Unreliable delivery of best practice care is a major component of medical error. Critically ill patients are particularly susceptible to error and unreliable care. Human factors analysis, widely used in industry, provides insights into how interactions between organizations, tasks, and the individual worker impact on human behaviour and affect systems reliability. We adopt a human factors approach to examine determinants of clinical reliability in the management of critically ill patients. We conducted a narrative review based on a Medline search (1950-March 2010) combining intensive/critical care (units) with medical errors, patient safety, or delivery of healthcare; keyword and Internet search 'human factors' or 'ergonomics'. Critical illness represents a high-risk, complex system spanning speciality and geographical boundaries. Substantial opportunities exist for improving the safety and reliability of care of critically ill patients at the level of the task, the individual healthcare provider, and the organization or system. Task standardization (best practice guidelines) and simplification (bundling or checklists) should be implemented where scientific evidence is strong, or adopted subject to further research ('dynamic standardization'). Technical interventions should be embedded in everyday practice by the adjunctive use of non-technical (behavioural) interventions. These include executive 'adoption' of clinical areas, systematic methods for identifying hazards and reflective learning from error, and a range of techniques for improving teamworking and communication. Human factors analysis provides a useful framework for understanding and rectifying the causes of error and unreliability, particularly in complex systems such as critical care.