Cadaveric studies have shown that intervertebral discs and ligaments are most vulnerable to injury when loaded simultaneously in bending and compression. The purpose of the present experiment was to measure bending and compressive stresses acting on the lumbar spine during forward bending and lifting activities, and to identify those aspects of lifting which increase the risk of injury. Twenty-one men and eighteen women lifted objects from the ground while the following parameters were varied: the angle of the knees ('squat' lift or 'stoop' lift), the mass of the object, its bulk, its distance in front of the feet, its distance away from the sagittal plane, and the speed of movement. Spinal compression was assessed by measuring the peak extensor moment generated by the back muscles and fascia during the lift. Extensor moment was calculated from the EMG activity of the erector spinae muscles, using corrections for muscle length, contraction velocity and electro-mechanical delay. The bending moment ('bending torque') acting on the intervertebral discs and ligaments was quantified by comparing dynamic measurements of lumbar flexion, obtained with the '3-Space Isotrak', with the normalised bending properties of cadaveric lumbar spines. Results showed that stoop lifting reduced the peak extensor moment by about 10% compared to squat lifting, but increased the bending torque by about 75%. Extensor moment and bending torque both increased substantially with increasing mass, bulk and distance from the feet. Non-sagittal plane lifts increased the bending torque by about 30%. The fastest lifts increased peak extensor moment by 60% but did not increase bending torque. We conclude that complex spinal loading during lifting tasks depends as much on the speed of movement, and the size and position of the object lifted, as on its mass. Analyses of spinal loading which consider only compressive forces do not give a full indication of the risk of injury to the intervertebral discs and ligaments.