The output from the central nervous system to muscles may be rhythmic in nature. Previous recordings investigating peripheral manifestations of such rhythmic activity are conflicting. This study attempts to resolve these conflicts by employing a novel arrangement to measure and correlate rhythms in tremor, electromyographic (EMG) activity and muscle vibration sounds during steady index finger abduction. An elastic attachment of the index finger to a strain gauge allowed a strong but relatively unfixed abducting contraction of the first dorsal interosseous (1DI). An accelerometer attached to the end of the finger recorded tremor, surface electrodes over 1DI recorded EMG signals and a heart-sounds monitor placed over 1DI recorded vibration. This arrangement enabled maintenance of a constant overall muscle contraction strength while still allowing measurement of the occurrence of tremulous movements of the finger. Ten normal subjects were studied with the index finger first extended at rest and then contracting 1DI to abduct the index finger against three different steady forces up to 50% of maximal voluntary contraction (MVC). Power spectral analysis of tremor, EMG activity and muscle vibration signals each revealed three frequency peaks occurring together at around 10 Hz, 20 Hz and 40 Hz. Coherence analysis showed that the same three peaks were present in the three signals. Phase analysis indicated a fixed time lag of tremor behind EMG of around 6.5 ms. This is compared with previous measurements of electromechanical delay. Other experiments indicated that the three peaks were of central nervous origin. Introducing mechanical perturbations or extra loading to the finger and making recordings under partial anaesthesia of the hand and forearm demonstrated preservation of all the peaks, suggesting that they did not originate from mechanical resonances or peripheral feedback loop resonances. It is concluded that, at least for a small hand muscle, there exist not one but a number of separate peak frequencies of oscillation during active contraction, and that these oscillations reflect synchronization of motor units at frequencies determined within the central nervous system. It is proposed that the multiple oscillations may be a means of frequency coding of motor commands.