The purpose of this investigation was to evaluate a series of program variables used to modify the time a subject spent attending in an experimental restaurant setting. Selective behaviors of a mildly retarded, female young adult were evaluated, based on combinations of three treatment variables (social reinforcement, token reinforcement and response cost) in modifying the time spent attending to task. Further, the addition of an intermittent schedule of cost contingency, delayed token exchange, delayed performance feedback, and a final weekly pay check were evaluated, and resulting levels of response maintenance were investigated. Combinations of social reinforcement, tokens and response cost were found less effective in increasing percent of time spent attending to task than were simultaneous application of all treatment variables. Relationships were established between maintenance of appropriate attending behavior and the selective withdrawal of both token reinforcement and cost contingency. Data collected suggested that an intermittent schedule of response cost, weekly token exchanges, minimal feedback on performance earning and weekly checks, in that order, were effective in maintaining response gains. Follow-up measures indicated that treatment gains were maintained after all contingencies were removed. Limitations of the present investigation and future areas of research were outlined. Descriptors: Response cost, token economy, social reinforcement, response maintenance, mentally retarded, vocational training.