Human error causes or contributes to considerably over half of all aviation mishaps. This report describes a 7-year study of aircraft incident data conducted in an attempt to further our understanding of the phenomenon of human error. The study of incidents as a surrogate for aircraft accidents is relevant only if incidents constitute a population or universe of which accidents are a subset. This assumption has been examined in a study of over 35,000 reports of aviation incidents collected from 1976-83 by the NASA Aviation Safety Reporting System. One-third of the reports involve conflicts among aircraft. The most common single-aircraft anomalies in flight involve altitude or track deviations. The most common controller errors involve failure to coordinate traffic with other elements of the air traffic control system. Analysis of these reports indicates that both human and system factors contributing to human errors can be identified. Many other incidents involve shortcomings specifically of the human, rather than of the system. Failures of control are rare in this series, but failures of decision-making and cockpit resource management are frequent. Boredom, complacency and ennui appear to underlie some failures, while very high workloads are associated with others. These data indicate that at least several categories of aircraft accidents involving operational and human factors are, in fact, subsets of populations of incidents containing the same elements. The environment in which an incident occurs is extremely important in determining its outcome. It is concluded that aviation incident reports are a necessary and important instrument in safety surveillance.