A high incidence of intraoperative awareness during cardiac surgery has been reported. Midlatency auditory evoked potentials (MLAEP) have been used recently as an indicator of awareness. In the current study, memory for information presented during anesthesia was investigated using MLAEP as one experimental indicator in 45 patients scheduled for elective cardiac surgery. In all patients general anesthesia was maintained using high-dosage fentanyl (1.2 mg · h−1). In addition, the patients of group 1 (n = 10) received flunitrazepam (1.2 mg · h−1), the patients of group 2 (n = 10) isoflurane (0.6–1.2 vol%), and the patients of group 3 (n = 10) propofol (4–8 mg · kg−1. h−1). Group 4 (n = 15) served as a control, and those patients were assigned randomly to one of the anesthetic regimes. After sternotomy and before cardiopulmonary bypass, an audiotape, which included an implicit memory task, was presented to the patients of groups 1–3. Auditory evoked potentials were recorded while the patients were awake and during general anesthesia immediately before and after the audiotape presentation. Latencies of the brain stem peak V and the early cortical potentials Na and Pa were measured. Three to 5 days postoperatively no patient had a clear explicit memory of intraoperative events. However, there were statistically significant differences in the incidence of implicit recall among the groups. Five patients in the flunitrazepam-fentanyl group, 1 patient in the isoflurane-fentanyl group, 1 patient in the propofol-fentanyl group, and no patient in the control group showed an implicit memory of the intraoperative tape message. In the awake state, MLAEP showed high peak-to-peak amplitudes and a periodic waveform. In the patients with implicit memory postoperatively, MLAEP continued to show this pattern during general anesthesia. The early cortical potentials Na and Pa did not increase in latency or decrease in amplitude before or after the audiotape presentation. In contrast, in the patients without implicit memory, MLAEP waveform was severely attenuated or abolished. Na and Pa showed marked increases in latencies and decreases in amplitudes or were completely suppressed. In 9 patients, including all patients (7 of 9) with implicit memory, Pa latency increased less than 12 ms, and 21 of 23 patients without implicit memory showed a Pa latency increase of greater than 12 ms during anesthesia and the audiotape presentation. Therefore, the Pa latency increase of greater or less than 12 ms may provide sensitivity of 100% and specificity of 77% in distinguishing patients with implicit memory from patients without implicit memory postoperatively. When the early cortical potentials of MLAEP are preserved during general anesthesia, auditory information may be processed and remembered postoperatively by an implicit memory task.