Little evidence for group differences in 2D:4D ratios based on sexual orientation after adjusting for publication bias

Tools

Shiramizu, Victor and Bartoš, František and Jones, Benedict C. and Pollet, Thomas V. (2026) Little evidence for group differences in 2D:4D ratios based on sexual orientation after adjusting for publication bias. Hormones and Behavior, 180. 105921. ISSN 1095-6867 (https://doi.org/10.1016/j.yhbeh.2026.105921)

[thumbnail of Shiramizu-etal-HB-2026-Little-evidence-for-group-differences-in-2D-4D-ratios-based-on-sexual-orientation]
Preview
 Text. Filename: Shiramizu-etal-HB-2026-Little-evidence-for-group-differences-in-2D-4D-ratios-based-on-sexual-orientation.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo
Download (792kB)| Preview

Abstract

The ratio between the lengths of the second and fourth digits (2D:4D) has been proposed as a putative marker of prenatal androgen exposure and investigated for its potential link to sexual orientation. In a recent updated meta-analysis, Swift-Gallant et al. (2025) reported more male-typical digit ratios in same-sex-oriented women than in heterosexual women, while homosexual men showed more female-typical ratios than heterosexual men. Their analyses found no evidence of publication bias for female sexual orientation but did for male sexual orientation, with effects remaining statistically significant after adjustment. Here, we reanalyzed the dataset compiled by Swift-Gallant et al. (2025) using Robust Bayesian Meta-Analysis (RoBMA), a framework that models uncertainty across multiple sources, including the presence or absence of an effect, heterogeneity, and publication bias. Our analyses showed little evidence for group differences in 2D:4D ratios by sexual orientation: RoBMA indicated moderate evidence against differences between homosexual and heterosexual women, with moderate to strong evidence of publication bias, and for men, moderate evidence against a right-hand difference but a small left-hand effect moderated by publication status. Thus, when publication bias was modeled directly using Bayesian model averaging of complementary bias adjustments, the evidence for a difference in 2D:4D ratios according to sexual orientation was substantially weakened, with this evidence for publication bias not attributable to unmodeled heterogeneity. Overall, these findings challenge claims of reliable group differences in 2D:4D ratios based on sexual orientation and question assertions that publication bias in this literature is negligible or reversed.

ORCID iDs

Shiramizu, Victor ORCID logoORCID: https://orcid.org/0000-0003-1669-2163, Bartoš, František, Jones, Benedict C. ORCID logoORCID: https://orcid.org/0000-0001-7777-0220 and Pollet, Thomas V.;
CORE (COnnecting REpositories)