The methods of nonbronchoscopic lung lavage used for collection of samples of epithelial lining fluid (ELF) in intubated patients are poorly standardized and incompletely validated. In infants with lung disease requiring ventilatory support, we evaluated two techniques of small volume saline ravage for the collection of a specimen suitable for pulmonary surfactant analysis. We aimed to compare apparent origin of the return fluid obtained by each method, equivalence and agreement of the estimates of measured pulmonary surfactant concentration, and the relative strength of association between surfactant indices and lung dysfunction. Fifty-three contemporaneous paired samples of lung lavage fluid suitable for surfactant analysis were collected from 31 infants using tracheal aspirate (TA, 4 x 0.5 ml saline), and then nonbronchoscopic bronchoalveolar lavage (NB-BAL, 3 x 1 ml/kg). Return fluid from TA had higher mean ELF concentration of total protein and IgA secretory component (SC), and a lower surfactant protein A (SP-A) concentration than NB-BAL, indicating that the TA lavage was sampling ELF more proximally in the tracheobronchial tree (protein: TA 7.7 versus NB-BAL 4.7 mg/ml; SC: 21 versus 1.8 mu g/ml; SP-A: 9.8 versus 19 mu g/ml; all p < 0.01). Mean concentration of surfactant indices in ELF differed only for SP-A, but for all indices, paired values showed poor agreement on Bland-Altman analysis, highlighting the potential imprecision associated with small volume lung ravage. TA return fluid yielded estimates of surfactant indices which were at least equivalent to NB-BAL in prediction of the severity of lung dysfunction. We conclude that NB-BAL return fluid has more distal origin, but analysis of TA fluid may have equal validity in the estimation of indices of pulmonary surfactant. The results of individual estimates of ELF constituents in a single sample of lavage fluid should be interpreted with caution, even when standardized sampling techniques are employed.