Discovery Lab Answers

What the Global Consciousness Project Found

Coverage note first: The studies summarized here reflect what the ESP-Nexus library currently holds on this question. The library’s share of the complete published literature on field RNG and the GCP has not been measured, so treat this as a summary of what the library holds, not a settled account of the field as a whole.

The Global Consciousness Project (GCP), created and directed by Roger D. Nelson beginning in 1997, operates a worldwide network of hardware random event generators (REGs/RNGs) that stream data continuously to a central server. The core hypothesis is that when large numbers of people share intense, globally synchronized emotional or attentional responses to world events, the network’s output shows measurable departures from random expectation. Here is what the audited evidence base shows.

The Main Quantitative Picture

The evidence base contains 28 result rows spanning roughly 2001–2025. The metric mix is substantial — z-scores, raw p-values, standardized effect sizes, and correlations — and these cannot be pooled into a single “overall effect size.” Each metric must be read on its own terms.

Note: Mason (2007) and Radin (2025) carry negative z-values but are coded as “positive” direction by the original authors — the sign reflects below-chance drift that was interpreted as a structured departure, not a chance null. Their direction labels come from the source rows, not from this answer.

What the Pattern Shows

Across the formal GCP series, the cumulative z-scores are large and consistent in direction as the event count has grown — from z = 5.0 at 109 events (Nelson 2002) through z = 7.31 at 500 events (Nelson 2024). Nelson describes the 17-year composite as showing a departure on the order of one in a trillion against chance. The individual-event and smaller-sample studies (Shimizu, Radin, Mason) report more variable results but mostly in the same direction.

Wilson (2010) is the clearest exception: across 14 sessions, the pooled field RNG result was null (p = .559). That result sits in the same evidence base as the cumulative positives and must be read alongside them.

Unsettled Signals (Required Disclosure)

The evidence base carries two important disagreements that any honest reading must surface:

1. Mechanism is genuinely contested. The GCP’s own primary statistical analyst, Peter Bancel, concluded after more than a decade of investigation that the data better support a goal-oriented experimenter effect rather than collective consciousness. Bancel’s surrogate-event analyses — testing earthquakes, full moons, plane crashes, rock concerts, World Cup games, and religious observances as untested categories — all yielded null results, failing to support the collective-consciousness hypothesis. He argued that the experimenter’s freedom to select events and define their time windows creates a vulnerability that Decision Augmentation Theory can explain without invoking global consciousness. Roger Nelson’s interpretation and Bancel’s interpretation use the same data and arrive at different conclusions about what is driving the signal.

2. Directions disagree across studies. Although the large majority of rows are coded positive, one result (Wilson 2010) is null, and the Mason (2007) and Radin (2025) results involve negative z-values whose directional coding by the original authors requires careful reading of the source papers. The heterogeneity across single-event and multi-event designs is real.

Skeptical Critiques

What critics argue. A recurring methodological concern is that the GCP’s protocol allows post-hoc flexibility in event selection and window definition — critics argue this inflates the apparent hit rate. The broader replication-crisis context (the Open Science Collaboration’s 2015 large-scale replication project demonstrated that a substantial fraction of nominally significant psychological findings fail to replicate at expected effect sizes) raises the prior threshold for confidence in any small-effect literature, and the GCP’s effect sizes, while statistically large at the cumulative level, are small per event.

What the experimental data show. Bancel’s reanalysis — itself a form of internal skeptical scrutiny — used a Theil-Sen robust estimator and Monte Carlo testing and still returned a significant cumulative result (z = 1.98, p = 0.024), but reinterpreted its source as goal-oriented psi rather than collective consciousness. The surrogate-event null results Bancel reported are the sharpest empirical challenge to the collective-consciousness interpretation from within the project’s own data.

Analysis. The statistical signal in the cumulative GCP series has been confirmed by multiple analysts working on the same dataset, including Bancel and May. The disagreement is not about whether a departure from chance exists, but about what produces it. Independent replication by groups with no connection to the GCP’s event-selection process has not been established in the sources retrieved for this question. The Hagel (2005) and Shimizu (2010, 2012) results come from independent field-RNG experiments outside the GCP network, and those show positive results, though at smaller scale.

For a deeper look at Nelson’s work on this question, including the PEAR Lab and FieldREG precursors, see the Consciousness-Correlated Anomalies page and, for Bancel’s interpretive analysis, the Goal-oriented versus collective consciousness interpretation page.