In June 1998, the National Institutes of Health Autism Coordinating Committee (NIH/ACC) invited representatives of 13 major medical and other professional academies and associations and six national autism parent research organizations to review research data on screening and diagnosis of autism spectrum disorders. Ten review papers and more than 4,000 publications were consulted in this effort. This paper highlights some promising areas for research identified in this process. One of the highest priorities is the search for the ultimate diagnostic indicator, a biological marker(s), for example, genetic, metabolic, immunologic, neurologic, that will distinguish autism unequivocally from other developmental disabilities. In the interim, research on infant screening and diagnosis might lower the threshold age for diagnosis to 8-12 months. The role of sensory-motor disorders in early diagnosis needs further research. Earlier and better diagnosis of co-occurring, potentially treatable disorders, including epileptic and epileptiform disorders, has implications both for diagnosis and treatment. Pharmacogenetic and pharmacogenomic research strategies could help diagnose subtypes and responders versus nonresponders to potential treatments. Better endpoints and outcome measures are needed, including improved procedures for cognitive and neuropsychological testing, more knowledge about verbal and nonverbal communication milestones, and less invasive and more sensitive neuroimaging measures. Critical questions remain regarding regression after apparently normal development, and about possible environmental precipitants. Finally, field trials of the reliability and validity of screening and diagnosis using the newly developed practice guidelines are needed.