When reviewing the structural and volumetric magnetic resonance imaging (MRI) data for ASD, the variability and lack of consensus in the findings seems discouraging at first, but some patterns are emerging. The most consistent finding is that of accelerated brain volume growth in early childhood, reported to be about a 10% increase in brain volume, which seems to peak around 2-4 years of age. This early overgrowth is probably followed by a plateau, although the latter has not been confirmed by large longitudinal studies [2–10]. The enlargement seems to occur in both grey and white matter, with some, but not all, studies suggesting that early in childhood there is a disproportionate contribution by white matter to this volume increase [11, 12].
A crucial corollary to this question addressed in the literature deals with whether there are regional specificities to the abnormalities in brain volume. However, there is little consistency between studies. For example, abnormalities in the volume of the cerebellum have been reported since 1988. Courchesne et al. first reported decreased volumes of vermian lobules VI and VII in the majority of their group of patients with autism. Although subsequent studies were not all able to replicate those findings, a recent meta-analysis by Stanfield et al. in 2007  confirmed such a reduction. Similarly, abnormalities in the volume of the amygdalae have been reported in some, but not all studies [15–17]. Meta-analyses by Amaral et al. and Stanfield et al. suggest that age is an important factor, and that such enlargement is present only in young children with ASD. This is a crucial issue, particularly in light of the most consistent neuroanatomical difference- (the increased brain volume in young children) and is a strong argument against ignoring the age of the ASD participants in any of these analyses.
There is more consistency in volumetric studies of the corpus callosum, with multiple meta-analyses reporting decreased volume in ASD, suggesting decreased interhemispheric connectivity in this population [18, 14]. Analyses of the basal ganglia have suggested increased volume of the caudate in autism [19, 20], which has been correlated with severity of repetitive behaviors. The linking of the neuroanatomical findings to behavioral symptoms is vital to understanding the role of the structural changes in the aetiology of ASD. Investigations of structural measures of the cingulate have reported decreased size associated with decreased metabolic activity in ASD . Volumetric abnormalities have also been reported, particularly in the frontal lobes  and also the temporal [23, 24] and parietal lobes , thalamus , and brainstem .
The spread and heterogeneity of these findings suggest that autism is a widely distributed disorder affecting both grey and white matter. Newer technologies allow for the more sophisticated quantification of the structure of the grey and white matter structure. Specifically, cortical grey matter can now be described in terms of cortical thickness and surface area, the product of which produces an estimate for cortical grey-matter volume. These two measures are of particular interest. as they are hypothesized to be related to distinct neurobiological processes. Cortical thickness seems to reflect dendritic arborisation and pruning within grey matter , or changes in myelination at the interface of white and grey matter , whereas surface area varies with the degree of cortical folding or gyrification, and is thought to depend on division of progenitor cells in the periventricular area during embryogenesis . Investigation of differences in these measures of cortical grey matter may provide important indications of very early neuroanatomical developmental events in the ASD population. In addition, as cortical thickness has been shown to vary during childhood, with changes occurring regionally in a developmental progression [31, 32], detailed measures of cortical thickness can provide crucial information on cortical maturation and be an important index of an altered developmental trajectory in the brains of children with ASD.
Complementary studies of diffusion tensor imaging (DTI) are providing increasing insights into the structure of white matter. DTI measures water diffusion within a tissue. The tightly organized white-matter tracts restrict the diffusion of water, producing anisotropic diffusion, whereas water diffusion in the grey matter tends to be less restricted [33, 34]. In DTI, the shape of diffusion is represented by an ellipsoid with three eigenvectors describing the directions of the radii, and three eigenvalues describing their length. Diffusion along the primary eigenvector represents longitudinal diffusivity, and is thought to be related to axonal integrity, a complex construct that may include accumulation of cellular debris, disordered microtubule arrangement, aggregation of microfilaments, cellular swelling and decreased axonal transport . By contrast, diffusion across the two other small directions is related to radial diffusivity, and is thought to be a marker of myelin integrity, although the sensitivity of radial diffusivity to detect myelin disruption may be decreased in the case of co-occurring axonal injury [36–38]. Mean diffusivity refers to the average diffusivity along all axes, whereas the shape of the ellipsoid is represented by the calculation of fractional anisotropy (FA), a metric related to differences between the three eigenvalues. Thus, several measures can be extracted from DTI neuroimaging studies, which will contribute to a fuller understanding of differences in white-matter development in ASD.
