Figure 3 Fear learning in the human amygdala. (a) The outlined box contains the area of the medial temporal lobe that includes the bilateral amygdala. (b–d) Amygdala activation to the CS is seen bilaterally after fear conditioning (b) and observational fear learning (c), and unilaterally (d) in the left amygdala after instructed fear.
Neural systems of observational fear learning Despite the extensive evidence for observational fear learning across species, there is surprisingly little research in nonhumans investigating the underlying neural mechanisms. Behavioral findings indicating that observational fear learning draws on the same processes as fear conditioning predict a role for the amygdala. Amygdala lesions in monkeys confirm that this region is critical in acquisition and appro- priate display of fear in social and novel situations56. The hippocampus is important in socially mediated formation of food preferences57 and, under certain circumstances, social recognition memory in rodents58. In addition, lesions in the MPFC in rodents alter social behavior59. Still, the neural processes engaged in observational fear learning remain to be explored in nonhuman animals.
In humans, as in other animals, most of our knowledge about observational fear learning is drawn from behavioral experiments. Only recently have the neural mechanisms underlying this kind of learning been explored42,60. In an imaging study42, subjects watched a movie of another person expressing distress when receiving electric shocks paired with a CS. Later, subjects expected to receive shocks along with the same stimulus as that in the movie they just watched. However, no shocks were administered during the test stage to ensure that their representation of the US-CS pairing was based solely on vicarious experiences. As in previous fear conditioning studies, the bilateral amygdala was involved during both learning (observation) and expres- sion (test) of learned fear, strongly supporting the assumption that similar associative mechanisms and their underlying neural processes support both conditioned and observational fear learning (Fig. 3a–c).
In fear conditioning, the amygdala is believed to process and store representations of the CS-US contingency. Although the amygdala has an ancient evolutionary history, its interconnectedness to neocortex has increased substantially in primates. The basolateral complex in the
primate amygdala has strong reciprocal connections to visual cortex, in particular to the inferotemporal region that responds to face identity and to facial expression61. In addition, the basolateral complex is directly connected to the ventral part of the MPFC and indirectly with more dorsal regions of the MPFC62,63. These observations are compatible with the suggestion that the primate amygdala may be particularly prone to form associations between more complex socio- emotional stimuli, especially when they are visually represented.
The evidence indicates that, at least in primates, representation of fear learning through observation and classical conditioning may be rather similar within the amygdala. However, in spite of the many features shared between conditioned and observational fear, nonsocial and social forms of learning must differ in several fundamental ways, implying involvement of partially dissociable neural networks outside the amygdala. For example, a conspecific’s expression of distress may signal an imminent threat that serves as a US and elicits an immediate unconditioned response in the observer that is associated with a CS. However, this response is also mediated by the observer’s perception of the model, which can be influenced by more elaborated processes, such as emotional perspective taking and mental attributions. These, in turn, may be dependent on social factors, such as familiarity, relatedness, social status and interpersonal learning history. Indeed, in mice, observing a familiar, but not an unfamiliar, mouse experiencing pain enhances sensitization to pain at a later test time64. The intrinsic aversiveness of observing a conspecific in pain is evidenced by the willingness of monkeys to starve themselves if a shock is administered to a fellow monkey every time the observer attempts to eat65, but again, this altruistic behavior is influenced by familiarity and past experience of the conspecific65,66.
These studies hint at two interacting pathways mediating fear learning through observation. First, as suggested by work on observa- tional fear learning in primates39,41,42, a conspecific’s expression of distress can be intrinsically aversive, indicating that somatosensory representations may be primed by mere observation of another individual’s emotional display without necessarily being accompanied by higher order social cognition thought to be unique to humans. This point has been emphasized by proponents of mirror-neuron models of emotion perception and empathy. According to these accounts, shared neural representations of one’s own experiences of an emotion and perception of the corresponding emotion in another individual are critical to emotional understanding and to empathizing with others67,68. Second, in spite of a partial independence from higher cognitive functions, factors related to the social context can be involved in the regulation of basic emotional responding during observation and the resulting learning.
Studies in humans support these two interacting mechanisms22. First, stressing the independence of goals, expectations and social context, subliminally presented faces that signal threat, either by appearing angry or fearful69 or through previous pairing with an aversive stimulus, can elicit amygdala-mediated fear response in an observer23. On the other hand, affective responses to emotional faces and their recruitment of the amygdala depend on the context pro- vided70 and on cognitive appraisals by means of prefrontal brain systems17. Basic emotional responses to another’s distress are affected by interpersonal learning history and the goals of the observer. For example, an observer’s affective response to another’s distress depends on whether the other person is expected to cooperate or compete in a future interactive game51. Imaging studies indicate that a neural net- work, including the anterior insula and anterior cingulate cortex (ACC), that encodes the affective (as opposed to sensory), motivational and autonomic aspects of pain10,71,72 is also involved when people
SEPTEMBER 2007 NATURE NEUROSCIENCE