Sources checked. Sources checked. FieldREG I (1996) and FieldREG II (1998) primary PDFs now verified from pear-lab.com (correct filenames) plus mirrors. PortREG, Jeffers, and Alcock remain in the body on purpose.
For almost thirty years a lab in Princeton's School of Engineering asked a blunt engineering question: can the consciousness of a human operator leave a measurable fingerprint on a machine that is supposed to be random? Not a séance. Not a TED pitch. A reverse-biased noise diode, a cascade of nine thousand polystyrene balls, and a lot of statistics.
That was the Princeton Engineering Anomalies Research program—PEAR—founded in 1979 by Robert G. Jahn, then Dean of the School of Engineering and Applied Science, with Brenda Dunne as laboratory manager. Roger Nelson coordinated experimental work there for years. The lab closed its physical facilities at the end of February 2007. Its archives, successors, and arguments did not close with the door.
What follows digs through that record: the random-event generator (REG) databases, the random mechanical cascade, FieldREG group studies, remote and off-time protocols, remote perception, the theoretical language PEAR used about information and reality, the failed multi-lab PortREG replication, the York/Princeton optical-slit work with Stanley Jeffers, James Alcock's null-hypothesis critique, and the afterlife of the program in ICRL, Psyleron, and the Global Consciousness Project. The theme is practical: what, if anything, PEAR's numbers imply about how humans interact with technology, objects, and space-time. The findings are contested. The claims and the critiques both get room.
What the lab said it was doing
PEAR's own website frames the work as a scientific study of "consciousness-related physical phenomena" at Princeton from 1979 to 2007. The 2007 press statement puts the founding purpose in engineering language: "to study the potential vulnerability of engineering devices and information processing systems to the anomalous influence of the consciousness of their human operators." Two main experimental tracks ran for decades: anomalous human/machine interactions with random physical systems, and remote perception of distant sites and events.
That framing matters. It is why a dean of engineering could defend the program inside a university that, as Nature noted when the lab shut, was not eager to be defined by it. Jahn and Dunne said they closed the lab themselves after accomplishing what they set out to do—to determine whether the effects were real and to identify major correlates—not because a referee finally won. Nature's Philip Ball reported the closure in that light: fringe science that still drew a crowd, and a research program that refused the everyday words telekinesis and telepathy while circling the same neighborhood.
Alumni kept working. Dean Radin at the Institute of Noetic Sciences. Roger Nelson at the Global Consciousness Project. Technology company Psyleron. The International Consciousness Research Laboratories (ICRL), a 501(c)(3) consortium that PEAR treated as its forward vehicle. Entangled.org appears on the current pear-lab.com legacy page as a later mobile platform for related research. The brand changed. The question did not.
The REG: intention against a noise diode
The core machine was a microelectronic random event generator. Operators sat near it (or, later, far from it) and tried, under pre-stated intentions, to push the mean of binary output higher (HI), lower (LO), or leave a baseline (BL) without trying. No physical contact with the noise source. Redundant logging. Calibrations.
Jahn, Dunne, Nelson, Dobyns, and Bradish summarized twelve years of that work in the Journal of Scientific Exploration in 1997. The abstract is the line most people quote, and it is specific: correlations of the binary sequences with the stated intentions of some 100 individual operators; absolute effect sizes on the order of 10⁻⁴ bits of deviation per bit processed; over huge databases a composite effect exceeding 7σ (p ≈ 3.5 × 10⁻¹³). Remote operators and efforts timed differently from machine operation showed similar effect sizes and structural details. Fully deterministic pseudorandom sources—hard-wired or algorithmic—yielded null overall mean shifts and no other anomalous features.
Inside the paper the benchmark table is tighter. Over twelve years, 91 anonymous individual operators contributed. The primary figure of merit is the HI–LO separation: zμ = 3.81 (pμ = 7 × 10⁻⁵). High-intention and low-intention mean achievements are roughly 1.3 × 10⁻⁴ and –7.8 × 10⁻⁵ bits deviation per bit processed. The effect is tiny. The claim is that it compounds.
