This study confirms the presence of regionally specific sex differences in gray matter thickness over the human lifespan. It shows that the cortical ribbon is actually thicker in some brain regions in females, despite the fact that females tend to have smaller bodies, and smaller brains, including smaller overall gray and white matter volumes than do men. The regions in which cortices are most prominently thicker in females than in males are in the right hemisphere association cortices, particularly the temporal and parietal lobes. In these regions, the cortex is up to 0.45 mm thicker in women; permutation analyses confirmed that these effects were not attributable to chance. The nonsignificant age-by-sex interactions in these regions suggest that the sex differences in temporoparietal cortices are stable across the life span. Rather than scaling the imaging data or statistically controlling for overall brain size as other research groups have done, we paired female subjects individually with male subjects matched on age and brain volume in a subset of 36 of the original 176 subjects studied. Notably, sex effects in this subgroup in the temporal and parietal cortices were even more significant than they were in the larger group, confirming that differences in cortical thickness are not mediated by differences in overall brain size. The findings of increased significance in the matched sample, despite the considerably reduced statistical power, may be consistent with findings in other samples showing increases in regional gray matter volumes that emerge only after controlling for overall differences in brain size between the sexes (Gur and others 1999
; Goldstein and others 2001
The pattern of results from analyses where body height was statistically controlled in adult subjects was similar to the pattern of results from simple correlations between sex and cortical thickness in the entire sample and in the brain size–matched sample. That is, independent of body size differences between males and females, females still have thicker cortex than males in numerous lateral cortical regions. Whereas in no case did group differences extend beyond the lateral surfaces of the temporal, parietal, and frontal lobes, variation in the pattern of results occurred when brain volume and height were used to control brain and body size differences between the sexes. Specifically, when the whole group and brain volume–matched group were evaluated, right hemisphere effects were more prominent than left, but when height was controlled in adult subjects, the pattern of results was bilateral. Thus, while height and brain volume are correlated in this sample (r
= 0.32), and in other samples (r
= 0.55 [Baare and others 2001
]), there are apparent differences in the pattern of results whether brain size correction or body size correction is used. Age differences in the different subsets of subjects used for the simple correlation maps, the brain size–corrected maps, and the height-corrected maps could result in the different pattern of results. However, age-by-sex interactions were not significant in the lateral cortices in either hemisphere where gender effects were observed in the various analyses (as shown in and ). This means that regardless of the age evaluated, gender differences should be comparable, at least in the lateral cortices of the temporal, frontal, and parietal lobes. Inferring actual differences based on qualitative visual comparison of the different maps may be inappropriate anyway given that we did not statistically test the difference between the simple correlation maps and the maps where height or brain volume were controlled.
The regional pattern of increased thickness of cortical gray matter in females is similar to the findings reported in independent young adult samples using similar surface-based methods for studying cortical thickness (Luders and others 2006
; Im and others 2006
). In both of these previous reports, women were observed to have thicker parietal cortices than men. Our results are consistent with prior volumetric findings of increased parietal lobe gray matter in women (Nopoulos and others 2000
; Allen and others 2003
) and with findings of greater gray matter density in women detected with voxel-based analyses (Good and others 2001a
).Frederikse and others (1999)
, in contrast, reported increased volumes in inferior parietal lobes in men, although they did not measure gray matter separately, rendering their results difficult to compare with ours. In addition to the parietal lobe effects, we observed large regions of increased cortical thickness in women in the posterior temporal regions, consistent with prior studies that measured cortical thickness directly (Luders and others 2006
; Im and others 2006
) but not consistent with a prior voxel-based morphometry study (Good and others 2001a
). Age differences in the samples assessed could have led to discrepancies in findings. Our findings of nonsignificant age-by-sex interactions in the posterior temporal and inferior parietal regions where the main effects of sex were most robust for cortical thickness suggest that the sex differences in this region are independent of age in our subjects. Methodological differences between studies using conventional volumetric measures, voxel-based morphometry, and measures of cortical thickness using surface-based methods could also have produced inconsistencies in findings. Specifically, surface-based methods are likely to improve anatomical correspondence between subjects and, thus, may provide increased sensitivity to detect group differences in regions where anatomical variability can yield poorly matched anatomy from the signal-based averaging of images that is employed in voxel-based morphometry studies. Further, conventional volumetric studies are limited to evaluating sex differences in regions that can be visually identified and anatomically defined, which may not represent the actual boundaries of regions, such as those identified here that have the largest sex differences.
