Self-reported utilization of health care services is important in epidemiological studies and in health care planning, policy, and research, and the accuracy of such information is essential. This study assessed the validity of self-reported utilization of health care services in a randomly selected cohort of 500 community-dwelling men aged 40 to 79 years in Olmsted County, Minnesota. Men had previously completed a self-administered questionnaire that elicited information on utilization of health care services; questions included total inpatient hospital nights in the previous year, total ambulatory physician visits in the previous year, and physician visits in the previous 2 weeks. The complete community medical records of the men were reviewed, and the accuracy of self-reported information was evaluated by assessing the difference between self-reported utilization and actual medical record utilization as ascertained from the medical records, and the distribution of these differences. Exact agreement was 93% for inpatient hospital nights, 91% for ambulatory physician visits in 2 weeks, and 30% for ambulatory physician visits in the previous year. There was an increase in the variability of the difference between self-reported and medical record utilization with an increasing number of inpatient nights or ambulatory physician visits, and a significant bias toward underreporting with an increasing number of ambulatory physician visits in the previous year. These findings suggest that self-reported inpatient nights in the previous year and ambulatory physician visits in 2 weeks are reasonably accurate, but self-reported ambulatory physician visits in the previous year may be less accurate and likely to be biased toward underreporting at higher numbers of visits. This information should be taken into account by researchers and health planners using self-reported measures of utilization.