Individualized treatment packages were developed for 3 children with high-rate severe pica using a discrimination training paradigm and a behavioral assessment-based procedure known as empirically derived consequences. Children received empirically derived reinforcers for eating under appropriate stimulus conditions (i.e., eating food only from a plate and placemat that served as a discriminative stimulus) and empirically derived punishers for attempts to engage in pica. This treatment package resulted in marked reductions in pica and an increase in appropriate eating for all 3 children in a "baited" analogue condition. In addition, low rates of pica were maintained for 9 months for all 3 children. These results suggest that treatment effectiveness may be enhanced when behavioral assessment data are used to identify potent consequences.