We recruited 985 patients with COPD but without hypoxemia or other serious disease, treated them in a standard fashion, and followed them closely for nearly 3 yr. At the time of recruitment the patients were carefully characterized as to symptom severity, lung function, exercise tolerance, and quality of life, and studies of lung function were repeated during follow-up. Overall mortality was 23% in 3 yr of follow-up. Patient age and the initial value of the FEV1 were the most accurate predictors of death; when FEV1 before bronchodilator was used, the response to bronchodilators was directly related to survival, but this relationship became nonsignificant when postbronchodilator FEV1 was used as a primary predictor. After adjustment for age and FEV1, mortality was related positively to TLC, resting heart rate, and perceived physical disability, and related negatively to exercise tolerance. These relationships, though significant, were relatively weak. When standardized for age and FEV1, mortality in the present series was less than that of a previous series (4), and the same as that of hypoxemic patients with COPD who received continuous home O2 therapy. Changes in FEV1 with time averaged -44 ml/yr, but the standard deviation was large. Patients with low initial values of FEV1 showed relatively little further decline, probably indicating a survivor effect. In patients with well-preserved initial FEV1, rate of decline correlated negatively with bronchodilator response, symptomatic wheezing, and psychological disturbances.