This paper describes a method of detecting intracranial aneurysm and arteriovenous malformation (AVM) by analysing weak sounds produced by the blood circulation at the affected part. There is thought to be no turbulence in the normal cerebrovascular system, whereas abnormalities such as aneurysm and AVM sometimes cause turbulence in the blood flow. Thus, a small fraction of the flow energy might be converted into an acoustically detectable noise. For the detection of the sound, sensitive detectors must be applied close to the head since the sound is very weak, and, as in cerebrovascular diseases, the origin of the sound is usually concealed deep inside the hard shelter of the skull. The detection system we used had a gain of 40 to 50 dB greater than that of an ordinary stethoscope. The detection points were the teeth or forehead. Usually the sound started about 160 msec after the ventricular contraction and lasted for 100 to 400 msec. Its frequency component mostly ranged from 400 to 2000 Hz, but the spectrum profile changed according to the position and degree of abnormalities. The uttered sound was very random, which facilitated detection of the position of the sound origin by means of cross correlation methods using a pair of detectors. This method is completely noninvasive, causes no pain to the patient, and might be used even in mass examinations.