An Updated Meta-Analysis Of Post-Prl Esp Ganzfeld Experiments: The Effect Of Standardness

ABSTRACT: A number of meta-analyses have been published claiming that ESP ganzfeld experiments do or do not provide convincing evidence for psi. Hyman and Honorton agreed that a subset of 28 manual ganzfeld studies were collectively significant, although they disagreed whether the results could be explained by methodological artifacts. A subsequent series of automated ganzfeld experiments by Honorton at the Psychophysical Research Laboratories (PRL) overcame Hyman's objections and provided independently significant results. However, meta-analyses of post-PRL ganzfeld experiments by Milton and Wiseman and by Milton have revealed that the mean effect sizes (MESs) obtained in the earlier work were not maintained. These meta-analyses were criticized because they included experiments that obtained negative results using allegedly nonstandard ganzfeld protocols (cf. Schmeidler & Edge). Such evaluations of standardness are potentially biased, because they were made with knowledge of the associated ESP results, so th e present authors decided to obtain less biased estimates by having 3 psychology graduate students independently rate the Method sections of 40 post-PRL ganzfeld studies, sans results. A standard ganzfeld was defined for the judges by a description of general ganzfeld methodology from Bem and Honorton, plus the Method section of the primary report describing the PRL autoganzfeld studies. Additional guidelines focused the raters on the experimental procedures per se and discouraged the coding of minor deviations that both sides of the debate would consider within the bounds of standardness. The basic meta-analysis of the 40 post-PRL studies confirmed Milton's conclusion that they are now statistically significant (by Stouffer Z), but MESs fell outside the confidence limits of both the manual and PRL databases. Both the post-PRL and the manual ganzfeld databases, but not the PRL database, were found to be markedly heterogeneous. It was noted that previous comparisons between the databases were flawed because th e manual and PRL databases were restricted to studies in which ESP outcomes were defined as the proportion of direct ESP hits (DH), whereas the post-PRL database was not. For both the manual and post-PRL databases the DH studies produced markedly better results than the non-DH studies. Comparisons between the DH studies in each database and all studies in each database gave mixed results, with MESs from the post-PRL DH studies now falling within the confidence limits of the PRL studies but not the manual DH studies. Standardness ratings of the post-PRL studies were positively correlated with ESP ESs, p = .024, one-tailed, and ESs from standard studies (above the midpoint of the rating scale) were significantly more positive than those from nonstandard studies, p = .020, one-tailed. DH studies were more standard than non-DH studies, p = .005. Weighting ESs by standardness ratings and restricting the meta-analysis to standard studies raised the Stouffer Zs and MESs, such that MESs for the post-PRL studies fell more safely within the PRL confidence limits. They sometimes fell inside and sometimes outside the manual confidence limits, depending on the particular analysis. Weighting studies by sample size brought the MESs of the various databases into closer proximity, but these weighted studies were not fully meta-analyzed for this report.

COPYRIGHT 2000 Parapsychology Press
COPYRIGHT 2001 Gale Group