Summary The main objective of this project was to develop a measurement system consisting of various sensors and instruments for quantitative assessment of the physical exposure of manual materials handlers, and to test it in the laboratory and in the field. Such a system could be used to determine the efficacy of different prevention methods. A second objective was to develop a sampling strategy to optimize the field measurement of handlers’ physical exposure. Many authors have criticized the shortcomings in data sampling methods, which have perhaps weakened the credibility of studies on the correlation between physical exposure and physical injuries. In the first part of the project, the task was to determine the variables contributing significantly to physical exposure and to select the appropriate instrumentation. Exposure variables had to meet at least two criteria: (1) be recognized in the literature as having a significant effect on handlers’ risk of injury; (2) be able to be measured quantitatively in the field. A large number of variables met the first criterion, but only a few met the second one. This was to be expected, given the difficulties involved in field measurement in spite of recent major technological advances. It was found essential to have a system capable of measuring the 3D kinematics of the entire body, i.e., tracking the movements of the main segments—back, pelvis, head, arms, forearms, hands, thighs, legs and feet. Promising systems that have emerged recently include IMUs (inertial measurement units) made up of triaxial accelerometers, gyroscopes and magnetometers whose signals, combined by means of a Kalman digital filter, are able to determine the 3D kinematics (motion, velocity and acceleration) of a segment. Such a system, developed by Xsens Technologies, appeared particularly promising; however, like any instrument, its validity and reliability first had to be checked under laboratory conditions, and secondly it had to be tested in the field on a target group of materials handlers. Laboratory trials on 12 volunteers (9 men, 3 women) showed that the Xsens system was able to estimate most of the joint angles for the entire body during handling tasks to within a very acceptable 5° error margin against the benchmark measurement system (an Optotrak optoelectronic system). IMUs are therefore potentially capable of tracking worker motion during daily tasks, at least in an environment free of magnetic distortion. If the magnetometer data are affected by magnetic disturbance, the error increases but returns to an acceptable level 30 s after the disturbance ceases. Before the data can be considered valid, a minimum 30 s delay should therefore be allowed after any magnetic disturbance. The second part of the project consisted in field testing of the Xsens inertial measurement system on 10 handlers (9 men, 1 women) during their regular tasks. All the subjects handled at least one order (average 115 products transferred onto a pallet) for an average of 32 min. In the field, quantitative comparison with a benchmark instrument is difficult. Nevertheless, a qualitative comparison can be made between the segment motion of the Xsens avatar and the actual subject images recorded with a video camera. The differences between the avatar’s movements and the synchronized video images were observed by a researcher. Of all the observations assessed (total = 2,298 observations), 68% were judged acceptable, that is, the video image of the subject matched the image of the Xsens avatar. Most of the errors were caused by magnetic disturbances from the forklift, but it is still possible to measure worker kinematics in the field with an acceptable error level. Moreover, when full-body kinematics is combined with other information available in the workplace, such as the order list showing the weights of the products handled, loading on the back can be calculated. All these data, taken together, now make it possible to estimate the physical exposure of materials handlers. To meet the second objective, a sampling strategy was developed. This will prove useful for optimizing future data collection tasks and ensuring good sampling. However, compromises will be necessary between very high precision (with the number of prescribed subjects) and field realities in terms of cost/benefit. There are numerous potential uses for this type of system. For example, it will be possible to quantify the efficacy of an ergonomic intervention designed to reduce physical exposure. Novice handlers be monitored during their training period and receive feedback on their work methods. Finally, it will be easier to quantify dose-effect curves to support the development of new occupational safety standards.