The rôle of an anaesthetic incident reporting programme in improving anaesthetic safety was studied. The programme had been running for 4 to 5 years in three large hospitals in Hong Kong and more than 1000 incidents have been reported. The number of reports being made and frequency of the various categories of incident reported, did not alter during the study period. Sixty nine percent of incidents were considered to be preventable. Human error contributed to 76% of incidents and violations of standard practice to 30% of incidents. The programme was effective in its ability to detect latent errors in the anaesthesia system and when these were corrected, incidents did not recur. The frequency with which various contributing factors were cited did not decrease with time. With the exception of problems dealt with by specific protocol development, the study found no evidence that an increasing awareness of the problem of human error was effective in reducing this kind of problem.