Living near a busy road is associated with a dramatic increase in the risk of childhood autism, a study has shown.
Early exposure to traffic pollution, either in the womb or during the first year of life, more than doubled a child's chances of having the disorder, scientists found.
Children from homes with the highest air pollution levels were three times more at risk than those from the least exposed homes.
Experts described the finding as "important" but stressed it did not prove a causal link between pollutant chemicals and impaired brain development.
Autism, a wide-ranging condition that affects communication and social skills, is believed to be caused by a combination of genetic and environmental factors.
While a number of genetic variants are known to be linked to the disorder, the role played by the environment has been less clear.
Scientists in California set out to investigate a possible link between autism rates and traffic pollution.
The Childhood Autism Risks from Genetics and the Environment (Charge) study looked at data on 245 children without the condition and 279 affected by autism.
Air pollution records from the US Environmental Protection Agency were used to estimate exposure to nitrogen dioxide and small sooty particles, both produced from motor vehicle exhausts.
The researchers took into account how far mothers and babies lived from busy roads, and levels of pollutants in the air.
Lead scientist Dr Heather Volk, from the Keck School of Medicine at the University of Southern California, said: "This work has broad potential public health implications. We've known for a long time that air pollution is bad for lungs, and especially for children. We're now beginning to understand how air pollution may affect the brain."
The findings are published today in the latest online edition of the journal Archives of General Psychiatry.
Dr Volk said her team assessed a range of factors, including how far people lived from roads, traffic levels, meteorological factors such as wind direction, and information from air quality monitors.
She was especially concerned about exposure to small and very fine pollution particles produced by diesel engines known as PM10s and PM2.5s.
Previous studies have linked inhalation of the particles to heart and lung disease, cancer, and premature death.
"From studies conducted in the lab, we know that we can breathe in tiny particles and they can produce inflammation," said Dr Volk.
"Particles have varied composition, and there are many chemicals that can bind to them. The components of these particles could be hazardous to the brain."
In their paper the authors concluded: "Although additional research to replicate these findings is needed, the public health implications of these findings are large because air pollution exposure is common and may have lasting neurological effects."
However, British experts said the findings should be interpreted with caution.
Professor Emily Simonoff, from the Institute of Psychiatry at King's College London, said the way the research was conducted meant the possibility of bias could not be excluded. A number of factors had not been properly taken into account, including father's age and family history of autism.
"This is potentially an important finding and it is therefore essential to consider the strengths and limitations of the study," said Prof Simonoff.
"At present, pregnant women should continue to look after their health during pregnancy but should not be unduly concerned."
Sophia Xiang Sun, from the Autism Research Centre at Cambridge University, said: "Although traffic-related air pollution might be one of the contributing factors to the development of autism, other factors cannot be ruled out. These factors include second-hand smoking during pregnancy, medical conditions related to pregnancy, indoor air pollution, especially if the family has a history of mental disorders as autism is highly genetic.
"Further research is needed to investigate the potential association between traffic-related air pollution and autism, ideally a prospective study that monitors traffic-related air pollution with the control of indoor air pollution and smoking. Until further research is carried out we will not know definitely if the association is there and, if it is there, how direct and to what degree."
Professor Uta Frith, from University College London, said: "It seems to me very unlikely that the association is causal, rather than correlational.
"Rather than taking the results at face value, I would like to know what it implies to live near a highway. It could imply all sorts of disadvantages, any of which might be associated with increased risk of autism, and with increased risk of other disorders as well."