Previous investigations of the social behavior of handicapped preschool children (Journal of Applied Behavior Analysis, 1974, 7, 583–590; Journal of Applied Behavior Analysis, 1976, 9, 31–40; Journal of Applied Behavior Analysis, 1977, 10, 289–298) demonstrated that introduction of adult or peer confederate intervention agents produced substantial increases in levels of positive social behavior emitted by the subjects. In addition, it was observed that changes in rates of positive social behavior emitted by recipients of intervention tactics were accompanied by parallel changes in rates of positive social behavior emitted by interacting peers. However, with one limited exception, sudden removals of arranged intervention procedures were followed by immediate reductions in the levels of positive social behavior emitted by subjects and peers in each study. The current investigation was designed to examine the effects of response‐dependent removal of intervention procedures on the positive behavior of three socially withdrawn preschool boys. Interactive effects on the social behavior of classroom peers who did not receive adult prompts and contingent attention events were also examined. A combination of withdrawal of treatment and multiple baseline procedures was employed. The three target subjects received fixed numbers of prompts and contingent attention events during Intervention I and Intervention II, Phase 1 conditions. During Intervention II, Phase 2 conditions, prompts and contingent attention events were reduced on a response‐dependent basis for two subjects and on a response‐independent basis for the third subject. The results suggest that: (a) the intervention procedures produced marked increases in positive social behavior emitted by each subject; (b) response‐dependent fading and thinning, contrasted with response‐independent tactics, maintained levels of positive social behavior equivalent to those observed during Intervention I and Intervention II, Phase 1 conditions; (c) changes in positive and negative behaviors emitted by peers paralleled changes in positive and negative behaviors emitted by each subject; and (d) no “spillover” of treatment effects was noted for subjects during periods in which they were not direct recipients of intervention procedures.