Mechanisms underlying prospective associations of perceived control with frailty and other health outcomes are not well understood. In the present study we used 3 waves of data from the Health and Retirement Study (N = 2,127) to test potential psychological and biological pathways linking perceived control with frailty over an 8-year period, and whether 4-year change in control predicts frailty independent of initial control. Lower odds of increasing frailty were associated with higher initial levels of perceived control (odds ratio [OR] = .74, p < .001, 95% confidence interval [CI] [.65, .83]) and a more positive change in perceived control (OR = .82, p = .006, 95% CI [.73, .92]), independent of the personality traits neuroticism and conscientiousness. In cross-lagged mediation models, the association of initial perceived control with frailty was partially mediated by positive affect, negative affect, and self-rated health, but not C-reactive protein or allostatic load. Associations of perceived control with positive and negative affect were bidirectional, with mediation in both directions. Initial frailty status was not related to 4-year change in perceived control. Perceived control may affect frailty risk through influences on affective states as well as perceived health. Findings also extend evidence that changes in perceived control may be prognostic of future health outcomes in older adults.