Cerebrum Article

Animal Fear and Human Guilt

Harvard psychologist Jerome Kagan reviews animal research on the brain bases of fear and finds difficulties with the attempt to generalize to humans. He suggests that the search focus on the evolutionary changes that gave humans our unique brains and made our moral sensibility necessary for survival.

Published: October 1, 1998

Rats behave as if they have the capacity for fear of imminent harm, but there is little evidence to suggest that they (or any other nonhuman species) experience the human emotions we call pride, loneliness, joy, anxiety, guilt, shame, awe, jealousy, disgust, love, or moral outrage. It is not surprising, then, that scientists interested in discovering the biological bases for human emotions through the study of laboratory animals have focused on fear.

The hypothesis that fear of harm is the key to human symptoms of distress also derives support from world events. It is difficult to ignore the butchery in Bosnia and Rwanda, unprovoked murders in urban ghettos, and bombings on London and Belfast streets. These make all of us uneasy, and scientists seeking to explain our mood of uncertainty, and perhaps mute it, have placed fear of attack in a position of priority in the human psyche.

An historical factor in this focus on fear—one relevant to our discussion of animal models for human distress—is that Sigmund Freud made anxiety, which he conceptualized as unrealistic fear, the chief culprit in neurotic symptoms.

Finally, there is the popular assumption that fear interferes with human adaptation. This was not always a reigning premise. In Puritan America, where a feature of life’s assignment was to control exaggerated desires for fame, wealth, status, and sexual pleasure, fear was an ally. (John Bunyan regarded fear of God as a blessing because it permitted individuals to love the Deity.) The imperative to restrain greed, lust, competitiveness, and aggression makes human will a prerequisite talent.

Today, however, when our assignment is to gain more friends, obtain higher status, take risks for material gain, or seduce a lover, fear is an enemy—and human will is far less potent when fear is the demon to be tamed. Thus, when history altered the relative importance of fear and desire (and relegated will to the ash heap of ideas where Newton’s “ether” lies), many scholars decided that humans can and should be free of fear and anxiety—a premise that is one of the distinguishing illusions in contemporary Western thought.

Problems With Fear

Although the investigation of fear has yielded impressive insights, current research on conditioned fear in animals, and regarding it as a model for understanding all the facets of human anxiety, is beset with three difficulties.

First, the word “fear” is used in different ways by scientists who study animals and by laymen and clinicians. Neuroscientists use “fear” as a hypothetical construct to explain an empirical relation between two events: for example, that a rat will freeze in response to a light that earlier was associated with a painful electric shock. Psychiatrists, psychologists, and most citizens, on the other hand, use the same word to name a conscious experience in those who, for example, dislike crossing high bridges, confinement in closed places, large spiders, or traveling on airplanes. These two uses require different words.

Second, there is probably not a single fear state, but several, each with its own incentives, physiological profiles, and behavior, and each must be distinguished, in turn, from states of anxiety.

Third, fear of harm may actually be a less critical cause of most human distress than is guilt or shame.

Mapping the Circuits of Fear

Consider the search for the brain bases of fear. The first important discovery was by H. G. Kluver and P. C Bucy1, who in 1939 found that after surgically removing the anterior temporal areas of the brains of monkeys (which included the amygdala and adjacent structures now known to be important in memory), the animals were tamer and less fearful of novelty or danger. This unexpected outcome, now called the Kluver-Bucy syndrome, suggested that the amygdala was a necessary structure in an animal’s display of fearful responses.

Advances that made possible recording from single neurons and tracing the intricate pathways of axons have led to a richer understanding of the structures in the central nervous system that mediate fearful behavior. Remember though, that Kluver and Bucy studied monkeys’ fearful behavior toward unfamiliar places and animals, while most research on fear over the past few decades has involved conditioned fearful reactions. Although both are studies of “fear,” the phenomena are different.

Neuroscientists who have contributed to our understanding of conditioned fear include Joseph LeDoux, Michael Davis, Mortimer Mishkin, Michela Gallagher, Michael Fanselow, Lawrence Weiskrantz, James McGaugh, and Steven Meier. LeDoux’s experiments are typical.2 A rat is presented with a neutral stimulus (for example, a light or a tone) that is followed quickly by a brief, but presumably painful, electric shock. Immobility (freezing) is one of the natural biological reactions of rats to a stimulus that heralds a coming electric shock. Thus, after a number of pairings of a tone with a shock, the animal will freeze at the tone even if no shock is delivered. This is called “conditioned freezing.” If the unconditioned stimulus is a sudden loud noise rather than an electric shock, the natural biological reaction is a bodily startle. Furthermore, rats can be conditioned to display an enhanced—or “potentiated”—startle at a light that had been associated with a shock when the light is now followed by a loud noise.

