The use of movement registration for daily physical activity assessment was evaluated during a 7-day period in 30 free-living subjects. Body movement was registered with a Tracmor motion sensor consisting of a triaxial accelerometer and a data unit for on-line processing of accelerometer output over 1-min intervals. Average Tracmor output was correlated against four different energy estimates: 1) average daily metabolic rate (ADMR), determined with doubly labeled water; 2) ADMR-sleeping metabolic rate (SMR; determined in a respiration chamber); 3) (ADMR-SMR) per kilogram of body mass; and 4) the overall physical activity level (PAL = ADMR/SMR). The highest correlation was found for the relationship between Tracmor output and PAL (r = 0.58). After correction for Tracmor values arising from vibrations produced by transportation means, this correlation was improved to 0.73. There was no difference between Tracmor output and PAL in discriminating between overall activity levels with "low" (PAL < 1.60), "moderate" (1.60 < or = PAL < or = 1.85), and "high" (PAL > 1.85) intensity. It is concluded that the Tracmor can be used in free-living subjects to distinguish among interindividual as well as intraindividual levels of daily physical activity.