Age-dependent cross-level gradients of exercise-associated advantages in behavioural, latent, and neural efficiency

Executive function supports goal-directed behaviour and is particularly vulnerable to age-related decline. Long-term exercise habits represent a promising lifestyle factor for maintaining cognitive health, yet it remains unclear whether exercise-associated advantages are domain-general or selectively expressed as a function of task type, age, and level of analysis. Using a cross-sectional four-group design stratified by age (younger vs. older adults) and long-term exercise-habit status (exercisers vs. non-exercisers), we integrated behavioural-efficiency indices (inverse efficiency score, IES), latent-efficiency indices (drift-rate measures derived from hierarchical drift-diffusion modelling, hDDM), and neural-efficiency indices (activation-pattern similarity from task-based functional magnetic resonance imaging, fMRI) across five paradigms: the Multi-Source Interference Task (MSIT), Stroop, task switching, N-back, and relational processing. Long-term exercise habits were associated with task-specific efficiency advantages, with age-dependent patterns emerging primarily in behavioural outcomes and cross-level gradients. Older adults showed consistently lower behavioural, latent, and neural efficiency across tasks. Exercise-related behavioural advantages were most evident in older adults and were observed in tasks showing significant Age × Exercise interactions. In younger adults, exercise-associated advantages were not reliably expressed at the behavioural level, whereas cross-level gradient analyses suggested stronger expression at latent and neural levels. Mediation analyses indicated that latent efficiency statistically mediated the association between neural efficiency and behavioural efficiency across all tasks. Cross-level gradient analyses further showed that the distribution of exercise-related advantages differed by age: in younger adults, advantages increased from the behavioural to the neural level, whereas in older adults they were more robust at the behavioural and latent levels and comparatively smaller at the neural level. Together, these findings suggest that long-term exercise habits are associated with multi-level differences in cognitive processing efficiency, with systematic age-related variation in their cross-level expression.