The diagnostic efficacy of two methods--demonstration of seroconversion in paired sera and detection of antigen in clinical specimens--was evaluated in 183 children with respiratory syncytial, parainfluenza or adenoviral, or pneumococcal respiratory tract infection. Viral infection was diagnosed in 46 (37%) of the 125 cases by antigen assay alone, in 36 (29%) by antibody assay alone and in 43 (34%) by both methods. In respiratory syncytial viral infections, antigen assays were more often positive than antibody assays; 80% of the cases were antigen positive and 63% solely antigen positive. In parainfluenza and adenoviral infections, antigen assays were not as useful; a positive result was seen in 59% and 44% of cases, respectively. Pneumococcal infection was diagnosed in 25 (30%) of the 84 cases by antigen assay alone, in 54 (64%) by antibody assay alone and in only 5 (6%) by both methods. Thus nearly all pneumococcal infections were diagnosed by only one method. The efficacy of antigen and antibody assay was clearly dependent on the age of the patients. In infants less than six months of age, nearly all (27 (90%)) of the 30 infections were diagnosed by antigen detection; 25 solely by antigen detection. In older patients, antigen and antibody assays supplemented each other. We conclude that antigen detection should be used as the primary method for the diagnosis of viral or pneumococcal respiratory tract infection. In infants, antigen detection is the only reliable method of microbial diagnosis. In addition, antigen detection is the method of rapid microbial diagnosis. In pneumococcal infections, the sensitivities of antigen detection methods are not sufficient; this is an important area for further research.