The leading joint hypothesis (LJH) suggests distinct types of control (leading and subordinate) at different joints during multi-joint movements. Taking into account specific features of movements in Parkinson's disease (PD), the LJH predicts distinct effect of PD on control of leading and subordinate joints: impaired interaction torque (INT) regulation should be emphasized at the subordinate joints, and impaired generation of muscle torque (MUS) magnitude should be more pronounced at the leading joint. This prediction was tested by studying three tasks of horizontal shoulder-elbow movements in PD patients and age-matched controls: cyclic line drawing, cyclic point-to-point, and discrete pointing movements. Each task included movements in different directions, providing both shoulder-lead and elbow-lead control patterns. Torque analysis supported the prediction, specifically for Tasks 2 and 3 in which movement targets were chosen to emphasize the shoulder- and elbow-lead control patterns. Patients did not exploit INT for motion generation as successfully as controls did, but only at the subordinate joint. Underproduction of MUS by PD patients was more apparent at the leading than subordinate joint. The results support joint-specific effect of PD on movement control. They also suggest that dyscoordination of joint motions in PD stems predominantly from impaired control of subordinate joints, while bradykinesia is associated more with control of the leading than subordinate joint. Possible contribution of the revealed impairments in joint control to some other movement features in PD is discussed. The study demonstrates the efficiency of the LJH application for revealing changes in joint control caused by motor disorders.