Protected specimen brushing (PSB), combined with quantitative culture, is now recognized as one of the reference methods for diagnosis of nosocomial pneumonia. However, no criteria exist with which to assess the quality of the PSB sample. We studied numbers of inflammatory cells and bronchial cells per microscopic field (magnification: x500, objective x50) in cytospin preparations of PSB samples. Results of cell count and quantitative culture in a first study period were compared with those in a second study period, following adaptation of the PSB technique and collection of samples from more peripheral sites. The cellular content of samples from patients and controls was investigated. We examined 86 samples from patients with suspected nosocomial pneumonia and 15 samples from uninfected controls. The number of samples with a high cellular content was considerably greater in the second study period. No positive cultures were obtained from samples containing < 10 cells per field. The numbers of cells in samples from uninfected controls were comparable to the numbers in samples from patients. Our results indicate that absence of cells probably represents inadequate sampling. Negative PSB cultures with cytospin preparations containing < 10 cells per microscopic field should therefore be considered with caution, and resampling considered.