We conducted a nested case-control study utilizing cases of clinical pneumonia identified in a community-based prospective surveillance study of children under 3 years of age in order to test the Validity of a survey questionnaire. Three types of sex- and age-matched concurrent controls were selected from the surveillance population: acute respiratory infection (ARI) clinic controls, ARI community controls and healthy community controls. Survey interviews were scheduled at random for any of four consecutive 7-day periods after the diagnosis of the case. The questionnaire covered a 30-day recall period. The combination of cough with fast breathing or shortness of breath, and with fever, provided the highest positive predictive value for pneumonia. The sensitivity of some questions dropped when the interview took place more than 15 days after the diagnosis of the case. However, the utilization of a 15-day recall period did not increase the positive predictive value of the survey. We conclude that in this trained population under surveillance, a survey questionnaire utilizing a 30-day recall period and using the combination of cough, fast breathing or shortness of breath and fever to define episodes with a high likelihood of pneumonia, offers an acceptable tool for the monitoring and evaluation of respiratory control programmes. This questionnaire needs further evaluation in an untrained population and in other regions before it can be adopted for use in ARI control programmes. © 1994 International Epidemiological Association.