Read this topic

Imagine a worried parent in a neurology clinic asking whether a brain scan can finally tell them, for certain, whether their son has autism. The honest answer is no, not today, and not for any developmental disorder covered on this page. Structural and functional brain scans (imaging techniques that make pictures of the brain's anatomy or its activity) have found real, replicated differences between groups of children with autism, ADHD, dyslexia and several other conditions and groups of children developing typically. But every one of those differences is a group-level average built from research cohorts, with so much overlap between individuals that no single scan can sort one child into 'has it' or 'doesn't have it'. This page works through what neuroimaging has actually found in developmental disorders, and why none of it has replaced the clinical interview as the way these conditions are diagnosed.

Neuroimaging in this context means using a scanner, most often magnetic resonance imaging (MRI, which builds detailed pictures of brain structure) or functional MRI (fMRI, which tracks blood flow as a proxy for which brain regions are active during a task), to compare the brains of people with a developmental disorder against people without one. A developmental disorder, for this page, means a condition that begins in childhood and affects how the brain matures: autism spectrum disorder, attention-deficit hyperactivity disorder (ADHD), developmental dyslexia, Tourette syndrome, and a handful of single-gene syndromes such as fragile X syndrome, Williams syndrome and Rett syndrome. Researchers scan groups of affected and unaffected children, average the results, and report where the two groups' brains differ on average, a research finding about a population, not a test result for the one child sitting in the scanner.

Here is the trap almost every candidate falls into: assuming that because a difference is real and replicated, it must be usable to diagnose an individual. It is not, for a simple statistical reason. Every finding on this page shows enormous overlap between the affected group and the typically developing group: many unaffected children have brains that look exactly like the 'affected' average, and many affected children have brains that look exactly like the 'unaffected' average. A finding can be true at the population level and useless for one scan.

In current UK clinical practice, not a single developmental disorder in this page is diagnosed by MRI or fMRI. Diagnosis rests on developmental history, direct behavioural observation, and standardised criteria (ICD-11 or DSM-5), or, for the single-gene syndromes, a genetic test. Clinicians order a scan only when there is a specific reason to rule out a separate structural problem, such as a tumour or a brain malformation, never to make or confirm the developmental diagnosis itself.

If scans cannot diagnose these conditions, it is fair to ask why psychiatrists study them at all. The answer is that population-level imaging differences point toward the underlying biology, which circuits mature unusually, and when, and that biology can eventually guide better treatments even without ever becoming a diagnostic test. Two ideas recur across almost every finding here. First, timing: several disorders are associated with brains that grow or mature along an unusual timeline rather than an unusually shaped one.

Second, circuits rather than single spots: the differences cluster in specific loops rather than the whole brain at once. The cerebellum (the fist-sized structure at the back of the brain that fine-tunes movement, and, it turns out, some cognitive and social processing) and the basal ganglia (a group of structures deep in the brain, including the caudate nucleus, involved in movement and habit control) recur again and again across otherwise unrelated conditions.

Take autism spectrum disorder first. Structural MRI studies find early brain overgrowth: total brain volume tracks roughly 5 to 10 per cent larger than typical during the second to fourth years of life, a difference that narrows again later in childhood. The same literature repeatedly finds a smaller posterior cerebellar vermis (the midline strip of the cerebellum, specifically its lobules VI and VII) and an amygdala (the almond-shaped structure central to processing fear and social salience) that is enlarged in early childhood before normalising or even shrinking later.

Now consider ADHD. A landmark cohort study (Shaw and colleagues, 2007) tracked cortical thickness over time and found that children with ADHD reach their peak cortical thickness around three years later than typically developing children, especially in the prefrontal cortex (the front-most part of the brain responsible for planning and self-control), a delay in the maturation timeline, not a permanently different shape. The same literature finds total brain volume around 3 to 5 per cent smaller on average, plus a smaller caudate nucleus (part of the basal ganglia).

Developmental dyslexia shows a different kind of finding: functional MRI recorded specifically during reading tasks finds reduced activation in the left temporo-parietal cortex and the left occipito-temporal 'visual word form area' (the region that recognises written words as visual patterns), a functional signature of a network working differently, not a structural lesion on a resting scan.

Tourette syndrome is associated with a smaller caudate nucleus and altered activity across the cortico-striato-thalamo-cortical circuit (the loop connecting the cortex, the basal ganglia and the thalamus that normally suppresses unwanted movements).

Three single-gene or exposure-related syndromes round out the picture. Fragile X syndrome shows an enlarged caudate nucleus and a smaller posterior cerebellar vermis, the opposite direction of change from ADHD's own caudate finding. Fetal alcohol spectrum disorder shows a thin or sometimes completely absent corpus callosum (the thick band of fibres connecting the brain's two hemispheres) alongside a smaller cerebellum. Williams syndrome shows relatively reduced parietal-occipital volume against relatively preserved frontal and temporal-limbic volume, a pattern that tracks with its unusual profile of poor visuospatial skills alongside strong sociability and language.

None of the conditions above is confirmed by a scan, and that single fact is the highest-yield point in this whole page. Clinicians diagnose autism spectrum disorder and ADHD clinically: they take a detailed developmental history from parents or carers, observe the child's behaviour directly, and compare what they see against ICD-11 or DSM-5 criteria, often using a structured tool such as the ADOS-2 for autism. Educational psychologists identify developmental dyslexia through standardised reading, spelling and phonological-processing assessment, never a brain scan. Clinicians diagnose Tourette syndrome purely from the history of motor and vocal tics persisting over time.

The single-gene syndromes are the one genuine exception to 'no test confirms it': a specific genetic test confirms fragile X syndrome, Williams syndrome and Rett syndrome, not imaging, and a documented exposure history plus dysmorphology confirms fetal alcohol spectrum disorder. Doctors request a scan in this whole group only when they suspect a separate, treatable structural cause, for example unexplained regression, seizures, or an abnormal neurological examination, which investigates something else entirely, not the developmental disorder itself.

Management follows the clinical diagnosis, never the scan. For ADHD, NICE guidance (NG87) sets out a stepped approach: try environmental and behavioural strategies first, then, for moderate to severe impairment, offer stimulant medication such as methylphenidate as first-line pharmacological treatment. For autism spectrum disorder, NICE guidance (CG142) states plainly that no medication treats the core features; management centres on structured behavioural and educational interventions, communication support, and treating any coexisting condition, such as anxiety or sleep problems, on its own merits. Teachers manage developmental dyslexia with structured, evidence-based literacy teaching, particularly phonics-based programmes, rather than any medical treatment. Clinicians manage Tourette syndrome with psychoeducation and habit-reversal therapy first, reserving antipsychotic medication such as risperidone for tics causing significant distress or impairment. A multidisciplinary team manages the single-gene and exposure syndromes, addressing each syndrome's specific medical and developmental needs under genetics and paediatric services, never guided by an imaging finding.

Whenever a vignette describes a 'characteristic MRI finding' in autism, ADHD, or any other developmental disorder and then asks what confirms the diagnosis, the trap answer is the scan itself; the real answer is always the clinical or behavioural criteria, or the genetic test, for the named syndromes. Before you move on, ask yourself: could you explain to a friend why a real, replicated group difference on a brain scan still cannot tell you whether the one person in front of you has the condition?

Neuroimaging has found real, replicated group-level differences across developmental disorders, timing and circuit differences above all, but none of them is a diagnostic test: every developmental disorder here is diagnosed by clinical criteria, standardised assessment, or genetic testing, never by a scan.