1. Slow finger movements were analysed in normal human subjects with regard to kinematics and EMG activity of the long finger muscles. Surface EMG from the finger extensor and flexor muscles on the forearm was recorded along with angular position and angular velocity during voluntary ramp movements at single metacarpophalangeal joints. Angular acceleration was computed from the velocity record. 2. It was found that movements were not smooth but characterized by steps or discontinuities, often recurring at intervals of 100-125 ms, yielding velocity and acceleration profiles dominated by 8-10 Hz cycles. The discontinuities were manifest from the very first trial and thus not dependent on training. Their amplitude and amount varied between subjects but were relatively stable for the individual subject. 3. The 8-10 Hz cycles were seen with voluntary ramp movements of widely varying velocities, higher velocities being associated with larger steps recurring with the same repetition rate as the small steps of slow voluntary ramps. Maximal step amplitude observed was more than one order of magnitude larger than physiological tremor. 4. The individual 8-10 Hz cycle was asymmetrical in that decelerations usually reached higher peaks than the preceding acceleration, suggesting that the antagonist contributed with a braking action. Moreover, in very slow voluntary ramps, the movement cycles were often interspaced by periods of zero velocity, providing a highly non-sinusoidal velocity profile. 5. The EMG of the agonist and the antagonist muscles was modulated in close relation to the accelerations and decelerations respectively of the individual movement cycle. These modulations were present in both extensor and flexor muscles, although they were more consistent and usually more prominent in the former. 6. The findings indicate that a feature of slow finger movements was an 8-10 Hz periodic output to the muscular system, suggesting that slow finger movements are implemented by a series of biphasic force pulses, involving not only the shortening agonist muscle propelling the movement, but the antagonist muscle as well whose activity increased shortly after the agonist and contributed to a sharp deceleration of the individual step of movement. 7. It is proposed, as a hypothesis, that this biphasic motor output may reflect a similar organization of the descending motor command for slow finger movements. Hence, this command would include a series of biphasic pulses, concatenated at a rate of 8-10 per second and a pulse-height regulator capable of setting the size of the pulse and thus the overall speed of the movement.