Records were made of electromyographic (EMG) responses of both upper limb muscles and the corresponding elbow joint movements following sinusoidal (0.3 Hz) isometric displacement of the elbow joint itself. Two motor conditions were tested. Firstly, the subjects had to control elbow position and secondly control joint torque. Randomly timed, flexing or extending ramp impulses were induced at different displacement velocities and amplitudes. Following long duration displacements (> 100 msec) the recorded EMG responses could clearly be separated into 3 different components (M1-M3). The M1 component was of constant duration but M3 corresponded to the duration of the ramp displacement. It is proposed that the M1 component is "coded" by the acceleration signal and the M3 component by the velocity signal. Only the shape of the M2 component was dependent upon the actual motor condition. With the subjects controlling the elbow joint angle the M2 components in the arm flexor and extensor EMG responses exhibited a peak whose rate of rise was dependent on displacement velocity. However, when elbow torque was controlled by the subjects the M2 component exhibited a plateau whose amplitude was dependent on displacement velocity. The amplitude of the M2 component was significantly larger during position-control than during torque-control. We propose that the difference in the behaviour of the EMG responses may be achieved by the appropriate central regulation of gamma-motoneurone activity or, alternatively, by selective modulation of different receptor inputs between the two tasks.