Use of transcranial Doppler (TCD) to diagnose vasospasm has been criticised. We examined reproducibility of TCD middle cerebral artery (MCA) velocity measurements. Thirty-six healthy adult volunteers were recruited. Four operators, two experienced and two inexperienced, participated. MCA velocity was measured twice by one operator and once by a second operator. Mean (95% limits of agreement) interoperator agreement was 2.4(+/-36.7) cm/s. Experienced vs. inexperienced, inexperienced vs. inexperienced, and experienced vs. experienced operators were -2.8(+/-39.3), -5.6(+/-40.1), 1.8(+/-22.1) cm/s, respectively. Intraoperator agreement across all operators, experienced and inexperienced were -0.5(+/-16.9), -1.6(+/-19.3), 0.7(+/-13.7) cm/s, respectively. Interoperator limits of agreement for experienced operators were almost half that of inexperienced operators. Intraoperator reproducibility was much better, regardless of level of experience, but aberrant results did occur even with experienced operators. If TCD measurements are used to guide management it is essential that operators are adequately trained, and readings repeated before potentially harmful treatments are instituted.