A quantitative digest of the ESP-Nexus structured study database for Mediumship: 9 studies spanning 2007–2026, with every figure traceable to its source. ESP-Nexus reports what the studies found and takes no position on whether the effects are genuine.
Across 8 studies spanning 2007 to 2024, quantitative investigations of mediumship have consistently reported above-chance performance on controlled reading tasks. Effect sizes range from a Cohen's d of 0.5 to 0.75 in Beischel's laboratory work, while hit rates in other studies reached 0.65 (Tressoldi, 2022) and 0.80 (Rock, 2015). Several results carry strong statistical signals: Wahbeh (2023) reported a z of -6.77, and two Beischel studies returned p-values below .0001, with Tressoldi (2024) finding p = .01.
Despite this directional consistency, the evidence base carries notable limitations. The sample of 8 studies is small, and the figures provided do not include heterogeneity statistics, confidence intervals, or independent replications from research groups unaffiliated with the primary investigators, making it difficult to assess publication bias or generalizability. Methodological variation across blinding procedures and scoring systems further complicates direct comparison, leaving the cumulative evidence base suggestive but not yet conclusively characterized.
The evidence
| Paper | Reported finding | Effect / significance | Basis |
|---|---|---|---|
| Tressoldi et al. (2026), Preprint (v. 25/02/2026; journal not specified) | Consensus themes from 11 books. | — | k = 11 |
| Wahbeh et al. (2025), Journal of Scientific Exploration [source] | Six-session protocol: overall trance channeling outcome. | — | N = 6 |
| Wahbeh et al. (2023), Journal of Scientific Exploration [source] | Correspondence: channeled vs non-channeled answers. | z = -6.77, p < 5 × 10−5 | N = 70 |
| Connelly et al. (2023), Journal of Anomalous Experience and Cognition (JAEX) [source] | Thematic Analysis: Three executive-function-linked cognitive themes in 19 medium tutors. | — | N = 19 |
| Tressoldi et al. (2022), OMEGA—Journal of Death and Dying [source] | Proportion of intended reading identification. | hit rate 0.65 | N = 100 |
| Beischel (2021) | Quintuple-blind accuracy study: item-level accuracy. | ES 0.75, p = .004, hit rate 0.528 | N = 31 |
| Beischel et al. (2021), Journal of Near-Death Studies [source] | MANOVA: Major PCI dimensions across 4 conditions. | p < 1 × 10−4 | N = 4 |
| Rock et al. (2015), Journal of Scientific Exploration | Gimeno: Red Lights Sitter-Group Sessions. | hit rate 0.8 | N = 15 |
| Beischel et al. (2007), Explore [source] | Summary rating scores: intended vs. | ES 0.5, p = .007 | N = 16 |
Source: ESP-Nexus structured study database, 9 studies. 9 born-digital figures; OCR-derived figures awaiting source-verification are excluded from this digest. Effect sizes are shown only where a standardized metric is recorded.