Extraversion and ESP performance: a meta-analysis and a new confirmation - extrasensory perception

Journal of Parapsychology, The, Sept, 1998 by Charles Honorton, Diane C. Ferrari, Daryl J. Bem

Randomization. Studies received credit if random number tables, random number generators, or mechanical shufflers were used to randomize the targets. Studies using informal methods of randomization such as hand shuffling and dice throwing, or in which there was no randomization, received no credit. Reports failing to identify the method of randomization received no credit.

Duplicate recording. Studies reporting duplicate recording of targets and responses received credit. Reports that failed to describe their data recording procedures and those in which data recording was performed by a single experimenter, received no credit.

Duplicate scoring. Studies reporting duplicate checking of hits retrieved credit. Reports that failed to describe their scoring procedures and those in which scoring was performed by a single experimenter, received no credit.

We assessed two aspects of methodology that could result in a spurious correlation between ESP scores and measures of extraversion: advance specification of the criteria used to define extraversion and the order in which the ESP task and extraversion measure were administered.

A priori definition of extraversion. An inflated relationship between ESP and extraversion scores could occur if the investigator selected an "optimal" extraversion/introversion breakdown after observing the data, without correcting for multiple analysis. We coded studies as to whether the classification of extraversion/introversion appeared to be predetermined or post hoc. No credit was given for studies using nonstandard classifications (e.g., [ or -] 1 sd, median split, etc.) unless the report explicitly stated that the method of classification was preplanned.

Order of administration of extraversion and ESP measures. A spurious correlation between ESP scores and extraversion could arise if subjects' responses on the extraversion scale were influenced by knowledge of their performance in the ESP task. There is some evidence that subjects' responses to psychological tests may be influenced by feedback concerning their ESP performance (Palmer & Lieberman, 1975). We coded studies as to whether the extraversion scale was given before or after the ESP test. This information was available for 45 of the 60 studies.

Meta-Analysis of Correlation Coefficients

We combined correlations across independent studies using the procedures described by Hedges and Olkin (1985) and Rosenthal (1984). All statistics were converted to indices of association; t tests were converted to point-biserials and phi coefficients were computed from 2 x 2 contingency tables.(1) We estimated unreported correlations from the reported p values (with results reported only as "nonsignificant" obtained in 13 cases set equal to .00). The signs of the correlations were adjusted if necessary to insure that positive correlations reflect positive relationships between extraversion and ESP performance. The correlations were pooled across studies but within categories based on methodological features associated with the studies. The correlations were transformed to their Fisher's z equivalents, weighted by their df, and averaged. We determined the two-tailed significance levels and 95% confidence intervals (CIs; Hedges and Olkin, 1985). Finally, we conducted chi-square tests of homogeneity (Hedges and Olkin, 1985; Rosenthal, 1984) and transformed mean z's back to the r metric.