BackgroundThe bidirectional relationship between cognitive function and depressive symptoms has been extensively reported. However, the potential mechanisms are still not clear. We aim to longitudinally investigate whether physical activity mediates the bidirectional relationships between cognitive function and specific dimensions of depressive symptoms. MethodsData from 6,787 individuals aged ≥50 of 2014/15 (T1), 2016/17 (T2), and 2018/19 (T3) waves of the English Longitudinal Study of Ageing (ELSA). Cognitive function was assessed by domains of memory, orientation in time, and executive function. Physical activity was measured with the intensity and frequency of participation. Specific dimensions of depressive symptoms were assessed by the 8-item Center for Epidemiologic Studies Depression Scale, distinguishing between cognitive-affective and somatic symptoms. Cross-lagged panel models were used to investigate the mediating role of physical activity in the bidirectional relationships between cognitive function and two dimensions of depressive symptoms. ResultsPoorer cognitive function was indirectly associated with worse cognitive-affective symptoms (indirect effect = −0.002, 95%CI: −0.004, −0.001) through lower physical activity levels. Poorer cognitive function was also indirectly associated with worse somatic symptoms (indirect effect = −0.003, 95%CI: −0.006, −0.002) through lower physical activity levels, and the reverse mediation was observed as well (indirect effect = −0.002, 95%CI: −0.004, −0.001). LimitationsThere is no distinction between potential within-person and between-person effects. ConclusionsCollaborative interventions of physical activity are beneficial in protecting cognitive function and mental health in older adults.