The Decline of Effect Size in Psi Research: An Evidence-Based Commentary on Tressoldi and Storm (2024)

Abstract

We analyzed all available meta-analytic databases covering 10 protocols that are being used in psi research for a decline in effect size over years. From the total of 10 protocols seven showed a significant decline. An analysis using the first 20 publications (or the maximum number of publications) for each protocol gave a clear overall decline in effect size with publication sequence number with a specific protocol (rho = -0.176 N = 185 P = 0.008). The apparent contradiction of strong declines of Effect Size vs. Year when analyzed per protocol and no decline over-all is explained by the repeating pattern of rise and fall of results obtained by the different protocols with the more sensitive protocols with a larger mean effect size appearing later in years.

We also independently found a very robust decline in effect size with increasing number of trials, N, or power of the psi studies (rho = -0.397, N = 873, P < 10 ^-30). Since the power of studies also turned out to increase over the years we analyzed the temporal decline of ES over years controlling for N and found that the temporal declines were not a consequent of the independent decline effect of ES with increasing N.