SummaryEighty-three patients were given midazolam 0. I mg.kg-I by intramuscular injection as premedication before general anaesthesia with alfentanil-nitrous oxide. During anaesthesia patients were presented (through headphones) with either statements about common facts of some years ago (group A ) (n = 43) or new verbal associations, e.g. names offictitious, nonfamous people (group B) (n = 40). In a previous study with the same anaesthetic technique, but without premedication there was significant activation of implicit memory @ < 0.001). In this study we found no explicit or implicit memory for the auditory information presented during anaesthesia. Midazolam premedication can prevent implicit memory activation during alfentanil-nitrous oxide anaesthesia.