1. Tremor force was recorded during stationary isometric contractions of intrinsic hand muscles of normal subjects. Subjects maintained a steady force level between their thumb and forefinger for 30 s. The force level varied from weak (0.2 kg) to strong contractions (7 kg). These experimental conditions were the same as those in two preceding studies, where single motor-unit activity (14) and the correlation between the discharges of two simultaneously recorded motor units and physiological tremor (11) have been investigated. 2. Two alterations of the power spectra were observed at successively stronger contractions: increase of tremor amplitude and changes in the shape of the power spectrum. At all force levels, the power spectra of tremor force show the well-known decay of tremor amplitude from the lower to the higher frequencies with a local peak at 6--10 Hz. This peak does not show a significant change with respect to frequency when the force level is varied. It is shifted toward lower frequencies in a pathological condition (Parkinsonism) where the recruitment firing rates of the motor units are significantly lower than in the normal. 3. Higher frequencies (greater than 20 Hz) are barely present in the power spectrum during the very weak contractions. They become significant as the contractions become stronger. 4. The steep decay of the power spectrum toward higher frequencies has a similar slope (--43 dB/decade) as the reduction in amplitude of the unfused part of the muscle contractions with increasing stimulus rates (--38 dB/decade). The cutoff of the power spectrum above 25 Hz parallels the achievement of total fusion of muscle twitches above this rate. 5. The results are consistent with the hypothesis that the power spectrum over the range of 6--25 Hz is mainly caused by the unfused parts of the twitch contractions of motor units firing between recruitment (6--8/s) and total fusion of the twitches (25--30/s). The decline of the power spectrum toward higher frequencies can be explained by mechanical damping, which results from increasing fusion of the twitch contractions. The low-frequency part of the power spectrum is assumed to be the result of the slow force deviations produced by changes in the net output of the motoneuron pool. 6. These assumptions were supported by additional animal experiments where the number and rate of force-producing elements could be controlled. Bundles of ventral root filaments innervating cat soleus and gastrocnemius muscles were stimulated synchronously and asynchronously at a number of different rates. The force output of the strain gauge was recorded, filtered, and analyzed in the same way as the human force records. 7. Stimualtion of one nerve bundle at one fixed frequency led to a sharp peak in the power spectrum at that frequency plus peaks of decreasing height representing the harmonics of the stimulation frequency. The height of the peaks decreased at --37 dB/decade. 8...