The factor structure of the Autism Diagnostic Interview-Revised (ADI-R) algorithm items was examined using exploratory (EFA) and confirmatory factor analyses (CFA) factor methods. The ADI-R was completed for 1,170 youths and adults (ages 2-46). Results of EFAs indicated strong support for two-factor structure, with social communication and stereotyped behavior factors. CFAs computed in a holdout sub-sample indicated roughly equal support for the above described two-factor model and a three factor model separating peer relationships and play from other social and communicative behaviors. Multi-group CFAs suggested that both two and three factor models showed good stability across age, with only slight changes in factor relationships. These findings indicate that the current ADI-R structure be revised to more accurately reflect the relationships between sub-scales.