Nonanesthetic gases or vapors do not abolish movement in response to noxious stimuli despite partial pressures and affinities for lipids that would, according to the Meyer-Overton hypothesis, predict such abolition. We investigated whether nonanesthetics depress learning and memory (i.e., provide amnesia). To define learning, we used a "fear-potentiated startle paradigm": rats trained to associate light with a noxious stimulus (footshock) will startle more, as measured by an accelerometer, when a startle-eliciting stimulus (e.g., a noise) is paired with light than when the startle-eliciting stimulus is presented alone. We imposed light-shock pairings on 98 rats under three conditions: no anesthesia (control); 0.20, 0.29, and 0.38 times the minimum alveolar anesthetic concentration (MAC) of desflurane; or two nonanesthetics (1,2-dichloroperfluorocyclobutane and perfluoropentane) at partial pressures predicted from their lipid solubilities to be between 0.2 and 1 MAC. Desflurane produced a dose-related depression of learning with abolition of learning at 0.28 MAC. Perfluoropentane at 0.2-predicted MAC had the same effect as 0.28 MAC desflurane. 1,2-Dichloroperfluorocyclobutane at 0.5- to 1-predicted MAC abolished learning. Because nonanesthetics suppress learning but not movement (the two critical components of anesthesia), they may prove useful in discriminating between mechanisms and sites of action of anesthetics.