Conventional near infrared spectroscopy (NIRS), introduced by Jöbsis in 1977, can be considered as a reliable trend monitor for cerebral oxygenation. Quantisation, however, is complex and cumbersome. Recently a relatively simple system for cerebral oximetry (INVOS 3100, Somanetics Corporation, USA) was developed, measuring the regional oxygen saturation (rSO2) in the capillary bed of the cerebrum, presented as a numerical figure for easy interpretation. In this study a comparison was made between a conventional NIRS instrument and the new INVOS instrument, in order to obtain information about sensitivity and usefulness of the INVOS system. Changes in cerebral haemodynamics were induced by a moderate decrease of the arterial oxygen saturation (SaO2) and by varying the arterial carbon dioxide level (PaCO2). This will result in a higher (hypercapnia) or lower (hypocapnia) cerebral blood flow and subsequent change of both NIRS signals and INVOS signal. Healthy volunteers were used for this study. It was found that the steady state value for rSO2 was 70 +/- 6% (mean +/- SD). During the lowering of arterial saturation a poor correlation was found between rSO2 and SaO2 (r = 0.47). Increased cerebral blood flow induced by hypercapnia was detected by both conventional NIRS and the INVOS. Decreased cerebral blood flow induced by hypocapnia could only be detected by conventional NIRS. It was concluded that due to the variation in displayed rSO2 and the high amount of averaging in the algorithm the INVOS instrument does not yet provide more information than conventional NIRS.