To describe shoulder motion the sternoclavicular, acromioclavicular and glenohumeral joint centres must be accurately located. Within the literature various methods to estimate joint centres of rotation location are proposed, with no agreement of the method best suited to the shoulder. The objective of this study was to determine the most reliable non-invasive method for locating joint centre locations of the shoulder complex. Functional methods using pin mounted markers were compared to anatomical methods, functional methods using skin mounted markers, imaging-based methods using CT-scan data, and regression equations. Three participants took part in the study, that involved insertion of intracortical pins into the clavicle, scapula and humerus, a CT-scan of the shoulder, and finally data collection using a motion analysis system. The various methods to estimate joint centre location did not all agree, however suggestions about the most reliable non-invasive methods could be made. For the sternoclavicular joint, the authors suggest the anatomical method using the most ventral landmark on the sternoclavicular joint, as recommended by the International Society of Biomechanics. For the acromioclavicular joint, the authors suggest the anatomical method using the landmark defined as the most dorsal point on the acromioclavicular joint, as proposed by van der Helm. For the glenohumeral joint, the simple regression equation of Rab is recommended.