Discovery Lab Answers

Ray Hyman was a critic, not an experimenter — he did not run ganzfeld studies and has no personal experimental results to report. His contributions to the ganzfeld literature were analytical: meta-analytic critiques and methodological commentary on others’ data.

What Hyman actually did

Coverage note first: the ESP-Nexus library holds 57 of the 78 studies in the standard ganzfeld meta-analytic database. The 16 studies not yet held skew toward null results, so any summary drawn from these holdings runs more positive than the full published database. Keep that in mind when reading the evidence below.

The 1985 critique

Hyman conducted a meta-analysis of Honorton’s database of 28 ganzfeld studies. Rather than producing a positive result of his own, he argued that the database contained systematic flaws — in randomization procedures, judging protocols, and multiple-analysis practices — that he believed accounted for the apparent positive effect. His position was that the studies did not form a coherent collection and that the reported effect was an artifact of these methodological problems.

The 1986 Joint Communiqué

Hyman and Honorton then co-authored a joint statement — an unusual document in which a critic and a proponent agreed on what the existing database showed while continuing to disagree on what it meant. Their jointly stated conclusions were:

• The ganzfeld database showed an overall significant effect not reasonably explained by selective reporting or multiple analysis alone.
• They nonetheless disagreed on whether that effect constituted evidence for psi.
• They issued specific joint recommendations on randomization, judging, feedback, and analysis for future work.

The Communiqué did not represent a change in Hyman’s skeptical position — it was an agreement on the factual state of the database and on methodological standards going forward, not an endorsement of a psi interpretation.

The 1994 critique of the autoganzfeld

When Bem and Honorton published the autoganzfeld results, Hyman responded directly. His critique focused on several specific points:

• He argued that the significant overall hit rate in the autoganzfeld was driven entirely by dynamic (video clip) targets, and that static targets — which made up the large majority of sessions in the original ganzfeld database — failed to replicate the earlier effect.
• He challenged the comparability Bem and Honorton drew between dynamic video targets and the multiple-image View Master reels in the earlier database, arguing these were not equivalent.
• He raised concerns about experimenter-assisted judging procedures, arguing that experimenters’ subjective biases could enter the ranking process.
• He noted that the Juilliard student subsample, often cited as a highlight, involved only 20 participants and did not perform significantly differently from other participants when tested directly.

The 2010 commentary

In a comment on Storm, Tressoldi, and Di Risio, Hyman argued that meta-analyses of ganzfeld studies manufacture apparent homogeneity across studies that do not, in his judgment, form coherent collections. He stated that for the original ganzfeld database and the original autoganzfeld database specifically, he could “confidently state that they clearly do not form coherent collections”. He also noted that effect sizes for psi in every major research program decline over time and approach zero — a pattern he treated as evidence against a genuine anomaly.

Skeptical critiques and discussion

What Hyman argues. Across his publications, Hyman’s core methodological claim is that pooling studies via meta-analysis obscures genuine between-study heterogeneity and that the studies in these databases differ qualitatively in ways that make their combination misleading. On the autoganzfeld specifically, he argues the headline result collapses an effect confined to one target type with null results for the dominant target type.

What the experimental data show. The structured evidence for this question includes a number of pooled analyses that reach positive and statistically significant results — including the Storm (2010) 29-study homogeneous set, the Williams (2011) 59-study post-communiqué collection, and the Tressoldi (2021) and Tressoldi (2024) large-database meta-analyses — alongside null results from Milton (1999) on 30 post-1987 studies and null results from Lieb (2024) and Stevens (2004). Between-study heterogeneity is flagged as high (I² ≥ 50%) in two pooled results. Four result rows report null or below-chance effects. The metrics across studies are not comparable (hit rates, standardized effect sizes, z-scores, correlations), and the METRIC MIX rule prohibits pooling them into a single summary figure.

Analysis. Hyman’s critique that the autoganzfeld effect was carried by dynamic targets is documented in the retrieved excerpts, and the Broughton and Alexander attempted replication — which Hyman discusses — used adequate power for the dynamic-target effect size but obtained a substantially lower hit rate for those trials. Whether that constitutes a failed replication or a different experimental context is a question the sources leave open. The disagreement between Hyman’s heterogeneity claims and the proponent meta-analysts’ homogeneity claims has not been resolved by a jointly accepted re-analysis; the two sides apply different criteria for what counts as a coherent study collection. The decline-over-time pattern Hyman identified is consistent with some of the structured evidence here — notably the contrast between the earlier PRL autoganzfeld results and later independent replications — but that pattern’s interpretation remains disputed in the primary literature.

For more on the 1986 agreement and its legacy, see the ESP-Nexus spoke page Hyman–Honorton Joint Communiqué and Methodological Standards, and for the broader debate chronology, Ganzfeld Debate.