Lay diagnoses of death collected at burial sites were validated against two 'gold standards': the hospital discharge diagnosis of causes of death obtained by a surveillance of hospital deaths (including autopsy results) and the physician review of verbal autopsies (VAs) that were carried out for a sample of cemetery records. The diagnostic indicators of the lay diagnoses were then used to provide estimates of the share of AIDS-attribuTable mortality. The verbal autopsy results provide an independent estimate of the percentage of AIDS deaths. From a total of 21,274 burial records, 2546 hospital discharge diagnoses, 1480 outcomes of autopsies and 200 adult verbal autopsies were gathered over a period of 1 year starting from February 2001. Independent of the gold standard, lay diagnoses such as lung disease and cold have a specificity of about 90% and a combined sensitivity of about 55% in determining AIDS mortality. Without a significant loss in specificity, the sensitivity increases to 60-65% when diarrhoea, TB, herpes zoster and mental or nerve problem are included. We thus conclude that even in the presence of a reluctance to talk of HIV/AIDS, lay diagnosis of causes of death can be used for monitoring AIDS mortality. Lung disease and cold, in particular, have become well-known euphemisms for AIDS in the community. The share of AIDS deaths in the adult population (20-54) is estimated at 68%, without noticeable differences between men and women. Our results confirm the high impact of HIV/AIDS on mortality as was estimated by epidemiological projections for Addis Ababa.