Can a person change a random outcome just by wanting it to change? For nearly three decades a laboratory at Princeton ran that exact experiment, millions of times, and reported a tiny but statistically stubborn "yes." The honest answer today is more careful: the reported effect is real as a number in the data, but most scientists do not accept it as evidence that minds move machines, and later attempts to reproduce it have largely failed.
This is a contested corner of science, so it deserves a fair hearing on both sides. There were serious researchers, careful protocols, and results that would be impressive if they held up. There are also serious statisticians who looked at the same data and concluded that publication bias, loose analysis choices, and the sheer size of the datasets can explain the findings without any new physics.
This article walks through what a random event generator is, why anyone tested intention against one, what the Princeton Engineering Anomalies Research lab and the Global Consciousness Project reported, the strongest version of those findings, and the strongest version of the criticism. The goal is not to sell you a conclusion but to hand you the map.
What Is a Random Event Generator?
A random event generator, or REG, is a small electronic device that produces a stream of genuinely unpredictable bits, ones and zeros. Unlike a software "random" function, which is really a formula, a hardware REG draws its randomness from a physical process such as electronic noise or quantum tunneling. Over a long run it should land on roughly 50 percent ones and 50 percent zeros, the way a fair coin lands on heads about half the time.
That predictability of the average is exactly what makes the device useful for this kind of test. If you record millions of bits and the ratio drifts away from 50 percent by more than chance would allow, something is nudging it. Manufacturers work hard to make sure nothing does, because these generators are used in cryptography and security where true randomness matters.
The research question was simple to state. If you sit a person in front of an REG and ask them to "will" more ones than zeros, does the output shift, even slightly, in the direction they intend?
Why Researchers Tested Intention
The idea sits inside a much older question about mind and matter. Reports of psychokinesis, the supposed ability to influence physical systems with thought, go back a long way, but early tests using dice and other large objects were easy to fault. An REG offered something better: a fast, automated, statistically clean target that removed the human hands from the physical outcome.
The appeal was measurability. Instead of arguing about whether a die was weighted or a coin was thrown fairly, you could log millions of bits, apply standard statistics, and ask a yes or no question. If intention did nothing, the numbers would stay at chance. If it did something, even a whisper of an effect would eventually surface across a large enough dataset.
This is a different world from the acoustic and physiological claims we usually examine on this site, where at least there is a clear physical mechanism to point to. REG research starts from the opposite end: a measurable statistical anomaly in search of a mechanism that no established physics predicts.
The PEAR Experiments
The best-known program was the Princeton Engineering Anomalies Research lab, or PEAR, founded in 1979 by Robert G. Jahn, then Dean of Princeton's School of Engineering, with Brenda Dunne as laboratory manager. It ran until it closed in 2007, roughly 27 years of work.
The centerpiece was the "benchmark" REG experiment. Operators, ordinary volunteers rather than claimed psychics, sat in front of a generator and tried to push the output high, push it low, or leave it alone as a baseline. Across about 2.5 million trials contributed by 91 operators over twelve years, PEAR reported a small deviation in the intended direction. The effect size was tiny, on the order of a few parts in ten thousand, but because the dataset was so large the benchmark reached a Z-score near 4, and the combined database across all variations reached a Z-score around 6, with odds against chance running into the billions to one.
Stated that way it sounds decisive. The interesting story, as usual, is in the details, and PEAR itself was upfront that the effect was minuscule and correlated with subjective factors it could not fully pin down.
The Global Consciousness Project
PEAR's work led to a bolder spin-off. In 1998, Roger Nelson, who had coordinated research at PEAR, launched the Global Consciousness Project, or GCP. Instead of one person trying to influence one machine, the GCP asks whether large-scale shared human emotion might leave a fingerprint in randomness.
The project runs a network of around 70 hardware random number generators, nicknamed "eggs," scattered across the globe and streaming data to a central server. The hypothesis is that during moments of intense collective attention, a disaster, a celebration, a global broadcast, the combined output of the network should drift away from randomness more than it otherwise would. Over roughly 500 events across about 17 years, the GCP has reported a cumulative deviation on the order of seven standard deviations from chance.
It is a genuinely evocative idea, and it spread far beyond the lab. Much like the cultural life of the 432Hz versus 440Hz conspiracy, the "global mind" framing traveled through popular culture on the strength of the story as much as the statistics.
The Strongest Reported Findings
Presented at their most persuasive, the results look like this. Two independent programs, using different setups and different questions, both reported that randomness bent slightly in the presence of human intention or attention. Both accumulated odds against chance that are astronomically large. The protocols were automated, the data were logged by machine, and many runs were pre-specified. Supporters argue that no single ordinary error, a loose wire, a biased chip, a fudged number, plausibly produces the same signature across decades and continents.
