The purpose of the present undertaking was to investigate prognosis of patients with non-specific or rheumatoid arthritis and to study the value of different diagnostic criteria for rheumatoid arthritis (RA) at the beginning of the disease. During the years 1973-75, a total of 442 patients with recent (less than or equal to 6 months) arthritis were studied at the Rheumatism Foundation Hospital, Heinola, Finland. In 1982 the outcome of the 200 patients with the diagnosis of RA or non-specific arthritis was established. Their ages at the time of the first hospitalization were 16-77 years, mean 41. There were 63 men and 137 women. The follow-up period was 6-9 years, mean 7.6. The outcome was measured by joint score, function score, the sum of ESR and CRP, X-ray index, outcome index, which was composed of the preceding ones, ESR, and CRP. The outcome was poor in half of the patients. Ninety-eight signs and symptoms registered at the first hospitalization were compared with the seven facets of outcome using Pearson's correlation coefficient r. The significances of the correlation coefficients were tested by Student's t-test. From 10 to 39 variables correlated highly significantly (p less than 0.001) with the measures of outcome; however, most of the variables had only moderate correlations. At best 14 variables showed 0.40 less than r less than 0.58 when correlated with the outcome index. In conclusion, at the onset of an arthritic disease symmetrical polyarthritis in peripheral joints, serum rheumatoid factor, X-ray changes, morning stiffness, high ESR, and old age correlated best with a destructive joint disease. Plasma proteins as indices of non-specific inflammation mattered less. The relationships between 22 entry variables and the seven facets of prognosis were further evaluated by means of stepwise multiple regression analysis. The results were essentially the same as reached above. One or two variables, most often the number of diseased peripheral joints and serum Waaler-Rose test, always explained the variance better than the number of ARA criteria alone. To study the value of the diagnostic criteria, diagnosis of the patients was made in the following three ways: RA with five or more erosive joints (N = 78), RF-positive and erosive RA (N = 93), RF-positive and nonerosive or RF-negative and erosive RA (N = 125). The patients outside each of the diagnostic groups formed the corresponding control groups.(ABSTRACT TRUNCATED AT 400 WORDS)