Although serial transcranial Doppler measurements of blood flow velocity are of considerable clinical utility, their use assumes that the velocity signals are stable and unchanging during short time periods. Contrary to this assumption, the authors found significant variations in mean velocity signals in both normal subjects and patients, which may confound the interpretation of serial studies. Signals were continuously obtained for 5 to 10 minutes from the middle cerebral artery of 11 normal subjects and 18 patients (22 studies) with a variety of neurosurgical disorders. The average difference between the peak and the trough in the waves observed in the normal population was 11 +/- 4% (standard error of mean), and 5 of the 11 had at least one wave with a difference of more than 20%. The average difference between the peak and trough signals in the neurosurgical population was 14 +/- 13% (SEM) and 13 of the 22 studies showed at least one wave with a difference of more than 20%. These variations were consistently seen and may be related to similar variations in blood pressure or intracranial pressure waves. Whatever the origin, these variations should be recognized during the interpretation of transcranial Doppler signal in clinical practice.