During a 7-year period bronchoalveolar lavage (BAL) was performed as a routine diagnostic procedure at fiberoptic bronchoscopy in 172 consecutive patients with diffuse pulmonary lesions. In 42 patients, BAL was technically insufficient or data were incomplete. These patients were excluded. The remaining 130 patients consisted of 78 men and 52 women with a median age of 43 years (range 19-79); 59 were smokers. They were divided into 6 groups: I. sarcoidosis (n = 77); II. cryptogenic fibrosing alveolitis (n = 16); III. secondary fibrosing alveolitis (n = 7); IV. malignancy (n = 7); V. allergic alveolitis (n = 6); VI. miscellaneous (n = 17). Group VI was not included in the statistical evaluation, which involved only the 113 patients in groups I-V. BAL was performed in a segment of the right middle lobe with 150-200 ml isotonic saline. The return fluid (BALF) was filtered through two layers of cotton gauze, and total and differential cell counts were assessed. Median BALF return volume was 67% (range 35-90). Eighty percent of the procedures were performed by the same operator. Total cell count displayed no significant difference amongst the five diagnostic groups (p = 0.06). Differential cell count displayed differences amongst the groups respecting macrophages (p = 0.002), lymphocytes (p = 0.0004), neutrophils (p = 0.0001) and eosinophils (p = 0.04). Patients with sarcoidosis had a higher percentage of lymphocytes, patients with secondary fibrosis a higher percentage of neutrophils, and patients with cryptogenic fibrosis a higher percentage of eosinophils than the other groups. Malignant cells were observed in BALF in 14.3% of patients with malignant lesions. Among the patients with sarcoidosis, 75% had a lymphocyte-dominated BALF (> 10%) compared with 31% of the patients with cryptogenic fibrosis, 14% of the patients with secondary fibrosis, and 43% of the patients with malignancy. Dominance of neutrophils (> 10%) and/or eosinophils (> 5%) in BALF was observed in cryptogenic and secondary fibrosis. In most patients, BAL cannot provide a definite diagnosis, but may support the clinical suspicion of a specific diagnosis. In clinical practice, BAL seems to be of limited value in the diagnostic evaluation of radiologically detected diffuse, non-infectious pulmonary lesions.