1. Irregularities in the development of tension during the tonic vibration reflex of the soleus muscle of the decerebrate cat have been analysed into their frequency components. The reflex was recorded isometrically and elicited by longitudinal vibration, normally at 150 Hz. The amplitude of vibration was set so as to elicit a maximal reflex response, suggesting 1:1 driving of the majority of the Ia afferents at the frequency of vibration. 2. The resulting power spectrum regularly showed a well marked tremor peak separated by a trough from any slow irregularities. The predominant frequency of this tremor varied from 4 to 11 Hz in different preparations, with a mean of 7.4 Hz; on average, frequencies within 1.7 Hz on either side contained over half the power of the predominant frequency. Altering the frequency of vibration did not alter the distribution of tremor frequencies. 3. The root mean square value of the tension irregularities, over the range 4-14 Hz, varied from 12 to 110 mN in different preparations (median value, 23 mN); this was superimposed on mean active reflex tensions varying from 2 to 10 N. 4. The 'tremor' due to a single motor unit was estimated from spectral analysis of tetanic contractions of the whole muscle and decreased with increasing frequency of activation. Comparison of the single unit values with the tremor seen during vibration in the same preparations showed that equivalent amounts of tremor to the latter could typically have been produced by the continued synchronous contraction of about five 'average' motor units firing at the predominant tremor frequency. 5. When a tonic stretch reflex was present its tremor frequencies did not differ consistently from those of the tonic vibration reflex. On average, the tremor was smaller for the stretch reflex than for the tonic vibration reflex; the difference was usually slight and might have been related to the stretch refex tension being smaller. 6. Evidence was obtained that the tremor was not due to any insecurity of 1:1 driving of the Ia afferents by the vibration. First, the tremor did not increase when the amplitude of vibration was decreased sufficiently to ensure that the degree of 1:1 driving must have been reduced. Secondly, the introduction of a comparable 'artificial tremor' by sinusoidally oscillating the muscle at low frequency did not produce the e.m.g. response that would have been expected if the applied 'tremor' had been modulating the firing of the Ia or any other group of afferents. 7. It is concluded that the observed tremor cannot be attributed to 'oscillation in the stretch reflex arc', though without prejudice to the role of this mechanism under other conditions and especially when the recording is not isometric. However, the genesis of the tremor has not been established and much of it might result simply from the chance synchronization of motor units that are firing below their tetanic fusion frequency.