OBJECTIVE: To study the effect of an educational intervention on general practitioners' (GPs') ability to diagnose bronchial obstruction after clinical examination. DESIGN: Based on physical chest examination 11 GPs assessed the degree of bronchial obstruction by estimating the patient's predicted forced expiratory volume in one second (FEV1%). Half way in the study the GPs were taught new knowledge on associations between lung sounds and bronchial airflow. The agreements between estimated and measured FEV1% predicted before and after this educational intervention were compared. SETTING: 11 GPs in five health centres in northern Norway took part. PATIENTS: 351 adult patients were included in phase 1, and 341 in phase 2. MAIN OUTCOME MEASURES: Estimated and measured FEV1% predicted were compared in subgroups of patients according to clinical findings in phase 1 and 2. The effect of the intervention on the doctors' weighting of various chest signs could thus be evaluated. Kappa agreement and correlation between estimated and measured FEV1% predicted in both phases were determined. RESULTS: The agreement between estimated and measured FEV1% predicted increased from Kw (weighted Kappa) = 0.33 in phase 1 to Kw = 0.43 in phase 2 (95% confidence interval 0.35-0.52). The GPs laid more relevant emphasis on rhonchi in their estimates after the educational intervention, while too much weight was laid on uncertain chest findings in phase 2. CONCLUSION: The study shows a potential for better use of physical chest examination in the diagnosis of bronchial obstruction.