Where Exactly the Two Reports Part Company

The disagreement is not statistical and pretending otherwise obscures what is interesting about it. Both reviewers agree that the data show a regularity that chance does not explain. The split occurs at the next step, and it is a question about inference rather than arithmetic. Utts's position is that when an effect has appeared repeatedly, across laboratories and investigators, at a consistent magnitude, the burden shifts: to keep attributing it to an unspecified flaw is to protect a conclusion from evidence.

Hyman's position is that all the studies in the pool descend from a small number of related protocols developed by the same people, so replication across laboratories does not mean independence of design, and a shared design flaw yields a consistent effect indistinguishable from a consistent phenomenon. Hyman also made the concession his critics rarely quote, writing that the effects were too consistent to dismiss and that he could not name the specific flaw responsible. Utts made the concession her supporters rarely quote, that the work had produced no understanding of mechanism and nothing usable. Both reports were republished in the Journal of Scientific Exploration in 1996 with short replies. Neither author has changed position since.