Discovery Lab Answers

Global Consciousness Project

Coverage note: The studies in the ESP-Nexus library on this question span the GCP’s formal experimental record from 2001 through 2025. The library’s share of the complete published literature on this question has not been measured, so what follows is a summary of what the library holds, not a settled account of the field. The retrieved studies skew toward positive findings — this is worth holding in mind when reading the overall picture.

What the GCP Is

The Global Consciousness Project (GCP) is a long-running international experiment, created in 1997 and directed by Roger D. Nelson, that operates a worldwide network of hardware random event generators (REGs, also called RNGs, or informally “eggs”) streaming data continuously to a central server. The core hypothesis is that when large numbers of people share intense, emotionally coherent experiences — disasters, celebrations, global tragedies — the network produces statistically detectable departures from randomness. The project is framed as an empirical test of the proposition that globally shared human attention or emotion correlates with measurable non-random structure in physical random systems.

The network has grown to over 65 devices deployed at fixed sites worldwide, with data sampled once per second. Each pre-specified global event is registered before analysis, and results are accumulated into a cumulative composite statistic over the life of the project.

The Formal Experimental Record

The evidence table below carries the full set of statistics. In broad terms, the cumulative GCP composite across pre-registered events is the most-discussed result. Nelson (2020) reports a correlation-based bottom-line meta-result across 500 formal tests from 1998–2016. Nelson (2024) reports a compounded result across 500 pre-registered events that yields a z of 7.31 — a departure from random expectation at very low probability. Nelson (2014) characterizes roughly 15 years of data across 456 hypothesis tests in similar terms.

Smaller-scale field studies in the library extend the picture. Shimizu (2012) found a positive correlation between audience size and RNG output at a stadium. Shimizu (2010) found a positive result for an RPG hardware RNG under a “view condition.” Mason (2007) found a drift-adjusted positive result in a meditation hall setting. Nelson (2006) found both a variance-reduction signal across eight years and an anticipatory effect around large earthquakes. Etzold (2003) found a positive aggregate across 101 cleaned events. Hagel (2005) reported a self-described pattern of coincident deviations over five months.

The one null result in the library is Wilson (2010), which found no overall effect across 14 field RNG sessions (p = .559).

Radin (2025) reports a Principal Components Analysis composite across multiple GCP-related measures at midnight, and Radin (2017) reports a positive combined result across five experiments at the “Man Burn” event.

Cavanaugh (2022) examined US homicide rate trend changes in relation to a hypothesized consciousness-related intervention and reported a below-chance effect size (ES = –0.476 on Cohen’s f), flagged here as `dir=positive` in the database’s coding — readers should consult the original paper for the directional interpretation, as the effect-size sign and direction label are in tension in this row.

Nelson (2025) reports a positive combined result from a FieldREG study in the Khufu Subterranean Chamber.

The metrics across these studies are not comparable — the table mixes z-scores, raw p-values, Pearson r, and other correlation measures. The grounding policy prohibits averaging or pooling across them into a single “overall effect size.”

Unsettled Signals

The library contains one null result — Wilson (2010) — against a large body of positive results. The evidence directions are not uniform, and this deserves explicit acknowledgment rather than being buried.

Mechanistic disagreement is a live issue. Peter Bancel, the GCP’s primary statistical analyst for nearly two decades, concluded after sustained investigation that the GCP data show evidence for a goal-oriented experimenter effect rather than collective consciousness. His surrogate event analyses — testing eight categories of untested events including earthquakes, full moons, plane crashes, rock concerts, World Cup games, and religious observances — all yielded null results, failing to support the collective consciousness hypothesis. Bancel argues that the experimenter’s freedom to select events and define their time windows creates vulnerability to psi-mediated selection, which Decision Augmentation Theory could explain. This is a substantive interpretive split within the research program itself, not only from outside critics.

The GCP 2.0 is an updated iteration described by Nelson (2025) and Nelson (2023), expanding the device network, adding focus-city clusters of 20 devices each, and extending the research agenda to include smaller concentrated groups such as meditators.

Skeptical Critiques

What critics argue. The primary statistical objection is that the GCP’s protocol gives researchers meaningful degrees of freedom in event selection and window definition — which events to include, exactly when they start and end — and that this flexibility, whether or not intentional, could generate spurious significance. A related concern is publication or reporting bias: events yielding positive results may be more likely to enter the formal registry than null events. Critics also raise the question of whether the cumulative z-score reflects a genuine physical effect or an artifact of analyst choices compounded over hundreds of decisions across 25 years.

What the experimental data show. Bancel (2011) addressed the window-definition concern directly, applying a Theil-Sen robust estimator and Monte Carlo testing to the full dataset and finding a positive result (z = 1.98, p = 0.024) under that more conservative treatment. Bancel (2008) found a cumulative z of 4.55 across 236 events selected under strict network-stability criteria, which was intended to reduce analyst discretion. Nelson (2024) reports that the 500-event series was pre-registered, which constrains post-hoc selection at the event level. Wilson (2010), however, found no effect across 14 field sessions, illustrating that not all field RNG applications replicate the pattern.

Analysis. The disagreement between the collective-consciousness interpretation and Bancel’s goal-oriented experimenter-effect interpretation has not been resolved by a mutually agreed-upon reanalysis published by parties holding both views. Surrogate event testing — Bancel’s primary approach to distinguish the two hypotheses — consistently yields null results for untested event categories, which is the published anchor for his conclusion. Independent replication of the GCP paradigm by research groups with no prior involvement in the project has not been reported in the sources retrieved for this question.

For deeper treatment of Nelson’s full research arc, see Consciousness-Correlated Anomalies: Theory and Evidence and Mass Consciousness Events: September 11 and Global Emotional Coherence. For Bancel’s interpretive work, see Goal-oriented versus collective consciousness interpretation and Global consciousness detection through network anomaly analysis. Nelson’s full profile is at https://esp-nexus.org/scientists/roger-d-nelson/.