Purpose: This study analyzed the effect of epoch length on energy expenditure (EE) estimates and on the moderate-to-vigorous physical activity (MVPA) measure obtained from EE estimates during accelerometer-based assessment. Methods: Ten active students exercised on a treadmill for four 5-min bouts, using two effort intensities (running and walking) and two physical activity (PA) patterns (continuous or intermittent) wearing an accelerometer. The criterion measure of EE was achieved using indirect calorimetry. Data obtained from the accelerometer were reintegrated into 1-sec, 10-sec, 30-sec and 60-sec epochs. Results: During the running intermittent condition, EE and MVPA estimates from the accelerometer were no different from the criterion measure with 30- and 60-sec epoch lengths but they were different with shorter epoch lengths. During the walking intermittent condition, no difference was observed between EE estimates and the criterion measure, regardless of the epoch length. During the running continuous condition, EE estimated from the accelerometer was significantly lower than the EE obtained with the criterion measure, regardless of the epoch length. During the walking continuous condition, no difference was observed. Conclusion: During the intermittent running condition, longer epoch lengths gave the best EE and MVPA estimates. This conclusion is contrary to the current general view that shorter epochs are the most accurate for PA assessment. However, PA estimates are closer to an external load estimates whereas EE estimates are closer to the internal load estimate. Depending on the objective of their study, researchers should be aware of these findings.