A multiple-respo nse analysis of aversive learning was conducted in human subjects. For each subject, two pictorial stimuli were presented—one paired with electric shock. After training, the magnitude of the acoustic startle eyeblink reflex elicited in the context of the shocked picture increased dramatically and was significantly larger than for reflexes elicited during the non- shocked stimulus. Five different picture contents were tested in separate groups: Reflex potentia- tion was larger for pictures rated as pleasant than pictures rated as unpleasant. Conditioned re- sponses were also evident for skin conductance, heart rate, and affective judgments. Different systems reflected different aspects of the acquired fear response: Conductance change covaried with arousal, and startle probe magnitude varied with affective valence (pleasure). The neurophysi- ological implications of the data are elucidated, and parallels drawn between animal and human subject findings. Motivated by the clinical observation that anxiety patients often exhibit exaggerated startle reactions, Brown, Kalish, and Farber (1951) first demonstrated that the amplitude of the whole body startle response (in the rat) increased when a sud- den, loud noise occurred in the context of a cue previously paired with electric shock. This modulation of the startle re- sponse by an experimentally induced fear state—termed fear- potentiated startle by Michael Davis and his associates (see Davis, 1989, for a review)—has been frequently replicated with animal subjects. Furthermore, the neural pathways of fear po- tentiation, from the amygdala to the reticular formation, are well explicated (Davis, 1989), and positive associations have been demonstrated between elevated startle responses and other fear measures in animals, such as the degree of freezing (Leaton & Borszcz, 1985). Whereas the phenomenon of fear-potentiated startle is well established in the rat, parallel learning studies using the startle probe method with human subjects are rare—indeed, almost none have been reported in the past three decades. Two early experiments by Spence and Runquist (1958) and Ross (1961) used an airpuff as the startle probe. They measured the result- ing eyeblink reflex, which is among the fastest and most reli-