The AA attendance variable at posttreatment proved to be a better predictor of drinking outcome than the AA involvement variable. This result is at odds with findings like those of Montgomery, Miller and Tonigan (1995), perhaps due to the way that AA Involvement is measured. The largest component of AA Involvement is AA attendance. (In this sample, AA attendance and AA involvement correlated at above r = .80). When AA attendance was in the structural model, the variance attributable to AA Involvement was reduced.