Anatomical studies have revealed the brain structures needed for acquisition of a conditioned freezing or potentiated startle reaction to a stimulus. In addition to the receptors and neurons that deliver the sensory information to the brain, the animals must have a thalamus, basolateral and central nuclei of the amygdala, and projections from the amygdala to sites in the brain stem that produce the freezing or startle. If the conditioned reaction is a rise in heart rate or blood pressure, projections from the amygdala to the autonomic nervous system also must be intact.

Conditioned fearful behavior is not a new discovery, of course. More than 70 years ago John Watson and his assistant, Rosalie Rayner, made headlines by conditioning an infant to cry at the sight of a white rat by making a very loud sound when the rat was presented to the infant.3 (They also demonstrated that the infant’s fearful reaction would generalize to a white rabbit.) This experiment and others like it are one reason why, during most of this century, parents and therapists have regarded most childhood fears as the result of frightening experiences—for example, an encounter with a large barking dog producing a conditioned fearful reaction to other furry animals.

The conclusion: The neural structures underlying conditioned freezing in rats comprise the basic circuit for fear, and perhaps anxiety, in humans. This idea is supported by longitudinal observations of a large group of healthy children studied from the age of four months to seven-and a-half years. The four-month-old infants who combined high motor activity and frequent crying at visual, auditory, and olfactory stimulation (about 20 percent of the sample) were more likely than the rest of the sample to become fearful two-year-olds, emotionally subdued four-year-olds, and seven-year-olds with anxious symptoms like fear of the dark or large animals. It is likely that these highly reactive infants inherited a low threshold of excitability in the amygdala and its projections.4

A “Single Fear State?”

The research on fearful behavior in animals is impressive, but it is not obvious that activation of the circuit involving the thalamus, amygdala, and its projections defines the brain bases for a single fear state that varies only in intensity or form of expression.

First, as noted earlier, it is possible to condition freezing behavior in a rat by using a loud noise rather than electric shock as the unconditioned stimulus. Further, rats will show a potentiated bodily startle reaction to a sudden loud sound if they are first exposed to 20 minutes of bright light and never experience any electric shock.5 It is difficult to argue that a loud noise or 20 minutes of bright light generate the same brain and psychological state that is produced by a painful electric shock.

Second, close to a century ago, Ivan Pavlov conditioned dogs in his St. Petersburg laboratory to salivate at a sound by associating that sound with placement of food powder in the animal’s mouth. But Pavlov did not suggest that, when the dog salivated at the conditioned stimulus, it was in a state of hunger. Analogously, perhaps the conditioned stimulus of a tone that produces freezing does not create a psychological state of fear in the animal.

Third, the amygdala can be activated by a diverse set of events that many might call aversive—electric shock, an unpleasant odor, a loud noise, a bright light. One or more of a cluster of reactions occurs, including freezing, retreat, biting, and changes in heart rate and blood pressure. It is less obvious, though, that these brain events always create a fear state. Consider the following:

  • Pheromones secreted by a ewe in estrus excite olfactory structures and subsequently activate the amygdala. Projections from the amygdala to the hypothalamus lead to hormonal changes and increases in sexual arousal, not fear.6
  • If adults are exposed to extremely brief presentations (about 3/100 of a second) of faces with either a fearful or a happy expression, followed immediately by a longer presentation of a face that has a neutral expression, they report seeing only the neutral face and do not report feeling fearful or anxious. Yet, brain scans of these subjects, taken when the fearful faces were being presented, reveal a significant increase in activity in the amygdala.7
  • An adult’s report of feeling fearful or anxious usually does not correlate well with many biological indexes presumed to be associated with fear, such as changes in heart rate, blood pressure, galvanic skin response—or even activation of limbic sites assessed with scanning methodologies like functional Magnetic Resonance Imaging or Positron Emission Tomography.8

We may make a conceptual error, therefore, in assuming that the circuits that mediate the acquisition of conditioned freezing or enhanced startle in animals are fundamentally similar to the circuits that mediate the various emotional states of anxiety in humans. Psychiatrists and psychologists now suggest that there are distinctly different anxiety disorders, each with its own genetic profile, age of onset, and physiology.9 If humans are vulnerable to different anxiety states, it is unlikely that the circuit for conditioned fear in animals is the basis for all of them.

