A new study suggests that many children diagnosed with severe language disorders in the 1980s and 1990s would instead be diagnosed as having autism today and so the rise in the number of cases of autism may be related to changes in how it is diagnosed.

Professor Dorothy Bishop, a Wellcome Trust Principal Research Fellow at the University of Oxford, led a study which revisited 38 adults, aged between 15-31, who had been diagnosed with having developmental language disorders as children rather than being autistic.

Professor Bishop and colleagues looked at whether they now met current diagnostic criteria for autistic spectrum disorders, either through reports of their childhood behaviour or on the basis of their current behaviour. The results are published this month in the journal Developmental Medicine & Child Neurology.

Developmental language disorders, which include specific language impairment, are diagnosed when a child has unusual difficulty in his or her grasp of the spoken language, despite normal development in other areas. This may range from a child who has very limited ability to produce or understand spoken sentences, to one who does speak in long and complex utterances, but nevertheless has problem communicating effectively because of problems in conveying a point or grasping what others mean.

Autistic spectrum disorders, which include autism and Asperger syndrome, are developmental disorders affecting how a person communicates with and relates to other people and how they make sense of the world around them.

Participants in the study were drawn from a pool of children who had participated in a series of studies of developmental language disorder conducted during the period 1986 to 2003 and about whose conditions detailed information was known. All attended special schools or classes for children with language impairments, and would have been diagnosed by educational psychologists, paediatricians or speech therapists as having developmental language disorders and none had previously been diagnosed as autistic. However, when reassessed by Professor Bishop and colleagues using current criteria, around a quarter were identified as having autistic spectrum disorder.

In recent years, the criteria for diagnosing developmental language disorders and autism have changed. This has coincided with a marked rise in the rates of diagnosis of autism. According to the Special Needs and Autism Project, the figure until the 1990s was widely accepted as being about 5 people per 10,000; even using the narrowest definition of autism, this rose to almost 40 in 10,000 by 2006

There are two main hypotheses to explain this rise: the "autism epidemic" hypothesis and the "diagnostic substitution" hypothesis. Whilst the former says that the rise is genuine, the latter maintains that the true prevalence of the disorder is constant but that changes in diagnostic criteria mean that more children are being diagnosed as autistic. The latter theory is supported by a UK study(3) using the General Practice Research Database, which found that the rise in autism was mirrored by a decline in frequency of language disorders, and now by Professor Bishop's study.

"Our study shows pretty direct evidence to support the theory that changes in diagnosis may contribute towards the rise in autism," says Professor Bishop. "These were children that people were saying were not autistic in the 1980s, but when we talk to their parents now about what they were like as children, it's clear that they would be classified as autistic now.

"Criteria for diagnosing autism were much more stringent in the 1980s than nowadays and a child wouldn't be classed as autistic unless he or she was very severe. Now, children are being identified who have more subtle characteristics and who could in the past easily have been missed."

However, Professor Bishop cautions against using the results to suggest that the prevalence of autism is not genuinely rising.

"We can't say that genuine cases of autism are not on the increase as the numbers in our study are very small," she says. "However, this is the only study to date where direct evidence has been found of people who would have had a different diagnosis today than they were given fifteen or twenty years ago."