BACKGROUND: Supraspinatus deficiency is the most frequent and important problem associated to rotator cuff pathologies. It reduces shoulder stability and can lead to osteoarthritis. The goal of this study was to develop a numerical model of the shoulder to analyse the biomechanical consequences of this pathology. METHODS: A 3D finite element model of the shoulder was developed from a normal cadaver specimen. It included the scapula, the humerus and the major abduction muscles. Instead of the usual ball-socket assumption, which prevents the natural translation of the humerus, shoulder stability was actively achieved by muscles. A feedback algorithm was developed to synchronise muscle forces during abduction. The numerical algorithm was validated against an algebraic model, and the calculated muscle moment arms were compared to the literature. Two cases were considered: a normal shoulder and the same one without supraspinatus. FINDINGS: For the normal shoulder, the model predicted the initial upward migration of the humeral head. The maximal humerus translation occurred at 30 degrees of abduction and was 0.75 mm above its ideal centered position. Without supraspinatus, it was 1.6 times higher and the contact point in the glenoid fossa was more eccentric. For the normal shoulder, the maximal glenohumeral force was 81% of the body weight, at 82 degrees of abduction. Without supraspinatus, it increased by 8%, while the increase of muscle forces was 30%. INTERPRETATION: Supraspinatus deficiency increased the upward migration of the humerus, the eccentric loading, and the joint and muscle forces, which may cause a limitation of active abduction and degenerative glenohumeral changes (osteoarthritis and the rotator cuff tear).