Abstract

Argues that successful replication in parapsychology should not be equated with the achievement of statistical significance, whether at the .05 or any other level. The history of the definition of statistical significance in parapsychology is traced, and the problems with using hypothesis -testing results to define successful replications are outlined. Suggestions for alternative approaches to looking at experimental data include calculating statistical power before doing an experiment; using estimation instead of, or in conjunction with, hypothesis testing; and implementing ideas from Bayesian statistics.