Many Reactions, Many Brain Circuits

Scientists who believe that a conditioned freezing response is sufficient evidence to prove that the animal feels fear are treating fear as an essence—just as late nineteenth century physicists, following Sir Isaac Newton, regarded space, time, and matter as essences. Albert Einstein changed the question that these physicists had been asking from “What is the constitution of matter?” to “How do two moving observers experience an event?” Might it not likewise be profitable to replace the question “What are the brain bases for fear?” with “How do different species react to events that signal danger?”

The work of Michael Davis and his colleagues at Yale University provides another reason for doubting that there is a single basic fear state that varies only in intensity.10 As I noted, animals will show an enhanced bodily startle, called potentiated startle, at a loud sound if it is preceded by a light that earlier had been paired with an electric shock of moderate intensity. Surprisingly, however, rats show a smaller startle when the light has been paired earlier with a shock of very high intensity. If the magnitude of the startle reflects the intensity of fear, shouldn’t the startle be greater when the light has been associated with a more painful electric shock?

It turns out that, when the shock is of moderate intensity, the potentiated startle is mediated by projections from the amygdala to a set of neurons called the ventral periaqueductal gray, which descend to motor centers that mediate the startle reflex. When the intensity of shock increases, different neurons—the dorsolateral periaqueductal gray—are activated and, as a result, the amygdala is inhibited and the magnitude of potentiated startle is reduced. Humans, therefore, would report greater worry but show a smaller potentiated startle at the anticipation of intense, rather than moderate, levels of pain. If a grizzly bear were peering at you from 200 yards away, the ventral gray would be activated and your startle to a sudden loud sound would be intense. But if the bear were within inches of your face, the dorsolateral gray would be activated and your startle would be muted. This hypothetical scenario implies the existence of at least two different states of fear.

Hunger provides an analogy. The psychological and bodily states of a person who has not eaten for 12 hours are qualitatively different from the states of one who has not eaten for four days. The latter is not just hungrier than the former. In fact, many years ago, Elliot Valenstein and his colleagues showed that certain nuclei in the hypothalamus, then believed to be the origin of hunger, were actually the basis of a more diffuse state, one that—depending on the immediate context—led to different behaviors. If food was available, the rat ate; if water, the animal drank; and if small wood chips were on the floor of the cage, the animal would retrieve them.11

A similar story may be told for the amygdala and its projections. These brain structures participate in many qualitatively different states and can mediate varied behaviors depending upon the immediate context. For example, a fear reaction to an unfamiliar place (which is part of the Kluver-Bucy syndrome) may involve a brain structure called the bed nucleus of the stria terminalis. It is likely that the brain state that defines this fear of a novel or unfamiliar event is different from the state that produces conditioned freezing at a tone or a light. Some neurons in both the cortex and the brain stem are devoted solely to detecting the mismatch between two events; detection of a discrepancy between what one expects and what occurs is built into the oldest and most fundamental structures of the brain. This fear reaction to unfamiliar events requires maturation of the brain, and appears in birds at two to four days of age, in puppies at six to seven weeks, and in humans at seven to ten months. (That is why crying at strangers and at separation from a caretaker first appear in human infants during the latter half of the first year.)

The distinction between a fear reaction to an unfamiliar or novel event or to a conditioned stimulus associated with electric shock, on the one hand, and anxiety, on the other, is also useful. Anxiety, which can occur when one anticipates an undesirable event in the future, might be restricted to humans. There is no evidence that any animal can anticipate an undesirable event that might occur days or weeks ahead.

Furthermore, although the brain constrains the number and quality of possible human emotions, it is unlikely that any brain state is linked in a deterministic fashion to any specific human emotion. The specific thoughts of a person lying quietly in a PET scanner, for example, are not predictable from the pattern of metabolic activity revealed by the scanner. The causes of the person’s thoughts must be stated in psychological language. Just as a gene can contribute to more than one physiological process, so a particular brain state can be the foundation for more than one psychological state.

