Forced-choice ESP and Zener cards

Zener cards provide five discrete outcomes, so the null model is exact: each independent guess succeeds with probability 0.20. Scoring requires no interpretation. A run can be represented as a binomial count, and the full sequence can be audited for dependence and optional exclusions.

That simplicity helped laboratory parapsychology grow. It also made small procedural leaks consequential. Poor shuffling, sensory cues, recording errors, repeated testing of selected participants, and stopping after a good run can all move a hit rate without extrasensory information.

A meta-analysis can combine many individually weak experiments and estimate a small mean departure from chance. Its answer depends on which studies are located, how quality is coded, whether repeated reports share participants, and what model is used for unpublished work. Advocates and critics have made different choices at each step.

The disagreement is therefore not captured by counting supportive and sceptical papers. The central question is whether prospectively registered, well-concealed studies conducted by independent teams converge on the same effect size. That convergence has not emerged clearly enough to treat psi as an established signal.

The Psychic Self-Test commits cryptographically to each target before a guess and reveals the nonce afterward. That does not prove the generator is perfect, but it makes substitution after the answer detectable. The protocol fixes the run length and includes every completed trial in open data.

One public register will not settle a century of dispute. It can produce a clean, continuously accumulating estimate with a visible 20-percent line and no curator choosing which sessions become studies. That modest contribution is more defensible than promising a final verdict.