Objectives To examine the relationship between objectively measured physical activity (PA) and cognitive function in white and black older adults. Design Cross-sectional. Setting REasons for Geographic and Racial Differences in Stroke (REGARDS) study. Participants Older adults who provided valid data from accelerometer and cognitive function tests (N = 7,098). Measurements Accelerometers provided estimates of PA variables for 4 to 7 consecutive days. PA cut-points of 50 counts per minute (cpm) and 1,065 cpm were applied to differentiate between being sedentary and light PA and between light and moderate to vigorous PA (MVPA), respectively. Prevalence of cognitive impairment was defined using the Six-Item Screener (<4/6). Letter fluency, animal fluency, word list learning, and Montreal Cognitive Assessment (orientation and recall) were used to assess memory and executive function. Results Of 7,098 participants (aged 70.1 ± 8.5, 54.2% female, 31.5% black), 359 (5.1%) had impaired cognition within ±12 months of PA measurement. The average proportion of time spent in MVPA (MVPA%) was 1.4 ± 1.9%. Participants in the highest quartile of MVPA% (~258.3 min/wk) were less likely to be cognitively impaired than those in the lowest quartile (odds ratio = 0.65, 95% confidence interval = 0.43-0.97). MVPA% was also significantly associated with executive function and memory z-scores (P <.001). Similar analyses of proportion of time spent in light PA (LPA%) and sedentary time (ST%) showed no significant associations with cognitive function. Conclusion Higher levels of objectively measured MVPA%, rather than LPA% or ST%, were associated with lower prevalence of cognitive impairment and better performance in memory and executive function in aging people. The amount of MVPA associated with lower prevalence of cognitive impairment is consistent with meeting PA guidelines.