Discovery Lab Answers

Effect Sizes in the ESP-Nexus Library: Charles Honorton

Before reading the table below, one qualification is essential: the library’s share of the published literature on Honorton’s work has not been measured. What follows summarizes the five studies currently held in the ESP-Nexus library — treat it as a sample of what the library holds, not a settled account of the field.

A second structural point: the five rows report different, non-comparable metrics — a standardized effect size (Cohen’s h), a proportion index (PI), a raw hit rate, and a combined p-value without an effect size. These cannot be averaged or pooled into a single bottom-line figure; each must be read on its own terms.

Reading the table

Direction is consistently positive across all five rows — every result favors above-chance performance. That said, the picture is not uniform in magnitude.

The Bem & Honorton (1994) autoganzfeld pooled result carries the largest effect size by the proportion index metric (0.59 PI, 32% hit rate, z = 2.89), but PI is not directly comparable to Cohen’s h. The Honorton (1990) combined database yields Cohen’s h = 0.29 with a confidence interval of — a more widely interpretable standardized metric, and the one most amenable to cross-study comparison within its own row. The 1989 cumulation across 62 investigators produces a much smaller z/√N figure (0.033), which reflects a different aggregation approach rather than a weaker individual study. The 1978 corpus entry supplies only a combined p-value with no effect size, so its magnitude cannot be placed on any effect-size scale.

A scatter plot of effect size against publication year cannot be rendered for these rows: the results span three different effect-size metrics across five studies, and the rows do not share a single comparable standardized measure with associated publication years that would support a fitted trend.

Unsettled signals the table surfaces

The sources retrieved for this question include several points of active disagreement or heterogeneity that the table alone does not convey:

• Bem and Honorton (1994) report a significant negative correlation between study sample size and effect size across the ten autoganzfeld studies — larger studies tended to show smaller effects, a pattern consistent with either publication bias or regression toward the mean in smaller samples.
• Half of the 28 ganzfeld studies in the broader database reviewed by Bem and Honorton (1994) came from only two laboratories, raising questions about whether the effect generalizes across research groups.
• Hyman and Honorton (1986) jointly concluded the database showed a significant effect not explained by selective reporting, while explicitly disagreeing on whether that effect constitutes evidence for psi — the methodological agreement and the interpretive disagreement coexist in the same document.

For the full research record, including the autoganzfeld protocol details and the Hyman–Honorton methodological exchange, see Charles Honorton and the spoke page on Ganzfeld methodology and standardization.