Use of scored radiographs as an outcome measure can help estimate the progression of rheumatoid arthritis (RA). Radiographs not only provide permanent records with which to evaluate RA serially, but can also be randomised and blinded, a major advantage in clinical trials. Medline was searched for information about the principal methods of assessing joints affected by RA. Each technique was evaluated for its measurement properties, advantages, and limitations. The most commonly used methods are those devised by Sharp, Larsen, and van der Heijde/Sharp, and their variants. Methods based on the Sharp technique provide separate scores for erosion and for joint space narrowing. Larsen and variants, together with the Simple Erosion Narrowing Score (SENS) method, provide an overall score. Each method's measurement properties (feasibility, time consumption, etc) depend on the degree of detail it considers. Authors consistently recommend taking a posteroanterior view of hand and foot radiographs, and the use of trained raters. Intra- and interrater reliability values are generally higher than 0.70 (less often assessed by the intraclass correlation coefficient than the correlation coefficient). Sensitivity to change is calculated by several techniques (standardised response mean (SRM), adjusted SRM, minimal detectable change, smallest detectable difference). Most methods assessed with SRM reach a value of 0.80 or more. Standardised procedures are available for performing and reading radiographs in RA. The choice of scoring method depends on the time and staff available, and the required degree of reliability and sensitivity to change.