That is the whole controversy in one paragraph. If the composite is real, intention is leaving a statistical bruise on a physical noise process. If the composite is an artifact of selection, optional stopping, operator subsets, or publication practice, then the bruise is in the analysis, not the diode. PEAR's own pseudorandom null result is their strongest internal control: when the "randomness" is a deterministic algorithm, the intention effect they report disappears. That cuts both ways. It argues against a simple feedback or recording bug that would hit every source the same way. It also raises the bar for any later claim that does not show the same split.
Nine thousand balls: the random mechanical cascade
Electronic noise is one kind of object. A pinball-style cascade is another.
In 1988, Dunne, Nelson, and Jahn reported operator-related anomalies on a Random Mechanical Cascade (RMC): 9,000 polystyrene balls dropping through a matrix of 330 pegs into 19 collecting bins, counted photoelectrically, displayed as feedback, recorded on-line. Operators tried to shift the developing distribution right or left relative to a concurrent baseline. Twenty-five operators completed one or more series. Across 87 series (3,393 runs), the overall mean difference of right versus left efforts had a probability against chance of less than 10⁻⁴; 15% of individual series were significant at p < .05; 63% conformed to the intended directions.
The concatenated result was asymmetric in a way the apparatus did not explain: virtually all of the right-versus-left separation came from the left-versus-baseline separation. Variance excesses also stacked more on the left. Remote-operator variants, the paper says, again produced significant results similar to local work. Multiple-operator early results did not simply add the individuals together.
Whatever one thinks of the p-values, the design choice is the point for this article. PEAR did not only poke at microelectronics. They asked whether intention tracked a macroscopic, visible object stream—balls, pegs, bins—the kind of machine a shop floor understands. If the effect is only an REG software story, the cascade should have been quiet. PEAR reported it was not.
Distance, time, and remote perception
If intention can touch a local diode, does intervening distance or time kill it?
Dunne and Jahn's 1992 JSE paper on remote human/machine interaction says no, not over the ranges they tested. Operators separated by up to several thousand miles produced anomalous mean shifts "comparable in scale and character" to local work, on both microelectronic generators and the mechanical cascade. The paper reviews early local benchmarks: REG HI–LO composite z = 3.614 (p = 2 × 10⁻⁴) from 33 operators; RMC overall difference z = 3.891 (p = 5 × 10⁻⁵). The same paper summarizes PEAR's remote-perception database: some 336 trials, z = 6.355 (p = 10⁻¹⁰), with percipients describing targets from under a mile to more than 5,000 miles away, and with temporal separations from days before to days after a partner's visit—often with the description recorded before the target was even selected. No significant reduction of the anomalous effect with increased distance or time over those ranges, they report.
That is the space-time claim in PEAR's own voice. Not wormholes. Not a new force named in a press release. A statistical claim that information exchange between consciousness and random physical processes—and between a percipient and a remote site—does not fall off the way radio does. The 1997 twelve-year REG review repeats the remote and off-time similarity for the machine work. The 2005 "PEAR Proposition" paper folds those empirical features into theory: relative insensitivity to objective physical correlates including distance and time; oscillating sequential patterns; male/female discrepancies; irregular replicability; and a call for models that give consciousness a "proactive role in the establishment of objective reality."
Read that last phrase slowly. It is the fabric-of-reality language. It is also a theoretical leap past what a HI–LO z-score can prove. PEAR was explicit that current science had no accepted mechanism. Critics were explicit that irregular replicability is what you expect when the null is true.
FieldREG: groups, no joystick
Lab operators sat down and pushed HI or LO on purpose. FieldREG asked a different question: does a portable random generator, running quietly in the room while a group of people share attention or emotion, drift off chance even when nobody is aiming at the machine?
