The feasibility of a portable accelerometer equipped with a three-directional sensor for the assessment of physical activity and the consequences for energy expenditure was examined under laboratory conditions and during normal daily life. Heart rate monitoring was also conducted to allow comparison of both techniques. In the laboratory study 16 healthy subjects performed a number of specified exercises within a range of activity levels that may be expected in normal life. Accelerometer output was compared with energy expenditure measured by continuous respirometry. A linear relationship was found between accelerometer output and energy expenditure for the pooled data. The standard error of estimate is 79.1 J.min-1.kg-1. In the field study four subjects were observed during a week under free living conditions. Energy expenditure was calculated from food intake registered over the whole period. Energy expenditure calculated from accelerometer output and heart rate exceeded the energy intake figures by 30% and 33%, respectively. Possible explanations for this discrepancy are discussed. Despite this discrepancy, accelerometer output appeared to correlate highly with energy intake (r = 0.99, P less than 0.025), which suggests accurate performance of the accelerometer under free living conditions. The heart rate method gave much poorer results in estimating individual energy expenditure.