Purpose: It is important to examine how childhood physical activity is related to adult physical activity in order to best tailor physical activity-promotion strategies. The time- and resource-intensive nature of studies spanning childhood into adulthood means the understanding of physical activity trajectories over this time span is limited. This study aimed to determine whether childhood domain-specific physical activities predict domain-specific physical activity 20 years later in adulthood, and whether age and sex play a role in these trajectories.

Methods: In 1985, 6412 children of age 9–15 years self-reported frequency and duration of discretionary sport and exercise (leisure activity), transport activity, school sport and physical education (PE) in the past week and number of sports played in the past year. In 2004–2006, 2201 of these participants (aged 26–36 years) completed the long International Physical Activity Questionnaire and/or wore a Yamax pedometer. Analyses included partial correlation coefficients and log-binomial regression.

Results: Childhood and adult activity were weakly correlated (r = −0.08–0.14). Total weekly physical activity in childhood did not predict adult activity. School PE predicted adult total weekly physical activity and daily steps (older females), while school sport demonstrated inconsistent associations. Leisure and transport activity in childhood predicted adult leisure activity among younger males and older females, respectively. Childhood past year sport participation positively predicted adult physical activity (younger males and older females).

Conclusions: Despite modest associations between childhood and adult physical activity that varied by domain, age and sex, promoting a range of physical activities to children of all ages is warranted.