The ganzfeld experiments

The full file

The ganzfeld experiments are not a single event but a body of laboratory work, run from the mid-1970s onward, and argued over in print for nearly forty years. Unusually for this shelf, the experiencers did not make the record at all. The people who tested them did, along with the critics who checked that testing, and most of the argument was conducted in one peer-reviewed journal.

What was reported

In 1974 and 1975 three researchers, Charles Honorton in the United States, William Braud and Adrian Parker, independently adapted a perceptual technique from the 1930s to test for telepathy. Bem and Honorton's 1994 paper describes the set-up. A receiver reclines in a soundproofed room with halved ping-pong balls taped over the eyes, a red light directed at them and white noise in headphones. In another isolated room a sender concentrates on a randomly chosen picture or film clip. For about 30 minutes the receiver describes whatever imagery comes to mind, then rates four candidates, the target and three decoys, without knowing which is which. If the target is rated highest the session scores a hit, so chance expectation is 25 per cent.

The claim was that receivers scored above that. Hyman's review of the early literature covered 42 studies in 34 reports dated from 1974 to 1981; Honorton's competing analysis used the 28 that gave hit rates, and found a combined result far beyond chance.

The record and where it is held

The record that puts this case on the shelf is the experimenters' and critics' own published work. In 1985 the Journal of Parapsychology printed Hyman's critical meta-analysis and Honorton's reply. In 1986 it printed their joint communiqué, in which the two agreed on "an overall significant effect in this data base that cannot reasonably be explained by selective reporting or multiple analysis", while still differing over whether it showed psi. They set out stricter standards for future work: security against sensory leakage, tested randomisation, correction for multiple analyses and full reporting.

From 1983 Honorton's Psychophysical Research Laboratories in Princeton ran the computer-controlled "autoganzfeld" to those standards, until loss of funding closed the laboratory in September 1989. The studies were published in full in the Journal of Parapsychology in 1990. Bem and Honorton then brought them to Psychological Bulletin in 1994: 240 receivers, 354 sessions, and in the primary analysis 106 hits in 329 sessions, a 32 per cent hit rate. Honorton died on 4 November 1992, before the paper was accepted. The 1994 paper and Hyman's comment on it also survive as scans in the CIA's declassified STARGATE files, stamped approved for release in 2000.

What investigators concluded

Hyman's 1994 comment accepted that the autoganzfeld was methodologically superior to earlier work but called its randomisation tests inadequate. He found that hit rates rose with how often a target had appeared in the database: targets used once or twice scored 27 per cent, close to chance, and targets shown three or more times scored 36 per cent. He asked for independent replication before anyone concluded that psi had been demonstrated.

Julie Milton and Richard Wiseman supplied one in 1999, a meta-analysis of 30 studies from seven laboratories following the same guidelines. The pooled effect was close to zero (Stouffer z = 0.70, p = .24), and they concluded the ganzfeld did not then "offer a replicable method for producing ESP in the laboratory." Lance Storm and S. Ertel replied in 2001 by pooling those studies with earlier databases into 79 studies; they found a significant effect and judged the ganzfeld replicable. In 2010 Storm, Tressoldi and Di Risio reported 29 ganzfeld studies from 1997 to 2008 with a mean effect size of 0.142, stronger with "selected" participants such as meditators and believers.

The leading conventional explanation

The leading conventional explanation is that the pooled effect is an artefact of how studies are chosen, combined and randomised, together with possible sensory leakage in particular laboratories. Hyman (2010) argued that the 2010 analysis manufactured "apparent homogeneity and consistency by eliminating many outliers", and that a well-powered study meeting its own recipe had already failed. Rouder, Morey and Province (2013) re-analysed the same data with Bayes factors. They found odds of about 6 billion to 1 for psi on the data as presented, but noted that studies randomised by hand showed larger effects than those randomised by computer. With those dropped and some omitted studies added, the figure fell to at most 330 to 1, which they judged unpersuasive given no plausible mechanism and likely unpublished failures. Wiseman, Smith and Kornbrot (1996) separately asked whether sounds from the sender's room could have reached the autoganzfeld experimenter.

The explanation's strength is that each step is documented: the effect shrinks under stricter selection and computer randomisation, and the independent 1999 database showed nothing. Its weakness is that no one has shown a specific flaw producing the hits in the best-controlled studies, and the autoganzfeld's own randomisation was by an electronic noise generator.

What remains disputed

What remains disputed is which studies belong in the count. Storm, Tressoldi and Utts replied to Rouder in 2013 that their case "is upheld" under the same Bayesian method. Each side's result depends on inclusion rules, outlier decisions and how much weight to give older studies, and neither has conceded. The dispute is now about meta-analytic method as much as about telepathy, and it has no agreed resolution.

Sources

• Does psi exist? Replicable evidence for an anomalous process of information transfer (archived) — Daryl J. Bem and Charles Honorton, Psychological Bulletin 115(1), 4–18, 1994 (copy hosted at Columbia University). The procedure, the Hyman–Honorton debate, the joint communiqué quotation, the autoganzfeld numbers and dates, Honorton's death.
• Anomaly or artifact? Comments on Bem and Honorton (archived) — Ray Hyman, Psychological Bulletin 115(1), 19–24, 1994; CIA release CIA-RDP96-00789R003200100001-5. Randomisation criticism and the target-frequency pattern.
• Does psi exist? Lack of replication of an anomalous process of information transfer (archived) — Julie Milton and Richard Wiseman, Psychological Bulletin 125(4), 1999 (University of Hertfordshire record). The 30-study null result and conclusion.
• Does psi exist? Comments on Milton and Wiseman's (1999) meta-analysis — Lance Storm and S. Ertel, Psychological Bulletin 127(3), 2001. The 79-study database.
• Meta-analysis of free-response studies, 1992–2008 (archived) — Storm, Tressoldi and Di Risio, Psychological Bulletin 136(4), 2010. 29 studies, effect size 0.142, selected participants.
• Meta-analysis that conceals more than it reveals (archived) — Ray Hyman, Psychological Bulletin 136(4), 2010. The homogeneity criticism.
• A Bayes factor meta-analysis of recent extrasensory perception experiments (archived) — Rouder, Morey and Province, Psychological Bulletin 139(1), 2013. The 6 billion and 330 to 1 figures and the randomisation finding.
• Testing the Storm et al. (2010) meta-analysis using Bayesian and frequentist approaches: reply to Rouder et al. (2013) (archived) — Storm, Tressoldi and Utts, Psychological Bulletin 139(1), 2013. The proponents' reply.
• Exploring possible sender-to-experimenter acoustic leakage in the PRL autoganzfeld experiments (archived) — Wiseman, Smith and Kornbrot, Journal of Parapsychology, 1996 (repository record only). The sound-leakage question.
• Ganzfeld ESP (archived) — Adrian Parker, Psi Encyclopedia, Society for Psychical Research, 2017, updated 2026. The 1974–75 origins with Honorton, Braud and Parker, and the 1986 communiqué's publication.