Abstract. Severity of glenohumeral arthropathy following shoulder instability is often graded radiologically on the basis of osteophyte size according to Samilson and Prieto. Intra- and inter-observer reliability of this popular grading system was determined using standard radiographs of 20 arthritic cadaveric humeri. Using two observers on two occasions, reliability of the grading system using the antero-posterior radiograph varied between fair and poor (kappa coefficients 0.070.33). This study shows that the Samilson and Prieto grading of glenohumeral arthropathy is unreliable and, in clinical use, when magnification and rotation is less easy to control, it is likely to be even more so. Another validated classification system must therefore be found.