To evaluate the reproducibility of the interpretation of radiographs for the diagnosis of scaphoid fractures, 134 radiographs (60 consecutive patients) were blindly observed by a resident radiologist, consultant radiologist, resident trauma surgeon and consultant trauma surgeon. The results of radiographs were analysed using kappa (kappa) statistics and stratified according to the results of the corresponding bone scan. In 23 patients the bone scan was positive for fracture of the scaphoid. Irrespective of training and experience, the kappa of the comparison of the scores of the radiographs between any two observers did not exceed 40 per cent. After stratification of the outcome of the bone scan, the kappa did not increase significantly. A panel of three experienced observers gave their opinion of all radiographs. The results were compared with the bone scan results. We conclude that radiographs in suspected scaphoid fracture cannot be used as a reliable diagnostic approach because of the low inter-observer agreement in the interpretation, irrespective of the experience and training of the observer.