Nelson, Bradish, Dobyns, Dunne, and Jahn reported the first formal cut in 1996 (Journal of Scientific Exploration, 10(1), 111–141). Portable REGs with software to index continuous binary sequences ran under protocol in ten venues—ICRL meetings, healing intensives, CUUPS gatherings, an academy session, a humor conference, an SSE Council business meeting, a Marfa Lights outing, and more. Each venue broke naturally into sessions, presentations, or days. The authors took the most extreme data segment from each of the ten applications, corrected for multiple sampling (Bonferroni and a scale-corrected variant), and reported a collective probability against chance of 2 × 10⁻⁴. All but one application—the SSE Council meeting—contributed positively. Five of the ten independently approached or cleared a conventional p < .05 bar after that conservative correction. Composite Z corresponding to the final p was 3.540. Effect-size estimates, they said, sat in the same ballpark as laboratory REG work (on the order of a few parts in ten thousand per bit; time-based sizes averaging about 0.3).
The design twist that matters for this article: in most cases the participants were only peripherally aware of the box and had little understanding of what it was doing. There was no pre-stated group intention on the output. PEAR's interpretation stayed speculative—a possible consciousness "field" that orders random physical processes when attention, cohesiveness, or shared emotion runs high. Their own control checks (random resampling of the field streams, and laboratory calibrations run through the same extreme-segment procedure) did not compound the way the active venues did.
FieldREG II followed in 1998 (JSE 12(3), 425–454). From eighteen exploratory applications (twelve already in the first paper), the team stated a testable split: environments that foster intense or profound subjective resonance should show larger mean deviations than more pragmatic, mundane assemblies. Then came sixty-one further applications—twenty-one hypothesis-based formal replications and forty more explorations. On the formal replications, the resonant subset compounded to a probability against chance of 2.2 × 10⁻⁶; the mundane subset sat at 0.91. That is PEAR's own contrast, not a critic's: when they predicted "deep shared engagement," the composite moved; when they predicted ordinary business, it did not.
Hold the caveats next to the headlines. FieldREG I still selects extreme segments inside each venue before correcting; a skeptic can argue about that selection even after Bonferroni. FieldREG is not the same protocol as the intentional HI–LO lab series, and it does not erase PortREG's failed primary mean-shift replication. What it does add, if you take the papers at their word, is a second human–machine claim: not only can a single operator try to lean on a diode, but a room full of people who are not trying may still leave a statistical fingerprint when the emotional weather is thick. That is either a bridge from PEAR's bench to Nelson's later Global Consciousness Project, or another place where analysis choices did the heavy lifting. The primary PDFs are now in hand either way. Read them before you pick a side.
PortREG: the replication that did not land
In 1996 a consortium formed at Freiburg, Giessen, and Princeton to pursue mind/machine anomalies together. Their first joint project was a replication of prior Princeton REG intention effects. The 2000 JSE report is the document skeptics point to, and PEAR co-authored it.
Design: each of the three labs collected 250 sessions of 3,000 trials (200 binary samples per trial), from 227 human operators total. Identical noise-source equipment. Similar protocols and analysis. Primary agreed criterion: magnitude of the HI–LO separation.
Primary result, in the authors' own abstract: overall HI–LO mean separations proceeded in the intended direction at all three laboratories, but "the overall sizes of these deviations failed by an order of magnitude to attain that of the prior experiments, or to achieve any persuasive level of statistical significance." Secondary, interior "structural" anomalies—trial-level standard-deviation reductions, irregular series-position patterns, dependencies on feedback type or run length—were reported beyond chance in composite. The authors' interpretation: the change from systematic intention-correlated mean shifts to a "polyglot pattern of structural distortions" shows inadequate understanding of the basic phenomena and a need for more sophisticated experiments and models.
That is not a victory lap. It is PEAR putting a failed primary replication in print. Hold onto it. Any fair reading of the program has to.
