In a randomised, controlled study alternate day prednisolone with an initial high dose phase ("prednisolone only series") has been compared with cyclophosphamide plus alternate day low dose prednisolone ("cyclophosphamide-prednisolone series") in 43 patients with previously untreated fibrosing alveolitis (five patients had received prednisolone in minimal dosage). In the prednisolone only series prednisolone 60 mg daily was given for one month and then reduced by 5 mg a week to 20 mg on alternate days or the minimum dose to maintain early improvement. Patients in the cyclophosphamide-prednisolone series received 100, 110, or 120 mg cyclophosphamide daily (depending on body weight) plus 20 mg prednisolone on alternate days. Treatment was continued indefinitely, or changed to the alternative regimen if the patient deteriorated, failed to improve, or developed drug toxicity. For response to treatment (as judged by change in breathlessness score, radiographic appearance, and lung function) patients were classified as improved, stable, or deteriorating. Deaths from cryptogenic fibrosing alveolitis were also analysed. Improvement had occurred at one or more assessments in seven of the 22 patients in the prednisolone only series and in five of the 21 patients in the cyclophosphamide-prednisolone series. At three years, however, only two of the 22 patients in the prednisolone only series were still improved and three stable, compared with one and seven of the 21 patients in the cyclophosphamide-prednisolone series (three of the seven had stopped treatment because of toxicity). Life table analysis suggested better survival in patients in the cyclophosphamide-prednisolone series but this was not significant. At three years 10 of 22 patients in the prednisolone only series had died compared with three of 21 patients in the cyclophosphamide-prednisolone series. With death or failure of first treatment regimen as outcome there was a significant advantage to the patients having cyclophosphamide-prednisolone. This advantage was explained in part by the better lung volumes in this group on admission. After allowance had been made for total lung capacity (TLC), no other factor was predictive of outcome. Analyses of subgroups according to TLC on admission showed that patients with a TLC below 60% predicted did badly and those with a TLC of 80% or more predicted did well with both regimens. Patients with an initial TLC of 60-79% predicted did better with the cyclophosphamide-prednisolone regimen. Side effects were uncommon in both series and those due to cyclophosphamide resolved when treatment was stopped. The combination of cyclophosphamide with prednisolone may be an alternative to prednisolone alone with an initial high dose phase. Many patients, however, failed to respond to either treatment.