Background. Cerebral state index (CSI) has recently been introduced as an intra-operative monitor of anaesthetic depth. We compared the performance of the CSI to the bispectral index (BIS) in measuring depth of anaesthesia during target-controlled infusion (TCI) of propofol. Methods. Twenty Chinese patients undergoing general anaesthesia were recruited. CSI and BIS, and predicted effect-site concentration of propofol were recorded. The level of sedation was tested by Modified Observer's Assessment of Alertness/Sedation Scale (MOAAS) every 20s during stepwise increase (TCI, 0.5ugml-1) of propofol. The loss of verbal contact (LVC) and loss of response (LOR) were defined by MOAAS values of 2-3 and less than 2, respectively. Baseline variability and the prediction probability (PK) were calculated for the BIS and CSI. The values of BIS05 and CSI05, BIS50 and CSI50, BIS95 and CSI95 were calculated at each end-point (LVC and LOR). Results. Baseline variability of CSI was more than that of BIS. Both CSI and BIS showed a high prediction probability for the steps awake vs LVC, awake vs LOR, and LVC vs LOR, and good correlations with MOAAS values. Conclusion. Despite larger baseline variation, CSI performed as well as BIS in terms of PK values and correlations with step changes in sedation.