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

The chi-square homogeneity test assesses the consistency of study outcomes, providing a quantitative index of replicability. A set of studies are exact replicates if their effect sizes are identical. They are homogeneous if the variability of effect sizes can be explained by sampling error. A significant but nonhomogeneous effect indicates the presence of moderating variables and homogeneity tests are used to identify moderating variables by subdividing studies into smaller, methodologically similar subgroups (Hedges, 1987).

RESULTS

We retrieved 60 independent studies contained in 35 publications by 17 independent investigators. The studies were reported over a span of 38 years, between 1945 and 1983. The data base comprises 2,963 subjects. Forty-five studies involve ESP card-guessing tasks or similar forced-choice tasks. Fourteen studies employed free-response ESP tasks, and one involved a remote physiological influence task.

The unweighted correlations range from -0.44 to .91. Figure 1 shows a stem-and-leaf display (Tukey, 1977) of the correlation coefficients. (The 13 studies that were assigned r's of zero because of insufficient information are omitted). Unlike other methods of displaying frequency distributions, the stem-and-leaf plot retains the numerical data precisely. (Turned on its side,

the stem-and-leaf plot is a histogram.) Each number includes a stem and one or more leaves. For example, the stem .2 is followed by leaves of 0, 3, 3, 8 representing r's of .20, .23, .23, and .28.

Figure 1. Stem-and-leaf frequency distribution of correlation coefficients
between extraversion and ESP performance (N = 47).

-.4        4
-.3        6
-.2        5 2
-.1        7 4 1
-.0        9 2 0 0
.0         0 5 8 9
.1         0 0 1 2 2 7 8 8
.2         0 3 3 8
.3         0 0 4 4 4 5 7 8 8
.4         0 6 6 9 9
.5         3 4 8
.6         6 7
.7
.8
.9         1

The meta-analysis is summarized in Table 1. The study grouping is identified in the first column. Columns two through four show, respectively, the number of studies, investigators, and subjects. The mean weighted effect size (r) is shown in column five and columns six and seven show, respectively, the lower and upper 95% (CI) for the effect size. The cumulative z score and its associated two-tailed p value are shown in columns eight and nine. The chi-square homogeneity statistic is presented in the last column.

Overall Results

The results for all 60 studies are shown in the first row. The mean weighted r is 0.09 (r = 4.63, p = .000004, two-tailed). The 95% CI is an r from .05 to .12. While significant, the study effect sizes are nonhomogeneous ([[[Chi].sup.2].sub.59] = 126.21, p [less than] .05). The second row shows the same analysis, omitting the 13 studies that were assigned r's of zero.

Forced-Choice Studies

The ESP/extraversion correlations for the 45 forced-choice (FC) studies are presented in row three of Table 1. These studies were contributed by 13 independent investigators and include 2169 subjects. The [TABULAR DATA FOR TABLE 1 OMITTED] mean weighted r is .06 (z = 2.86, p = .0042, 95% CI from .02 to .11). The FC correlations are significantly nonhomogeneous ([[Chi].sup.2].sub.44] = 92.82, p [less than] .05).