Following brain retraction with graded local compression pressure (LCP), somatosensory evoked potential (SEP) to median nerve stimulation and cortical blood flow (CoBF) by a hydrogen clearance method were studied at the primary somatosensory area in mongrel dogs. LCP was raised by 10 mmHg every 30 minutes. As the first step, a basic experiment of the hydrogen clearance method was performed, and its reliability and usefulness were confirmed. Studies included changes of SEP, CoBF and histological examination following reduction of cerebral perfusion pressure (CPP). According to the different methods of producing reduction of CPP, dogs could be grouped into three. Group 1 consisted of 9 dogs in which CPP was reduced by exsanguination. Group 2 included 11 dogs which had normal saline infused into the cisterns magna. Group 3 was composed of 11 dogs which had local brain compression. In the exsanguination group, there was a threshold type relationship between N(1) amplitude of SEP and CoBF. N(1) amplitude started to be suppressed when CPP decreased to 40-50 mmHg (CoBF decreased by 35 % ), and was suppressed by 50 % when CPP became 30-40 mmHg, and it was finally abolished when CPP became 20-30 mmHg (CoBF decreased by 55 % ). There were no abnormal histological findings. This suggests that neuronal dysfunction is dependent only on reduction of cerebral blood flow. In the cisterns magna infusion group, changes of N(1) amplitude were quite different from those of the exsanguination group. N(1) amplitude started to decrease as cisterna magna pressure (CMP) increased despite CoBF remaining unchanged. N(1) amplitude had decreased by 50 % when CMP arose to 70 mmHg (CPP, 70-80 mmHg), and flattened when CMP was raised to over 110 mmHg (CCP, below 30 mmHg). It is assumed that neuronal dysfunction relied on not only CoBF reduction but also compression of nerve fibers secondary to dilation of ventricles and subarachnoid space. Local compression decreased both N(1) amplitude and CoBF more markedly than the other two groups. N(1) amplitude decreased as soon as local compression was added, was suppressed by 50 % at 50 mmHg of LCP (CPP, 90-100 mmHg), and flattened when LCP was elevated over 90 mmHg (CPP, below 60 mmHg). This suggests that the immediate, severe occurrence of neuronal dysfunction was due to both disturbance of cerebral circulation and destruction of neuronal architectures by brain distortion following mechanical injury. This study supports the concept that brain spatulae should be applied with caution during brain retraction in neurosurgical procedures because both cerebral circulation and neuronal function are easily disturbed.