Autism doesn’t have a single cause. It develops from a combination of genetic predispositions, prenatal conditions, and differences in how the brain wires itself during early development. Current estimates from the CDC put autism prevalence at about 1 in 31 children, and the scientific picture of what drives it has become significantly clearer over the past two decades. The short answer: genes do the heavy lifting, but the prenatal environment and early brain development shape the outcome.
Genetics Play the Largest Role
Autism is one of the most heritable neurodevelopmental conditions known, with heritability estimates ranging from 70% to 90%. That means the vast majority of what determines whether someone is autistic comes down to their DNA. But “genetic” doesn’t necessarily mean “inherited from a parent.” The genetic landscape of autism is surprisingly complex.
Common genetic variants, the kind shared widely across the population, account for roughly 49% of autism liability. These aren’t rare mutations. They’re ordinary genetic differences that, when enough of them stack up in the right combination, shift brain development in a direction associated with autism. No single common variant has a large effect on its own. It’s the accumulated weight of many small contributions.
Then there are rarer genetic events. De novo mutations, meaning new genetic changes that appear in a child but aren’t present in either parent, contribute an estimated 16% to 19% of autism liability. These include spontaneous deletions or duplications of DNA segments and point mutations in single genes. Rare inherited variants and X-linked variants add a few more percentage points each. Researchers have identified risk-associated DNA variations spread across all 23 pairs of human chromosomes, involving hundreds of different gene locations out of the roughly 20,000 genes in the human genome.
What all these genetic factors tend to have in common is their effect on brain development: how neurons migrate to the right locations, how synapses form, how brain circuits get refined. The specific genes involved often govern the basic construction and maintenance of neural connections.
How the Autistic Brain Develops Differently
One of the clearest physical differences in autism involves synaptic pruning, the process by which the brain trims back excess connections during childhood to make its circuits more efficient. In typical development, the brain massively overproduces synapses in the first few years, then prunes roughly half of them away. In autism, this pruning process appears to be reduced, leaving an overabundance of synaptic connections. Researchers have described this as a “deficit of pruning” rather than a problem of too many connections being built in the first place.
This difference shows up early. MRI studies of infants who were later diagnosed with autism reveal that the amygdala, a brain region involved in processing social and emotional information, grows too rapidly between 6 and 12 months of age. This overgrowth happens before the behavioral characteristics of autism fully emerge. The faster the amygdala grew during infancy, the more social difficulties the child showed at age two. Notably, these infants showed no measurable cognitive differences at 6 months, but experienced a gradual decline in cognitive scores between 6 and 24 months.
Structural differences also appear in the cerebellum, which plays a role in motor coordination, sensory processing, and some aspects of social cognition. Studies consistently find reduced volume of cerebellar gray matter and loss of a specific type of neuron called Purkinje cells in autistic individuals. The way sensory-related brain structures communicate with the cerebellum also differs, which helps explain the sensory sensitivities many autistic people experience.
What Happens Before Birth Matters
The prenatal environment can influence whether someone with a genetic predisposition toward autism actually develops it. One well-studied pathway involves the maternal immune system. When a pregnant person experiences significant inflammation from infection or other immune activation, their immune signaling molecules can cross the placenta or trigger the placenta itself to produce inflammatory signals. These molecules normally help guide how neurons migrate and how early brain circuits get wired. When their levels are disrupted, the fetal brain’s development can be altered in ways that affect synaptic formation, connectivity, and neurotransmitter systems.
This doesn’t mean that getting a cold during pregnancy causes autism. The research points to sustained or intense immune activation, not routine minor illness. The mechanism involves a disruption of the balance between pro-inflammatory and anti-inflammatory signaling at a critical window of brain development.
Certain medications taken during pregnancy also carry measurable risk. Prenatal exposure to anti-seizure medications roughly doubles the overall risk of autism. Valproate, a medication historically used for epilepsy and mood disorders, carries the highest risk among these, nearly tripling it. Other anti-seizure medications like carbamazepine and oxcarbazepine show smaller but still statistically significant increases. These drugs affect how genes are expressed during fetal development by altering chemical tags on DNA-packaging proteins, which can disrupt normal brain formation during a critical window.
Parental Age and Other Risk Factors
Both maternal and paternal age at conception independently affect autism risk. Mothers aged 35 or older have about a 30% higher likelihood of having an autistic child compared to mothers aged 25 to 29, after adjusting for other factors. Fathers aged 40 or older carry about a 40% increased risk compared to fathers in the 25 to 29 range. The combination compounds: firstborn children of two older parents are about three times more likely to be autistic than third-or-later-born children of younger parents.
The likely explanation involves de novo mutations. Sperm cells in particular accumulate new genetic mutations with each passing year, and older eggs may also carry more epigenetic changes. This fits neatly with the broader genetic picture, where spontaneous mutations play a meaningful role in autism liability.
Epigenetics: Where Genes and Environment Meet
Epigenetics refers to changes in how genes are read and expressed without altering the DNA sequence itself. Think of it as the difference between having a book and having certain pages bookmarked or highlighted. The text doesn’t change, but which parts get read does. In autism, both genetic and environmental factors can shift these epigenetic patterns in ways that affect brain development.
Two main types of epigenetic change are relevant. DNA methylation involves chemical tags that silence or activate genes. Histone modification changes how tightly DNA is wound around its packaging proteins, making genes more or less accessible. Both too much and too little of either modification can disrupt the same neurodevelopmental pathways, following what researchers describe as a U-shaped pattern. Environmental exposures like certain industrial chemicals (bisphenol A, lead) can alter DNA methylation, while medications like valproate cause excessive histone modification. The result in either direction is the same: disrupted gene expression during critical periods of brain formation, potentially affecting synaptic development, neural connectivity, and neurotransmitter balance.
Why No Two Autistic People Are Alike
The sheer number of contributing factors explains the enormous range of autistic experiences. Hundreds of gene locations are involved, each contributing a small effect. The prenatal environment adds its own layer of variability. The timing of any disruption to brain development matters enormously, since a few weeks’ difference can determine which brain circuits are affected. Two autistic people might share the diagnosis but have almost entirely different genetic profiles, different patterns of brain connectivity, and very different strengths and challenges.
This is also why autism runs in families without following a simple inheritance pattern. A parent might carry a modest load of common risk variants and never be diagnosed, then have a child who inherits those variants plus picks up a de novo mutation that tips the balance. Or two parents might each contribute a different set of common variants that, combined in their child, cross a threshold neither parent approached individually. The genetics are genuinely polygenic, meaning many genes each contribute a small piece, rather than one or two genes acting as an on/off switch.