Discovery Lab Answers

Coverage first

The ESP-Nexus library holds 57 of the 78 studies in the standard ganzfeld meta-analytic database (the Tressoldi & Storm 2024 GZMA v13 set), plus the published abstract of one more. The 16 it is still missing skew toward null results, so any summary drawn from these holdings — including this one — runs somewhat more positive than the full published database. Read what follows as a description of the audited subset, not a settled verdict on the whole field.

The pattern over time

The audited rows do not show a single monotonic trend; they show a body of positive pooled estimates punctuated by prominent null episodes. Rather than “the effect size grew” or “shrank,” what the rows document is a fairly stable standardized effect size in the low-0.1 range across the pooled meta-analyses, with one well-known interruption and persistent between-study heterogeneity.

Because the rows report several non-comparable metrics — some are standardized effect sizes (z/√N, Hedges g, Cohen’s h), some are raw hit rates, one is a correlation, one a z-only — they cannot be averaged into a single “overall” figure. Below I group by metric and read the pattern qualitatively.

The standardized pooled meta-analyses (the closest thing to a like-for-like time series)

Reading across the z/√N rows — the metric that recurs most and is genuinely comparable — the pooled estimate sits at roughly 0.09–0.14 from the 2001/2010 databases through Tressoldi’s 2021 and 2024 pooled figures. That is not a large drift up or down over the span these rows cover; it is a persistent small positive effect on that scale. The Cohen’s-h values (Harris 2024, and the individual studies below) run higher, but h is a different metric and should not be lined up against the z/√N numbers.

The Milton and Wiseman episode

The single most cited discontinuity is the late-1990s null. Milton (1999)’s set of 30 new studies (1987–1997) pooled to essentially zero (ES 0.013, z = 0.70, p = .24, dir = null). One row (Tressoldi 2010 relaying Milton & Wiseman (1999)‘s own figures) documents that meta-analysis and reports it as p = .041 — but those are the original authors’ figures as re-derived inside the holding paper, not a new independent result. When Milton’s set was updated through March 1999 (including Dalton 1997), the same author’s cumulation returned to significance (ES 0.038, z = 2.28, p = .011). So the “flat” result and its partial recovery both come from the same author’s successive cumulations — worth stating precisely rather than as “mixed.”

Individual studies and hit-rate estimates

The single-study and hit-rate rows cluster around a ~31–34% direct-hit rate against a 25% chance baseline (Pooley 2023: 0.32, k=41; Williams 2011: 0.31, k=59; Storm 2010: 0.322; Dalton 1995 PRL: 0.314), with higher-scoring outliers in selected-population studies (Dalton 1997 creative groups: 0.47; Symmons 1997: 0.412).

Unsettled signals (must be surfaced)

• Directions disagree. Not every row is positive: Milton (1999)’s 30-study set, Lieb (2024)’s confirmatory receiver hit-rate (0.3125, z = 0.83, p = .199), Lau (2004)’s six long studies (0.30), and Stevens (2004)’s pooled index (p = .33) are flagged null. Four rows report a null or below-chance result.
• Heterogeneity is high. Two of the pooled results carry I² ≥ 50% between-study heterogeneity — meaning the studies inside a given meta-analysis are not estimating one clean common value, which complicates any single “the effect size is X” statement.
• One row is internally noted as a relayed figure (Milton & Wiseman via Tressoldi 2010), attributable to the original authors rather than the holding paper.

Bottom line

Across the audited subset, the ganzfeld pooled effect on the comparable z/√N scale has stayed small and positive — roughly 0.09–0.14 in the large unified databases — rather than clearly rising or falling over fifty years, with a conspicuous late-1990s null (Milton 1999) that a later cumulation partly reversed, and continuing heterogeneity and periodic null results throughout. And again: the 16 studies this library is missing lean null, so even this stable-and-positive picture is a touch more favorable than the full published database would render.

For the site’s own synthesis, see the Phenomena section (ganzfeld) and the single-study breakdowns under Studies.