To assess the validity of accelerometry in measuring daily physical activity, the energy consumption calculated by accelerometry, with respiratory gas analysis as a reference, was evaluated in 45 non-athletes during various exercise tests. Subjects were required to (1) walk on a treadmill ergometer at various speeds, (2) walk on a treadmill ergometer at a fixed speed and with a stride of 20% more or 20% less than that when walking freely, (3) walk on a treadmill ergometer at a fixed speed wearing either sneakers or leather-soled shoes, and (4) cycle on a bicycle ergometer. There were strong linear relationships between the measurements during the progressively graded treadmill test, with an overall Pearson correlation coefficient of 0.97. The mean estimated difference ranged from -0.77 to 0.27 kcal/min and the coefficients of variation from 13.2% to 22.2%. However, the difference between the methods was not negligible for individual subjects. Accelerometry overestimated energy expenditure during short-step walking, and underestimated it during long-step walking. No significant difference in energy expenditure was found according to the type of shoes worn. Cycling activity was not recorded by accelerometry. Accelerometry is a reasonably accurate and feasible method for evaluating the physical activities of non-athletes, and could be a common tool for epidemiological research and health promotion despite its limitations. (Circ J 2002; 66: 751 - 754)