METHODS An experimental study of fatiguing isometric trunk extension was conducted to investigate the spinal loading estimated from an electromyogram-assisted biomechanical model. OBJECTIVE To evaluate the sensitivity of the model outputs to two crucial assumptions: electromyogram-force relationship and method of gain estimation. BACKGROUND In the proposed electromyogram-assisted models of the trunk, the nature of the electromyogram-force relationship and the wide variation in reported muscle gains can result in a wide variation in estimates of spinal loading. Given the absence of any gold standard for validation of muscle forces, the delineation of confidence intervals for the estimated loads has become critical. METHODS Ten subjects performed a fatiguing isometric trunk extension while the net muscular torque output and trunk muscular activity were measured. An electromyogram-assisted model was used to estimate the torque output and spinal loading. Linear and nonlinear erector spinae electromyogram-force relationships and three methods for gain estimation were investigated: constant gain determined from an initial maximum extension exertion, constant gain based on the fatiguing exertion, and a time-varying gain from the fatigue test. RESULTS The predicted torque was not sensitive to the electromyogram-force relationship; the nonlinear model produced 10% lower estimates of peak spinal compression force and 14% higher estimates of peak anterior shear force. The gain determined from an initial calibration exertion underestimated the external torque and underpredicted the peak compression force by 20%, compared with gains calculated in the fatigue test. CONCLUSIONS The nature of the electromyogram-force relationship and of the method for estimating the gain significantly affect the outcomes of an electromyogram-assisted model of spinal loading.