Parapsychological researchers have long been interested in exploring if there are any factors which might relate to why some people report having more psi experiences in their everyday life than do others. Similarly, while most experimental work is done with volunteer subjects who have not been chosen on the basis of their supposed psi ability, it has been observed that some people appear to do better in experimental psi tests than others. One approach to examining possible reasons for these observed differences has involved exploring the relationship between various personality factors and psi ability.
Two meta-analyses of studies which have looked for correlations between performance on a psi task and different personality traits will be discussed here. One of these involved studies which looked for a relationship between a person's opinion of psi and their own psi abilities with their psi test performance. Research examining what has come to be known as the sheep/goat effect, supported the hypothesis that in experimental psi tests those with positive attitudes (``sheep'') tend to score above chance, and those with negative attitudes (``goats'') below chance.
Lawrence  conducted a meta-analysis of the 73 published studies examining the sheep/goat effect. These studies were conducted by 37 principal investigators, and involved over 4,500 subjects who completed over 685,000 trials. The overall effect size per trial is small ( r = 0.029), but highly significant over these studies which involved a large number of procedural manipulations and potential modifying variables. The combined Stouffer z = 8.17, p = . Using seven different measure of study quality, Lawrence found that effect size did not covary with study quality. A file-drawer estimate (Rosenthal's ``fail-safe N'') revealed that 1726 unreported studies with null results (i.e., 23 unreported studies for each of the 73 reported ones) would be required to reduce the significance of the database to chance expectancy.
This database has used a wide range of different sheep/goat scales, ranging from single questions to more lengthy questionnaires. The means of determining belief have also varied, with most focusing upon previous personal psi experiences, self-evaluation of personal psi ability, opinions regarding one's ability to display psi ability in the specific testing situation and/or one's general attitudes towards such phenomena. Lawrence found there was no overall relationship between effect size and the type of measure used, from which he concluded that the sheep/goat effect was quite robust regardless of how it was measured.
Another personality trait that has been studied in relation to psi performance is extraversion/introversion. Honorton, Ferrari and Bem  conducted a meta-analysis on the 60 published studies examining this relationship. Prior to this meta-analysis, descriptive reviews of this database had concluded that extraverts performed better than introverts on psi tasks (Eysenck, , Palmer , Sargent ). However, the ability of meta-analysis to identify flaws and modifying variables led to a different finding in the meta-analysis. While the meta-analysis did find a significant overall effect ( r = .09, combined z = 4.63, p = 0.000004), the effect sizes were non-homogeneous. The studies were divided into smaller groups according to various procedural variables in order to discover the source of the non-homogeneity. The authors separated the 45 studies using forced-choice procedures from the 14 studies using free-response methods. Once again, significant but non-homogeneous, effects were found (forced choice: r = .06, combined z = 2.86, p = 0.0042; free-response: r = .20, combined z = 4.82, p = 0.0000015). A further division of these two groups of studies examined whether testing subjects individually or in groups had any impact on the outcomes. This analysis revealed that of the forced-choice studies, 21 studies had tested subjects individually, resulting in a significant, but non-homogeneous effect ( r = .15, combined z = 4.54, p = 0.000006). In the 24 forced-choice studies where participants were tested in groups, there was no significant effect ( r = .00, z = --0.02), although there was homogeneity .
A flaw analysis showed that the significant effect in the forced-choice database was entirely due to 18 studies in which the extraversion measure had been given after the ESP test, the significance of this correlation being due to 9 of these studies in which the subjects knew how they had performed on their psi task before they completed the extraversion questionnaire. This finding raises the strong possibility that the correlation was due to psychological, as opposed to paranormal, factors. Thus the previous descriptive reviews which had found a significant, positive relationship between extraversion and psi-scoring had failed to uncover the inconsistency in the degree to which this effect was present in these studies, and the flaw which lead Honorton, Ferrari and Bem to conclude that the relationship in forced-choice studies would appear to be artifactual.
In the subset of 14 free-response extraversion studies, a significant ( r = .20, combined z = 4.82, p = 0000015) but non-homogeneous effect was obtained. Dividing the studies according to individual or group testing procedures revealed that the 2 studies employing group testing were responsible for the non-homogeneity. The results for the 12 studies which testing subjects individually show homogeneity and a significant correlation ( r = .20, combined z = 4.46, p = 0.0000083). Eleven of the studies documented the presentation order of the psi test and extraversion questionnaire. In all of these studies, the extraversion questionnaire was given prior to the ESP test, thereby avoiding the potential problem of subject's knowledge of their ESP results influencing the way that they completed their extraversion questionnaire. These 11 studies show a significant and homogeneous extraversion/ESP correlation ( r = .21, combined z = 4.57, p = .000005).
After completing the extraversion/ESP meta-analysis, Honorton et al. examined the autoganzfeld database to see if they could confirm the relationship. For the 221 autoganzfeld trials for which they had extraversion data, they obtained a significant ESP/extraversion correlation ( . This finding is consistent with those from the free-response extraversion meta-analysis.