Autism regression typically happens around 20 months of age, based on a meta-analysis of 28 studies. The broader window spans the second and third years of life, though cases have been documented anywhere from 12 months to nearly 7 years old. About one in five children with autism experiences regression, meaning they develop normally for a period and then lose skills they had already gained.
The Typical Age Window
Most children who experience autism regression do so between their first and third birthdays, with the average onset clustering around 20 months. This timing is significant because it falls right in the middle of a period when toddlers are typically building vocabulary rapidly and becoming more socially engaged. Parents often describe a child who was saying words, making eye contact, and responding to their name, then gradually or suddenly stopping.
The range is wider than many people expect. While the second year of life is the most common period, regression has been documented as late as 81 months (nearly 7 years). These later cases tend to look different and can overlap with a historically separate diagnosis called childhood disintegrative disorder, which involves a longer period of normal development followed by a more severe and widespread loss of skills across multiple areas, not just language and social behavior.
What Skill Loss Looks Like
Language is the most commonly lost skill. A child who had a handful of words or short phrases may stop using them entirely. But regression rarely affects language alone. Parents frequently notice that their child stops responding to their name, loses interest in social games like peek-a-boo, makes less eye contact, and becomes less engaged with people around them. Some children also lose motor skills or self-care abilities they had started to develop, though this is less common than social and communication losses.
The speed of regression varies. Some parents describe it as happening almost overnight, while others notice a slow fade over weeks or months. In the gradual cases, it can be harder to pinpoint exactly when the change began, which is one reason parents sometimes disagree with clinicians about whether regression actually occurred. Video evidence from home recordings has helped researchers confirm that many parent reports of regression are accurate, even when the timeline is fuzzy.
How Common Regression Is
The first large population-based analysis of regression in autism found that at least 20 percent of children with autism experience a measurable loss of skills. Out of the children studied, 21 percent had clear documentation of skill loss. This number may actually be an undercount, since regression that happens slowly can go unnoticed or unreported, and some children may have had subtle developmental differences before the regression that weren’t recognized at the time.
This means regression is a significant minority pattern, not the norm. The majority of children with autism show early signs from infancy that build gradually rather than appearing after a period of typical development. Still, one in five is common enough that any parent noticing a loss of previously acquired skills in a toddler should take it seriously.
What’s Happening in the Brain
The biological mechanisms behind regression aren’t fully understood, but several lines of evidence point to disruptions in how the brain manages its own connections. During normal development, the brain’s immune cells actively prune unnecessary connections between neurons to strengthen the ones that matter most. In autism, this pruning process appears to go wrong. Animal studies have shown that a temporary decrease in these immune cells leads to pruning defects, which are strongly associated with social difficulties.
Inflammation in the brain also plays a role. When the developing brain encounters certain triggers, its immune cells can become overactivated, releasing inflammatory signals that amplify tissue damage rather than resolving it. This inflammatory state may contribute to the loss of skills during regression. Abnormal energy production within cells (specifically in the mitochondria, the structures that power each cell) is one of the most common metabolic issues found in people with autism, and it may make the brain more vulnerable to these inflammatory cascades.
Changes in how genes are read and expressed, without altering the DNA itself, are another area of interest. Children with autism show differences in a chemical process called methylation that affects gene activity. These differences appear to be concentrated in regions that control brain immune cell development and inflammation, suggesting a connection between genetic regulation, immune function, and the neurological changes that underlie regression.
Regression vs. Childhood Disintegrative Disorder
Before 2013, a separate diagnosis called childhood disintegrative disorder (CDD) existed for children who developed normally for at least two years and then lost skills dramatically. CDD was folded into the broader autism spectrum diagnosis in the DSM-5, but the pattern it described is genuinely different from typical autism regression. CDD begins later, follows a longer stretch of completely normal development, involves more severe and widespread skill loss, and historically carries a worse prognosis. Seizures are more common in CDD than in typical autism regression, though a clear underlying medical cause is rarely found. Children with CDD who don’t develop seizures or show seizure-like brain activity tend to have better outcomes than those who do.
Long-Term Outcomes After Regression
One of the most reassuring findings for parents is that regression doesn’t necessarily predict worse long-term outcomes. A population-based birth cohort study comparing 168 autistic adults, some who had experienced regression and some who hadn’t, found no significant differences in IQ at age five, IQ in adulthood, rates of epilepsy, or social functioning scores. The effect sizes were small enough that any real difference between the groups, if it exists at all, is likely modest.
This contradicts some older research. A 1998 study found that children with regression had lower IQ at age six, higher rates of epilepsy, and weaker language development in adulthood. However, even that study found no difference in overall social adaptation in adulthood. The more recent population-based data suggests that regression itself is not a reliable predictor of how a child will do over the long term. Early intervention, the child’s broader developmental profile, and the supports they receive likely matter more than whether skills were lost before they were regained.