A meta-analysis of forced-choice experiments comparing clairvoyance and precognition

This meta-analysis examined a database of studies published in the main parapsychology journals from 1935-1997 that compared outcomes of precognition and clairvoyance trials under relatively similar experimental conditions. Both the precognition and clairvoyance studies had a statistically significant cumulated overall effect but there was no evidence to suggest that clairvoyance worked better than precognition, with the mean effect sizes (z/N-1/2) of the two types of ESP in the 22 study pairs being very similar at 0.010 for precognition and 0.009 for clairvoyance. There were no statistically significant correlations between the presence of procedural safeguards and effect size and hence no suggestion that methodological problems had played any strong and obvious role in the overall effects, although the small database would be expected to provide relatively low statistical power for detecting any such effects. None of the planned analyses examining the effects of potential moderator variables upon effect size were statistically significant but a post-hoc ANOVA indicated a statistically significant interaction, F(1, 16) = 15.04, p = .001. between whether the trials were precognitive or clairvoyant and how the two types of trial were interspersed (separated into studies or mixed within a study). Being a post-hoc result and one of many analyses performed, this finding may not be meaningful. Equally, however, the general lack of significant findings does not conclusively indicate a lack of genuine relationships, given the low statistical power in the database. In case low power had in fact been a problem in identifying moderator variables, contrast between groups in effect size alone, regardless of the statistical significance of the difference was used as a criterion to identify promising variables to examine or exploit in future research. These variables and suggestions for future research are discussed.