La revisión por pares halla que las afirmaciones sobre «sexo cerebral» se basan en estadísticas defectuosas

Métodos defectuosos y conclusiones exageradas sobre el «sexo cerebral»: una revisión por pares de Kurth et al. (2022)

Comentario y análisis Sexo, biología y deporte

Authors: Colin M. Wright, Adam M. Rochussen, Samuel G. Stagg
Journal: Theory and Society
Year: 2026
Source: link.springer.com/article/10.1007/s11186-026-09747-2

Summary

A widely cited 2022 brain-scan study by Kurth and colleagues reported that the brains of 24 transgender women, before hormone treatment, fell between those of men and women, though closer to men. This 2026 critique argues the key difference from men only looked real because of lenient statistical choices: testing in one direction only, and not adjusting for making three comparisons. With standard adjustments, the difference from men could easily be chance, though the difference from women still holds. Small groups and one unusual result make it fragile. Sexual orientation, known to relate to brain structure, was never analysed though the data existed. The critique is plausible but uses only published figures, was accepted nine days after it was sent in, and its lead author is a Manhattan Institute fellow on the journal's advisory board.

Key findings

El estudio de 2022 comparó a 24 hombres, 24 mujeres y 24 mujeres transgénero que aún no tomaban hormonas, usando un modelo informático que puntúa los cerebros en una única escala de masculino a femenino. Concluyó que los cerebros de las mujeres trans se desplazaban de su sexo de nacimiento hacia su identidad de género.
El estudio solo probó una diferencia en una dirección. Al probar ambas direcciones, como es habitual, la diferencia clave entre mujeres trans y hombres se debilita. Al ajustar por el hecho de que se hicieron tres comparaciones, podría deberse fácilmente al azar.
La diferencia mayor entre mujeres trans y mujeres se mantendría tras estos ajustes. Así que la disputa es sobre si los cerebros de las mujeres trans difieren de los de los hombres, no sobre si difieren de los de las mujeres.
Con solo 24 personas por grupo, el tamaño de cualquier diferencia es inestable. Un resultado inusual en el grupo de hombres nunca se comprobó volviendo a ejecutar el análisis sin él.
La prueba que el estudio usó para justificar su método estadístico es débil con grupos tan pequeños. Los críticos dicen que al menos deberían haberse presentado también métodos que hagan menos suposiciones.
De las 24 mujeres trans, 6 se sentían atraídas por hombres y 18 por mujeres. La orientación sexual está vinculada a la estructura cerebral, y los autores originales tenían estos datos pero no los tuvieron en cuenta.
Cada crítico pidió repetidamente los datos originales y no obtuvo respuesta, por lo que la crítica se apoya en las cifras publicadas. Piden que se publiquen los datos y que otros repitan el estudio.

Abstract

Kurth et al. (2022) concluded that the brain anatomy of transgender women is shifted away from their biological sex and toward their gender identity. We evaluate whether the study’s evidence supports this conclusion by focusing on its statistical methods and treatment of sexual orientation in its analyses. The principal finding relies on a very small sample, a weak justification for parametric analyses, one-tailed post hoc tests, no correction for multiple comparisons, and limited robustness testing. The study also did not properly account for sexual orientation as a potential neuroanatomical confound despite having the relevant participant data to do so. These methodological choices consistently bias the results toward smaller p-values and stronger apparent support for the authors’ stated hypothesis. The evidence presented therefore does not justify the study’s headline conclusion.

Conclusion

The analyses presented do not adequately support the paper's headline conclusion that transgender women's brains are shifted toward their gender identity. The evidence used to support that conclusion is weakened by a series of methodological limitations and poor choices, including a small sample size, parametric statistical methods, one-tailed post hoc hypothesis tests without correction for multiple comparisons, and the absence of several robustness analyses. The study also does not address sexual orientation as a potential confound, despite the authors possessing both the neuroanatomical and sexual orientation data needed to evaluate this possibility. These methodological choices and omissions consistently strengthen the perceived support for the authors' stated hypothesis while weakening confidence in the reported findings. Given the paper's unusually broad influence in neuroscience, medicine, journalism, law, and public policy, the methodological issues warrant careful consideration before the study can be cited as evidence supporting the hypothesis, and the authors welcome a response from the original authors and encourage public release of the anonymized participant-level data so that the findings can be independently replicated and evaluated.

Methodological limitations

No new data or re-analysis: The critics could not get the original data, so they cannot show whether accounting for sexual orientation or removing the unusual result actually changes the finding. These points remain ideas, not proof.
Which adjustment to use is debatable: The adjustment the critics apply for multiple comparisons is a strict one. The original authors stated in advance which direction they expected, which some statisticians accept as a reason to test in one direction. How much weight to give the adjusted result is a judgement call.
Narrow focus: The critique looks only at this one 2022 study. It does not cover the wider brain-scan research on transgender people, which includes larger studies with mixed results.
Only checks the statistics: It does not judge the computer model itself (what it was trained on, or whether a single 'brain sex' scale makes sense) or how the brain scans were processed, which the critics set aside 'for the sake of space'.
Very fast review: The paper was received on 19 August 2026 and accepted on 28 August 2026, an unusually short time for checking by outside experts.
Author ties to the journal and a think tank: Lead author Colin Wright is a Manhattan Institute fellow, has been paid as an expert witness on sex and gender identity, and sits on the journal's advisory board (he says he was not involved in handling the paper). The Manhattan Institute paid the fee to make it free to read.
Published in a sociology journal: It appeared in a sociology journal rather than a brain science or statistics journal, where specialist reviewers might have looked harder at the brain-scan methods.
No reply yet: Kurth and colleagues had not published a response at the time of review, so the critique has not been tested in debate.

detrans.ai

Summary

Key findings

Abstract

Conclusion