Biochemical evaluation of iron depletion due to bleeding is difficult in rheumatoid arthritis (RA) patients with inflammatory anaemia. The authors have tried to assess the usefulness of serum ferritin (SF) level measurements--considered a reliable index of iron stores--by comparing the values obtained with true iron stores, as determined by Perls' test on bone marrow. The study involved 50 RA patients with anaemia and raised erythrocyte sedimentation rate (ESR). The poor correlation (r = 0.51) observed between the two tests in this population was mainly due to differences in ESR. SF levels and Perls' test results correlated better in patients with an ESR above 50 mm/h. (r = 0.69) than in patients with an ESR above 50 mm/h. (r = 0.44). The mean ESR values were higher than those found in patients without inflammatory syndrome. SF levels, therefore, are unreliable in patients with inflammatory syndrome, and no practical conclusion concerning iron stores can be drawn from the finding of a normal SF level. In the present study, only SF levels higher than 250 ng/ml were associated with positive Perls' tests. The behaviour of ferritin in the presence of inflammation may explain the results obtained.