Jeffers, optical slits, and Alcock's null
Stanley Jeffers and Michael Ibison took a different cut: could operators reduce fringe contrast in a Young's double-slit interference pattern—an experiment that puts intention next to a textbook quantum demonstration?
Their 1998 JSE paper reports York University and Princeton University runs. York participants were invited to exert mental effort; Princeton participants were invited simply to reduce fringe contrast. Terminal results: York Z = −0.481 (null); Princeton Z = 1.654, just significant at the 5% level. The authors discuss instruction differences (means-directed versus goal-directed) and treat the Princeton result as marginal. Jeffers' broader program, as Alcock recounts it, included null outcomes Jeffers himself did not wave away.
James E. Alcock's 2003 Journal of Consciousness Studies essay, "Give the Null Hypothesis a Chance," is the clearest mainstream counterweight in this file. Alcock argues that parapsychology often fails to let a failed prediction strengthen the null. On Jeffers: a conceptual replication that came back null (with only a marginal Princeton double-slit result Alcock and Jeffers both find unconvincing) has been neglected inside the field—another refusal to give the null a fair hearing. On PortREG: neither Princeton nor the German labs found anything of significance on the hypotheses under test; post-hoc "structural" anomalies, Alcock writes, are how a success of sorts is wrestled from the jaws of the null. His bottom line on replication: parapsychologists have not produced a demonstration that researchers in general can reliably repeat, and failures are ignored, explained away, or reinterpreted as exotic properties of psi.
You do not have to adopt Alcock's metaphysics to take the methodological punch. PortREG's primary endpoint missed. Jeffers' York data sat on the null. Those are load-bearing facts for anyone still citing the twelve-year composite as settled engineering risk.
After the door closed: ICRL, Psyleron, GCP
PEAR's February 10, 2007 press release said the physical lab would finish at the end of February. Oversight of ICRL—founded in 1996 as a multi-disciplinary, multi-nation consortium—would carry the educational and research mission. Jahn and Dunne would advise Psyleron, a Princeton-area firm building REG tools for research and public exploration. Both would continue with the Society for Scientific Exploration. More than fifty publications stayed on the PEAR site; Margins of Reality (Harcourt, 1987) stayed in print; a PEAR Proposition DVD summed the history.
The Global Consciousness Project sits beside that legacy, not inside the closed lab. Nelson's GCP funding executive summary (archived copy of the Princeton noosphere site) describes an independent collaboration affiliated with IONS: continuous data from physical REGs at about 65 locations, one 200-bit trial per second per node since August 1998, with a hypothesis that engaging world events correlate with anomalous structure in the network. The summary's primary formal-event claim: a roughly 4-sigma accumulation over about 186 rigorously defined tests (odds cited as 70,000 to one). Nelson's bio line in that document: GCP Director from 1997; Princeton / PEAR from 1980 to 2002. pear-lab.com lists Nelson's GCP work and Radin's IONS work as continuations by alumni.
Treat GCP numbers as GCP's own formal-series claim, not as a PEAR lab result and not as a settled physics result. The same replication and multiple-testing questions that hang over PEAR hang here, louder, because the "events" are chosen from the news of the world.
Implications: technology, objects, space-time
Strip the brand names. What is the practical implication set, stated without hype?
On technology. PEAR's founding sentence was about vulnerability of devices and information systems to operator consciousness. If the twelve-year REG composite and the cascade results are even partly real, then "random" subsystems in safety-critical or cryptographic settings are not only hardware problems; they are human-factors problems of a kind no DO-178 checklist lists. If PortREG and Jeffers are the better guide, then the vulnerability was never there at engineering scale, and the lesson is different: long-running anomaly programs can manufacture composite significance that multi-lab pre-registered primaries do not reproduce. Either way, the engineering takeaway is discipline. Pre-state the primary endpoint. Publish the miss. Do not promote the diode effect while soft-pedaling the consortium failure.
