Princeton Engineering Anomalies Research (PEAR)

The full file

In 1979 Robert G. Jahn, then dean of Princeton's School of Engineering and Applied Science, founded a small laboratory to test whether human intention could alter the behaviour of physical devices. Princeton Engineering Anomalies Research, known as PEAR, ran for twenty-eight years in the engineering building, managed by Brenda Dunne and paid for almost entirely by private donors, among them James McDonnell of McDonnell Douglas. Unusually for the field, its last major experiment was a strict replication, run with outside partners, whose primary result its own founders published as a failure.

What was reported

PEAR's main instrument was the random event generator: an electronic noise source sampled into strings of binary bits, summed into "trials" of 200 bits, with a chance mean of 100. Volunteer "operators", unpaid and anonymous, sat by the machine and tried in turn to push its output higher, push it lower, or leave it alone, the three intentions recorded in advance. Jahn set out the programme for an engineering audience in "The persistent paradox of psychic phenomena: an engineering perspective", in the Proceedings of the IEEE in 1982.

Over the following years the laboratory reported a small but persistent shift in the intended direction. Its benchmark database, as summarised in 2000, came from 91 operators over twelve years: 522 series and nearly two and a half million trials, with a separation between high and low intentions of about 0.02 on the 100-point scale and a Z-score of 3.81. That is a few parts in ten thousand, made statistically significant only by the size of the database.

The record and where it is held

The case is on this shelf because of a record its subjects did not write: the multi-laboratory replication. In 1996 PEAR formed the Mind/Machine Interaction Consortium with the Freiburg Anomalous Mind/Machine Interactions group at the Institut für Grenzgebiete der Psychologie und Psychohygiene, Freiburg, and the Giessen Anomalies Research Project at Justus-Liebig-Universität Giessen. Each laboratory collected 250 sessions of 3,000 trials on identical PortREG noise sources, with 227 operators in all, under an agreed primary criterion: the high-minus-low separation. The twelve authors, from all three groups, published the result in the Journal of Scientific Exploration 14(4), pp. 499–555, in 2000. A copy is hosted by the International Consciousness Research Laboratories, the organisation that succeeded PEAR.

What investigators concluded

The consortium's abstract is plain. The separations went the intended way at all three sites, but "failed by an order of magnitude" to match the earlier experiments, or to reach "any persuasive level of statistical significance". In the paper's tables the pooled separation is about 0.003, with a Z-score of 0.60; Princeton's own share is about 0.002, Z of 0.17. The authors called this "an empirical paradox". They reported instead "interior structural anomalies", such as depressed trial-level standard deviations and irregular series-position patterns, which they judged well beyond chance, though none was the agreed test. They suggested that the simpler equipment, computer-screen feedback, or a changed operator pool might explain the drop.

Stanley Jeffers, a physicist at York University in Toronto who had run experiments of this kind with others in the field, wrote in the Skeptical Inquirer in 2006 that "even the PEAR group was unable to reproduce a credible effect". He also argued that PEAR's baseline data, the condition in which operators tried to do nothing, fell outside the laboratory's own significance limits, which would suggest the device was not behaving randomly.

Jahn and Dunne closed the laboratory in February 2007. Nature's news report of 13 February quoted their statement that they had "accomplished what we originally set out to do 28 years ago". The Daily Princetonian reported that the equipment had been dismantled, much of it bought by other researchers, and that donor deaths had left the laboratory unable to renew it; Jahn spoke of "a law of diminishing returns". The Daily Pennsylvanian, reporting on 19 February, said instead that the laboratory would close at the end of the year.

The leading conventional explanation

For random-generator studies in general, the leading explanation is small-study and publication bias. In 2006 H. Bösch, F. Steinkamp and E. Boller combined 380 such studies in Psychological Bulletin and found "a significant but very small overall effect size", with effect sizes "strongly and inversely related to sample size". A simulation showed that this pattern "could in principle be a result of publication bias". Bösch and Boller had been Freiburg members of the consortium, so the meta-analysis was independent of PEAR but not of the replication. For PEAR itself, the plainest reading is that an effect this small, found mainly in one laboratory's database, shrank towards zero when the same laboratory and two others tested it under an agreed criterion.

The explanation has limits. Publication bias across many studies does not by itself explain the shift in PEAR's own benchmark database. Jeffers's baseline argument points to a fault in the device or the analysis but does not identify one.

What remains disputed

Proponents did not accept the bias reading. Dean Radin, Roger Nelson, York Dobyns and Joop Houtkooper replied in the same journal that Bösch and colleagues had wrongly assumed effect size to be independent of sample size, and called selective reporting "an implausible explanation". Whether the benchmark shift reflects an undetected technical or statistical artefact, and whether the structural anomalies in the replication mean anything beyond the many comparisons made, have not been settled in print.

Sources

• Mind/Machine Interaction Consortium: PortREG Replication Experiments (archived) — R. Jahn et al., Journal of Scientific Exploration 14(4), 499–555, 2000 (ICRL copy). Consortium membership, design, 227 operators, prior PEAR database, primary result, structural anomalies, authors' explanations.
• Examining psychokinesis: the interaction of human intention with random number generators — a meta-analysis — H. Bösch, F. Steinkamp and E. Boller, Psychological Bulletin 132(4), 497–523, 2006. Abstract: 380 studies, small effect, inverse relation to sample size, publication bias.
• Reexamining psychokinesis: comment on Bösch, Steinkamp, and Boller — D. Radin, R. Nelson, Y. Dobyns and J. Houtkooper, Psychological Bulletin 132(4), 529–532, 2006. Proponents' rejection of the publication-bias reading.
• The persistent paradox of psychic phenomena: an engineering perspective (archived) — R. G. Jahn, Proceedings of the IEEE 70(2), 136–170, 1982. Citation details only (Crossref metadata).
• The PEAR Proposition: Fact or Fallacy? (archived) — Stanley Jeffers, Skeptical Inquirer 30(3), 2006. Replication failure and the baseline critique.
• When research goes PEAR-shaped (archived) — Philip Ball, Nature news, 13 February 2007. Closure by Jahn and Dunne, their statement, private funding, ICRL.
• Weird science loses its home (archived) — The Daily Princetonian, February 2007. Founding, donors, dismantling, sale of equipment.
• Anomalous Princeton lab shuts its doors — Rebecca Kaplan, The Daily Pennsylvanian, 19 February 2007. Retirement to ICRL and a later closing date.
• Robert George Jahn ’51 *55 (archived) — Princeton Alumni Weekly memorial. Deanship from 1971, PEAR 1979–2007, death in November 2017.
• Princeton Engineering Anomalies Research (PEAR) Laboratory (archived) — Roger Nelson, Psi Encyclopedia, Society for Psychical Research, 2017. Insider's account of staff, funding and equipment; used as an index.