Physical activity plays an important role in the prevention of many serious diseases. To develop targeted strategies that encourage physical activity, knowledge of stability of physical activity levels over time is essential. The aim of this study was to examine tracking of leisure time physical activity in adults in Northern Norway during three decades. We followed 5432 women and men who attended the Tromsø Study in 1979-1980, as well as repeated examinations after 7 and 28 yr. Baseline age was 20-54 yr (mean age=35.8 yr). Physical activity was assessed by self-administered questionnaires. Tracking of physical activity, defined as maintenance of relative rank of physical activity level, was estimated by Spearman correlation coefficient and by weighted κ statistics. Tracking in terms of predictability of later values from earlier measurements was analyzed by generalized estimating equations. A higher-than-expected proportion of subjects maintained their physical activity level from examination 1 to 2 (58%) and 3 (53%). κ statistics showed agreement of 0.41 and 0.29, respectively. Belonging to a specific physical activity level at baseline increased the odds of belonging to the same category at later examinations (sedentary odds ratio (OR)=3.9 (95% confidence interval (CI)=3.5-4.4), moderately active OR=2.2 (95% CI=2.0-2.4), active OR=2.9 (95% CI=2.6-3.3), and highly active OR=14.0 (95% CI=8.7-22.5)). Being physically active in young adulthood increased the odds of being physically active later in life (moderately active OR=3.4 (95% CI=3.0-3.9), active OR=5.4 (95% CI=4.6-6.4), and highly active OR=13.0 (95% CI=7.4-22.8)). This study showed tracking of leisure time physical activity during 28 yr in a cohort of adults.