What the Judging Procedure Was Actually Doing

The point on which Targ's work stands or falls is not whether the transcripts read impressively, which they often do, but what the blind judging established. In the SRI design a judge received a set of transcripts and a set of site descriptions and had to pair them, and a better-than-chance pairing rate was the published result.

The force of the Marks and Kammann critique of 1978 was that pairing is a task about relative order as much as about content: if the transcripts carry any incidental marker of the sequence in which they were produced, and the sites were visited in a known order, a judge can score well while learning nothing about any site. This is not an accusation of fraud and Marks and Kammann did not make one; it is a claim that an uncontrolled channel existed. Targ's answer has been that the cues were trivial, that removing them in later series did not remove the effect, and that critics have concentrated on the earliest and weakest studies because the later ones are harder to attack. Both positions are still argued in print. The 1995 government review found the statistical effects across the whole programme reliable and did not resolve what produced them, which is roughly where the question has stayed.