henhao
Sign in
NeuroscienceCerebral Cortex
Peer-reviewed journal article

Sex Differences In The Adult Human Brain: Evidence From 5,216 UK Biobank Participants

Stuart James RitchieiD, Simon R Cox, Xueyi SheniD, Michael Vincent LombardoiD, Lianne Maria ReusiD, Clara AllozaiD, Mathew A. HarrisiD, Helen L. Alderson, Stuart Hunter, Emma NeilsoniD, David C M LiewaldiD, Bonnie Auyeung BiD, Heather Clare WhalleyiD, Stephen M. LawrieiD, Catharine R. GaléiD, Mark E. BastiniD, Andrew Mark McIntoshiD, Ian J. DearyiD

Abstract

Sex differences in the human brain are of interest, for example because of sex differences in the observed prevalence of psychiatric disorders and in some psychological traits. We report the largest single-sample study of structural and functional sex differences in the human brain (2,750 female, 2,466 male participants; 44-77 years). Males had higher volumes, surface areas, and white matter fractional anisotropy; females had thicker cortices and higher white matter tract complexity. There was considerable distributional overlap between the sexes. Subregional differences were not fully attributable to differences in total volume or height. There was generally greater male variance across structural measures. Functional connectome organization showed stronger connectivity for males in unimodal sensorimotor cortices, and stronger connectivity for females in the default mode network. This large-scale study provides a foundation for attempts to understand the causes and consequences of sex differences in adult brain structure and function.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
855
FWCI
88.29
Field percentile
top 1%
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (22 Jan 2018) · CC BY 4.0

Across 5,216 UK Biobank adults, males had larger brain volumes and surface areas, females thicker cortices and stronger default-mode connectivity; male variance was generally greater.

Earlier regional studies of brain sex differences were underpowered (meta-analysed region-level studies averaged 130 participants), white matter diffusion data were scarce, and sex differences in brain variance had never been tested in anyone older than 20 years. This single-scanner UK Biobank analysis provides multimodal reference estimates of both mean and variance differences across structural, diffusion and resting-state functional MRI.

Key results

  1. ShownMales had larger global, subcortical and regional cortical volumes and surface areas, often with large effects. Adjusting for total brain volume or height shrank many of these differences without removing them all, and some regions reversed to favor females.

    Cohen's d −1.41 (total), −1.28 (grey), −1.49 (white matter volume); mean d −0.70 over fourteen subcortical volumes; mean d −0.67 (volume) and −0.83 (surface area) over 68 regions. After total brain volume (TBV) adjustment, 44/68 regional volumes were non-significant and 13 were larger in females; height adjustment attenuated global and subcortical d by 71.3% on average (total volume d −0.42). · Sex differences in overall and subcortical brain…; Sex differences in subregional brain volume…

  2. ShownFemales had thicker cortex in most regions and higher white matter orientation dispersion (OD) in every tract. Males showed higher fractional anisotropy (FA) in most tracts.

    Thickness was significantly higher in females in 47 regions (mean d 0.22, range 0.07 to 0.45); males were thicker only in the right insula. 46/68 regions still favored females after TBV adjustment and 34/68 after height adjustment. Across 22 tracts, FA average d −0.19 (cortico-spinal d −0.54); OD average d 0.30. · Sex differences in subregional brain volume…; Sex differences in white matter microstructure

  3. ShownMales showed greater variance in global and subcortical volumes, in regional volume and surface area, and in most FA tracts. Cortical thickness showed almost no variance difference.

    Average variance ratio (VR) 0.82 for overall and subcortical volumes (VR below 1 means greater male variance). Regional VR was significant in 64/68 regions for volume (0.88 to 0.67) and 66/68 for surface area (0.88 to 0.65). Thickness: only the left medial orbitofrontal differed (VR = 1.19, female greater). FA average VR 0.92, but cortico-spinal VR = 1.17 (p = .0003). · Sex differences in overall and subcortical brain…; Sex differences in subregional brain volume…; Sex differences in white matter microstructure

  4. ShownMore than half of resting-state functional connections differed by sex. Females had stronger connectivity within the default mode network (DMN); males had stronger coupling among sensorimotor, visual and rostral lateral prefrontal regions.

    811 of 1,485 connections (54.7%) differed significantly by sex among 55 independent component analysis (ICA) nodes in 4,004 participants (absolute β 0.071 to 0.447 for females, 0.071 to 0.519 for males). Weighted degree was higher in females for posterior cingulate/precuneus, medial prefrontal and temporo-parietal junction nodes, and higher in males for sensorimotor and visual nodes. · Resting-state fMRI (rsfMRI); Sex differences in resting-state functional…

  5. SuggestedIn cross-sectional mediation models, global volume and surface area accounted for most of the small male advantage in verbal-numerical reasoning but much less of the reaction-time difference. Cortical thickness mediated very little.