John Cheever,12 who died in the second half of this century, and Alice James,13 who died a century earlier, are two writers who probably inherited similar, if not identical, vulnerabilities for a dysphoric, melancholic mood. Cheever, whose beliefs about human nature were shaped when Freudian theories were popular, assumed that his depression was the result of childhood experiences. As a consequence, he relied on drugs and psychotherapy to lighten his mood. When James suffered her depression at age 19, however, popular opinion assumed that women, especially educated women, were susceptible to depression because they lacked sufficient psychic energy. James concluded, therefore, that she had inherited her melancholia and certain that she could not change it, decided that she wished to die. Thus, even if Cheever and James had inherited the same genetic diathesis, they could not have experienced identical states of depression since their cultural contexts were different. Vernon Mountcastle, an eminent neuroscientist, stated this more generally, in the careful language of Analytical Philosophy, when he wrote that “every mental process is a brain process, but not every mentalistic sentence is identical to some neurophysiological sentence.”

In short, there are good reasons why the concept of a single fear state resting on one particular brain circuit should be replaced with a family of terms for related, but distinct states.

The Biological Process of Moral Judgement

Another reason to question the idea that conditioned fear in animals is a useful model for the varieties of human distress is the possibility that shame and guilt are more critical bases for human psychological symptoms than fear.

The central nucleus of the amygdala, necessary for acquiring a conditioned fear reaction, became progressively smaller with evolution, as the basolateral nucleus of the amygdala and the prefrontal cortex, and their connections, were enhanced. These evolutionary changes in human brain anatomy may have been accompanied by a subordination of fear of pain to shame and guilt as more important bases for feelings of unease and, in some cases, psychiatric symptoms. Most Americans are neither fearful nor anxious, for example, when they mingle with large crowds in a European city they have never visited. But this easy strolling among strangers in an unfamiliar place could not occur in monkeys, gorillas, or chimpanzees, for whom fear of novel objects and places is a salient characteristic.

Late in their second year, all children begin to show an appreciation of proper and improper behavior. The first signs can be observed in every home in every culture. They occur at the time that children are becoming consciously aware of their intentions and are able to infer what others are thinking. When a two-year-old child knocks a glass on the floor, she will glance at a parent with a wary face. Her emerging recognition that she did something wrong, based on the maturation of her brain, may be as profound as the changes when humans split off from their primate ancestors some five million years ago.

The human capacity for moral evaluation and its accompanying emotions of shame and guilt took from our primate ancestors a keen sensitivity to the voice, face, and actions of others, but added four unique abilities:

  1. inferring the thoughts and feelings of others
  2. application of the categories “good” and “bad” to events in the world and to the self
  3. regular reflection on past actions
  4. the cognitive capacity (by six or seven years of age) to know that a particular act that violated a moral standard could have been suppressed.

This combination, appearing in no species but ours, created a biologically novel psychological process: We evaluate ourselves as a consequence of meeting or violating our personal standards. Primatologists offer rich anecdotes of chimpanzee behavior to persuade us that these animals may possess rules, but honesty forces them to admit that they have never seen a guilty chimpanzee. And they will not, for guilt is not a possible state for chimpanzees.

Guilt and Shame—A Better Model for Human Distress

The syndrome called post-traumatic stress disorder, usually studied in war veterans, had been assumed for many years to be the consequence of an extreme fear reaction to the possibility of being killed. Now it appears that post-traumatic stress disorder is more frequent in soldiers who participated in an atrocity or felt guilty for surviving a friend than in soldiers who came very close to death but did not witness an atrocity or lose a buddy.14, 15 Similarly, women who have been raped are more likely to develop post-traumatic stress disorder than women of the same age and social class who have been robbed at gun point and threatened with death.16 A rape victim who wonders if any aspect of her behavior invited the attack will be more vulnerable to guilt than one who is robbed, or attacked by an animal. Guilt, rather than fear, appears to be central in both cases.

The state created by an imminent attack from an intruder is physiologically and psychologically different from the state of self-blame that occurs when a person violates his ethical standard on loyalty to a close friend. Anyone inclined to collapse the two feeling states into a single, overarching category probably would also treat as similar the state of a refugee who has had no food for two weeks and that of a monk who refuses to eat for two weeks to protest religious persecution.

If these distinctions prove valid, deeper understanding of guilt and shame might point the way to a fuller model of human distress than one restricted to conditioned freezing or potentiated startle. Perhaps that is why Hippocrates, Galen, and other ancient students of human nature nominated melancholia, rather than fear, as the fourth basic human temperament, and why Robert Burton entitled his book, “The Anatomy of Melancholy,” not “The Anatomy of Fear.”

