Blog · Global Coherence
For two decades a worldwide network of hardware random number generators listened for moments when humanity's attention converged — and reported a cumulative deviation with odds of a trillion to one against chance. That number is real. So are the critiques, including one from the project's own senior analyst. Here is the whole story, told the way Ashta tells everything: hypothesis under test, never proof.
The story begins at the Princeton Engineering Anomalies Research lab — PEAR — founded in 1979 by Robert Jahn, then Dean of Engineering at Princeton, with lab manager Brenda Dunne. For nearly three decades PEAR reported tiny statistical correlations between human intention and the output of electronic random event generators. The university tolerated the lab rather than embraced it — the relationship was famously strained — and PEAR closed in February 2007. Whatever you make of its claims, PEAR set the methodological question that still matters: if intention influences a random physical system at all, the effect is so small that only massive, disciplined accumulation of data could ever reveal it.
Roger Nelson was PEAR's Coordinator of Research from 1980 to 2002. In the 1990s he took the generators out of the lab — "FieldREG" studies placed portable devices at ceremonies, concerts and group events, looking for deviations during moments of shared attention. Prototype global analyses followed: a January 1997 synchronized meditation, and the funeral of Princess Diana in September 1997. In August 1998, Nelson launched the Global Consciousness Project (GCP): a permanent, worldwide network of hardware random number generators — nicknamed "eggs" — streaming data continuously to a central archive, waiting for the world's attention to converge.
One point of honesty before anything else, because the popular name gets it wrong: the GCP was never a Princeton University program. It lived for years at the noosphere.princeton.edu web address because Nelson was Princeton staff, but it was an independent, privately funded collaboration of several dozen researchers, supported in part by the Institute of Noetic Sciences. "Princeton's random number experiment" is a fair description of the lineage, not the letterhead.
Each egg is a hardware random number generator based on quantum-indeterminate electronic noise, with XOR post-processing to strip any first-order bias. Every second, each device sums 200 random bits — expected mean 100 — and sends the trial to the archive. The network grew from 3 nodes in 1998 to a plateau of roughly 65–70 nodes across all continents by 2004.
The discipline that makes the GCP interesting is the hypothesis registry. Before examining the data, the team registered a prediction: a named world event, a fixed time window, and a fixed statistic. The primary measure — "network variance" — is the squared Stouffer's Z across all eggs each second, accumulated over the event window; it is mathematically equivalent to asking whether the devices became slightly correlated with one another. Recurring events like New Year's Eve reused fixed specifications. The formal analyses, on the project's own account, contain no post-hoc element — an approach Ashta recognises, because pre-registration is the backbone of our own protocols.
The formal series ran from August 1998 to December 2015: 513 registered events. The composite result across all of them is a cumulative deviation of about 7 sigma — Z = 7.31, odds against chance quoted at more than a trillion to one. The most famous single entry is September 11, 2001, where the pre-specified window showed a significant deviation, later published in Foundations of Physics Letters. The widely repeated claim that the eggs "predicted 9/11" — deviations beginning hours before the attacks — comes from exploratory analyses the project itself does not count as formal evidence.
Two honest qualifications, both made by the GCP's own investigators. First, the effect is tiny: most individual events are not significant on their own, and the project's FAQ concedes that any single "success" might be chance and any single null might hide a real signal. The claim rests entirely on accumulation across hundreds of replications. Second, a statistical correlation — even a strong one — is not a mechanism, and no physical mechanism has ever been proposed that survives scrutiny.
Independent analysts May and Spottiswoode reanalysed the 9/11 data and found the result lived inside its chosen time window: a few minutes shorter, or half an hour longer, and it is consistent with chance. NASA astrophysicist Jeffrey Scargle argued the 9/11 claim should not be accepted unless Bayesian and frequentist analyses agree — and they did not. Skeptic Robert Todd Carroll pressed the unfalsifiability problem: if deviations before an event count as precognition and deviations after count as response, what result could ever count against the hypothesis? He also noted how soft the event registry could be — alongside disasters and mass meditations sit entries like a home-run record and a television finale. Physicist Robert Park's deeper objection was that diffuse statistics are simply the wrong instrument for so extraordinary a claim.
The most important critique came from inside. Peter Bancel, the project's own senior physicist-analyst, spent years testing the structure of the anomaly and published his conclusion in 2017: the data, in his assessment, contain a real statistical oddity — but its structure "does not support the global consciousness proposal." It behaves less like a planetary field and more like a goal-oriented effect tracking the experimenters themselves. Nelson published a considered rebuttal, and by 2007 had already conceded to New Scientist that the data do not permit saying with any accuracy what, if anything, the eggs respond to. When a project's founder and its chief analyst disagree in print about what twenty years of data mean, the honest summary is: unresolved.
The formal experiment ended in 2015. The network kept recording exploratory data until April 3, 2026, when a hardware failure at the hosting service ended active collection; the archives remain public at global-mind.org. A successor — GCP 2.0 — has been stood up between 2021 and 2023 by the HeartMath Institute's Global Coherence Initiative, with Nelson, Dean Radin and Rollin McCraty involved, planning some 4,000 quantum RNG sensors worldwide, with over 1,400 deployed so far. Notably, its research agenda now explicitly includes the Bancel question: field effect, or experimenter effect? Ashta is not affiliated with the GCP, HeartMath or IONS — we describe their work because it is the essential history of this question, not because we endorse their conclusions.
What do we take from it? The GCP is the founding example of intention-at-scale research done in the open: registry, fixed windows, public data, published critiques. It is also a two-decade lesson in what such research still lacks — adversarial analysis built in from the start, and controls that can separate a genuine field effect from the people running the experiment. Those are precisely the safeguards Ashta builds into its own group-intention tests and its mass-participation experiment: certified hardware RNGs, pre-registered protocols, sham controls, frozen analysis, and a null as the default expectation. The GCP asked a magnificent question. Our job is to ask it in a way that produces an answer either side would have to accept — including a flat no. See also: the Power of Eight and Rupert Sheldrake's experiments, the other originating claims in our research lineage.
