In the treatment of patients with pain, measures related to (pain) behaviour are of major importance. Ambulatory activity monitoring can be used to obtain insight into actual behaviour. This study was designed to validate the Activity Monitor (AM), an instrument based on long-term ambulatory monitoring of accelerometer signals, to assess several physical activities during normal daily life. Ten failed back surgery (FBS) patients performed a number of functional activities in and around their own houses. During the measurements, continuous ambulatory registrations of accelerometer signals were made, based on four body-mounted accelerometers (one on each upper leg, two on the trunk). Video recordings made simultaneously with the measurements were used as a reference. The continuous output of the AM (postures, transitions, dynamic activities) was compared with visual analysis of the videotapes. The overall results showed an agreement between AM output and video analysis of 87% (inter subject range: 83-88%). The maximal error in the determination of the duration of activities was 0.3%. The overall number of dynamic periods was determined well (AM: 359; video: 368), while the number of transitions was slightly overestimated (AM: 228; video: 205). The results when using the three-sensor version of the AM were somewhat less accurate (overall agreement from 87% to 82%). The AM appeared to be a valid instrument to quantify aspects of behaviour of FPS patients, such as duration of activities and number of transitions. This new technique of ambulatory measurement of mobility activities seems to be a relevant and promising extension of the techniques currently used in the evaluation of pain treatment.