Objective: To develop a taxonomy describing patient safety events in general practice from reports submitted by a random representative sample of general practitioners (GPs), and to determine proportions of reported event types. Design: 433 reports received by the Threats to Australian Patient Safety (TAPS) study were analysed by three investigating GPs, classifying event types contained. Agreement between investigators was recorded as the taxonomy developed. Setting and participants: 84 volunteers from a random sample of 320 GPs, previously shown to be representative of 4666 GPs in New South Wales, Australia. Main outcome measures: Taxonomy, agreement of investigators coding, proportions of error types. Results: A three-level taxonomy resulted. At the first level, errors relating to the processes of healthcare (type 1; n = 365 (69.5%)) were more common than those relating to deficiencies in the knowledge and skills of health professionals (type 2; n = 160 (30.5%)). At the second level, five type 1 themes were identified: healthcare systems (n = 112 (21.3%)); investigations (n = 65 (12.4%)); medications (n = 107 (20.4%)); other treatments (n = 13 (2.5%)); and communication (n = 68 (12.9%)). Two type 2 themes were identified: diagnosis (n = 62 (11.8%)) and management (n = 98 (18.7%)). The third level comprised 35 descriptors of the themes. Good inter-coder agreement was demonstrated with an overall κ score of 0.66. A least two out of three investigators independently agreed on event classification in 92% of cases. Conclusions: The proposed taxonomy for reported events in general practice provides a comprehensible tool for clinicians describing threats to patient safety, and could be built into reporting systems to remove difficulties arising from coder interpretation of events.