The purpose of this study was to teach two mentally retarded employees to perform tasks associated with their jobs. Typically, they served meals during lunch and dinner hours; however, they failed to wipe counters, and check and restock supplies. A self-instructional package, similar to the package introduced by Bornstein and Quevillon (1976), was incorporated to improve their performance in these areas. A multiple-baseline design was utilized in which training was sequentially introduced across subjects. The subjects were taught to ask questions about which tasks needed to be completed, answer their questions, and guide their performance in the task by self-instructing. The self-instructional strategies were introduced just prior to lunch and dinner (30-min period) and during the first 30 min of the lunch and dinner service periods, respectively. During training, both subjects demonstrated that they were capable of rehearsing each of the self-instructional strategies in a specific sequence. Following training, both subjects performed at least one of the three tasks (i.e., wiping counters, checking supplies, restocking supplies) during 84 and 88% of the observed work periods. Additionally, the results of the intervention indicated that both subjects spent more time working and met or exceeded nonprobationary, nonhandicapped coworkers' production standards (social validation measure). The results of this study are compared with those reported by others. Future areas of research are indicated.