Discovery Lab Answers

Total participants across forced-choice ESP experiments

Coverage first: this reflects the forced-choice ESP studies currently in the ESP-Nexus library, not the complete published literature on this question. We have not measured how much of that literature the library holds, so treat what follows as a summary of what we hold — not a settled account of the field.

On the specific question — a single pooled participant count and pooled hit rate

The audited evidence base for this question does not contain a single pooled participant total across “every forced-choice ESP experiment,” nor a single pooled hit rate with a confidence interval computed across the whole set. The rows are separate meta-analyses that pool different study sets, report different, non-comparable metrics, and cannot be summed or averaged into one bottom-line figure:

A few things to note about why these can’t be collapsed:

• Only one row (Milton 1999) reports a participant count — roughly 86,000 in a mass-media ESP effect pooled over 8 studies, and that row is null / below-chance (z = −1.6). The other rows report study-level effect sizes without a participant total in the evidence base.
• The metrics differ: Storm & Tressoldi (2023) and Steinkamp (1998) report a standardized z/√n effect; Rock (2023) reports a Cohen’s d. Averaging a z/√n against a Cohen’s d would produce a number that means nothing.
• No pooled hit rate with a confidence interval across the full forced-choice corpus exists in these rows. The CIs shown are per-meta-analysis, not a single grand pooled interval.

So the direct answer is: the library does not hold a single “total participants / pooled hit rate / CI” figure for the entire forced-choice corpus. It holds several separate meta-analyses, each pooling its own set.

What the pattern looks like across the set

Read as a whole, the directions disagree:

• Storm & Tressoldi (2023) report a small but significant positive combined effect over 1987–2022 (141 studies, ES ≈ 0.02, z = 8.52), and describe the database as homogeneous with a slight incline — no evidence of decline over 36 years.
• Rock (2023) reports a larger positive Cohen’s d (0.39) for a narrower set (forced-choice clairvoyance/precognition, exceptional subjects, normal state).
• Steinkamp (1998) reports a small positive z/√n effect (0.01) across 22 study pairs.
• Milton (1999), the one row carrying a participant count, is null / below-chance (ES = −0.0046, z = −1.6) — the pooled mass-media ESP effect.

Unsettled signals you should weigh

• Directions disagree — positive in three rows, null/below-chance in Milton (1999).
• High between-study heterogeneity (I² ≥ 50%) is flagged in one pooled result.
• The largest participant pool in the set (Milton’s ~86,000) is the one that returned no effect.

If you want the fuller forced-choice synthesis on the site, see the Forced-Choice Precognition and GESP Frameworks page, which covers preregistered forced-choice work including a 2025 Psi@Home study whose primary hypotheses were not confirmed.