Babies don’t develop autism after birth from something that happens in the outside world. The condition begins taking shape during pregnancy, as the brain forms in the womb. Genetics account for roughly 83 to 87% of autism risk, making it one of the most heritable neurodevelopmental conditions known. The latest CDC data from 2022 puts autism prevalence at about 1 in 31 children, up from 1 in 150 in 2000, largely reflecting broader screening and updated diagnostic criteria.
Autism Starts During Prenatal Brain Development
The clearest physical evidence that autism originates before birth comes from a 2014 study published in the New England Journal of Medicine. Researchers examined brain tissue from children with autism and found disorganized patches in the cortex, the brain’s outer layer. During normal fetal development, neurons in the cortex organize into six distinct layers. In 91% of the autism samples, markers for several of these layers were missing, compared to just 9% of samples from children without autism.
These disorganized patches weren’t spread across the entire brain. They appeared in small, focal areas roughly 5 to 7 millimeters long, spanning multiple cortical layers. Because this layered architecture forms during the second and third trimesters, the disruption clearly traces back to prenatal life. As Thomas Insel, then director of the National Institute of Mental Health, put it: “We can presume this reflects a process occurring long before birth.”
Genetics Carry Most of the Risk
Twin studies consistently show that autism is heavily genetic. When one identical twin has autism, the other twin has it at dramatically higher rates than fraternal twins do, producing heritability estimates around 83 to 87%. That means the vast majority of what determines whether a child develops autism comes down to the genetic instructions they inherit or, in some cases, spontaneous changes in their DNA.
Those spontaneous changes, called de novo mutations, play a surprisingly large role. These are genetic alterations that appear for the first time in a child and aren’t present in either parent’s DNA. Research from large-scale exome sequencing studies estimates that de novo mutations in protein-coding genes contribute to about 30% of autism cases in families with no prior history of the condition. For girls diagnosed with autism, de novo coding mutations account for roughly 45% of cases. Gene-disrupting mutations (the type that break or silence a gene entirely) occur at nearly twice the rate in children with autism compared to their unaffected siblings.
In families that already have multiple members with autism, inherited gene variants carried by both parents play a bigger role. In families with no history, a spontaneous mutation is more often the trigger. This helps explain why autism can appear in any family regardless of background.
Prenatal Environment Also Plays a Role
While genetics set the stage, certain conditions during pregnancy can shift the odds. These aren’t causes on their own but rather factors that interact with a child’s genetic vulnerability.
Maternal immune activation is one well-studied pathway. When a pregnant person experiences significant infection or inflammatory illness, the immune response produces signaling molecules that can cross the placenta and reach the developing fetal brain. These signals can interfere with how neural stem cells multiply, how young neurons migrate to their correct positions, and how early brain circuits wire together. The result is altered cortical architecture, the same kind of disorganization seen in postmortem autism brain studies.
Certain medications taken during pregnancy also carry measurable risk. Valproic acid, an anti-seizure medication, is the most clearly documented example. Children exposed to valproate in the womb had an absolute risk of about 4.4% for autism spectrum disorder, and the risk was nearly three times higher than for unexposed children even after accounting for the mother’s epilepsy. Other common anti-seizure medications like lamotrigine and carbamazepine did not show the same increased risk, which points to something specific about valproate’s effects on fetal brain development rather than epilepsy itself.
Parental Age and Premature Birth
Advanced parental age at conception is a consistent risk factor. A meta-analysis found that every 10-year increase in a father’s age raises autism risk by about 21%, and the same increase in a mother’s age raises it by about 18%. The most likely explanation is that older parents accumulate more spontaneous mutations in their reproductive cells over time, increasing the chance of de novo mutations in their children.
Being born prematurely (before 37 weeks) also increases the likelihood. Children born preterm are about 3.3 times more likely to receive an autism diagnosis than children born at full term, with diagnostic assessments finding a pooled prevalence of roughly 6% in preterm populations. Premature birth may compound existing genetic risk by exposing an immature brain to stresses it isn’t yet equipped to handle, including inflammation, oxygen fluctuations, and disrupted growth signaling during a critical window of cortical development.
Folic Acid Lowers the Odds
One of the few modifiable factors with strong evidence is folic acid supplementation. A meta-analysis found that taking folic acid during early pregnancy was associated with 43% lower odds of a child being diagnosed with autism. The protective threshold was at least 400 micrograms daily from food and supplements combined, which aligns with standard prenatal vitamin recommendations. Both folic acid alone and folic acid combined with other vitamins and minerals showed a reduced risk. Early pregnancy appears to be the sensitive window, likely because folic acid supports the rapid cell division and neural tube formation happening in the first trimester.
Why It’s Not One Single Cause
Autism doesn’t follow a simple pattern of one gene or one exposure producing the condition. Instead, it emerges from a convergence of genetic predisposition and prenatal environmental influences acting on the developing brain across multiple stages of pregnancy. Some children carry high-impact de novo mutations that are sufficient on their own. Others inherit a combination of common gene variants from both parents that collectively push past a threshold. Still others have moderate genetic risk that tips over in the presence of prenatal inflammation, certain medication exposures, or prematurity.
This layered causation is why autism runs in some families but appears unexpectedly in others, why it varies so widely in severity, and why no single prenatal test can predict it. What the evidence consistently shows is that by the time a baby is born, the neurological differences that lead to autism are already in place.