The possibility of causing instability in the stretch reflex has been examined in three different human muscles: biceps, first dorsal interosseous (FDI) of the hand and digastric. Tremor recorded as fluctuation of isometric force was compared with that occurring during contraction against a spring load. The spring compliance was selected to make the natural frequency of the part in each case appropriate for oscillations in the short latency stretch reflex. A computer model of the whole system was used to predict the frequency at which oscillations should be expected and to estimate the reflex gain required in each case to cause sustained oscillations. Estimates were computed of the autospectra of the force records and of the rectified surface EMG signals and of the coherence functions. Normal subjects showed no evidence of a distinct spectral peak during isometric recording from any of the three muscles. However, in anisometric conditions regular oscillations in force occurred in biceps, but not in FDI or digastric. The oscillations in biceps at 8-9 Hz were accompanied by similar oscillations in the EMG which were highly coherent with the force signal. The results are consistent with the presence of a strong segmental stretch reflex effect in biceps and weak or absent reflex in FDI. Digastric is known to contain no muscle spindles and therefore to lack a stretch reflex. In two subjects who volunteered that they had more tremor than normal, but had no known neurological abnormality, there was a distinct peak in the force spectrum at 8-9 Hz in biceps and FDI in isometric conditions with coherent EMG activity. The peak increased in size in anisometric conditions in biceps but not in FDI. This component appears to be of central rather than of reflex origin. No equivalent component was found in digastric records. The results are discussed in relation to the possible role of the short latency stretch reflex in the genesis of physiological tremor in different muscles.