This fMRI study was undertaken to test whether the pathophysiological mechanism of mirror movements in hemiparetic stroke patients involves activation of the unaffected motor cortex. We studied 16 control subjects and 51 stroke patients. fMRI was performed at 1.5 T using a finger flexion-extension movement paradigm. The incidence of bilateral primary sensorimotor cortex activation was significantly increased during movements of the affected hand of stroke patients who showed mirror movements. Moreover, the incidence of bilateral primary sensorimotor cortex activation increased with the severity of mirror movements and primary sensorimotor cortex was activated bilaterally in all patients who showed sustained mirror movements. We conclude that the motor cortex activation on the non-stroke side is associated with mirror movements and is correlated with the severity of mirror movements. It seems that the pathophysiological mechanism of sustained mirror movements in stroke patients involves the unaffected motor cortex.