Abstract Background Physical activity (PA) is important to maintaining functional independence. It is not clear how patterns of change in late-life PA are associated with contemporaneous changes in physical performance measures. Methods Self-reported PA, gait speed, grip strength, timed chair stand, and leg power were assessed in 3,865 men aged ≥ 65 years at baseline (2000–2002) and Year 7 (2007–2009). Group-based trajectory modeling, using up to four PA measures over this period, identified PA trajectories. Multivariate linear regression models (adjusted least square mean [95% confidence interval {CI}]) described associations between-PA trajectories and concurrent changes in performance. Results Three discrete PA patterns were identified, all with declining PA. Linear declines in each performance measure (baseline to Year 7) were observed across all three PA groups, but there was some variability in the rate of decline. Multivariate models assessing the graded response by PA trajectory showed a trend where the high-activity group had the smallest declines in performance while the low-activity group had the largest (p-for trend < .03). Changes in the high-activity group were the following: gait speed (−0.10 m/s [−0.12, −0.08]), grip strength (−3.79 kg [−4.35, −3.23]), and chair stands (−0.38 [−0.50, −0.25]), whereas changes in the low-activity group were the following: gait speed (−0.16 [−0.17, −0.14]), grip strength (−4.83 kg [−5.10, −4.55]), and chair stands (−0.53 [−0.59, −0.46]). Between-group differences in leg power trajectories across PA patterns were not significant. Conclusions Declines in functional performance were higher among those with lower PA trajectories, providing further evidence for the interrelationship between changes in PA and performance during old age.