Three articular indices for measuring disease activity are compared. In a cross sectional study the Thompson articular index (a modified Lansbury index) correlated better with laboratory variables than the Ritchie articular index or a swollen joint score (Thompson 0.74-0.77; Ritchie 0.57-0.58; swollen joint count 0.51-0.59). In a longitudinal study it was shown that the Thompson articular index is sensitive to detect changes of disease activity. We suggest that this index is appropriate for clinical trials.