A simple non-invasive method to improve the reliability of bacteriological analyses of sputum was evaluated in 25 patients with chronic bronchitis. This method consists of placing dental cotton swabs between the cheek and the gum and under the tongue at the level of the excretory salivary glands. Once these plugs are in place, the sputum is collected by coughing under the control of a physiotherapist. The quantitative bacteriological data from sputum collected by the "protected" and "unprotected" methods were compared with those from bronchial secretions collected by transtracheal aspiration, used as reference method. Oropharyngeal cells and commensal bacteria were isolated at a significantly lower frequency in sputum collected by the protected method. Considering the predominant pathogenic bacteria, an excellent correlation was obtained between transtracheal aspiration and sputum in 80% of cases when sputum was collected by protected method and in 68% of cases by unprotected method. In four patients, the transtracheal aspiration was sterile, while a potentially pathogenic organism associated with a high leukocyte numeration was identified in sputum. All the potential pathogens isolated in the positive transtracheal aspiration were recovered in the protected sputum (100%), but they could be associated with a non-predominant commensal flora (11 cases) or a potentially misleading pathogenic bacteria (2 cases).