Conventional radiographic scoring methods in rheumatoid arthritis (RA) are designed to quantify progression and disregard any improvement. Reparative changes observed during longterm followup of RA have rarely been described as healing phenomena. Healing may become increasingly important with the availability of more potent antirheumatic drugs. We investigated whether radiologic healing phenomena can be identified by different observers in a blinded fashion. Healing phenomena were defined as (1) reappearance (and sclerosis) of the cortical plate, (2) partial or complete filling in of an erosion, or (3) subchondral bone sclerosis with osteophyte formation (secondary osteoarthritis). Pairs of radiographs of hands, wrists, and forefeet [taken 2 to 8 (mean 4.8) yrs apart] of 34 patients were selected from longterm studies: 24 sets with healing phenomena and 10 with progressive disease without healing. The radiographs were blinded and read in pairs in random order by 3 observers unaware of the patients selected for the study. One observer read the set a second time after 8 weeks, resulting in a total of 4 observations. Thirty-eight joints were rated 0 to 5 depending on the amount of surface destruction. In addition an attempt was made to identify joints with healing phenomena. The data were analyzed with help of descriptive statistics such as means, standard deviations, and frequency tables. Out of 1292 joints scored at the second time point, 74 had healing phenomena. These joints were identified in a mean of 89% (95/89/88/82%, respectively) in the 4 observations performed by the 3 readers. Patients without healing phenomena were correctly identified by all observers. In 54 joints less typical healing phenomena were seen in all observations. Sixty-two joints with healing in addition to the 74 joints were seen in 3 observations, 76 in 2, and 127 in one observation. All observers agreed that 1090 joints had no healing phenomenon. The 24 patients with healing had a slight reduction in the Ratingen score, while the 10 patients without healing showed a moderate progression. In the group with healing an increase and decrease in the score occurred in the same patients at different joints. In patients with RA followed for several years healing or reparative changes of erosions can be observed. These phenomena can be identified by different observers with high sensitivity even when the observers are blinded to chronological sequence of the films.