• AA;
  • mediation;
  • social support



This study tested the hypothesis that the relationship between Alcoholics Anonymous (AA) involvement and reduced substance use is partially explained (or ‘mediated’) by changes in social networks.


This is a naturalistic longitudinal study of the course of alcohol problems.


Study sites were the 10 largest public and private alcohol treatment programs in a northern California county.


Three hundred and seventy-seven men and 277 women were recruited upon seeking treatment at study sites.


At baseline and 1-year follow-up, we assessed alcohol consequences and dependence symptoms, consumption, social support for abstinence, pro-drinking social influences and AA involvement.


In the structural equation model, AA involvement was a significant predictor of lower alcohol consumption and fewer related problems. The size of this effect decreased by 36% when network size and support for drinking were included as mediators. In logistic regression models predicting abstinence at follow-up, AA remained highly significant after including social network variables but was again reduced in magnitude. Thirty-day abstinence was predicted by AA involvement (OR = 2.9), not having pro-drinking influences in one's network (OR = 0.7) and having support for reducing consumption from people met in AA (versus no support; OR = 3.4). In contrast, having support from non-AA members was not a significant predictor of abstinence. For alcohol-related outcomes other than abstinence, significant relationships were found for both AA-based and non-AA-based support.


The type of social support specifically given by AA members, such as 24-hour availability, role modeling and experientially based advice for staying sober, may help to explain AA's mechanism of action. Results highlight the value of focusing on outcomes reflective of AA's goals (such as abstinence) when studying how AA works.