This article discusses methodological concepts and challenges underlying the interpretation of changes in plain radiographs of the joints of patients with rheumatoid arthritis. A series of consensus conferences (OMERACT [Outcome Measures in Rheumatology]) has resulted in the formulation and execution of a research agenda to harmonise reading and interpretation of films. This is important in the light of the increasing evidence that drugs can impact on the progression of joint damage. In these conferences, methodological issues have been divided according to applicability tenets summarised in the OMERACT Filter of Truth, Discrimination, and Feasibility. To pass the Filter, a measure must measure what it is supposed to measure (Truth), must discriminate between clinically relevant states (Discrimination) and be feasible in terms of costs and interpretability. 'Truth' issues include the choice of joints, the view and other technical specifications of the radiograph, such as which abnormalities to score, the level of aggregation of the information, culminating in the choice of the scoring system. 'Discrimination' issues include reproducibility and sensitivity to change. The current research agenda includes items such as defining a criterion for 'no relevant progression', comparison between time ordered and randomly ordered reading, further comparison of methods and subscores, and methodology around missing values.