Postoperative chest infection was studied prospectively in 73 patients in order to evaluate standard laboratory methods of sputum examination and to relate the results to the patients' clinical state and to antibiotic therapy. When a culture medium selective for haemophilus was used in addition to unselective media, homogenisation of the specimen gave no advantage. Laboratory and clinical findings usually corresponded well. Profuse growths of Streptococcus pneumoniae or Haemophilus influenzae were clearly associated with clinical evidence of chest infection but other Gramnegative bacilli and Staphylococcus aureus much less so. Coliforms were more prominent after antibiotic therapy.