This study was concerned with the effect of delays of reinforcement on the maintenance of previously learned work behaviours in retardates. Responses to delays appear to be idiosyncratic with no apparent ordering of effects between most and least delay, as they were used in this study. Production levels of all subjects were raised on some conditions, the conditions varying from subject to subject. The increase in production above the level the subjects worked at when not on the reward system, was more than sufficient to cover the cost of the programme. It was also found that when subjects were in the condition most suited to themselves errors were eliminated and irrelevant responses reduced although there were no contingencies operating on this aspect.