    Sex differences were d −0.18 (reasoning) and −0.22 (reaction time). For reasoning, volume and surface area mediation was >82%, versus 7.1% and 5.4% for thickness in the two sample halves. For reaction time, total brain and white matter volume mediated >27%; other measures <15.3%. In multi-region reasoning models, superior temporal volume and surface area mediated 29.1% and 18.4%. · Relation of neurostructural differences to…

Methods, data and reproducibility

Cross-sectional analysis of age- and ethnicity-adjusted MRI phenotypes (FreeSurfer cortical parcels, FSL subcortical volumes, DTI and NODDI tract metrics, resting-state fMRI partial-correlation networks), using Welch t-tests, variance-ratio F-tests and false discovery rate correction, plus brain-size and height adjustment and mediation models of cognition.

  • UK Biobank imaging participants (2,750 female, 2,466 male; age 44-77 years), all scanned on one Siemens Skyra 3T in Manchester · human · n = 5,216 · UK Biobank application 10279
  • FreeSurfer v5.3 Desikan-Killiany parcellations (68 cortical regions: volume, surface area, thickness) from first-release T1 scans, visually quality-checked · human · n = 3,875
  • FSL-derived overall (total, grey, white matter) and FIRST subcortical volumes supplied as UK Biobank imaging-derived phenotypes · human
  • Diffusion MRI tract-average FA and NODDI orientation dispersion for 22 of 27 AutoPtx white matter tracts · human
  • Resting-state fMRI partial-correlation network matrices over 55 group-ICA nodes (100-component parcellation built on 4,162 participants) · human · n = 4,004 with age and ethnicity data
  • Cognitive tests at the imaging visit: verbal-numerical reasoning (13 items) and reaction time (4 scored trials) · human
Methods & tools
FSL (FNIRT, FAST, FIRST, MELODIC, MCFLIRT), FreeSurfer v5.3 with Desikan-Killiany atlas, SurfStat MATLAB toolbox (vertex-wise regression), BEDPOSTx and PROBTRACKx tractography, AutoPtx FSL plugin, AMICO (NODDI modelling), FSLNets (ridge-regularised partial correlation), Liewald-Cox Heatmapper, R p.adjust (FDR) and glm, psych r.test (Fisher r-to-z), BayesFactor ttestBF, LASSO regression with structural equation mediation models
Code
https://github.com/daducci/AMICOhttps://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLNetshttp://www.math.mcgill.ca/keith/surfstathttp://surfer.nmr.mgh.harvard.edu/http://www.fmrib.ox.ac.uk/fsl
Data & accessions
http://www.ukbiobank.ac.uk/http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=2367http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=1977http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=9028http://www.fmrib.ox.ac.uk/datasets/ukbiobank/index.htmlhttps://biobank.ctsu.ox.ac.uk/crystal/docs/brain_mri.pdf
Software & environment
FreeSurfer v5.3; SurfStat on MATLAB R2014a; R with glm, p.adjust, psych (r.test) and BayesFactor (ttestBF); UK Biobank preprocessing as described in Brain Imaging Document version 1.3.
Reproducibility gaps
No link to the authors' own analysis code; data were accessed under UK Biobank application 10279; adjustment methods, mediation equations, LASSO details and cognitive-test descriptions are only in Supplemental Materials not included here.
Limitations (6)
  • UK Biobank covered only ages of roughly 45 to 75 and had a 5.47% response rate, over-representing degree holders, so the sample is unlikely to fully represent UK males and females. · stated by the authors
  • Many female participants may have been going through or past menopause (possibly with hormone replacement therapy), and sex hormones, developmental social experience and nonlinear age effects were not modelled. · stated by the authors
  • Cognitive tests were psychometrically weak (13-item reasoning test, 4 scored reaction-time trials), and the three-variable cross-sectional mediation models cannot establish causality, so the cognitive findings are preliminary. · stated by the authors
  • Low rates of psychiatric disorders in UK Biobank prevented linking brain sex differences to clinical diagnoses, and many other imaging measures (e.g., radial diffusivity, gyrification, white matter hyperintensities) were not examined. · stated by the authors
  • Latent general FA and OD factors lacked measurement invariance across sexes, which weakens interpretation of mediation through these factors and of between-sex comparisons at the general-factor level. · noted in this summary
  • The resting-state sex model adjusted only for age and ethnicity; no head-motion or brain-size covariates are reported, so differences in motion or head size could contribute to the connectivity contrasts. · noted in this summary

Who should read:Read this if you need large-sample normative effect sizes (means and variances) for adult brain sex differences across structural, diffusion and resting-state MRI, or must decide how to handle brain-size adjustment.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISLuo '26Wierenga '20Huang '26Forde '20Gardner '26Yang '23
77% semantic matchNeuroscience
Audrey C. Luo, Meike D. Hettwer, Aaron F. Alexander-Bloch+22 more·bioRxiv·Nov 6, 2026