This study examined whether attendance and quality of participation in parent management training predicted treatment response. Data were from 445 parents (55% minority, 62% single; almost all of low socioeconomic status) who had 1st-grade children with severe conduct problems. Quality of participation in weekly parent groups was based on group leader ratings. Parent outcomes were based on interviewer ratings, behavioral observations, parent reports, and teacher ratings. Results of hierarchical linear models suggested that few family characteristics predicted attendance in this efficacy trial and that attendance was not related to changes in parenting over the year. However, several family characteristics predicted quality of participation in parent management training, and this in turn predicted changes in parental perceptions, warmth, physical punishment, and school involvement. From a clinical perspective, these findings suggest that it is not enough to get parents to attend sessions; it is also necessary to facilitate their active engagement in the therapeutic process. (Contains 4 tables.)