Purpose: This longitudinal study examines the effect of social participation on mortality and cognitive impairment for the elderly in Taiwan. Methods: Data were from an elderly population panel in Taiwan 'The Survey of Health and Living Status of the Elderly' 1993-1999. Social participation was defined as paid/unpaid job and participating in volunteer and social groups. Logistic regression analysis was used for predicting the influence of social participation to 6-year mortality and cognitive function impairment, by controlling of socioeconomic status and health-related variables. Results: The elderly with continuous paid work were more likely to participate in social groups. Having paid or unpaid work at the baseline year could lower the risk of mortality six years later, especially for men. Having unpaid work was associated with a higher risk of impaired cognitive function compared to that of non-workers. Participating in a religious group reduced the risk of mortality for women and participating in political groups reduced the risk of impaired cognitive function for men. Discussion: Participating in some types of social activities may reduce mortality or cognitive function impairment in the elderly. However, different activities, gender roles and uneven opportunities for work and social group participation may affect the risk of these endpoints.