Normal voluntary movements are considered to be of two kinds, or to involve two components, (i) a ballistic or "open-loop" type, which are preprogrammed and executed without reference to current sensory information and (ii) a corrective or "closed-loop" type, whose course or termination are regulated by such information. In a previous paper it was suggested that Parkinsonism disrupted the first kind of movement, but intention tremor did not. In the present paper three experiments designed to test this hypothesis are described. Subjects were tested on an acquisition-tracking task using an oscilloscope display and joystick control, and measurements were made of the duration, velocity and error of their initial movements to acquire the target. Parkinsonian movements were found to be considerably different from normal in that (a) most movements by this group lasted longer than the reaction time for their initiation, as if including some secondary correction in their execution, (b) the rate of movement was not varied for different amplitudes (so keeping the duration fairly constant) as in normal subjects, but rather movements of all amplitudes were made at a constant slow rate, so that duration increased markedly with the larger steps, (c) error increased disproportionately as the velocity of movement increased; in particular any movements completed in one reaction time or less tended to be wildly inaccurate, (d) removing either the target or the response marker from the screen at the beginning of a movement had a significant effect, making it shorter in duration and smaller in amplitude than those usually produced with both markers visible all the time. Parkinsonian subjects showed no improvement in performance with repeated attempts at one movement over a whole sequence, so their deficit appears to be stable even after practice on a known fixed task. These results are interpreted as supporting the hypothesis that Parkinsonism interferes with the generation of accurate ballistic action which are characteristic of normal skilled movement. Tremor subjects in general resembled normal control subjects in their initial acquistion movements, but their accuracy was less with the larger steps.