On objects. FieldREG put a quiet box in rooms where nobody was aiming at it—rituals, conferences, a humor stage, a business meeting—and reported drifts that tracked emotional weather more than joysticks. The cascade put intention next to matter you can watch. Balls. Pegs. Bins. That is closer to a shop, a track, a hangar than a psychology survey. PEAR reported right/left separations and a stubborn left-side asymmetry. Critics will say any mechanical system has bias channels. Fair. The right response is more hardware characterization, not a vibe. Objects do not care about our metaphors. They care about forces we can name.
On space-time. Remote and off-time machine protocols, plus the remote-perception database, are where PEAR pressed hardest on the fabric of reality. Their theoretical papers (including the 1986 Foundations of Physics quantum-mechanics-and-consciousness line of work and the 2005 Proposition) argue for models in which consciousness participates in how outcomes become actual—complementarity language, information language, not cartoon telekinesis. Distance and delay, they say, did not kill the effect sizes. That is either a clue that we have the wrong map of information, or a clue that the residual is not a physical coupling at all. Present science does not have a mechanism that makes PEAR's preferred reading obligatory. Present science also does not get to pretend the databases were never published.
The honest middle is uncomfortable, which is usually where the work is. Tiny reported effect sizes. Huge claimed composites. A failed primary multi-lab replication. A null optical program at York. A lab that closed on its own schedule and left a nonprofit, a gadget company, and a global REG network behind. Human beings will keep leaning on machines, and machines will keep returning noise. Whether that noise ever leans back is still, on the public record, a fight—not a footnote.
What to take from it
PEAR asked whether intention leaves a mark on random technology and on the perception of distant places. Their best internal summaries say yes, at roughly one part in ten thousand per bit, compounding over vast runs, including at a distance. Their best collaborative replication says the primary mean-shift claim did not reproduce at persuasive size. Jeffers' York slits sat on the null; Princeton's were marginal. Alcock's demand still stands: give the null hypothesis a chance, especially when your own consortium primary misses.
We do not have to settle the metaphysics to settle the habit. Pre-register the endpoint. Separate exploration from confirmation. Keep the cascade as honest as the diode. And when a lab bets decades on the edge of mind and matter, read the papers that hurt the story as carefully as the ones that feed it.
The noise is still there. So is the operator. What we owe both is precision.
---
Sources
Primary sources
- Princeton Engineering Anomalies Research (pear-lab.com). About / legacy overview: PEAR 1979–2007; Jahn; Dunne; alumni Radin (IONS) and Nelson (GCP); Psyleron; Entangled.org. https://www.pear-lab.com/
- PEAR publications index. https://www.pear-lab.com/publications
- Jahn, R. G., Dunne, B. J., Nelson, R. D., Dobyns, Y. H., & Bradish, G. J. (1997). Correlations of random binary sequences with pre-stated operator intention: A review of a 12-year program. Journal of Scientific Exploration, 11(3). PDF: https://www.pear-lab.com/pdfs/1997-correlations-random-binary-sequences-12-year-review.pdf
- Dunne, B. J., Nelson, R. D., & Jahn, R. G. (1988). Operator-related anomalies in a random mechanical cascade. Journal of Scientific Exploration, 2(2), 155–179. PDF: https://www.pear-lab.com/pdfs/1988-operator-related-anomalies-rmc.pdf
