Radiographs are a suitable outcome measure in patients with rheumatoid arthritis. They reflect the history of the joint pathology and provide a permanent record necessary for serial evaluation of the disease. Great care should be taken to overcome technical problems with radiographs to ensure that good quality films are available to score. Many scoring methods have been described ranging from a global score for the whole patient to the more sophisticated methods of scoring erosions and joint space narrowing in a selected number of joints. These latter abnormalities give additive information and are the most important features in scoring radiographs in rheumatoid arthritis. An overview of the most important methods is given with an emphasis on four selected methods: the Larsen method, a modification of this by Rau and Herborn, the Sharp method and a modification of this by van der Heijde. All four methods produce sufficient intra- and inter-observer reliability. Although data are scarce, the Sharp method and its modifications seem the most sensitive methods for detecting changes over time. However, these are more time-consuming than the Larsen method or its modifications. Depending on the type of study a choice can be made between the two types of methods. For clinical trials where small differences are important, the (modified) Sharp method seems the most appropriate. In working with large data sets, time might be a more crucial factor and, therefore, the (modified) Larsen method could be chosen.