The joints of 30 rheumatoid arthritis patients were assessed by one observer for signs of inflammation. Computer analysis was then used to calculate 70 different articular indices for each patient. Pearson correlation coefficients were calculated between serum C-reactive protein levels and the articular indices. The results show that: findings in a restricted set of examined joints were equivalent to those in a more complete set; the simultaneous presence of joint tenderness and swelling yielded higher correlation than did either variable alone; and joint "weighting" for size yielded higher correlation than did simple counts.