The aims of the present study were to examine the role of epoch length (time sampling interval) on locomotive and non-locomotive physical activity (PA) estimates and the relationships with obesity in healthy adults in each sex. Subjects were 308 Japanese women and 183 men aged 20-to-64years old. Time in locomotive and non-locomotive light PA (2.0 ~ 2.9METs; metabolic equivalents) and moderate-to-vigorous PA (MVPA, 3 METs or more) and step counts were assessed using a triaxial accelerometer (Active style Pro HJA-350IT, Omron Healthcare Co., Ltd) presented in 60s and 10s epochs. Body mass index (BMI) by height and body weight was calculated. There were significant differences between results by epoch lengths in both locomotive and non-locomotive PA. In particular, the difference was remarkable for non-locomotive PA. Time in each MVPA for women and men when presented in 10s epochs were significantly higher. Time in total MVPA was different by nearly 30min for women and nearly 20min for men. Moreover, time in non-locomotive light PA when presented as 60s epochs was significantly higher in both women and men than that when presented as 10s epochs, and time in total light PA showed an obvious difference of nearly 50min higher in 60s epochs for women. Conversely, the relationships between PA with 60sec and 10sec estimates and BMI were weak, even though over half of PA variables showed significant correlations. These findings suggest that non-locomotive PA evaluated by the triaxial accelerometer was obviously affected by epoch length. The finding also showed gender differences.