Several recent prospective analyses involving community-based populations have demonstrated a protective effect on survival for frequent attendance at religious services. How such involvement increases survival are unclear. To test the hypothesis that religious attendance might serve to improve and maintain good health behaviors, mental health, and social relationships, changes and consistencies in these variables were studied between 1965 and 1994 for 2,676 Alameda County Study participants, from 17 to 65 years of age in 1965, who survived to 1994. Measures included smoking, physical activity, alcohol consumption, medical checkups, depression, social interactions, and marital status. Those reporting weekly religious attendance in 1965 were more likely to both improve poor health behaviors and maintain good ones by 1994 than were those whose attendance was less or none. Weekly attendance was also associated with improving and maintaining good mental health, increased social relationships, and marital stability. Results were stronger for women in improving poor health behaviors and mental health, consistent with known gender differences in associations between religious attendance and survival. Further understanding the mechanisms involved could aid health promotion and intervention efforts.