Robert G. Jahn

Jahn's career before 1979 was in electric propulsion, the physics of accelerating a spacecraft by ionising a propellant rather than burning it, and he wrote the textbook that trained a generation in the subject. That standing is what allowed the next part. When a graduate student asked to look at claims about mental influence on machines, Jahn opened a small laboratory to test it, and he justified it to a wary administration in engineering terms rather than psychical ones: modern systems depend on hardware random number generators, and if a human operator could bias such a device even slightly, that would be an engineering fact about the reliability of those systems whatever its ultimate explanation.

The framing determined the whole research programme. PEAR did not hunt for gifted psychics; it used ordinary volunteers, called them operators, and looked for a minute statistical shift accumulated over enormous numbers of binary trials, on the reasoning that a real but tiny effect would only become visible in bulk. It also determined how the work was received, since housing the laboratory inside a school of engineering meant its output was reviewed largely by people with no training in experimental psychology, which critics identified early as a structural weakness.

The laboratory reported a persistent departure from chance across its cumulative database, characterised as an effect on the order of one part in ten thousand, together with the claim that individual operators differed reliably from one another. Criticism concentrated on three points: that the aggregate result was driven disproportionately by a small number of operators rather than distributed across participants, that the documentation and controls fell short of what psychology journals required, and that the analyses were not pre-specified.

The National Research Council review of 1988 was unfavourable. The decisive episode came around 2000, when PEAR joined two German groups, at Giessen and at Freiburg, in a consortium replication using PEAR's own hardware and software under one shared protocol; all three sites produced trends in the expected direction that were not statistically significant, a result Jahn and Dunne discussed at length and which most readers outside the laboratory took as a failure to confirm.

A later meta-analysis of the whole random-generator literature by Bösch, Steinkamp and Boller in 2006 found the reported effect shrank as sample sizes grew, the standard signature of publication bias. Princeton allowed the laboratory to close on Jahn's retirement in 2007. He and Dunne continued through the International Consciousness Research Laboratories, and Jahn maintained to the end that the effect was real and small rather than absent.