Discovery Lab Answers

Ganzfeld ESP Results Over Time

Coverage note 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 not yet held skew toward null results, so any summary drawn from these holdings — including this one — runs more positive than the full published database. Keep that in mind when reading the pattern below.The broader picture across all 28 result rows in the evidence table follows.

The early period: positive signals accumulating (1980–1994)

Sargent (1980) reported a positive result from 30 trials. Through the mid-1980s, Charles Honorton led the Psychophysical Research Laboratories program using the autoganzfeld system — designed and built by Rick E. Berger (Berger & Honorton, 1986) — which automated target selection and judging to address earlier protocol criticisms. Honorton (1990) pooled 11 series from that program and found a positive overall result across 358 sessions. Utts (1991), reviewing the same autoganzfeld database, arrived at essentially the same pooled figures.

Studies targeting creative or artistically selected populations also produced positive results in this period. Schlitz (1992) tested 20 Juilliard participants and found a hit rate of 0.50. Dalton (1997) pooled four creative-group studies and found a hit rate of 0.47 across 128 sessions. Radin (1994) reported a positive hit rate of 0.41 in a 32-trial creative-population study. This creative-participant pattern was an active area of investigation through the 1990s.

Bem (1994) pooled the autoganzfeld primary data (329 sessions, 10 series) and found a positive result.

A key disagreement: Milton & Wiseman (1999)

Milton (1999) conducted a meta-analysis of 30 new ganzfeld studies published between 1987 and 1997 — studies conducted outside the PRL autoganzfeld program — and found a near-zero, non-significant result (z = 0.70, p = .24, 1,198 trials). This is one of the clearest null signals in the table and represents a genuine discontinuity: the same procedure, applied by independent labs post-communiqué, did not replicate the PRL effect in that window. When Dalton’s 1997 Edinburgh study was added to the pool, Milton (1999) updated the cumulation to 39 studies and the result shifted to positive (z = 2.28, p = .011, 1,460 trials) — showing how sensitive the pooled figure was to individual large-N contributions.

Larger pooled analyses from 2001 onward: positive but heterogeneous

Storm (2001) unified the older and newer databases into a 79-study pool and found a positive grand pooled effect. Storm, Tressoldi & Di Risio (2010) meta-analyzed 20 studies across 4,196 trials and found a positive result. Storm (2010) separately analyzed a homogeneous 29-study set and found a positive result across 1,498 trials. Williams (2011) compiled 59 post-communiqué studies independent of PRL — 2,832 sessions — and found a positive pooled hit rate of 0.31.

Tressoldi (2021) applied a frequentist random-effects model (REML, Knapp-Hartung correction) to 113 effect rows across the full ganzfeld database and found a positive overall Hedges’ g — a small effect — with a confidence interval that excluded zero.

Pooley (2023) used a binomial hierarchical model on 38 studies (three influential studies removed) and found a positive intercept with a confidence interval that did not include zero.

However, the UNSETTLED SIGNALS the evidence table flags are real and must not be glossed over:

• Two of the pooled analyses report high between-study heterogeneity (I² ≥ 50%), meaning the studies are not measuring a uniform effect — results differ across laboratories, protocols, and populations by more than chance variation alone would predict.
• Three result rows report null or below-chance outcomes, of which Milton’s 30-study meta-analysis is the most consequential.
• Lau (2004), pooling six long ganzfeld studies, found a cumulative hit rate of 0.30 — at chance for a 1-in-4 design.

A recent pre-registered contribution

Watt (2020) contributed the first study to a registration-based prospective meta-analysis of ganzfeld ESP studies, using a precognition design at the Koestler Parapsychology Unit. The study found a positive direct-hit result (22/60, hit rate 0.37, z = 1.94, p = .03). A precognition design minimizes some target-leakage concerns and was proposed as useful for labs without a separate sender room.

Metric caution

The 28 result rows in the evidence table use several different, non-comparable metrics — standardized effect sizes (Cohen’s h, z/√n, Hedges’ g), raw hit rates, and a proportion index. The chart below covers only the z/√n studies, which share a single comparable scale. No single pooled “bottom-line” figure can be honestly derived across all rows, and the table should be read column by column within a metric, not averaged across metrics.

Skeptical critiques

What critics argue. Milton and Wiseman (1999) argued that when PRL autoganzfeld studies are set aside and only independent replications are examined, the ganzfeld effect disappears — their 30-study meta-analysis found a non-significant result. The methodological claim underlying this was that the PRL effect might not generalize to other laboratories and procedures. Wackermann, Pütz & Allefeld (2008) raised a different class of critique: that the theoretical rationale linking the ganzfeld state to psi (the “internal attention state” and “noise reduction” hypotheses) is not well grounded in established psychophysiology of the ganzfeld phenomenon, and that experimental findings contradicting the hypnagogic hypothesis have been underaddressed by the parapsychology community.

What the experimental data show. When Dalton’s 1997 Edinburgh study was included, Milton (1999) updated the cumulation and the pooled result shifted to positive (z = 2.28). Storm (2010) analyzed a homogeneous 29-study set and found a positive result. Williams (2011) compiled 59 post-communiqué studies independent of PRL and found a positive pooled hit rate of 0.31. Tressoldi (2021) found a positive Hedges’ g across 113 effect rows with a confidence interval excluding zero. On the psychophysiology side, Wackermann et al. (2008) documented that prolonged ganzfeld exposure produces complex hallucinatory percepts with identifiable EEG correlates (alpha attenuation, thalamo-cortical activity), but noted these findings sit in tension with the assumptions built into ganzfeld ESP paradigms.

Analysis. The dispute between the Milton & Wiseman (1999) null result and the subsequent larger positive pooled analyses was never resolved by a single decisive experiment — it was partially absorbed into larger databases that themselves show high heterogeneity. The question of whether effect size varies systematically with laboratory, experimenter, or participant population remains open: the high I² in two pooled analyses means the studies are not interchangeable. The psychophysiological critique Wackermann et al. raised — that the state the ganzfeld produces may not match the theoretical model used to justify the ESP paradigm — has not generated a published programmatic response that the sources here document.

For a broader synthesis of ganzfeld research on the site, see Phenomena and Studies.