Variables, such as peak and accumulated moments and spine compression forces, have been shown to be risk factors for occupational low back pain. Estimates of these forces during prolonged, dynamic, asymmetric tasks using biomechanical models is complex and time-consuming. A simple technique for continuous measurement of these variables over a prolonged period is needed to measure the distribution of spinal loading during both sagittal plane lifts and complex asymmetrical jobs. The aim of this study was to determine whether a linear normalization of erector spinae EMG to spine compression force, called compression normalized EMG (CNEMG), could be used to estimate spinal loading for simulations of asymmetrical occupational tasks. The estimates of spine compression force obtained using the normalized EMG are presented in the form of an amplitude probability distribution function and are compared with estimates of a three-dimensional biomechanical model. The per cent time a worker spends above particular levels of spinal loading of interest, such as the NIOSH action limit for compression, are displayed. Five males performed simulated occupational tasks. The exposure time at a specific level of spine compression force for a combination of three tasks, estimated by CNEMG, was, on average, within 6.5% of the time calculated by the biomechanical model. However, if the task combination was dominated by an axial twisting moment, then the difference was, on average, 13.4%. The difference in magnitude of spine compression at a specific probability was, on average, 14.9% and when axial trunk twist dominated, 30.7%. It is concluded that CNEMG can estimate probability at a specific level of spine compression force when the task combination is characterized by a predominant extensor moment in the sagittal plane. Estimates of spine compression at a specific probability, and estimates obtained during task combinations dominated by an axial twisting moment, are poor.