In air pollution epidemiology, estimates of long term exposure are often based on measurements made at one fixed site monitor per area. This may lead to exposure misclassification. The present paper validates a questionnaire-based indicator of ambient air pollution levels and its applicability to assess their within-area variability. Within the framework of the SAPALDIA (Swiss Study on Air Pollution and Lung Diseases in Adults) cross-sectional study (1991), 9,651 participants reported their level of annoyance caused by air pollution on an 11-point scale. This subjective measure was compared with annual mean concentrations of particulate matter less than 10 microm in diameter (PM10) and nitrogen dioxide. The impact of individual factors on reported scores was evaluated. Nitrogen dioxide concentrations at home outdoors (measured in 1993), smoking, workplace dust exposure, and respiratory symptoms were found to be predictors of individual annoyance scores. Regression of population mean annoyance scores against annual mean PM10 and nitrogen dioxide concentrations (measured in 1993 and 1991, respectively) across areas showed a linear relation and strong correlations (r>0.85). Analysis within areas yielded consistent results. The observed associations between subjective and objective air pollution exposure estimates suggest that population mean scores, but not individual scores, may serve as a simple tool for grading air quality within areas. Reported annoyance due to air pollution should be considered an indicator for a complex environmental condition and thus might be used for evaluating the implementation of environmental policies.