Answers are grounded in this paper and cite their sources.
Reversed publication bias—the idea that politically sensitive findings may be selectively suppressed in favor of null effects—has recently gained attention in public and online discussions. Roth et al.'s (2015) meta-analysis of the association between intelligence and school grades (ρ = 0.54) has been frequently cited as supposed evidence, because its funnel plots appear to show larger correlations in studies with smaller sampling error. However, this study demonstrates that the pattern is entirely spurious. Reanalysis of the original data reveals that the asymmetry arises from the use of the conditional standard error of the correlation coefficient, which depends on the observed value of r and mechanically induces funnel-plot skew. When more appropriate methods, such as Fisher's z -transformation with unconditional standard errors, are applied, the asymmetry disappears and Egger's test becomes nonsignificant, t (238) = −1.41, p = .160. A complementary simulation study further confirms that conditional-error weighting can generate strong false signals of reversed publication bias and inflate total effect-size estimates even when no bias is present. Overall, these findings provide no evidence for reversed publication bias in research on intelligence and school grades. Using conditional standard errors of raw correlation coefficients in meta-analyses should be completely avoided. • No evidence of reversed publication bias in Roth et al. (2015) . • Funnel plot asymmetry stems from using conditional SEs against raw correlations. • Asymmetry disappears when using Fisher z-transformed correlations. • Weights based on conditional SEs inflate the meta-analytic total effect size. • Conditional SEs of correlations should be avoided in meta-analyses.
No takes yet. Share an insight, caveat, or question.
A 2026 study studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: