The effect of modeling and reinforcement on the peer-praising behavior of six preschool children was investigated. Prior to the experiment, six children were divided into three dyads. Each dyad met with the experimenter throughout the study for a daily 7-min play session. Following baseline, a training package consisting of an instruction to praise, adult modeling or praising, and token reinforcement was applied in a multiple-baseline design across the three dyads. In the initial training condition, only a stereotypic praise statement was modeled. In a subsequent training condition, diverse, descriptive praise models were presented. Several types of generalization probes across peers were interspersed throughout the study. Training was successful in teaching generalized praising skills to preschool children. However, some generalization probes suggested that training the response only in specifically praiseworthy situations may be more appropriate, given the situation-specific characteristics of this behavior.