Abstract‘Audit’ is now in widespread use in our National Health Service, but there is little documentation of improvement resulting from audit. If it is to be used to identify mistakes in clinical practice, we must be prepared to admit them, or to have our work evaluated by others; such openness is uncommon. Understandably, doctors are concerned about possible litigation, under‐ mining of their authority, and/or interference with their methods of practice. Furthermore, few are willing to comment adversely on the work of others.There is confusion about the best method(s) of audit. Outcome and process audit serve different purposes. Outcome audit measures the effect of care on patients' health, but in individual cases a poor outcome may result after impeccable medical care and a good outcome after poor care. Outcome audit requires a large patient group, is costly, and is of value for a limited number of conditions. Appropriate standards are needed to judge the results; it may be fallacious to compare outcomes in different settings, and with different patient groups. Furthermore, outcome audit is of little value for auditing the care given by individual doctors, and this limits its value in clinical education.Process audit deals with the appropriateness of clinical actions, on the assumption that they affect outcome. It can detect poor performance when outcome audit would be unlikely to identify poor outcome. For common or well‐defined problems, process audit can make use of clear criteria, for example agreed protocols, and this may have immediate benefit for individual patients. When there is no agreed protocol the overall quality of care can still be audited against relatively explicit criteria, if there is agreement on the relevant ‘principles of clinical practice’, i.e. the rules which should guide the clinical management of individual patients. Clearly these should cover the collection, recording and analysis of patient data; planning for diagnosis, monitoring, treatment and patient education, and steps to be taken when there is uncertainty about the best course of action. Performance in these areas can be assessed by reviewing the patients' notes, but most clinical records are inadequate for this purpose.Process audit can be applied to all clinical problems, and is the method of choice for assessing the actions of individuals. Its educational value is self‐ evident. It allows the identification of deficiencies, and the provision of feedback to correct them. Furthermore, it provides for a continuum of audit through the undergraduate and postgraduate years of a doct