Reliability in the assessment of radiographic abnormalities in the sacro-iliac joint was investigated by rheumatologists and radiologists experienced in the reading of such films. This study differs from previous ones in that films were masked to reduce cueing from the radiographic appearance of other pelvic bone and joint structures. Agreement was greatest (greater than or equal to 80%) when only the presence or absence of abnormality was considered. When corrected for chance, agreement on the individual radiographic abnormalities was frequently poor or modest. The use of reference radiographs reduced absolute disagreement and had a modest influence on absolute agreement. The scoring of films for epidemiologic purposes should preferably be delegated to senior radiologists using standard reference films. Nevertheless, there remains a considerable margin of error in the interpretation of sacroiliac joint radiographs which is of clinical and epidemiological importance considering the weight that radiographs of this joint carry in diagnostic rheumatology.