Objective: The latest definitions of autism spectrum disorders (ASDs) were specified in "DSM-IV-TR" in 2000. "DSM-5" criteria are planned for 2013. Here, we estimated the prevalence of ASDs and autism according to "DSM-IV-TR," clarified confusion concerning diagnostic criteria, and evaluated "DSM-5" draft criteria for ASD posted by the American Psychiatry Association (APA) in February 2010. Method: This was an epidemiological study of 5,484 eight-year-old children in Finland, 4,422 (81%) of them rated via the Autism Spectrum Screening Questionnaire by parents and/or teachers, and 110 examined by using a structured interview, semi-structured observation, IQ measurement, school-day observation, and patient records. Diagnoses were assigned according to "DSM-IV-TR" criteria and "DSM-5" draft criteria in children with a full-scale IQ (FSIQ) greater than or equal to 50. Patient records were evaluated in children with an FSIQ less than 50 to discover diagnoses of ASDs. Results: The prevalence of ASDs was 8.4 in 1,000 and that of autism 4.1 in 1,000 according to "DSM-IV-TR." Of the subjects with ASDs and autism, 65% and 61% were high-functioning (FSIQ greater than or equal to 70), respectively. The prevalence of pervasive developmental disorder not otherwise specified was not estimated because of inconsistency in "DSM-IV-TR" criteria. "DSM-5" draft criteria were shown to be less sensitive in regard to identification of subjects with ASDs, particularly those with Asperger's syndrome and some high-functioning subjects with autism. Conclusions: "DSM-IV-TR" helps with the definition of ASDs only up to a point. We suggest modifications to five details of "DSM-5" draft criteria posted by the APA in February 2010. Completing revision of "DSM" criteria for ASDs is a challenging task. (Contains 1 figure and 11 tables.)