Advocates also point out that the effects, while tiny, were consistent in direction more often than a pure fluke should allow, and that a small real effect is exactly what you would expect if consciousness interacts with matter only weakly. In this reading, the whisper is the discovery.
That is the case at its strongest, and it deserves to be stated clearly before we take it apart.
The Main Criticisms
Now the other side, which is where mainstream science lands. The first problem is the size of the effect. In PEAR's data the proportion of hits in the intended direction was about 50.02 percent against the 50 percent expected by chance, and the parapsychologist Dean Radin put the overall effect at just under 51 percent. An effect that small becomes "significant" only because the dataset is enormous, and at that scale even faint, non-paranormal biases can dominate.
The second problem is dependence on a few data points. The science writer John McCrone noted that a single PEAR volunteer, often called Operator 10, was involved in about 15 percent of the trials yet accounted for roughly half of the total excess hits. When that one operator's data were set aside, the reported effect fell toward chance. A result that leans on one person is not a stable foundation.
The third problem is the analysis freedom in the GCP. Critics including Edwin May and James Spottiswoode reexamined the September 11 data and found that the apparent deviation existed mainly within the particular time window that had been chosen, not in the raw record. Because there are no fixed, objective rules for which events count or exactly when an event's window begins and ends, the door is open to reading meaningful patterns into noise after the fact. In 2017 Peter Bancel, himself part of the GCP effort, concluded that the data fit a goal-oriented or experimenter-driven pattern better than a genuine "global consciousness" signal. This is the same discipline of separating pattern from proof that we apply to claims about 432Hz and plants or water.
Replication and Statistical Interpretation
The deciding issue in modern science is whether independent teams can reproduce a result, and here the record is unkind. Physicist Stanley Jeffers at York University ran PEAR-style experiments and got chance results. A formal three-lab consortium, sometimes called the PortREG replication, brought together PEAR and German groups at Giessen and Freiburg, using PEAR's own software and hardware and a shared protocol. It reported positive but non-significant trends and failed to reproduce the original benchmark effect in its primary measure. More recent pre-registered attempts, the gold standard because the analysis is locked in before the data arrive, have likewise failed to find the effect.
The broadest look came in 2006, when Holger Bösch, Fiona Steinkamp, and Emil Boller published a meta-analysis of 380 intention-versus-RNG studies in the journal Psychological Bulletin. They did find a small, statistically significant overall effect. But they also found that smaller studies reported larger effects than big ones, that the results were extremely inconsistent from study to study, and that this exact pattern is the classic signature of publication bias, the tendency for striking results to get published while null results sit in a drawer. Their careful conclusion was that the anomaly could not be taken as clear evidence of genuine psychokinesis.
That is the crux. Two explanations fit the same numbers. One is a faint new force of mind over matter. The other is that many studies, imperfect selection of what gets reported, and flexible analysis can manufacture a tiny signal out of nothing. When a smaller and more rigorous study can make the effect vanish, the ordinary explanation is the stronger bet, a standard we hold to across our coverage of what the evidence for 432Hz actually shows.
The Honest Conclusion
So can human intention affect random events? On the current evidence, there is no reliable, reproducible demonstration that it can. The REG experiments produced real numbers and were run by real scientists, and dismissing them as fraud or foolishness would be unfair and inaccurate. But the effects are vanishingly small, they lean heavily on particular operators and particular analysis choices, and they have not survived strict independent replication. Extraordinary claims need especially firm evidence, and this evidence is not firm.
Why does the work still fascinate people? Partly because the question touches something deep, the wish that attention and care might reach beyond our own skin. Partly because the story is beautifully simple to tell. And partly because a "seven sigma" headline sounds unanswerable until you learn what a few parts in ten thousand really means. The REG saga is a case study in how honest researchers, sincere volunteers, and rigorous-looking statistics can still point at a mirage. That lesson is worth keeping whatever you believe about consciousness.
None of this diminishes the everyday value of focus and intention in your own life. Setting an intention before you meditate or listen to music does not require any effect on a random number generator to be worthwhile. If you like using sound as a cue to settle your attention, tools such as the 432 Player let you retune your favorite tracks, though the benefit there is the ritual and the listening, not a paranormal one.
The Takeaway
The REG experiments are a rare thing: a fringe claim tested seriously, at scale, for decades. Treat the reported effects as a genuine puzzle in the data rather than proof of mind over matter. The strongest findings are tiny, fragile, and unreplicated, and the simplest explanation, ordinary statistical artifacts and selective reporting, has not been ruled out. Stay curious, respect the researchers who did the work, and keep the burden of proof where it belongs.