The Hyman-Utts statistical debate (1995)

The 1995 American Institutes for Research review of the Stargate Project was structured around two independent statistical analyses. AIR contracted both. The first was written by Jessica Utts, a professor of statistics at the University of California, Davis. The second was written by Ray Hyman, a psychologist at the University of Oregon and a longtime professional critic of parapsychology. Utts and Hyman were asked to look at the same datasets. They agreed in advance to publish their conclusions side by side.

Utts wrote that the laboratory results from SRI and the later Science Applications International Corporation work were statistically significant, that the simpler methodological objections had been ruled out by the better experiments, and that the existence of the effect, whatever its mechanism, should be considered established. She estimated combined effect sizes comparable to those accepted in mainstream social-psychology research.

Hyman conceded the statistics. He agreed that the SRI and SAIC experiments had successfully ruled out cueing, fraud, and sensory leakage in the better trials. He disagreed sharply with Utts about what that meant. The right standard, he argued, was not statistical significance but successful replication across independent labs, of which there had been very little. Without a known mechanism and without independent replication, the effect should be considered unproven.

The two reviewers' written exchange continued after the AIR report was published. In 1996, Hyman published a critique in the Journal of Parapsychology under the title "Evaluation of the program on anomalous mental phenomena." Utts replied in the same issue. They have continued to refer to each other's arguments in subsequent papers up to the present day.

The Utts-Hyman exchange is the standard reference point for academic discussions of remote viewing. It is cited by both defenders and critics of the program, who often disagree about which reviewer's reading the actual evidence supports.