OBJECTIVE: This review focuses on how patients' recall of their stay in the ICU can be modified pharmacologically. DATA SOURCES: Computerized MEDLINE and PAPERCHASE searches of English- and foreign-language published research from 1966 to 1995, bibliographies, pharmaceutical and personal files, and conference abstract reports. STUDY SELECTION: All abstracts from uncontrolled and controlled clinical trials were reviewed. DATA EXTRACTION: Study design, population, results, and safety information were retained. Efficacy conclusions were drawn from controlled trials. DATA SYNTHESIS: Patients without cerebral injury may recall mental and physical discomfort during their stay in the ICU. All benzodiazepines produce amnestic effects, but the short duration of action, lack of long-acting metabolites, and potent amnestic effects make lorazepam and midazolam preferable in this setting. Infusions of propofol for conscious sedation produce concentrations below those required for consistent amnesia. Opioids generally do not produce amnesia; however, end-organ failure and use of high doses of opioids may increase plasma concentrations to levels that produce impairment of learning and various degrees of amnesia. High infusion rates of ketamine may be required for satisfactory amnesia and pain control (with coadministration of benzodiazepine). Barbiturates and haloperidol do not impair memory in patients who are not critically ill. Antihistamines and anticholinergics that do not penetrate the central nervous system do not produce amnesia. Flumazenil may induce recall. CONCLUSIONS: Patients may remember their stay in the ICU, depending on the type of injury and the drug therapy. Of the drugs presented, benzodiazepines most reliably provide anterograde amnesia, whereas ketamine and propofol exhibit dose-dependent effects on memory.