We investigated the reliability and predictive validity of several measures of the social environment using a sample of 127 adults with type II (non-insulin-dependent) diabetes. Of particular interest was a revised scale of family support for performing diabetes self-care behaviors. Across four different aspects of the diabetes regimen, it was consistently found that regimen-specific measures of family support differentiated subjects who were low, medium, or high on adherence better than global family-support scores. Social-environment measures were generally successful in predicting a subject's level of regimen adherence after accounting for variance attributable to demographic factors. The family-support measures were the strongest and most consistent predictors of adherence, but measures of stress and medical-care satisfaction and the interaction between stress and the other social-environment measures also often improved the prediction of a subject's adherence status. These results support the usefulness of measures of the social environment specific to diabetes, and research to investigate other aspects of the social context of regimen adherence is recommended.