Abstract

Argues that the concept of randomness is significantly more complex than is generally appreciated. Research in parapsychology is examined critically in light of certain problems that derive from this complexity. Classical inferential statistics are not appropriate for many parapsychology experiments, and the p values reported in such experiments are often much too low. The use of special research designs, statistical analyses, and publication procedures is recommended for parapsychology and, by extension, for any research domain where effect sizes are very small or the events of interest are very rare. The proposed design calls for experimental (with psi active) and control (with psi absent or blocked) trials to be mixed unpredictably during the experiment.