13569).

What the paper reports

Under ganzfeld stimulation, participants achieved a 39% direct-hit rate (43 hits in 110 trials against a 25% chance baseline), with the pre-specified sum-of-ranks measure at SOR = 225, p = .000012. In the ordinary waking remote-viewing condition the hit rate was 27.5% (30 / 110), SOR = 253, p = .034. The combined ganzfeld effect was z = 4.22 with an effect size ES = 0.40 (computed as z/√n).¹

The between-condition contrast, the actual question the design was built to answer, only approached significance: t(39) = 1.86, p = .065. The published claim rests on the within-condition ganzfeld result, not on a clean ganzfeld-beats-waking comparison.

How it was run

• Counterbalanced, repeated-measures design; each participant completed both a ganzfeld and a waking remote viewing trial. Primary pre-specified outcome: sum of ranks by an independent judge.
• Precognitive framing: the target was randomly selected by a Target Experimenter after the mentation was produced, using Excel’s randomization function. The Target Experimenter had no contact with the participant.
• Targets drawn from two pools of 20 Google Earth locations, grouped into discriminable sets of four; two pools prevented target repetition across a participant’s two trials.
• Ganzfeld arm: reclining chair, eye shields, white noise, a 40W red light at roughly one metre, about seven minutes of progressive relaxation, then about thirty minutes of ganzfeld exposure.
• Experiment 1 sample: convenience sample of 40 (28F, 12M; mean age 26.2, range 18–54), not selected on belief or experience. The ganzfeld condition combined to 110 participants across the series.
• Collusion control by design: the laboratory is sound-attenuated but not electromagnetically shielded, which the authors acknowledge; the precognitive timing removes the target’s existence during mentation, closing that path.

Results, as reported

The paper reports no confidence intervals for the effect size, so none are stated here. The waking hit rate is given as 27.5% in the paper; 30 of 110 is 27.3%, a small rounding discrepancy carried verbatim from the original.

Confirmatory versus exploratory. Each experiment pre-specified two primary confirmatory tests: the ganzfeld and the waking sum of ranks against mean chance expectation. In Experiment 1 the ganzfeld condition was significant (z = 1.77, p = .038) and the waking condition was suggestive but not significant (z = 1.63, p = .052); the combined three-experiment figures are in the table above. A secondary first-place-rank (direct hit) analysis in Experiment 1 was non-significant in both arms (ganzfeld 35%, binomial p = .10; waking 30%, p = .29). The individual-difference predictions (hypotheses H3 to H8 in Experiment 1, with transliminality, openness, and dissociation measures substituted in Experiments 2 and 3) were also pre-specified, but were unrelated to performance and did not replicate. The authors set all tests 2-tailed and, given the confirmatory intent, applied no multiple-comparison correction. The genuinely post-hoc analyses are the combined between-condition comparisons, where the picture is mixed: a 2×2 mixed ANOVA gave a significant main effect of condition (F(1,47) = 4.906, p = .032) and an Experiment 3 Rosenthal-Rubin test was significant (Zdiff = 1.97, p = .024), but the authors’ preferred within-participant comparisons were marginal or null (t(39) = 1.86, p = .065; t(39) = .45, p = .657; Experiment 3 t(29) = 1.799, p = .082). The mixed ANOVA is confounded by a carry-over effect: participants scored higher in whichever condition they completed second. Some Experiment 2 figures differ from earlier conference proceedings because scoring errors were corrected.

Eleven-dimension audit

Pre-registration

Hypotheses and the primary outcome measure were pre-specified in the paper. No external registration repository is named for Experiment 1; treat as internally pre-specified rather than independently registered.

Randomization

Target chosen by Excel’s random function after mentation, within a precognitive design. Adequate as described; no raw randomization audit trail is reported.

Sensory leakage

Strong by design: with no target in existence during mentation, real-time leakage and collusion are neutralized. The lack of electromagnetic shielding is acknowledged by the authors.

Blinding

The independent judge ranked the four candidate locations against each mentation and was blind to the target’s identity within the set, which is the basis of the sum-of-ranks outcome. The judge was not masked to the experimental condition, however: ratings were made and analysed separately for the ganzfeld and waking conditions, so condition was known at judging. That is a limitation for the between-condition comparison the study was built to make. The participant experimenter was also present during both mentation and feedback, a route for expectancy. The study is not double-blind.

Optional stopping

Three combinable experiments with no explicit stopping rule stated. Post hoc combination across experiments warrants caution.

Outcome measure

Sum of ranks pre-declared as primary, consistent with the group’s prior practice. The 39% direct-hits figure is reported alongside but secondary.

Effect size

Combined ganzfeld z = 4.22, with an effect size ES = 0.40 (computed as z/√n), framed by the authors as replicable. The absolute effect is modest (39% against 25%). The paper reports no confidence intervals for the effect size, so none are stated here.

Multiple comparisons

Individual-difference tests ran across twelve consciousness sub-dimensions (the PCI subscales in Experiment 1, with transliminality, openness, and dissociation substituted in Experiments 2 and 3). The Experiment 1 associations did not replicate in Experiments 2 and 3, as the authors themselves report. The authors treated these individual-difference analyses as exploratory, with the intention of confirming any signal in the later experiments, and on that basis made no correction for the multiple tests; they set all tests 2-tailed even though several hypotheses were directional, a conservative choice.

Internal replication

The ganzfeld effect was present in all three experiments per the authors; the individual-difference correlations were not.

External replication

Builds on the Roe & Flint (2007) pilot (SOR = 42, p = .008)² and sits within a contested wider ganzfeld and remote viewing literature.

Transparency

Open access under CC BY, with full participant instructions reproduced. A data-availability statement is present: the data are available through the Psi Open Data repository (SPR), at open-data.spr.ac.uk.

The adversarial record

• Roe & Flint (2007): the precursor ganzfeld remote viewing pilot this series was built to follow up with a proper waking comparison.
• The wider remote viewing base is contested. The paper itself cites the Marks & Kamman (1980) critique of early SRI work on randomization and transcript-editing grounds, alongside the Tart, Puthoff & Targ (1980) rebuttal.
• Reading against the study: Experiment 1 was not externally pre-registered, the participant experimenter was unblinded during the session, and the ganzfeld-versus-waking contrast only approached significance (p = .065). The within-condition ganzfeld result carries the claim, not the comparison the study set out to make.
• The repeated-measures design carries an order/carry-over effect: participants scored higher in whichever condition they completed second (the paper reports better within-condition performance when a condition came second). Counterbalancing across trials does not fully remove this, and it confounds the combined between-condition analyses.