Nuclear magnetic resonance (NMR) imaging shows promise in the measurement of human cerebral blood flow (CBF) in that nonradioactive indicators may be used. Our earlier investigations with trifluoromethane (FC-23) gas have shown that this compound can be used to safely and effectively measure CBF in anesthetized animal models. In this Phase I dose-escalation study we set out to determine the maximal tolerated concentration (MTC) of FC-23 in normal healthy male volunteers and to assess its feasibility as an NMR indicator. Five subjects were exposed in a blinded fashion to escalating concentrations of FC-23 between 10% and 60%, randomly interleaved with exposures to both room air and 40% nitrous oxide. On each study day, the subjects breathed the test gas for eight pulses of 3 min each with 2-min clearance periods between the pulses. The subjects underwent intensive physiologic and neurobehavioral monitoring throughout the study period. The first subject experienced an anesthetic response to 60% FC-23, and the second subject experienced "discomfort" and requested discontinuation at the initiation of 40% FC-23. The MTC was subsequently determined to be 30% FC-23 (all subjects tolerated the gas), although a small (37.6 vs. 40.5) but statistically significant retention of carbon dioxide was found (p = .003). When one subject received 30% FC-23 during an NMR imaging study, a pronounced anesthetic effect with intolerable hyperacusis was demonstrated. Human studies of FC-23 have been discontinued in our laboratory.