Active Support, a package of procedures which includes activity planning, support planning, and training on providing effective assistance, was introduced in five community residences serving 19 adults with severe mental retardation following a multiple baseline design. Real-time observational data were collected on the level of assistance residents received from staff and their engagement in activity. Active Support was shown in a companion paper (Jones et al., 1999) to increase the levels of assistance residents received and their engagement in activity. Increased assistance was particularly experienced by the behaviorally less able and the disparity in activity between the more and less able was reduced. In the analysis presented here, the effectiveness of assistance was evaluated before and after Active Support training by calculating the likelihood of engagement occurring given the occurrence of assistance. This likelihood was represented by the statistic, Yule's Q. Yule's Q significantly increased following Active Support training, an increase that was maintained at follow-up. The increased effectiveness of assistance was related to other research findings on the relationship between staff: resident interaction patterns and resident behavior.