Discovery Lab Answers

Evidence for Ganzfeld Telepathy

Coverage qualification — read this first

The ESP-Nexus library holds 57 of the 78 studies in the standard ganzfeld meta-analytic database (Tressoldi & Storm 2024, GZMA v13). The 16 studies currently missing from the library skew toward null results, so any summary drawn from these holdings — including this one — runs more positive than the full published database. The figures below are accurate extractions from the held studies, but they are a sample, not a settled account of the whole field.

What the ganzfeld paradigm is

In a ganzfeld telepathy experiment, a “receiver” sits in a state of mild sensory attenuation — halved ping-pong ball halves over the eyes bathed in red light, white noise through headphones — while a distant “sender” concentrates on a randomly selected target (typically a video clip or image). The receiver describes ongoing mental imagery, then ranks four possible targets (one correct, three decoys) at session end. Chance performance on a four-alternative forced-choice is 25%. The question is whether hit rates measurably exceed that baseline.

The quantitative picture

The evidence base is large but heterogeneous, and the metrics are not uniform across studies. The table below organizes the audited results by scope and year. Because different studies report different measures (standardized effect sizes using different formulas, raw hit rates, z-scores alone), no single pooled figure spans all rows — that would mix non-comparable quantities.

Pattern across the evidence

Across the pooled meta-analyses, hit rates cluster in the 0.31–0.34 range against a 25% chance baseline, and standardized effect sizes (where directly comparable, using the same z/√n metric) run roughly 0.07–0.14 across the larger syntheses. These are small effects by conventional standards, but they emerge with high statistical confidence in the larger databases — Tressoldi (2024)’s 78-study figure carries p=.0009, and Storm (2001)’s 79-study synthesis reaches p in the 10⁻⁹ range.

Individual studies and smaller syntheses tell a more varied story. Milton and Wiseman’s 1999 main analysis — 30 new ganzfeld studies conducted after the 1986 Joint Communiqué — returned a null result (ES=0.013, p=.24). That same paper’s updated cumulation, adding the Dalton 1997 data, shifted to marginal significance (ES=0.038, p=.011). Lau (2004)’s synthesis of six long-format studies and the Lieb (2024) confirmatory pre-registered trial (48 sessions, p=.199) also did not depart from chance. Stevens (2004) similarly found a null on 100 trials.

The unsettled signals are real and must not be minimized:

• Reported directions disagree across studies — the evidence base contains genuine nulls alongside significant positives, not just small positives at varying confidence levels.
• Two of the pooled analyses in the library carry high between-study heterogeneity (I² ≥ 50%), meaning the variation in outcomes across studies exceeds what sampling error alone would predict. The source of that heterogeneity — procedural differences, participant selection, laboratory, sender/receiver design — has not been resolved.
• Pooley et al. (2023) found that a review period during the target-judging phase was associated with worse outcomes, a procedural moderator that cuts against simple accumulation of all ganzfeld results.

Milton and Wiseman (1999)’s figures, as relayed in Tressoldi (2010), should be noted carefully: that meta-analysis is the primary source of the null-era result and is attributed to those authors, not to Tressoldi.

Moderators and correlates

Pütz et al. (2008) used a covert design in which receivers did not know they were in a psi experiment, obtaining a hit rate of 0.325 (p=.039) — relevant to the question of whether demand characteristics drive results. Dalton (1997) found substantially elevated hit rates (0.47) in creative populations using dynamic video targets. Marcusson-Clavertz and Cardeña (2011) found that high hypnotizability, combined with altered-state experience during the ganzfeld, was associated with psi-hitting in a small study — though correlations for other predictors (belief, prior psi experiences, dissociation) were non-significant. Storm and Tressoldi (2009) found no evidence of a decline effect across the free-response database over a 44-year period, and no differences by modality type (telepathy, clairvoyance, precognition) or target type.

Skeptical critiques

What critics argue. Milton and Wiseman’s 1999 meta-analysis in Psychological Bulletin [15 in the evidence table] was the principal post-PRL critical intervention: analyzing 30 ganzfeld studies conducted after Honorton’s original communiqué, they reported that the effect did not replicate (ES=0.013, p=.24), and argued that the PRL autoganzfeld result had not generalized to other laboratories.

What the experimental data show. Storm (2001) and Storm (2010) subsequently assembled larger databases — 79 and 29 studies respectively — and reported significant positive effects (p=7.78×10⁻⁹ and p=2.13×10⁻⁸). Williams (2011) independently assembled 59 post-communiqué studies and obtained a combined hit rate of 0.31 (p=8.59×10⁻¹⁴). Milton and Wiseman’s own updated cumulation (adding the Dalton 1997 data) shifted to p=.011. Harris (2024)’s 38-study synthesis reports ES=0.27 (Cohen’s h, z=7.1). Pre-registered confirmatory trials have returned mixed outcomes: Watt (2020) found a positive result (hit rate 0.37, p=.03 on 60 trials), while Lieb (2024) did not depart from chance on 48 confirmatory trials (p=.199).

Analysis. The dispute over the PRL generalization turned on which studies were included and which pooling method was used — a database-construction and moderator question as much as an empirical one. Milton and Wiseman used 30 studies; later syntheses added more studies and reached different conclusions. The null result in Lieb (2024) — a fully pre-registered confirmatory design — has not yet been absorbed into a new major meta-analysis. Between-study heterogeneity (I² ≥ 50% in two pooled analyses) means that a single pooled figure, positive or null, does not characterize all contributing studies; which procedural factors drive that variance remains an open research question.

For a deeper treatment of how the autoganzfeld system was designed and how methodological controls evolved, see the ESP-Nexus page on Automated Ganzfeld and Psi-Detection Methodology, and the Ganzfeld ESP evidence digest in the phenomena section.