Two methods of assessing changes in hand and wrist radiographs, the Larsen Index and Amos Index, were compared and contrasted in 48 patients with rheumatoid arthritis (RA). X-rays were taken before and after 1 year of treatment with remission-inducing drugs and scored independently by three observers using both methods. Single readings of radiographs using the Larsen Index were highly reproducible (intra- and inter-observer correlations greater than 0.90 on all occasions), but reproducibility fell when changes were determined from paired films (intra- and inter-observer correlations 0.47-0.82). Variability was greatest in the proximal interphalangeal and wrist joints. The Amos Index, which assesses erosive progression and not the appearance of single films, was more reproducible (intra- and inter-observer correlations 0.72-0.94); it would therefore be better able to detect small changes in groups of patients. The scoring systems were significantly correlated (r = 0.63; P less than 0.001), and with both different observers scored the films at the same "level'. The Amos Index is simple, reproducible, and does not require standard films. It is the assessment of choice.