The experimental protocol was divided into seven phases: desensitization, reliability, pre-intervention, inter¬vention, postintervention, interim period, and follow- up. Coding of playground behavior took place during all phases except the interim period, and experimenters were governed by three policies: (a) avoid interaction with teachers or children to the extent possible, (b) stand on the fringes of the playground and remain as unobtrusive as possible, and (c) avoid prolonged eye contact directed toward any single child or teacher. Coding took place during daily morning playground ses¬sions. If children were unable to play outside because of inclement weather (17% of potential coding days), data collection was canceled; all decisions to stay indoors during playground time were made by the preschool staff, independent of the research team. Data collection was occasionally canceled because of special events (e.g., local fire department visit in their fire trucks; 4% of potential coding days). Each of the seven protocol phases is described below in turn.