BackgroundPhysical activity is associated with improved health and function in older adults, yet most older adults are sedentary. Loneliness is associated with decreased physical activity at the cross-section, but longitudinal studies are scarce. We examined longitudinal associations between loneliness and physical activity—and whether they were modified by marital status and network size (the number of children, relatives, and friends a person interacts with at least once a month).MethodsWe analyzed data from 1,931 older adults without dementia at baseline from the Rush Memory and Aging Project with a mean follow-up of 4.8 years (mean age 79.6 ± 7.7, 74.9% women). Loneliness was assessed using the de Jong Gierveld Loneliness Scale. Physical activity was assessed as the frequency with which participants engaged in five categories of activities (e.g., walking, gardening, calisthenics, bicycling, and swimming). Linear mixed effects models examined associations between baseline loneliness and change in physical activity over time after adjusting for demographics, depressive symptoms, global cognition, disability, network size, marital status, social support, and social and cognitive activities. We assessed for effect modification by marital status and network size.ResultsAssociations between loneliness and physical activity differed by marital status. In widowed individuals, baseline loneliness was associated with a 0.06 h/week greater decrease in physical activity per year compared to those who were not lonely (p = 0.005, CI -0.1, 0.02)—which equaled a 150% decrease in physical activity per year. Loneliness did not predict a statistically significant decrease in physical activity in married or unmarried individuals.DiscussionLoneliness is associated with decreased physical activity in widowed older adults and should be considered in the design of interventions to prevent or slow the decline in physical activity and promote healthy aging.