A parametric model of the human reflex torque response to a large-amplitude, constant angular velocity elbow extension was developed in order to help quantify spasticity in hemiparetic stroke patients, and to better understand its pathophysiology. The model accounted for the routinely observed leveling of torque (i.e., a plateau) at a mean angular increment of 51 degrees +/- 10 degrees s.d. (n = 98) after the initial rise. This torque "plateau" was observed in all eight subjects, and in 98 of 125 trials across 25 experimental sessions. The occurrence of this plateau cannot be explained by decreases in elbow flexor moment arms during elbow extension. Rather, the plateau is attributable to a consistent leveling in muscle activation as confirmed both qualitatively from recordings of rectified, smoothed electromyograph (EMG) activity, and quantitatively using an EMG coefficient model. A parametric model was developed in which the pattern of muscle activation in the stretch reflex response of elbow flexors was described as a cumulative normal distribution with respect to joint angle. Two activation functions, one related to biceps and the other to brachioradialis/brachialis, were incorporated into the model in order to account for observations of a bimodal angular stiffness profile. The resulting model yielded biologically plausible parameters of the stretch reflex response which may prove useful for quantifying spasticity. In addition, the model parameters had clear pathophysiological analogs, which may help us understand the nature of the stretch reflex response in spastic muscles.