This critical review examined a series of 10 controlled studies in which traditional operant behavioral procedures were compared with more recently developed normalized interventions for teaching language to young children with autism. Main characteristics of the older treatments include highly structured direct teaching sessions of discrete trials, teacher initiation, artificial reinforcers, and response shaping. Normalized interventions consist of loosely structured sessions of indirect teaching with everyday situations, child initiation, natural reinforcers, and liberal criteria for presentation of reinforcers. The main conclusion was that in all eight studies with language criterion responses, normalized language training was more effective than discrete-trial training. Furthermore, in both studies that assessed parental affect, normalized treatment yielded more positive affect than discrete-trial training.