The amount and pattern of physical activity that benefits cognitive health remain unclear.Participants from the UK Biobank cohort who had a full week of accelerometer-based moderate-to-vigorous physical activity (MVPA) and light physical activity (LPA) data were included in the analysis. The data for dementia diagnosis were collected from 2006 to 2024. Associations between the incidence of all-cause dementia, Alzheimer's disease (AD), vascular dementia (VaD), and PA amounts and patterns were assessed using Cox proportional hazards regression models. The analysis included 1) comparing MVPA gradients with reference group performing less than 150 min/week; 2) classifying MVPA patterns as effective intensive (≥300 min/week with ≥50 % of MVPA in 1-2 days), effective regular (≥300 min/week not up to effective intensive), and ineffective (<300 min/week); 3) performing stratified analyses by age, sex, and APOE ε4 carrier status; and 4) evaluating the association between LPA and dementia risk among participants classified as ineffective MVPA levels.91,512 individuals (mean [SD] age, 56.03[7.8] years; 55.9 % female) were included. Compared with participants performing <150 min of MVPA per week, those accumulating 150-299 min/week, whether through concentrated (1-2 days) or regular pattern, did not show significantly lower dementia incidence. However, accumulating >300 min/week of MVPA was associated with a reduced risk. When stratified at 300 min/week of MVPA, hazard ratios for dementia were 0.73 (95 % CI: 0.60-0.89) for the weekend pattern and 0.79 (95 % CI: 0.64-0.98) for the regular pattern. For ineffective MVPA, engaging in >840 min/week of LPA was associated with lower dementia incidence.Accumulating >300 min/week of MVPA, whether concentrated within 1-2 days or distributed evenly across the week, was associated with a decreased risk of dementia. Additionally, higher levels of LPA partially compensated for low MVPA in lowering dementia risk.Copyright © 2025. Published by Elsevier Masson SAS.