|| Checking for direct PDF access through Ovid
In voxel-based morphometry studies, the larger the sample size, the fewer the number of group differences reported.The literature consists of mostly small-scale studies, whose results are likely to be inflated or false positive.Most published results show little or no replication across independent studies.The only group difference that is robustly replicated is a smaller brain volume in dyslexic individuals (d = 0.4).Higher methodological standards are necessary to discover true neuroanatomical differences.Investigations into the neuroanatomical bases of developmental dyslexia have now spanned more than 40 years, starting with the post-mortem examination of a few individual brains in the 60s and 70s, and exploding in the 90s with the widespread use of MRI. The time is now ripe to reappraise the considerable amount of data gathered with MRI using different types of sequences (T1, diffusion, spectroscopy) and analysed using different methods (manual, voxel-based or surface-based morphometry, fractional anisotropy and tractography, multivariate analyses…). While selective reviews of mostly small-scale studies seem to provide a coherent view of the brain disruptions that are typical of dyslexia, involving left perisylvian and occipito-temporal regions, we argue that this view may be deceptive and that meta-analyses and large-scale studies rather highlight many inconsistencies and limitations. We discuss problems inherent to small sample size as well as methodological difficulties that still undermine the discovery of reliable neuroanatomical bases of dyslexia, and we outline some recommendations to further improve this research area.