Cortical grey-matter studies
Hadjikhani et al. were the first to examine cortical thickness in adults with ASD using in vivo MRI and in a regionally specific manner, with a vertex-by-vertex approach across the whole brain. A number of regions were found to have thinner cortex in adults with ASD relative to controls, including regions in areas known to be important for social cognition and the mirror neuron system. By contrast, Hyde et al. found increases in cortical thickness in an autism group relative to matched controls in regions from all four lobes, in areas important for the key areas of impairment seen in autism. Increased cortical thickness was also seen in primary sensory areas, and a small number of regions had thinner cortex in the autism group.
In studies including children only, Hardan et al. found significantly thicker cortex in the whole brain overall for boys with autism, with similar findings in the temporal and parietal lobes. In a follow-up study , the investigators reported that children with ASD had significant decreases in total grey matter with age relative to the control children, as well as decreases in cortical thickness with age. However, only the difference in occipital cortical thickness remained significant after adjusting for multiple comparisons, and group differences disappeared after using IQ as a covariate. By contrast, with a sample ranging in age from 10 to 65 years, Raznahan et al. found that in regions showing an age-by-group interaction, there was no relation between age and cortical thickness in the ASD group, whereas there were decreases in cortical thickness with age in the controls. In addition, cortical thickness in typically developing children was increased relative to children with ASD at younger ages, but decreased relative to the ASD group at older ages.
Considerably fewer studies have investigated brain sulcation patterns, although atypicalities in sulcal position and depth have been reported in children and adolescents with ASD, with the most marked effects seen in the younger children [43, 44]. Increased gyrification index in left frontal lobe in children but not adults, and decreased cortical folding with age in ASD relative to controls, were reported by Hardan et al.. However, a recent study of surface area reported no age-by-group interactions for this measure .
Thus, most studies of cortical thickness in ASD have reported increases in values relative to typical controls, although exceptions have also been reported, with trends possibly suggesting that such differences are more pronounced in younger children, and that age-by-group interactions may be evident. In addition, there has been considerable variability in the techniques used. Advances in automated cortical surface analyses [46–48] allow fine-grained examination of regional and sub-regional differences in cortical thickness, and investigation of developmental changes in more detail. Application of these techniques may yield greater consistency in findings for comparisons of children and adults with and without ASD. Lastly, sample heterogeneity may have accounted for the heterogeneity of results as not all cohorts had been assessed with standardized diagnostic instruments (Autism Diagnostic Observation Schedule (ADOS) and Autism Diagnostic Interview, Revised (ADI-R) consistently, and some did not match for IQ, although recent findings did find an association between IQ and cortical thickness [49, 50]. Studies of cortical surface are still in their infancy, but given their potential to describe differential neurodevelopmental events, hold great promise.
There have been several studies examining FA in white-matter tracts in children and adults with ASD. Lower FA in a group comparison has typically been interpreted to reflect decreased organization and coherence within fibre tracts. Although the locations have varied (orbitofrontal, medial prefrontal, temporal lobe, corpus callosum, cingulate cortex, arcuate fasciculus, ILF, uncinate fasciculus, cerebellar outflow tracts, internal capsule), and different techniques have been used, including tractography and voxel-based techniques, most studies have found evidence of reduced FA in various regions in children and adults with ASD compared with control groups [51–66]. Consistent with the variable findings in other structural brain measures in this field, some recent studies have also found evidence of regions of increased FA in ASD, in samples of young children  and adolescents . However, because FA represents the relative magnitudes of parallel and perpendicular diffusion (shape of ellipsoid), investigation of the individual eigenvalues are needed to better interpret any differences in DTI that may exist between ASD and control groups. Recent studies have begun to report the sources of these differences in radial, axial and mean diffusivity. Widespread increases in mean diffusivity have been described [69–71], suggesting a poorly organized white matter, although little consensus exists on the relative contribution of axial versus radial diffusivity in this population. In fact, abnormalities in radial diffusivity have been reported [59, 60, 72], implicating the myelin component of white matter in ASD pathophysiology. Despite the variability in the brain areas in which differences are reported in DTI measures between individuals with and without ASD, many groups have found evidence of atypical white-matter measures, lending credence to these findings [51–71]. However, more detailed studies analysing the more specific measures from DTI, such as radial and axial diffusivity, carefully controlling for age, may allow a more reliable picture to emerge that will further help our understanding of the aetiology of these abnormalities and the link between these changes in the brain structure and associated behavioral profiles in the ASD population.
In summary, structural studies of grey and white matter suggest an abnormal developmental trajectory of brain growth, with evidence of poorly organized white matter, increased cortical thickness and atypicalities in gyration patterns, possibly implicating abnormalities in neuronal migration, cortical organization and myelination in ASD. The abundance of inconsistent findings in the published literature on autism might reflect differences between study populations, such as age, level of impairment, and presence of medical and behavioral comorbidities in the selected groups.