Seven decades have elapsed since Leo Kanner described the syndrome he termed early infantile autism. Over this time, and particularly over the past two decades, noteworthy changes have occurred in how the condition is conceptualized. Here we provide an overview of these changes, beginning with a brief discussion of the significance of classification in general before discussing Kanner's original paper and subsequent changes. We touch on relevant issues, such as comorbidity, dimensional aspects of diagnosis and screening, and the complex issue of diagnosis relative to eligibility for services. Approaches to diagnosis have tended to swing from emphasizing overarching groups (lumping) to focusing on potentially distinct subgroups (splitting). Autism raises particular problems given the broad range of syndrome expression over age and developmental level. The most recent revision of the American Psychiatric Association's diagnostic taxonomy marks a significant departure from its predecessor and has been the focus of much debate. It remains unclear which of the currently existing categorical approaches will ultimately be most widely applied. We hope to convey a sense of areas in which consensus has been achieved and areas of continued controversy.