The aim of this study is to: (1) compare the association between various levels of physical activity (PA) and mortality; and (2) examine the potential modifying effect of depressive symptoms on the PA-mortality associations. Previous large scale randomized studies rarely assess the association in conjunction with modifying effects of depressive symptoms. In this study, participants consisted of 624 (mean age = 77.35 years) non-institutionalized elderly from the Americans' Changing Lives Longitudinal Study. Depression was measured using the Center for Epidemiological Studies' Depression Scale. Participants in gardening, walking, and sports were first classified into four PA frequency levels, "never," "rarely," "sometimes," and "often." Those who self-reported "often" engaged in activities of gardening and walking and had reduced odds of mortality of 77% and 83%, adjusted odds ratio (ORadj) = .23 and .17, 95% confidence interval (CI) = .09-.59 and.07-.41 when compared to those who reported "never." However, mortality risk was not linked to sports activity. The modifying effects of depressive symptoms on PA (depressive symptoms x PA) were then tested, PA was not associated with increased risk for mortality for gardening (parameter estimates, PE = -.03 +/- .62, p = .958), and for walking (PE = .04 +/- .57, p = .948). Elderly people who engaged in gardening and walking might have protection effects on later risk of mortality. Depressive symptoms showed negative modifying effects that prevent PA predicting later mortality.