Emotion-specific activity in the autonomic nervous system was generated by constructing facial prototypes of emotion mllscle by muscle and by reliving past emotional experiences. The autonomic activity produced distinguished not only between positive and negative emotions, but also among negative emotions. This finding challenges emotion theoi'ies that have proposed autonomic activity to be undifferentiated or that have failed to address the implications of autonomic differentiation in emotioll. For almost a century scientists have argued about whether or not activity in the autonomic nervous system (ANS) is emotion-specific. Some of the most influential cognitive theories of emotion (1, 2) presume undifferentiated autonomic arousal despite a number of reports of emotion-specific autonomic activity (35). We now report evidence of such specificity in an experiment designed to remedy methodological problems that have lessened the impact of previous studies: (i) A broad sample of six emotions was studied, rather than the two or three that are typical. (ii) Verification procedures were instituted to maximize the likelihood that each sample contained only the single target emotion and no other. (iii) A sufficiently broad sample of autonomic measures was obtained to enable differentiation of multiple emotions, with appropriate statistical protection against spurious findings due to multiple dependent measures. (iv) Autonomic measures were taken from the onset of emotion production continuously until it was terminated. More typical measures taken before and after production of an emotion may completely miss shortlived target emotions, (v) Multiple eliciting tasks were used with the same subjects. (vi) Professional actors (N = 12) and scientists who study the face (N = 4) served as subjects to minimize contamination of emotion samples by extraneous affect associated with frustration or embarrassment. We studied six target emotions (surprise, disgust, sadness, anger, fear, and happiness) elicited by two tasks (directed facial action and relived emotion), with emotion ordering counterbalanced within tasks, During both tasks, facial behavior was recorded on videotape, and second-by-second averages were obtained for five physiological measures: (i) heart rate-measured with bipolar chest leads with Redux paste; (ii) leftand (iii) right-hand temperatures-measured with thermistors taped to the palmar surface of the first phalanges of the middle finger of each hand; (iv) skin resistance-measured with Ag-AgCI electrodes with Beckman paste attached to the palmar surface of the middle phalanges of the first and third fingers of the nondominant hand; and (v) forearm flexor muscle tension-measured with AgFig. I. Frames from the videotape of one of the actor's performance of the fear prototype instructions: (A) "raise your brows and pull them together," (B) "now raise your upper eyelids," (el "now also stretch your lips horizontally, back toward your ears." 1208 S(,IFN(,F vm ??1 AgCl electrodes with Redux paste and electronic integration of the electromyogram. The directed facial action task comprised six trials; in each a nonemotional expression was performed and followed by an emotion-prototypic expression, that is, an expression that theory and evidence indicate universally signals one of the target emotions (6). Subjects were not asked to produce an emotional expression but instead were told precisely which muscles to contract (Fig. 1). Their attempts to follow these instructions were aided by a mirror and coaching (by P.E.). The nonemotional expression comprised two actions not included in any of the emotional expressions to control for ANS changes associated with making any facial movement. Expressions were held for 10 seconds. This task resembles a traditional emotion posing task (in which, for example, subjects are asked to look fearful), but improves on it by precisely specifying for the subject, and for the experimenter's subsequent verification, the exact set of muscle movements that is required. Video records offacial expressions were measured (7) to ensure that autonomic data would be included in the analyses only if the instructed set of actions had been made; 86.5 percent of the data were used. In the relived emotion task, subjects were asked to experience each of the six emotions (in counterbalanced orders) by reliving a past emotional experience for 30 seconds. This task resembles traditional imagery tasks, but more specifically focuses on reliving a past emotional experience. After each trial, subjects rated the intensity of any felt emotion on a scale from 0 to 8. Autonomic data were used only when the relived emotion was felt at the midpoint of the scale or greater and when no other emotion was reported at a similar strength; 55.8 percent of the data were used. Change scores were computed for each emotion on each task (directed facial action: averaged data during emotional face minus that during nonemotional face; relived emotion: averaged data during relived emotion minus that during the preceding 10-second rest period). The experiment was analyzed in a 2 by 2 by 6 (actors versus scientists by task by emotion) multivariate analysis of variance. Our hypothesis that there are autonomic differences among the six emotions was supported [emotion main effect, F(25 , 317) = 2.51, P < O.OO1J. There were differences in emotion-specific autonomic patterns between the two eliciting tasks [task by emotion interaction, F(2S, 62) = 2.0, P = 0.014J. 16 SEPTEMBER IYRl Skin ~Hi9h: High-temperature ~ Anger / Low: Heart Fear rate Sad \ Low: Happy Disgust Surprise Fig. 2. Decision tree for discriminating emotions in direction facial action task. The nature of the emotion-specific ANS activity was explored with t-tests within significant univariate effects. Two flndings were consistent across tasks: (i) Heart rate increased more in anger (mean calculated across tasks ± standard error, +8.0 ± 1.8 beats per minute) and fear (+8.0 ± 1.6 beats per minute) than in happiness (+ 2.6 ± 1.0 beats per minute). (ii) Left and right finger temperatures increased more in anger (left, +O.IO°C ± 0.009°; right, +0.08°C ± 0.008°) than in happiness (left, -O.07°C ± 0.002°; right, -O.03°C ± 0.002°). In addition to these differences between the negative emotions of anger and fear and the positive emotion of happiness, there were important differences among negative emotions. In the