We have studied the effect of nitrous oxide on bispectral index (BIS), calculated from a bipolar encephalogram. Inhalation of 70% nitrous oxide resulted in loss of consciousness in all healthy volunteers (n = 10) but no change in BIS. Brief inhalation up to 1.2% sevoflurane also resulted in loss of consciousness in volunteers (n = 5), but with sevoflurane, BIS decreased. BIS and the haemodynamic effects of adding nitrous oxide were also measured during coronary artery bypass surgery in patients (n = 10) receiving midazolam and fentanyl infusions. Measurements were made after 0%, 33%, 66% and 0% nitrous oxide, just before skin incision and after sternotomy. Nitrous oxide caused no change in BIS. BIS may indicate a sufficient hypnotic depth to prevent awareness during surgery, but our study demonstrated that pharmacological unconsciousness-hypnosis can also be reached by mechanisms to which BIS is not sensitive. Thus BIS is a sufficient but not a necessary criterion for adequate depth of anaesthesia or prevention of awareness.