The functional significance of physiological tremor - the high frequency (8-12 Hz), low amplitude oscillation that occurs during the maintenance of steady limb postures - is not known. Often tremor, perhaps because of its pathological manifestations, is considered a source of unwanted noise in the system, something to be damped out or controlled. An examination of the phase relationship between tremor and rapid voluntary finger movement in normal subjects suggests a very different view. In four experiments in which tremor displacement and accompanying electromyographic activity were simultaneously monitored, we show a clear and systematic relationship between tremor and movement initiation. Empirically obtained frequency distributions of tremor peak-to-movement initiation time were most closely aligned to a probability density function (derived via numerical integration techniques) that assumed movements were initiated when the muscle-joint system possessed peak momentum. This relationship - evaluated by Chi-square goodness-of-fit tests - was evident regardless of whether the movements were self-paced (Experiments 1 and 2) or in response to an auditory reaction time signal (Experiments 3 and 4). The addition of a load to the finger in Experiments 2 and 4, though tending to reduce tremor frequency, did not prove disruptive, nor did a fractionated reaction time analysis reveal any significant inertial contribution to the maintenance of the phase relationship. These data are consistent with an emerging view that the motor control system is sensitive to its own dynamics, and suggest that under certain conditions normal physiological tremor is a potentially exploitable oscillation intrinsic to the motor system.