Objective To examine rest–activity circadian rhythm (RAR) and cognitive decline in older men. Design Longitudinal. Setting Osteoporotic Fractures in Men (MrOS) and ancillary Outcomes of Sleep Disorders in Men (MrOS Sleep) studies. Participants MrOS and MrOS Sleep participants (N=2,754; mean age 76.0 ± 5.3). Measurements The Modified Mini‐Mental State examination (3MS) was used to assess cognition at baseline (2003–05) and follow‐up examinations (2005–06 and 2007–09). Wrist actigraphy was used to measure 24‐hour activity counts at baseline. RAR variables included amplitude (strength of activity rhythm), mesor (mean activity level), pseudo F‐statistic (overall circadian rhythm robustness), and acrophase (time of daily peak activity). Results After an average of 3.4 ± 0.5 years, men with lower amplitudes, mesors, and pseudo F‐statistics had greater decline in 3MS performance (amplitude: –0.7 points Q1 vs –0.5 points Q4, p<.001; mesor: –0.5 points Q1 vs –0.2 points Q4, p=.01; pseudo F‐statistic: –0.5 points Q1 vs –0.3 points Q4, p<.001). Lower amplitudes and pseudo‐F statistics were associated with greater odds of clinically significant cognitive decline (≥5‐point decrease) (amplitude Q1 vs. Q4: odds ratio (OR)=1.4, 95% confidence interval (CI)=1.0–1.9; pseudo‐F statistic Q1 vs Q4: OR=1.4, 95% CI=1.0–1.9). Men with phase‐advanced acrophase had greater odds of clinically significant cognitive decline (OR=1.8, 95% CI=1.2–2.8). Results were adjusted for multiple confounders. Conclusion Several parameters of disrupted RAR (lower amplitude, pseudo F‐statistic, mesor, phase‐advanced acrophase) were associated with greater cognitive decline in older community‐dwelling men. These findings contribute to a growing body of evidence suggesting that altered RARs are associated with cognitive decline in older adults. J Am Geriatr Soc 66:2136–2143, 2018.