The purpose of this study was to describe the agreement between two hospitals on either side of the Atlantic Ocean in reading first day plain radiographs of suspected scaphoid fractures. Two groups of observers, one North American and one European, consisting of observers at various levels of training were compared. Kappa statistics were used to determine inter- and intra-observer agreement. The receiver-operating characteristics (ROC) curves and area under this curve (AUC) for each observer were calculated to determine test performance. Scaphoid radiographs of 80 consecutive patients seen with clinically suspected scaphoid fracture were included in the study. The results of the bone scan were used as the reference standard. There is an acceptable agreement between the USA and Europe for both radiologists and orthopaedic surgeons at various levels of training (kappa 0.45-0.88). Intra-observer agreement is acceptable as well (kappa 0.46-0.86). Considering all normal or equivocal radiographs, there is slight to moderate agreement at all levels of training in and between both centres and specialities (kappa 0-0.64). Overall test performance was acceptable for all consecutive radiographs (AUC 0.65-0.82) and poor for the normal radiographs (AUC 0.46-0.65). We could not detect differences in performance due to location of the readers' hospital or institution of training. The use of more sensitive techniques should be restricted to patients with negative or equivocal radiographs in clinically suspected scaphoid fractures.