Discovery Lab Answers

Dream Telepathy Experiments — Data Tables

Coverage note: The figures below come from studies currently in the ESP-Nexus library. The library’s share of the published literature on dream telepathy has not been measured, so treat this as a summary of what the library holds rather than a settled account of the field. The direction of any sampling bias is unknown.

The structured evidence retrieved for this question spans 24 studies and 26 result rows, covering a range of telepathy paradigms — telephone-based caller identification, SMS guessing, video-monitored trials, and laboratory dream/altered-state protocols. Two distinct metrics appear across the rows — raw hit rates and standardized effect sizes — and because these are different, non-comparable measures, they cannot be averaged or pooled into a single “overall” figure.

What the hit-rate studies report

Chance expectation in four-choice designs is 0.25; in the telephone caller-identification paradigm (four callers) it is also 0.25. Most hit-rate rows sit above that baseline, but Stedall (2025) — a preregistered test with 117 participants — returned a hit rate of 0.486 with p = .76, a clear null. Rey (2010), testing Zener-card telepathy across 2,400 trials, also returned a null (p = .26).

What the standardized effect-size studies report

The standardized effect sizes in the positive rows span a wide range — from 0.03 (Sheldrake 2009, large-N pooled) to 0.85 (Silberstein 2024, very small N = 10). Schmidt (2009) illustrates the pattern well: combined across three studies the result crossed the conventional threshold (p = .02), but Study 1 alone was null (p = .23). Vernon (2020) also returned a null on its primary test despite a nominally positive Cohen’s d, because the confidence interval includes zero.

Dream and altered-state context

Source addresses laboratory dream telepathy specifically. REM-awakening protocols produce dream recall in a high proportion of trials, which makes them advantageous for matching dream content to targets — though the retrieved excerpt does not supply specific effect sizes for dream studies separate from the broader telepathy literature. Baptista, Derakhshani, and Tressoldi (2014) review dream studies alongside ganzfeld, forced-choice, and remote-viewing paradigms in their broader anomalous-cognition synthesis.

Unsettled signals — what the evidence does not resolve

The evidence base contains genuine disagreement that must be stated plainly:

• Null results exist alongside large positive effects. Stedall (2025), the only preregistered trial in these rows, returned a null. Rey (2010) returned a null across 2,400 trials. Schmidt (2009) Study 1 was null. These sit alongside rows with p-values in the 10⁻¹⁵ to 10⁻¹⁹ range.
• Effect sizes are highly heterogeneous. Standardized effects range from 0.03 to 0.85 — a spread that makes any single summary figure misleading.
• Much of the positive evidence comes from one research group. A large proportion of the positive hit-rate rows are from Sheldrake studies across 1999–2024, which limits what can be said about independent replication.
• Metric mix. Because hit rates and standardized effect sizes measure different things and come from different designs, no single pooled bottom-line figure is supported by these rows.

Skeptical critiques

What critics argue. The meta-analytic literature on telepathy paradigms has faced recurring methodological objections. Within the ganzfeld literature — the closest laboratory analogue to dream telepathy — Milton and Wiseman (1999) conducted a meta-analysis of 30 post-PRL studies (1987–1997) and reported a null result (Z = 0.70, p = .24), arguing that the PRL findings had not been replicated and that the ganzfeld paradigm did not provide evidence for psychic functioning.

What the experimental data show. Baptista, Derakhshani, and Tressoldi (2014) note that the Milton and Wiseman result triggered substantial debate, with subsequent analyses examining whether the post-PRL corpus was comparable in design to the PRL studies. The retrieved rows show that Sheldrake (2025), pooling k = 23 studies, reports a positive standardized effect with a narrow confidence interval. However, Stedall (2025) — explicitly preregistered — returned a null on its primary hypothesis test despite 604 trials and 117 participants.

Analysis. Schmidt (2009) is notable for showing within a single research program that combined results can cross threshold while individual studies do not — a pattern consistent with either a small real effect that requires aggregation or with publication-related aggregation artifacts. Independent preregistered replication outside the primary research groups producing the positive results remains sparse in these rows; Stedall (2025) is the clearest example of one, and it was null.

For broader context on how ESP-Nexus audits individual studies in this area, see How Studies Are Audited, and for the full study listing by topic, Studies.