Records obtained for 578 (96 %) of the 602 patients certified as having died from tuberculosis in Hong Kong during a 1-year period were reviewed by an expert independent assessor in London. In his opinion 14 (2 % of the 578) had never had tuberculosis, and in 28 (5 %) it was not possible to decide whether or not the patient had ever had tuberculosis. Tuberculosis had been no more than a contributory factor in causing death in 44 (8 %), in 29 (5 %) it was irrelevant, and in 33 (6 %) it was not possible to determine what role it had played. The remaining 430 (74 %) patients were considered actually to have died from tuberculosis, 307 from active disease and 123 from the late effects of inactive disease. If all 578 patients had died from tuberculosis this would represent a death rate for the survey year of 13.1 per 100 000 of the population. The rate falls to 9.7 per 100 000 if based on the 430 patients considered by the assessor to have died from tuberculosis, a reduction of 26 %. Since the annual death rate from active disease (6.9 per 100 000 in this survey) is a better index of the current efficiency of an antituberculosis programme than the total death rate, it is suggested that efforts should be made to obtain and publish separate annual rates for deaths from active and from inactive tuberculosis. An analysis of potentially avoidable delays and failures in diagnosis showed that these were mainly due to the patient's delay in seeking or accepting advice, or inadequate investigation, particularly failure to examine the sputum, by the unofficial (non-governmental) medical services. The majority of failures in management were attributable in whole or in part to the patient.