The conditions subsumed under the term ‘autistic spectrum disorder’ (ASD) vary in severity at the time of initial presentation; they also vary in expression over time. Unfortunately, the most widely adopted reference standard for the diagnosis of ASD, the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), does not quantify severity of expression of ASD at the time of initial diagnosis, nor can it be used to document longitudinal changes in symptom expression, other than to state in a ‘yes/no’ fashion that an individual might no longer meet criteria for diagnosis. In clinical practice, measures intended to quantify the severity of ASD, such as the Childhood Autism Rating Scale, the Gilliam Autism Rating Scale, or the Autism Diagnostic Observation Schedule (ADOS), are typically given once, rather than serially (unlike IQ or academic achievement testing, which are commonly given every 2 or 3 years, if not annually). Finally, the diagnosis of ASD is generally made without reference to the affected individual's level of general intelligence. Most scales for assessing ASD do not take the participants' chronological age or level of general development into account, although there are exceptions. These clinical practices constrain our thinking about ASD. Here I present a schema for thinking about ASD along three dimensions: (1) severity of atypical symptoms, (2) level of general cognitive ability or comorbid mental retardation, and (3) age. This schema is not meant as a diagnostic instrument in and of itself. Rather, it is intended as a frame of reference within which to locate the scores of existing instruments, and as a way of conceptualizing ASD. I suggest that this frame of reference will lead to improved clinical care for individuals of all ages with ASD, and might point the way toward further researchable questions in areas pertaining to the etiology, educational management, and epidemiology of autism and related disorders.