Autism and the environment: what studies can and cannot show
Autism and environmental exposures: what epidemiology can and cannot establish, why association is not causation, and how genes and environment interact.
By Manouchehr Hessabi, MD, MPH
Few topics in public health are misframed as often as autism and the environment. The public conversation tends to swing between two confident extremes: that autism is entirely genetic and the environment plays no part, or that some single exposure causes it. Neither claim is what the evidence supports, and the distance between them is where the actual science lives.
This is an explainer about what research on autism and environmental exposures can and cannot establish. It is educational and not a substitute for personal medical advice, and it does not recommend any test, product, or course of action. The goal here is narrower and more useful than a verdict: to show how an epidemiologist reads this literature, so that the next headline is easier to judge.
The question, framed correctly
Most arguments about autism collapse two very different questions into one. The first is what raises the chance of autism across a population. The second is what caused a particular child's autism. These are not the same question, and they do not have the same kind of answer.
Epidemiology, the study of how often conditions occur and what they track with, mostly answers the first. It works at the level of groups, not individuals. A study can report that a factor is more common among children later diagnosed with autism without being able to say that the factor produced any single diagnosis. Keeping those two questions apart is the first discipline the topic demands.
The honest starting position, the one the weight of the evidence supports, is this: autism has a strong genetic component, and researchers study environmental exposures as possible contributors that interact with genes, rather than as a single proven cause. Everything below is an unpacking of that sentence.
What the prevalence numbers do and do not mean
The numbers people cite most are prevalence figures, and they are routinely misread. In April 2025 the Centers for Disease Control and Prevention reported, through its Autism and Developmental Disabilities Monitoring (ADDM) Network, that about 1 in 31 eight-year-old children were identified with autism spectrum disorder for the 2022 surveillance year, up from 1 in 36 for 2020 (CDC MMWR, 2025).
A rising count is not automatically a rising rate of the underlying condition. The CDC report itself attributes much of the change to broader diagnostic criteria, improved and earlier identification, and reduced disparities in who gets evaluated, rather than solely to a true increase in how often autism occurs.
A small piece of vocabulary makes this clearer. Prevalence is how many people have a condition at a given time; incidence is how many new cases arise over a period. When the way a condition is defined and detected changes, prevalence can climb even if the real rate of new cases is stable. More children identified can mean better identification, a genuinely higher rate, or both at once. The numbers alone cannot separate those, which is exactly why careful reports name their own limits.
Genes first, then environment as a modifier
Any honest account of autism risk starts with genetics. Twin and large family studies have long pointed to a substantial genetic contribution to autism liability, meaning much of the variation in who is affected tracks with inherited factors. The precise figures vary by study and method, so the responsible summary is "substantial," not a single headline percentage.
Environmental research does not contradict that. It asks a more specific question: whether particular exposures shift risk on top of genetic susceptibility. This is the idea of gene-environment interaction, the recognition that genes and environment are not rival explanations but can act together, with an exposure mattering more in a child who is already genetically predisposed.
That framing is visible in how the field is organized. The National Institute of Environmental Health Sciences (NIEHS) supports a research portfolio on autism built around the exposome, the full set of environmental exposures a person encounters across early life, studied alongside genetic factors (NIEHS). The exposome concept exists precisely because no single exposure is treated as the answer.
What "associated with" actually means
The most important phrase in this entire literature is "associated with," and it is the one most often lost in translation. When a study reports that an exposure is associated with higher odds of autism, it means the two were observed together more than expected by chance in that sample. It does not mean the exposure caused the outcome.
The distinction between association and causation is not pedantic. Consider a simple example: families living near heavy traffic may differ from other families in income, access to care, age, and many other ways, several of which could independently relate to a diagnosis. An association flags something worth investigating. It is a signal, not a verdict.
The exposures researchers most commonly probe are familiar from the headlines: air pollution, certain prenatal exposures, advanced parental age, and birth complications. Each has been studied, and air pollution in particular continues to draw careful peer-reviewed attention, including work examining multiple pollutants together in large California populations (PMC, 2025). None of these, individually, is a settled single cause. They are threads in an active investigation, and treating any one of them as proven goes beyond what the studies claim.
Why these studies are genuinely hard
It helps to understand why the science is slow rather than to read the slowness as evasion. Three difficulties recur.
The first is confounding. Families differ along many dimensions that also track with exposure, and untangling the exposure of interest from everything that travels with it is hard. A study that does not account for a relevant difference can mistake that difference for the exposure's effect.
The second is measurement. Much exposure data, especially before birth, is reconstructed after the fact, which invites recall bias, the tendency for memory of past exposures to differ between families whose children were and were not diagnosed. Exposure that is mismeasured weakens or distorts any association built on it.
The third is time. The gap between a prenatal or early-life exposure and a later diagnosis can be years, and a great deal happens in between. Long latency makes clean attribution difficult even when an effect is real.
These are the reasons the field is moving toward the exposome and toward large, coordinated data efforts. In 2025 the National Institutes of Health launched a multi-study Autism Data Science Initiative to examine how genetic and environmental influences, including air pollution, interact in autism development (NIH, via Drexel University, 2025). Studying many exposures and genes together, rather than one at a time, is a direct response to how genuinely hard the one-at-a-time approach has proven to be.
How to read the next autism headline
The practical payoff of all this is a short checklist a careful reader can apply to almost any new claim.
- Is this a single study or a body of evidence? One study is a data point, not a conclusion.
- Does it report an association or a cause? The words used are usually a reliable guide.
- What population was studied, and does it resemble the group the headline generalizes to?
- What did the researchers control for? Strong studies name their confounders.
- Has the finding been replicated? Durable results tend to recur.
Run a headline through those questions and most overclaims fall apart on their own.
The responsible reading of the current science is that autism is complex and multifactorial. Genetics carry substantial weight, environmental exposures are studied as potential modifiers that interact with that genetic background, and the prevalence figures reflect a mix of real and diagnostic factors that no single number can disentangle. Good science holds that uncertainty honestly rather than resolving it prematurely in either direction.
For readers who want to go deeper into how environmental factors are studied in child neurodevelopment, the research on autism and the environment lays out the methods in more detail, and the underlying peer-reviewed work is collected in the publications. Nothing in this article is medical advice; questions about an individual child belong with that child's clinician.