Our moral sense and the accompanying emotions of guilt and shame are evolutionarily adaptive for a species that lives in groups and can harm others at any time, for these emotions constrain aggression and cruelty. Consider that the total number of acts of rudeness, vandalism, theft, rape, and murder that occurred yesterday throughout the world is infinitesimal when compared with the total number of opportunities each adult, from dawn to midnight, had to display any one of those behaviors. The ratio of antisocial acts actually committed over the total number of opportunities for such acts approaches zero. We are, as William Blake appreciated, both lion and lamb.

Early emergence of an appreciation of which actions were wrong may have enabled humans to survive. The control of aggression by an older sibling jealous of the attention given a younger one is necessary since the older child has both the opportunity and strength to hurt the younger. The frequency of seriously violent acts of that kind, however, is so tiny that each occurrence makes the front page of the local newspaper.

That insight should not breed indifference to the human capacity for hatred and cruelty; but neither should we deny the moral emotions that chance genetic mutations made possible between 100,000 and 200,000 years ago. Those genetic changes gave our species a frontal lobe large enough to enable us to harbor resentment, jealousy, and hostility long after our acute anger had passed. As a result, our species needed a psychological system to control aggression toward others. A human moral sense, like a spider’s web, is a unique product of evolution that has persisted because it ensures survival of our species.


  1. Kluver, H. & Bucy, P. C., 1939. Preliminary analysis of functions of the temporal lobes in monkeys. Archives of Neurology and Psychiatry 42: 979-997.
  2. LeDoux, J. E., 1992. Brain mechanisms of emotion and emotional learning. Current Opinion in Neurobiology 2:191-197; LeDoux, J.E., 1995. Emotion: Clues from the brain. Annual Review of Psychology 46:209-235.
  3. Watson, J. B. & Rayner, R., 1920. Conditioned emotional reactions. Journal of Experimental Psychology 3:1-14.
  4. Kagan, J., 1994. Galen’s Prophecy. New York: Basic Books.
  5. Walker, D. L. & Davis, M., 1997. Anxiogenic effects of high illumination levels assessed with acoustic startle paradigm. Biological Psychiatry 42:461-471.
  6. Perkins, A. & Fitzgerald, J. A., 1997. Sexual orientation in domestic rams. In L. Ellis & L. Ebertz (Eds.). Sexual Orientation. London: Praeger, pp. 107-127.
  7. Whalen, P.J., Rauch, S. L., Etcoff, N.L., McInerney, S. C., Lee, M. B., & Jenike, M. A., 1998. Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. The Journal of Neuroscience 18:411-418.
  8. Mountz, J. M., Modell, J. G., Wilson, M. W., Curtis, J. C., Lee, M. A., Schmaltz, S., & Kuhl, D. E., 1989. Positron emission tomographic evaluation of cerebral blood flow during state anxiety and simple phobia. Archives of General Psychiatry 46:501-514.
  9. Mineka, S., Watson, D., & Clark, L.A., 1998. Comorbidity of anxiety and unipolar mood disorders. Palo Alto, California: Annual Review of Psychology 49:377-412.
  10. Walker, D. L. & Davis, M., 1997. Involvement of the dorsal periaquedeuctal gray in the loss of fear-potentiated startle accompanying high foot-shock training. Behavioral Neuroscience 111: 692-702.
  11. Valenstein, E. S., 1964. Problems of measurement and interpretation with reinforcing brain stimulation. Psychological Review 71:415-437.
  12. Cheever, J., 1993. The Journal of John Cheever. New York: Ballantine Books.
  13. Strause, J., 1980. Alice James. Boston: Houghton-Mifflin.
  14. Yehuda, R., Southwick, S. M., & Giller, E. L., 1992. Exposure to atrocities and severity of chronic posttraumatic stress disorder in Vietnam combat veterans. American Journal of Psychiatry 149: 333-336.
  15. Bleich, A., Koslowsky, M/., Dolev, A., & Lerer, B., 1997. Posttraumatic stress disorder and depression. British Journal of Psychiatry 170:479-482.
  16. Valentiner, D. P., Foa, E. D., Riggs, D. S., & Gershuny, B. S., 1996. Coping strategies and posttraumatic stress disorder in female victims of sexual and nonsexual assault. Journal of Abnormal Psychology 105: 455-458.