When the Centers for Disease Control announced last year that an average of one out of 150 children had autism, it convinced many people that America was seeing an explosion of autism cases.

Before the 1990s, the official estimates were one autistic child out of every 2,000 to 5,000 children. Several autism advocacy groups took that as proof that some environmental toxin, such as mercury preservative in vaccines, had caused a huge spike in the autism numbers.

But there are now intriguing indications that most if not all of the autism increase is the result of broadening the criteria for the diagnosis and identifying children with autism who would have been labeled with a different diagnosis in the past.

A 2006 study in the journal Pediatrics found, for instance, that the national increase in identified autism cases in elementary schoolchildren between 1984 and 2003 had been paralleled by a similar decrease in the number of children labeled as retarded or learning disabled.

Paul Shattuck of Washington University in St. Louis, the lead author of the study, wrote last year that "in 44 of 50 states, the increase in autism was completely offset by a decrease in the prevalence of children considered 'cognitively disabled' or 'learning disabled.' "

In an interview this week, Dr. Shattuck said that because of changes in the definition of autism and how it is measured, it is impossible to know how much it may have increased from past years.

But his study certainly suggests that "diagnostic substitution" -- labeling someone as autistic today who would have been labeled as retarded 30 years ago -- is a substantial part of the picture.

Dr. Shattuck's study isn't the only one showing this trend.

In a 2004 study, Lisa Croen of the Kaiser Foundation Research Institute in California and her team found that the increase in children diagnosed with autism in that state between 1987 and 1994 was almost exactly paralleled by a decrease in those diagnosed with retardation.

Nancy Minshew, the director of the University of Pittsburgh's Center for Excellence in Autism Research, said last week, "I used to think there were more cases [than in past years], but I don't think so any more." She is now convinced that the higher numbers are "not an increase in the number of cases, but are an improvement in recognition."

In past decades, she said, it was often hard to get doctors or schools to diagnose higher-functioning children as having autism. They were often labeled as having "behavior difficulties."

Dr. Shattuck said other epidemiological studies have shown that the rate of severe autism has stayed steady at about one to two children per 1,000, so that the main part of the increase to an estimated six to seven children per 1,000 has come in the milder, higher-functioning forms of the disorder.

That points partly to the broader definition being used for what are called Autism Spectrum Disorders today, he said.

"When we talk about autism spectrum disorders," he said, "we're talking about kids who have very different symptoms. Some are severely retarded; some have high IQs; some have pathological shyness; others want to have contact but are socially awkward."

And when people say they don't remember seeing so many autistic children when they were growing up, or ask where all the adults with autism are, there are two possible explanations, Dr. Minshew said.

One is that many autistic children in the past were never sent to school. In what she called the "Forrest Gump era, you didn't even go to school, or you went to a totally separate school."

The other phenomenon was that some autistic children were labeled as schizophrenic, and many may have ended up in state hospitals or other institutions, she said.

There is even a kind of logic to that, Dr. Minshew said, because some of the hallmarks of schizophrenia -- behaving oddly, a lack of facial expressions, poor eye contact, speaking in a monotone and using fewer gestures than normal -- are "essentially the same" in both autism and schizophrenia.

David Mandell, an epidemiologist at the University of Pennsylvania medical school, recently surveyed the adult patients in Norristown State Hospital in Eastern Pennsylvania, nearly all of whom are labeled schizophrenic, and found that about 20 percent of them meet the behavioral criteria for being autistic.

While he believes misdiagnosis in the past explains a part of the increase in autism numbers, Dr. Mandell also believes the growth has been too great to be accounted for just by continuing genetic abnormalities.

"The increase is probably too fast to be genetics," he said, "so there probably is something that is environmental, but there is nothing to suggest it's the vaccines."