Discovery Lab Answers

Can I produce a single pooled number to compare?

No — and the reason is methodological, not evasive. The twelve ganzfeld meta-analyses ESP-Nexus holds do not report their results on one common metric, so there is no legitimate way to average them into a single “ESP-Nexus pooled effect size” to set against the published Storm–Tressoldi figures. The evidence set mixes:

• standardized effect sizes (z/√n, Hedges’ g, Cohen’s h) — eight rows,
• raw hit rates — two rows,
• a correlation-type metric — one row,
• **a Stouffer z only** — one row.

Pooling across those would mean averaging quantities that are not on the same scale. So instead of a fabricated bottom-line number, here is what the set actually shows and how the Storm–Tressoldi analyses sit within it.

What the Storm–Tressoldi meta-analyses themselves report

Several of the rows are the Storm/Tressoldi line of work, so the comparison is partly internal to their own series:

• Tressoldi and Storm (2024) — the Stage 2 registered-report culmination — pool k=78 studies and report a small positive standardized effect (ES ≈ 0.074, p ≈ .0009).
• Tressoldi and Storm (2021) report a comparable small positive standardized effect over k=113 (ES ≈ 0.088).
• Storm (2010) (the Psychological Bulletin free-response analysis) reports a larger standardized effect and a hit rate around 0.32 on a homogeneous 29-study set.
• Storm (2001) reports a hit rate near 0.31 over the unified 79-study database.

So within the Storm–Tressoldi corpus itself, the direction is consistently positive, but the magnitude shrinks markedly from the earlier syntheses to the pre-registered 2024/2021 analyses.

Do the other analyses agree?

Directionally, most do — but not all, and the spread is wide:

• Positive, larger effects: Honorton (1990) (ES ≈ 0.29 Cohen’s h) and Harris (2024) (ES ≈ 0.27 Cohen’s h) sit at the high end. Milton (1997) reports a significant Stouffer z (3.37).
• Positive hit-rate analyses: Pooley (2023) (hit rate ≈ 0.32, k=41) and Williams (2011) (hit rate ≈ 0.303, k=59) land close to the Storm hit-rate figures.
• The null: Milton and Wiseman (1999) report a null on the 30 post-Honorton studies (ES ≈ 0.013, p ≈ .24) — the well-known break in the meta-analytic trend. Notably, when Milton (1999) later updated the same cumulation through March 1999 (adding Dalton 1997a), the effect became small-but-significant (ES ≈ 0.038, p ≈ .011). Tressoldi (2010) reanalyzed the Milton–Wiseman database and reported a positive correlation-type result. So the same underlying data supports different conclusions depending on scope and method — a live disagreement, not a settled point.

Unsettled signals you should weigh

• Directions disagree: eleven rows are positive, one (Milton–Wiseman 1999) is null.
• Heterogeneity is high (I² ≥ 50%) in two of the pooled results — meaning the studies are not estimating one stable effect, which itself weakens any single pooled summary.
• Magnitude declines over the series: the earlier databases (Honorton 1990, Storm 2001/2010) show larger standardized effects than the pre-registered 2021/2024 analyses.

Bottom line

Restricting to one coherent metric at a time: the hit-rate analyses cluster around 0.30–0.32 against a 0.25 chance baseline, and the standardized-effect analyses are all positive but range from ~0.01 (null) to ~0.29. On both metrics the Storm–Tressoldi results are consistent in direction with the broader set, and their hit-rate figures agree closely with Pooley and Williams. Where they diverge is scale: their own pre-registered 2024/2021 effects are at the small end, well below Honorton’s and Harris’s, and the Milton–Wiseman null remains an unreconciled outlier. So: broadly concordant in direction, not in magnitude, with one genuine null and high heterogeneity — no single pooled number can honestly stand for the whole set.

For the site’s own synthesis, see Tressoldi & Storm (2024) and Replication and Meta-Analysis Debates.