We studied memory of health scenarios for end-of-life support decisions and stability of life support preferences. Psychology students (n = 36, age M = 27.25, SD = 6.21, 64% females) were administered the Life Support Preferences Questionnaire to assess their memory of six medical scenarios with different prognoses, care treatment, and end-of-life support choices. Recall, recognition, confidence and decision stability were assessed immediately and one month later. Correct recall decreased and incorrect recall increased from immediate to long-term recall, F(2, 68) = 74.38, p < .001, η2p = .69. In recall, participants spontaneously gave false information consistent with prior knowledge of illnesses and medical scenarios. Participants who had suffered a disease or serious accident did worse on correct recall, F(1, 34) = 6.59, p = .015, η2p = .16, and had more errors, F(1, 34) = 4.68, p = .038, η2p = .12, than participants who had not. In the recognition test there were no differences between hits and false alarms, showing the difficulty in discriminating between true and false contents. Confidence was greater for hits than for false alarms, F(1, 34) = 10.86, p = .002, η2p = .24, but this subjective measure did not seem to be a good predictor of accuracy because confidence was greater than the mean value for hits (p = .001, d = 1.74) and for false alarms (p = .001, d = 0.96). Long-term memory was quite poor and biased, but life support preferences did not change much.