- Dunne, B. J., & Jahn, R. G. (1992). Experiments in remote human/machine interaction. Journal of Scientific Exploration, 6(4), 311–332. PDF: https://www.pear-lab.com/pdfs/1992-experiments-remote-human-machine.pdf
- Jahn, R., et al. (2000). Mind/Machine Interaction Consortium: PortREG replication experiments. Journal of Scientific Exploration, 14(4), 499–555. PDF: https://www.pear-lab.com/pdfs/2000-mmi-consortium-portreg-replication.pdf
- Ibison, M., & Jeffers, S. (1998). A double-slit diffraction experiment to investigate claims of consciousness-related anomalies. Journal of Scientific Exploration, 12(4). PDF: https://www.pear-lab.com/pdfs/1998-double-slit-consciousness-experiment.pdf
- Jahn, R. G., & Dunne, B. J. (2005). The PEAR Proposition. Journal of Scientific Exploration, 19(2), 195–245. PDF: https://www.pear-lab.com/pdfs/2005-pear-proposition.pdf
- Jahn, R. G., & Dunne, B. J. (1986). On the quantum mechanics of consciousness, with application to anomalous phenomena. Foundations of Physics, 16, 721–772. Lab PDF: https://www.pear-lab.com/pdfs/1986-quantum-mechanics-consciousness.pdf — Springer record: https://link.springer.com/article/10.1007/BF00734564
- PEAR press statement announcing laboratory closure (Feb. 10, 2007). https://pearlab.icrl.org/press-statement.html
- PEAR Future / ICRL, Psyleron, SSE. https://pearlab.icrl.org/future.html
- International Consciousness Research Laboratories. https://www.icrl.org/
- pearlab.icrl.org home archive portal. https://pearlab.icrl.org/
- Alcock, J. E. (2003). Give the null hypothesis a chance: Reasons to remain doubtful about the existence of psi. Journal of Consciousness Studies, 10(6–7), 29–50. PDF: https://www.sgha.net/library/Alcock-editorial.pdf
- Ball, P. (2007, Feb. 13). When research goes PEAR-shaped. Nature news. https://www.nature.com/news/2007/070212/full/news070212-6.html
- Nelson, R. D. Funding The Global Consciousness Project (executive summary; ~65 REG nodes; ~4σ / ~186 formal tests; Nelson PEAR 1980–2002). Wayback archive of https://noosphere.princeton.edu/papers/pdf/GCP.exec.summary.pdf — https://web.archive.org/web/20150906120328/https://noosphere.princeton.edu/papers/pdf/GCP.exec.summary.pdf
- Society for Scientific Exploration. https://www.scientificexploration.org/
- Nelson, R. D., Bradish, G. J., Dobyns, Y. H., Dunne, B. J., & Jahn, R. G. (1996). FieldREG anomalies in group situations. Journal of Scientific Exploration, 10(1), 111–141. PDF: https://www.pear-lab.com/pdfs/1996-fieldreg-anomalies-group-situations.pdf (mirrors: https://global-mind.org/papers/pear/fieldreg1.pdf )
- Nelson, R. D., Jahn, R. G., Dunne, B. J., Dobyns, Y. H., & Bradish, G. J. (1998). FieldREG II: Consciousness field effects: Replications and explorations. Journal of Scientific Exploration, 12(3), 425–454. PDF: https://www.pear-lab.com/pdfs/1998-fieldreg-ii-consciousness-field-effects.pdf (mirrors: https://global-mind.org/papers/pear/fieldreg2.pdf ; https://icrl.org/wp-content/uploads/2020/02/1998-fieldreg-ii-consciousness-field-effects.pdf )
Flagged / alternate routes
- FieldREG: earlier guessed filenames (
1996-fieldreg.pdf,1998-fieldreg-ii.pdf) returned HTML error shells. Correct pear-lab.com paths load as real PDFs:1996-fieldreg-anomalies-group-situations.pdfand1998-fieldreg-ii-consciousness-field-effects.pdf. Also verified via global-mind.org, gcp2.net, ICRL WordPress, and Wayback. Livenoosphere.princeton.edu/papers/pear/fieldreg*.pdfstill failed at recheck. ResearchGate and Google Scholar returned 403 to automated fetch; Semantic Scholar API returned 429; CiteSeerX download 404. Body now includes FieldREG from the primary PDFs. - Live noosphere.princeton.edu root and live GCP exec-summary URL still unreachable; GCP claims verified from the Wayback PDF.
- New York Times Feb. 10, 2007 PEAR story returned HTTP 403; Nature's Feb. 13, 2007 piece is used for closure context.
