Academic journal article The Journal of Parapsychology

A Meta-Analysis of Forced-Choice Experiments Comparing Clairvoyance and Precognition

Academic journal article The Journal of Parapsychology

A Meta-Analysis of Forced-Choice Experiments Comparing Clairvoyance and Precognition

Article excerpt

Parapsychologists have long been interested in the question of whether clairvoyance and precognition are really different phenomena (e.g., Morris, 1982). It has been suggested that apparent precognition might actually consist of a person sampling the present environment via real-time ESP and extrapolating from the information to make an informed prediction about future events (see, e.g., Mundle, 1978). If this hypothesis is correct, then clairvoyance studies would be expected to result in higher effect sizes than precognition studies because of the extra calculational step involved in the latter type of ESP task. Conversely, it has been proposed that what appears to be a clairvoyance task may not involve real-time information acquisition but rather consists of precognition of the feedback of the target's identity later received by the participant (e.g., Carington, 1945, p. 91). In this case, the effect sizes of the two types of ESP task would be expected to be the same and would be expected to be influenced by the same moderator variables, all other things being equal. There could, of course, be various additional models of clairvoyance and precognition that would have different predictions about the relative strengths of the effect sizes. However, only two papers to date (Milton, 1998; Tart, 1983) have reported comparisons of effect sizes between the two types of studies.

Tart (1983) presented an unusual meta-analysis of 85 forced-choice ESP studies. He discarded all studies that were not statistically significant at an alpha of .05 and also excluded all but the most successful subset of data from any study that broke down the overall result into individual participant data or individual condition or run data. By assessing the bit rate per trial, he found that real-time ESP (clairvoyance or telepathy) studies outperformed the precognitive studies (p [less than] 5 x [10.sup.-4]). However, the reverse held in Milton's (1998) meta-analysis of 78 free-response ESP studies where the mean effect size (z/[N.sup.1/2]) for the 6 precognition studies (0.34) was higher than that of both the telepathy studies (0.18) and the clairvoyance studies (0.08).

Although the results from Tart's (1983) and Milton's (1998) papers appear somewhat contradictory, both results have problems of interpretation. Milton's (1998) one-way ANOVA to test the effect size differences between precognition, clairvoyance, and telepathy was nonsignificant. The number of precognitive studies is too small to be representative.' of the latent population of such studies or to give an accurate estimate of their true effect size, which may be much lower than the observed figure. The confidence intervals around the estimated effect sizes of the three types of ESP overlapped considerably, making it easily possible that the true effect size of the population of precognition studies is below that of the real-time ESP studies, rather than above as the point estimates suggest. Tart's (1983) result is also inconclusive, but for different reasons. In that paper, the study selection criterion may have introduced a bias. If real-time and precognitive studies had the same effect size but precognitive studies tended to have a larger number of trials than real-time ones, the higher level of statistical power in the precognition studies would have enabled more precognition studies with lower effect sizes to reach the .05 alpha cutoff. This in turn would have meant a decrease in the average effect size for the precognition studies. Tart did not report study sizes and this artifact may be a strong possibility, given that precognition trials can be easier to run than clairvoyance trials and hence easier to collect in large numbers. Moreover, if more real-time than precognition studies reported breakdowns of data into subsets, Tart's policy of including only the "peak performance" data from each study would have similarly inflated real-time effect sizes relative to precognitive effect sizes. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.