Insulin resistance is common in the general population and tends to cluster with glucose intolerance, dyslipidaemia and high blood pressure. The importance of the insulin-resistant phenotype for the assessment of cardiovascular risk and response to intervention is increasingly being recognized. Therefore, there is a need for an accurate and reproducible method for measuring insulin resistance in vivo. The euglycaemic insulin clamp is currently the best available standard technique. It provides steady-state measures of insulin action and is easily combined with a number of other investigative methods (tracer dilution, limb catheterization, indirect calorimetry, positron emission tomography and nuclear magnetic resonance scans). Whereas homeostatic model assessment uses fasting plasma glucose and insulin concentrations to derive indices of insulin sensitivity and secretion from a mathematical model, other techniques are based on the exogenous infusion of glucose or insulin, or both, either under steady-state (the insulin suppression test) or under dynamic conditions (insulin tolerance test, intravenous glucose-tolerance test with minimal model analysis, and constant infusion of glucose with model assessment). This article recalls the principles of insulin action, with special reference to the concept of clearance and the equivalence of different approaches to estimating this function. Merits and disadvantages of the various techniques are then concisely reviewed, with emphasis on their relative feasibilities and reliabilities. Recent developments and future trends are mentioned. Criteria for choice and some reference data are given to aid the clinical investigator.