Since its development more than twenty years ago, non-invasive near-infrared-spectroscopy (NIRS) has been widely used to monitor cerebral oxygenation. Despite of its growing number of users, the diagnostic value of near-infrared spectroscopy still remains unclear, especially in case of acute brain injury and long-term neuromonitoring, necessary during intensive care therapy. To evaluate quality and sensitivity of NIRS measurements compared to invasive ICP-, CPP- and regional brain tissue--pO2 (p(ti)O2) monitoring, 31 patients, suffering from severe brain injury due to subarachnoid hemorrhage or severe head injury, were studied. NIRS measurements were only possible in 80% (using the INVOS oximeter) and in 46% (using the CRITIKON monitor), while good data quality was obtained in 100% from ICP, CPP and p(ti)O2. Major reasons for the failure of NIRS measurements were: (1) a wet chamber between sensor and skin, (2) galea hematoma or (3) subdural air after craniotomy. Different tests were performed to compare the sensitivity of regular oxygen saturation (NIRS) with the sensitivity of invasively determined p(ti)O2. Only induced hyperoxia (FiO2 = 1.0) revealed a significant correlation between both parameters (r = 0.67, p < 0.01). Lower or no correlation was found after changing paCO2 and administration of mannitol. The high failure rate and the limited sensitivity does not make the clinical use of near-infrared spectroscopy suitable as a part of neuromonitoring after acute brain injury at the present time.