Objective This study tested longitudinal associations between absolute levels of perceived partner responsiveness (PPR; how much people perceive that their romantic partners understand, care for, and appreciate them), daily negative affect reactivity and positive affect reactivity, and all-cause mortality in a sample of 1,208 adults for three waves of data collection spanning 20 years. We also tested whether longitudinal changes in PPR predicted mortality via affect reactivity. Methods Data were taken from the National Survey of Midlife Development in the United States. PPR was assessed at waves 1 and 2, affect reactivity to stressors was assessed by daily diary reports at wave 2, and mortality status was obtained at wave 3. Results Mediation analyses revealed absolute levels of PPR at wave 1 predicted wave 3 mortality via wave 2 affective reactivity in the predicted direction, but this did not remain robust when statistically accounting for covariates (e.g., marital risk, neuroticism), beta = .004, 95% confidence interval = -.03 to .04. However, wave 1-2 PPR change predicted negative affect (but not positive affect) reactivity to daily stressors at wave 2, which then predicted mortality risk a decade later (wave 3); these results held when adjusting for relevant demographic, health, and psychosocial covariates, beta = -.04, 95% confidence interval = -.09 to -.002. Conclusions These findings are among the first to provide direct evidence of psychological mechanisms underlying the links between intimate relationships and mortality and have implications for research aiming to develop interventions that increase or maintain responsiveness in relationships over time.