Bronchoalveolar lavage (BAL) cell counts are used to assess 'alveolitis' in patients with interstitial lung diseases (ILD) but inflammatory cells from airways can contribute to the differential cell count. To determine what BAL volume samples airway cells in patients with ILD we measured the proportion of bronchial epithelial cells (BECs) in four successive 25 ml aliquots in a single lung subsegment in 23 patients with ILD (cryptogenic fibrosing alveolitis (CFA) four, rheumatoid lung (RL) three, asbestosis (ASB) 11, sarcoidosis (SARC) five). Cells recovered from the first two 25 ml lavages exhibited higher proportions of BECs (15 +/- 4% and 9 +/- 2% respectively) than those from the remaining two aliquots (3 +/- 1%, 3 +/- 1%, each P less than 0.01), suggesting that the first 50 ml BAL preferentially samples airway cells compared to the second 50 ml BAL. To evaluate airway and alveolar inflammatory cell proportions in ILD we performed two separate 50 ml BALs (samples I and II) in a single subsegment in 38 patients with ILD (CFA seven, RL five, ASB 19, SARC seven) and measured the proportions of recovered cells in each sample separately and combined. Seven control individuals were also studied. Sample I contained 1-67% (mean 26 +/- 3%) of the total recovered cells. Neutrophil (PMN) proportions were higher in sample I compared to sample II in CFA (20 +/- 6 vs 8 +/- 2%), RL (30 +/- 9 vs 8 +/- 2%) and ASB (12 +/- 2 vs 7 +/- 1%), P less than 0.05 for each, but were similar in samples I and II in patients with SARC (3 +/- 1 vs 2 +/- 1%) and controls (2 +/- 1 vs 2 +/- 1%). In combined samples (I + II), absolute PMN proportions were up to 8% higher than in sample II alone whereas absolute lymphocyte proportions were up to 8% less than in sample II alone. These data suggest that separate processing of the fluid recovered from the first 50 ml BAL in ILD patients provides information on the location of inflammatory cells and improves the accuracy of BAL cell counts.