The cellular bases for the thicker cortices in women compared with men cannot be determined using current in vivo imaging technologies. Nevertheless, the pattern of regional differences in thickness across the sexes may be consistent with postmortem findings of increased neuronal density and increased cortical volumes in the posterior temporal cortex of women (Witelson and others 1995
; Harasty and others 1997
). Regional cortical thickening in women may also be consistent with the profile of cognitive differences long observed between the sexes, particularly the female advantage on language tasks that may be attributable to their thicker cortices in posterior perisylvian language regions. Gur and others (1999)
have suggested that more cortical gray matter in women may provide a computational advantage (compared with possible differences in white matter, which would affect speed of information transfer). This hypothesis may also be consistent with our findings, although we should note that thicker cortices were most prominent in the right hemisphere, which is usually nondominant for language.
A thicker cortex, however, may not necessarily be better than a thinner one. Studies of normal development, for example, have consistently shown cortical thinning to occur with age as part of normal brain maturation (Jernigan and others 1991
; Giedd and others 1999
; Sowell and others 2004
). Thinning of frontal and parietal cortices in normally developing children, moreover, is associated with improvements in performance on language tasks (Sowell, Delis, and others 2001
; Sowell and others 2004
). Cortical thinning during childhood and adolescence is thought to derive both from progressive changes in myelination (Yakovlev and Lecours 1967
; Benes and others 1994
) and from regressive changes, such as synaptic pruning (Huttenlocher and de Courten 1987
). Both of these cellular changes are thought to improve computational speed and efficiency as redundant synapses are eliminated and oft-used cortical circuits are insulated with myelin. The parietal cortex subserves visuospatial functions, and thus, thinner cortices in men within temporoparietal regions, if it indeed arises from greater pruning and myelination during development and more efficient computational processing, may contribute to the superior visuospatial skills of men. This hypothesis is testable through longitudinal studies in which detailed cognitive assessments would accompany detailed morphological studies. Differences in the rates of cortical maturation between males and females, and relationships between cortical maturation and changes in cognitive function, could be assessed across the cerebrum.
Most of the analyses in this report were focused on sex effects on the entire sample of 176 subjects collapsed across the age range from 7 to 87 years. The nonsignificant interactions of age with sex in regions where sex differences were large (i.e., posterior temporal and inferior parietal regions) suggest that sex effects were stable throughout the lifespan. Interactions between the quadratic age term and sex were significant in bilateral dorsal frontal and temporal regions, indicating that sex differences in cortical thickness vary depending on the age of the participant and that sex must be considered when evaluating age effects in these regions. Furthermore, these age-by-sex interactions could account for discrepancies in reported sex differences in these regions depending on the age composition of the groups in the study. Our findings suggest that the effects of aging may be more prominent in males in dorsal frontal and temporal cortices. However, the combined linear and quadratic interactions were complex and confirm only that sex differences in these regions are not stable over the wide range of ages studied. Our results may be consistent with those of others who have failed to detect age-by-sex interactions in adult populations (Raz and others 1997
; Lemaitre and others 2005
), although to our knowledge no other reports have evaluated age-by-sex interactions from childhood through old age.
As can be seen in , the quadratic regression line curves slightly upward in the oldest subjects. Gray matter is unlikely to actually increase after the eighth decade, rather, the oldest subjects studied were likely not representative of all individuals of that age. In other words, only individuals with above-average cognitive and physical functions were likely capable of participating in the study, and more frail and less cognitively adept subjects did not volunteer. This is probably the case for most studies of normal aging that require on-site participation. Unfortunately, we do not have the data to test this explanation in the current sample. Future studies should include detailed measures of cognitive functioning and health status to help determine whether the most elderly subjects actually differ from their younger counterparts.
We also mapped for the first time differences in brain size between males and females. Permutation analyses showed that male brains are significantly larger than female brains in every region assessed, and our maps showed that the differences are more pronounced at the frontal and occipital poles (up to approximately 6 mm increased DFC-H in males) and less pronounced on the lateral surfaces of the brain at the extremes of the temporal and parietal lobes (up to approximately 2 mm increased DFC-H in males). The possible cognitive correlates of these local size differences may be less transparent than those described for cortical thickness, given that cognitive differences observed between the sexes are not functions typically associated with anterior frontal and occipital lobes.
The regions where the differences in size were smallest are the same locations where sex differences in cortical thickness were most pronounced. Occasionally in imaging studies, artifacts are observed where MRI signal intensity suffers falloff in brain regions that are furthest from the center of the field of view. This signal falloff can affect segmentation of gray and white matters and decrease volumes of measured gray matter in the cortex; therefore, it could perhaps also artificially decrease cortical thickness in these regions in men, whose lateral temporal lobes extend further from the center of the field of view than do the corresponding cortices in women. The regions where male brains are largest, however, are not the regions where sex differences in cortical thickness were statistically significant. More importantly, we conducted sex analyses for cortical thickness in a subgroup of subjects who were carefully matched, male-for-female, on total brain volume and age. The pattern of sex effects on cortical thickness in these analyses agreed with the pattern found in the larger group of subjects, where differences in brain size between sexes were a potential confound. Thus, we are confident that the observed differences in localized cortical thickness are independent of localized differences in brain size.