The ability of subjects with multidirectional shoulder instability to use proprioception to complete a series of upper limb repositioning tasks was investigated. Twelve subjects with multidirectional instability and 12 control subjects were blindfolded and instructed to use proprioception to reproduce a self-selected target position as accurately as possible. Subjects completed 10 repetitions for each limb using three distinct upper limb movements: overhead reaching, scapular plane pointing, and humeral external rotation with abduction. A three-dimensional video motion analysis system tracked limb position and determined spatial hand position error. Subjects with multidirectional shoulder instability showed significantly greater hand position error than control subjects. No hand position error differences were found between the symptomatic and the asymptomatic limbs of subjects in the instability group. Inter-repetition error for subjects in both groups improved significantly during the first three movement cycles. These results suggest that after movement initiation, dynamic proprioception was a factor in improving hand position accuracy in both groups, but to a lesser degree in subjects with multidirectional instability. Consequently, subjects with multidirectional instability may have a reduced capacity to use proprioception to refine and control the motor output of the upper limb.