Honorton’s 1978 framework specified four conditions a psi input must satisfy before it can be consciously recognized: detection (mediation through ongoing experience such as imagery), discrimination (sufficient impact to stand out from concurrent inputs), retention (storage and report before sensory contact with the source), and confirmation (objective correspondence over repeated trials sufficient to rule out chance).^[1] He argued these criteria explain the patterning of spontaneous cases: roughly two out of three reported psi experiences occur in dreams or other non-ordinary states, where sensory functioning is reduced and attention turns inward, and “crisis” cases predominate because they are noticed, retained, and confirmed.^[1] The psi-conducive states he documented shared four antecedents: somatic relaxation, reduced sensory functioning, sufficient cortical arousal to sustain awareness, and deployment of attention toward internal mentation.^[1] On this analysis he urged the field to shift from asking what conditions are necessary for the occurrence of psi to the empirically tractable question of what conditions are necessary for its detection.^[1]

Eight experimenters contributed sessions to the autoganzfeld database, including Honorton, Rick Berger (who implemented the system), Mario Varvoglis, Marta Quant, Patricia Derr, Ephraim Schechter, Diane Ferrari, and visiting scientist Marilyn Schlitz; Honorton trained several of them personally.^[4] The 241 participants (100 men, 141 women) ranged in age from 17 to 74 (mean 37.3) with a mean of 15.6 years of formal education. Recruitment came mainly from colleague referrals (24%), media presentations about PRL (23%), friends or acquaintances of staff (20%), and referrals from other participants (18%). Belief in psi was strong (mean 6.20 on a 7-point scale); 88% reported personal experiences suggestive of psi, and 80% had training in meditation or other internal-attention techniques, a self-selected sample, not a random one, which constrains generalization of the hit rates to broader populations.^[4] The work was funded by the McDonnell and Fetzer foundations, with additional grants to Honorton from the American Society for Psychical Research and the Parapsychology Foundation supporting the final write-up.^[4][6]

In his 1974 review, dreaming accounted for 37–65% of spontaneous ESP reports across four international surveys, and dream-mediated cases were more often rated informationally complete (85%) than waking cases (under half).^[3] Of twelve card-guessing studies comparing hypnotic and waking conditions, nine showed significant condition differences. His own hypnotic-dream study with Stump produced 46% correct target identifications against a 25% chance expectation, with two independent confirmations and a partial replication reported by Parker and Beloff.^[3] In a partial sensory-deprivation study, subjects reporting relaxed, inwardly focused states scored 76% hits against a 50% baseline, while externally oriented subjects scored 46%.^[3] In the confidence-call studies, 82% of subjects receiving accurate feedback improved their proportion of correct confidence calls, versus 44% of those given false feedback.^[3]

Honorton’s eight Maimonides series (1974–1978) yielded significant retrieval rates in seven, with a combined estimate of P = 7.9 × 10⁻⁸ across recognition and recall methods.^[1] By 1978 he counted 26 separate studies in 11 laboratories comprising more than 1,000 sessions by over 500 subjects, of which 14 (54%) reported overall significant psi rates against the 1.3 expected by chance at the 5% level; 17 independent replication attempts in 10 laboratories included 7 significant at the .01 level or lower.^[1] He acknowledged the unreported-failure problem explicitly and computed that even an assumed 140 unreported null studies would leave the observed result significant; he also argued publicly that all ganzfeld studies should be reported in full whether significant or not.^[1] Two comparative studies (Braud, Wood and Braud; Honorton’s Series 6) reported significant scoring in ganzfeld groups and chance scoring in matched non-ganzfeld controls.^[1]

Between February 1983 and September 1989, 241 volunteers contributed 355 sessions in 11 series (3 pilot, 8 formal). The 122 direct hits (34.4%) gave an exact binomial p = .00005 (z = 3.89), with a 95% confidence interval of 30%–39%; effect sizes were homogeneous across the 11 series (χ² = 16.25, 10 df, p = .093) and across all eight experimenters (χ² = 7.13, 7 df, p = .415).^[4] Dynamic targets: 190 sessions, 77 hits (40%, z = 4.62, p = 1.9 × 10⁻⁶); static targets: 165 sessions, 45 hits (27%, n.s.); the dynamic–static difference was significant (p = .002).^[4] Auditory leakage from the VCR soundtrack was the one cueing channel found to be partially open under artificial amplification; after a circuit modification verified by audio spectrum analysis (475 Hz–15.2 kHz), the final 62 sessions yielded 44% hits, and dynamic-target performance rose to 50%, so the artifact, once eliminated, did not account for the effect.^[4] RNG control samples totaling over 945,000 trials passed Kolmogorov-Smirnov uniformity testing (p = .577), and experimental target and judging-order distributions were uniform.^[4] Three series were unfinished when the laboratory closed; assuming chance for the missing trials still left the database significant (z = 3.61), addressing the optional-stopping concern that methodology research has documented as a driver of false positives.^[4][16] Bonferroni adjustment for the 15 reported significance tests left six of the nine individually significant outcomes significant, including the overall hit rate and the dynamic-target results.^[4] One caveat the authors disclosed: Series 302, terminated early at 25 sessions, left an imbalance in target frequencies, so its raw 64% hit rate was adjusted against an empirically computed 34.08% response-bias baseline before analysis.^[6]

