The validity of the Verbal Autopsy (VA) in death due to acute respiratory infection (ARI), was tested in 36 children who died by any acute infectious disease as stated by the necropsy diagnosis, at two public hospitals in Mexico City; the illness started at home. Clinical data obtained through VA were compared with diagnoses of necropsies, which were considered as "gold standard". The presence of dyspnoea for more than one day showed sensitivity of 0.69 and specificity of 0.74, while history of coughing showed a sensitivity of 0.61 and a specificity of 0.73. Combination of both clinical data improved specificity (0.83), but decreased sensitivity (0.54). Additional sources of diagnosis (a panel of assessors, the clinical record and the death certificate), also showed good sensitivity (0.69-0.77) and specificity (0.74-7.8). Focus on history of dyspnea and/or cough in children with an infectious syndrome should be emphasized, as a useful epidemiologic tool to determine children's mortality due to ARI in areas where diagnosis resources are constrained.