Early diagnosis and treatment of autism increases the likelihood that symptoms associated with the disorder can be alleviated. However, the behaviors of both typically and atypically developing young infants and toddlers are quite variable and often change as these children age. Studies have shown that this is the case for same-aged children who are diagnosed with autistic disorder (AD) or other Pervasive Developmental Disabilities (PDDs). Therefore, an early accurate assessment of AD or PDD may be problematic. Moreover, instruments used to make the diagnosis may not be as reliable as desired. Statistics employed to evaluate diagnostic accuracy and behavioral stability of instruments' or clinicians' assessments suggest that their diagnoses have been only moderately successful. In addition, the statistics themselves have limitations that would suggest that the measures of diagnostic accuracy and behavioral stability implemented may not be as effective as they would seem. A resolution to these problems is proposed. © 2012 Wiley Periodicals, Inc.