Across PRL’s three Apple-computer PK games, effects in silent-RNG conditions (no real-time feedback) tended to be stronger than in feedback conditions, contrary to the assumption that feedback is necessary for micro-PK effects.^[15] Among the eight participants who served in all three experiments, silent-condition hitting in Psi Invaders was significant (t = 5.76, 7 df, p = .001), with all eight showing positive effect sizes (p = .0039, exact binomial). Extraversion/introversion interacted with condition: a significant positive correlation between introversion and silent-condition effect size was found in Psi Invaders (r = .649, p = .006, one-tailed), confirming a prediction from the earlier Volition data, and MBTI “feeling” types showed higher silent-condition effect sizes than “thinking” types (r = .54, p = .046), paralleling the feeling-type advantage in PRL’s ganzfeld First-Timers series.^[15] These were exploratory, small-sample analyses.^[15]

The database comprised 309 studies in 113 publications by 62 senior authors (1935–1987), nearly two million trials, more than 50,000 subjects; studies by S. G. Soal and Walter J. Levy were excluded on reliability grounds. Overall combined z = 11.41 (p = 6.3 × 10⁻²⁵); 30% of studies were significant at the 5% level, contributed by 40 different investigators. Rosenthal’s fail-safe N was 14,268, a ratio of 46 unreported null studies per reported study would be required to nullify the result. A 10% trim removing extreme outliers reduced the combined z to 6.02, still highly significant.^[7] Each study was coded on eight quality criteria (preset sample size, preplanned analysis, randomization method, randomness controls, automated and duplicate recording, automated and duplicate checking); quality did not correlate with effect size (r = .081, n.s.), effect size stayed constant across the half-century while quality improved significantly, and high-quality studies did not score lower than low-quality ones.^[7] The moderator analysis found selected subjects outperforming unselected (mean ES .051 vs .008, p = .001), individual testing outperforming group testing, a feedback gradient (none through trial-by-trial, r = .231), and a decline over longer time intervals confined to unselected subjects. In the “Optimal” subgroup, combining selected subjects, individual testing, and trial-by-trial feedback, seven of eight studies were independently significant, versus none of the nine “Suboptimal” studies; the Optimal studies also had significantly higher quality ratings.^[7]

In the First-Timers data, participants referred by colleagues or other participants showed significant performance (52.4% direct hits, p = .0064, N = 21), MBTI feeling types scored 50% direct hits (p = .0038, N = 28) while thinking types scored at chance (18.2%, N = 22), and the number of types of self-reported ESP experiences correlated with performance (r = .306, p = .018).^[20] Honorton and Schechter’s resulting model combined prior psi testing, mental discipline practice, personal psi experiences, and feeling-perceptive MBTI classification; in PRL data all 6 participants fitting the four-factor model obtained direct hits, and the 28 fitting a three-factor version hit at 64%. Broughton, Kanthamani, and Khilji’s independent test at the Institute for Parapsychology drew on four series totaling 159 sessions with 144 participants (overall 26% hits, nonsignificant) and was framed explicitly as a confirmation attempt of the PRL findings.^[21] Bem and Honorton (1994) report the model as independently replicated in that laboratory.^[6] The Juilliard sample comprised 20 undergraduates (8 music, 10 drama, 2 dance): 10 of 20 direct hits (50%, p = .014, z = 2.20, effect size h = .52), musicians 6 of 8 (75%, p = .0042); performance significantly exceeded the general PRL population (t = 2.09, p = .038, two-tailed). Torrance creativity scores did not correlate simply with performance: the correlation was slightly negative with two extreme-scoring outliers included and positive (r = .41, p = .045) with them excluded, a pattern the authors flagged as uninterpretable without further data.^[8]

Hyman’s 1985 critique covered 42 studies from 34 reports (1974–1981). Honorton’s response analyzed the 28 studies reporting direct hits: 23 (82%) had positive z scores, 12 (43%) were independently significant at the 5% level, and the composite Stouffer z was 6.60 (p = 2.1 × 10⁻¹¹); assigning z = 0 to ten further studies that used the relevant judging procedure but did not report hit rates still gave Stouffer z = 5.67. By laboratory, 6 of 10 laboratories reported significantly positive outcomes (combined z = 6.16), and discarding the two most prolific laboratories left z = 3.67 across the remaining eight.^[6] Rosenthal’s file-drawer statistic for the 28 direct-hit studies was 423 unreported studies (about 15:1); a 1980 survey of parapsychologists had located only 19 completed but unreported ganzfeld studies, 7 of them significant.^[6] Both authors agreed there was an apparent clustering of significant studies with fewer than 20 trials, the basis of Hyman’s retrospective-study concern; the autoganzfeld answered it structurally by registering series status and sample size in software before data collection.^[6][4]

The 1988 National Research Council report, drawing on Hyman’s meta-analysis, concluded against scientific justification for parapsychological phenomena; Bem and Honorton document that the NRC report did not acknowledge the communiqué’s agreed points, and that the background paper the NRC commissioned from Harris and Rosenthal found the ganzfeld studies to be the only evaluated area regularly meeting basic requirements of sound experimental design, estimating an obtained accuracy rate of about one third against a chance rate of one quarter.^[6] Honorton, with Palmer and Utts, published a formal reply to the NRC study.^[4] Within parapsychology, philosopher Stephen Braude has questioned whether Honorton’s years of debate with Hyman were time well spent, while granting that the autoganzfeld results matched the earlier series’ significance levels, a dissenting view about strategy rather than about the data.^[26] The Hyman–Honorton standards continue to anchor methodological discussion of the paradigm, including in subsequent meta-analytic disputes over post-1989 replications.^[24]