Assessment of cerebral autoregulation is an important adjunct to measurement of cerebral blood flow for diagnosis, monitoring or prognosis of cerebrovascular disease. The most common approach tests the effects of changes in mean arterial blood pressure on cerebral blood flow, known as pressure autoregulation. A 'gold standard' for this purpose is not available and the literature shows considerable disparity of methods and criteria. This is understandable because cerebral autoregulation is more a concept rather than a physically measurable entity. Static methods utilize steady-state values to test for changes in cerebral blood flow (or velocity) when mean arterial pressure is changed significantly. This is usually achieved with the use of drugs, shifts in blood volume or by observing spontaneous changes. The long time interval between measurements is a particular concern in many of the studies reviewed. Parallel changes in other critical variables, such as pCO2, haematocrit, brain activation and sympathetic tone, are rarely controlled for. Proposed indices of static autoregulation are based on changes in cerebrovascular resistance, on parameters of the linear regression of flow/velocity versus pressure changes, or only on the absolute changes in flow. The limitations of studies which assess patient groups rather than individual cases are highlighted. Newer methods of dynamic assessment are based on transient changes in cerebral blood flow (or velocity) induced by the deflation of thigh cuffs, Valsalva manoeuvres, tilting and induced or spontaneous oscillations in mean arterial blood pressure. Dynamic testing overcomes several limitations of static methods but it is not clear whether the two approaches are interchangeable. Classification of autoregulation performance using dynamic methods has been based on mathematical modelling, coherent averaging, transfer function analysis, crosscorrelation function or impulse response analysis. More research on reproducibility and inter-method comparisons is urgently needed, particularly involving the assessment of pressure autoregulation in individuals rather than patient groups.