We evaluated the Vaccine Adverse Event Reporting System (VAERS), the spontaneous reporting system for vaccine-associated adverse events in the United States, as a public health surveillance system, using evaluation guidelines from the Centers for Disease Control and Prevention. We found that VAERS is simple for reporters to use, flexible by design and its data are available in a timely fashion. The predictive value positive for one severe event is known to be high, but for most events is unknown. The acceptability, sensitivity and representativeness of VAERS are unknown. The study of vaccine safety is complicated by underreporting, erroneous reporting, frequent multiple exposures and multiple outcomes.