INTRODUCTION: Associations between heavy lifting and low back pain have prompted attempts to assess spinal loading in the work-place. Direct measurement of spinal compression, by inserting a pressure-sensitive needle into the intervertebral disc, is not suitable during dynamic tasks, so more indirect methods have been developed. In one approach, force-plate and movement analysis data are incorporated into a l inked-segment model which calculates net moments and joint reaction forces. A second approach involves measuring the electromyographic (EMG) activity of trunk muscles in order to directly quantify extensor moment and compressive force. Linked-segment models cannot detect antagonistic (stabilising) muscle forces, whereas EMG techniques are unable to detect inertial forces acting along the axis of the spine. The present study compared estimates of spinal loading using both techniques in order to quantify the forces h idden from each.METHODS: Eight men lifted boxes, weighing 6.7kg and 15.7kg, from the ground, using four techniques, and two speeds. During each lift, rectified and averaged EMG signals from the erector spinae muscles were recorded at 60Hz from surface electrodes placed bilaterally at T10 and L3. Simultaneous recordings were made of vertical ground reaction force, using two Kistler force plates, and of body movements, using a 4-camera Vicon system. After completion of the lifts, static calibrations of EMG against extensor moment were carried out, and the effects of muscle length and contraction velocity on the EMG-extensor moment relationship were corrected for (1). The linked-segment model (2) was used to estimate the net moment at L5-S1 as well as that component of the joint reaction force which accelerates the spine along its axis. The latter represents the inertial t hrust from the legs and is unrelated to back muscle activity and so is undetected by the EMG technique.RESULTS: The EMG technique indicated that peak extensor moments at L5-S1 ranged from 196-334Nm. These were 17-24% higher than the net moments obtained using the linked segment model, and corresponded to compressive forces of approximately 2.5-4.8kN. The t hrust from the legs contributed only 1-4% of the total compressive force on average, and was highest in rapid and heavy bent-legs lifts. This thrust was linearly related to the vertical ground reaction force (r=0.86).DISCUSSION: The generally good agreement between the two techniques increases confidence in the predictions of both. Axial inertial forces h idden from EMG recordings were less than 4% of the peak spinal compressive force, and were predicted accurately from force-plate recordings. Increased extensor moments predicted by the EMG technique may be attributable to antagonistic muscle activity, although other explanations are possible.CONCLUSION: EMG techniques are suitable for assessing fully dynamic spinal loading but additional force-plate data can slightly improve accuracy during more arduous lifts.