A paper in January's Archives of Pediatrics and Adolescent Medicine argues that the incidence has increased due to a change in the way that autism is defined. We have password controlled access to the Archives of Pediatrics and Adolescent Medicine: contact me if you can't see this password information page.
The BMJ reports the paper in a news item, and a debate is being pursued through its Rapid Responses.
I have just discovered a review in the latest Acta Paediatrica (Jan 2005, 2-15), by Michael Rutter, titled "Incidence of autism spectrum disorders: changes over time and their meaning". It may well not refer to the research I mention above, but it looks worth reading. We have online access to this journal (although the January 2005 issue is not there yet...) - use the Leicester e-link to link to it.