Rest tremor is one of the four main clinical features of Parkinson's disease (PD), besides rigidity, bradykinesia and postural instability. While rigidity, bradykinesia and postural instability can be explained with changes in neurotransmitter concentrations and neuronal activity in basal ganglia, the pathogenesis of parkinsonian tremor is not fully understood. According to the leading hypothesis tremor is generated by neurons or groups of neurons in the basal ganglia which act as central oscillators and generate repetitive impulses to the muscles of the body parts involved. The exact morphological substrate for central oscillators and the mechanisms leading to their activation are still an object of debate. Peripheral neural structures exert modulatory influence on tremor amplitude, but not on tremor frequency. We hypothesise that rest tremor in PD is the result of two mechanisms: increased activity and increased synchronisation of central oscillators. We tested our hypothesis by demonstrating that the reduction in rest tremor amplitude is accompanied by increased variability of tremor frequency. The reduction of tremor amplitude is attributed to decreased activity and poor synchronisation of central oscillators in basal ganglia; the increased variability of tremor frequency is attributed to poor synchronisation of the central oscillators. In addition, we demonstrated that the recurrence of clinically visible rest tremor is accompanied by a reduction in tremor frequency variability. This reduction is attributed to increased synchronisation of central oscillators in basal ganglia. We argue that both mechanisms, increased activity of central oscillators and increased synchronisation of central oscillators, are equally important and we predict that tremor becomes clinically evident only when both mechanisms are active at the same time. In circumstances when one of the mechanisms is suppressed tremor amplitude becomes markedly reduced. On the one hand, if the number of active central oscillators is very low, the muscle-stimulating impulses are too weak to cause clinically evident tremor. On the other hand, if central oscillator synchronisation is poor, the impulses originating from different central oscillators are not in phase and thus cancel out, again leading to reduced stimulation of muscles and reduced tremor amplitude. Our hypothesis is supported by our measurements on patients with PD and by experimental data cited in the literature. The proposed two mechanisms could have clinical implications. New medical treatments, which would specifically target only one of the proposed mechanisms (oscillator activity or synchronisation), could be effective in reducing tremor amplitude and thus supplement established antiparkinsonian treatments.