After stroke, hemiparesis will result in impairments to locomotor control. Specifically, muscle coordination deficits, in the form of inappropriately phased muscle-activity patterns, occur in both the paretic and nonparetic limbs. These dysfunctional paretic muscle-coordination patterns can adapt to somatosensory inputs, and also the sensorimotor state of nonparetic limb can influence paretic limb. However, the relative contribution of interlimb pathways for improving paretic muscle-activation patterns in terms of phasing remains unknown. In this study, we investigated whether the paretic muscle-activity phasing can be influenced by the relative angular-spatial relationship of the nonparetic limb by using a split-crank ergometer, where the cranks could be decoupled. Eighteen participants with chronic stroke were asked to pedal bilaterally during each task while surface electromyogram signals were recorded bilaterally from four lower extremity muscles (vastus medialis, rectus femoris, tibialis anterior, and soleus). During each experiment, the relative angular crank positions were manipulated by increasing or decreasing their difference by randomly ordered increments of 30° over the complete cycle [0° (in phase pedaling), 30°, 60°, 90°, 120°, 150°, 180° (standard pedaling), 210°, 240°, 270°, 300°, 330° (out of phase pedaling)]. We found that the paretic and nonparetic muscle phasing in the cycle systematically adapted to varied relative angular relationships, and this systematic relationship was well modeled by a sinusoidal relationship. Also, the paretic uniarticular muscle (vastus medialis) showed larger phase shifts compared with biarticular muscle (rectus femoris). More importantly, for each stroke subject, we demonstrated an exclusive crank-angular relation that resulted in the generation of more appropriately phased paretic muscle activity. These findings provide new evidence to better understand the capability of impaired nervous system to produce a more normalized muscle-phasing pattern poststroke.