Brain structure, scaling, and variability

Start with the measurement, not a male or female brain

Male brains are larger on average in raw global volume, while female cortices are thicker on average in many analyses. Those statements concern different measurements and neither describes every individual. In 5,216 UK Biobank participants, 2,750 female and 2,466 male, mean age 61.7 and range 44 to 77, raw male volumes and surface areas were generally larger and female cortical thickness was generally greater, with substantial overlap (Ritchie et al., 2018). The atlas does not resize its meshes to imitate those group means.

Regional structures do not scale linearly with total brain size. Simply dividing a regional volume by total volume can therefore create or erase an apparent difference. Williams and colleagues modelled nonlinear allometry for 629 measures in 40,028 UK Biobank participants, 21,142 female and 18,886 male. Sex associations occurred in 67 percent of measures, with median absolute standardized beta 0.13, while direction varied by measure, hemisphere, subregion, and segmentation method (Williams et al., 2021). This is the core reason every row names its measure and adjustment model.

Ritchie and Williams both use UK Biobank imaging and are not independent replications. The cohort is a healthier, less socioeconomically deprived volunteer sample than the general United Kingdom population and is concentrated in middle and later adulthood. Neither analysis treats sexual orientation as an explanatory variable. Large samples make small, model-dependent associations statistically stable; they do not make the distributions non-overlapping.

Volume, area, and thickness answer different questions

Cortical volume combines surface area and thickness, but those components have partly distinct developmental and geometric bases. A region can therefore show one direction for volume, another for thickness, or no detectable difference in area. The highlighted mesh names the gross structure only. It does not display the measured thickness, area, volume, cell number, or tissue microstructure.

Lee and colleagues reported reproducible cortical patterns in HCP, N=1,085 (592 female, 493 male, ages 22 to 37), and an independent UK Biobank cohort, N=669 (375 female, 294 male, ages 45 to 50), then compared them with clinical cohorts totalling 313 participants (Lee et al., 2026). As of 20 July 2026, the publisher’s downloadable article still carries an unedited early-access warning. It corroborates the cortical territories here but is not the sole source for a core row, and its claims must be rechecked if the edited record changes.

Variability is not a mean difference

The ENIGMA mega-analysis combined MRI data from 16,683 healthy people ages 1 to 90, 47 percent female. Measure-specific analytic samples were smaller: the highlighted subcortical comparisons included 13,696 people, 7,141 female and 6,555 male. Greater male between-person variance appeared for every analysed subcortical volume, every cortical surface-area measure, and 60 percent of cortical-thickness measures (Wierenga et al., 2022). A variance result says that one distribution is more dispersed. It does not say that every male value is more extreme, that the male mean is larger, or that one person’s anatomy can be classified without error.

Asymmetry is regional, not one global score

Saltoun and colleagues analysed 37,441 UK Biobank participants, 19,904 female and 17,537 male, and found several distributed structural asymmetry patterns rather than one dominant whole-brain lateralisation axis. Sex was one of many phenotypic associations. Their regional results included female leftward shifts in occipital pole, ventral striatum, and thalamus, with female rightward shifts in planum temporale and tapetum (Saltoun et al., 2023). Highlighting a bilateral mesh cannot show the sign or magnitude of an asymmetry index.

The planum temporale makes uncertainty visible. Saltoun’s regional association and an earlier meta-analysis do not give the same simple story. The meta-analysis found no reliable female-male difference in planum-temporale asymmetry, dichotic listening, or functional imaging of language (Sommer et al., 2008). This null does not deny every regional asymmetry association; it prevents a broad claim that one sex has a more lateralized language system.

What structure cannot establish

These are population-average anatomical measurements, not maps of personality, cognition, desire, skill, diagnosis, or social behaviour. In two independent datasets, multivariate sex-related brain and behavioural scores were only weakly associated after accounting for brain size, and causality remained unresolved (van Eijk et al., 2021). A highlighted region therefore marks where anatomy was measured. It does not infer psychological function from structure or classify the person viewing the atlas.

Reviewed status: Owner review was completed by Taig Mac Carthy on 21 July 2026 for publication. Reviewed status records editorial approval; it is not clinical endorsement.