extension of the fear conditioning model to instructed fear, there was robust activation of the left amygdala, which correlated with the physiological expression of fear learning (Fig. 3d). Activation of the left insular cortex also correlated with expression of learning. The insular cortex is a critical component for conveying a cortical repre- sentation of pain to the amygdala94 and for subjective awareness of physiological states72. The verbally mediated learning is likely to have resulted in an abstract cortical representation of the potentially painful shock, which may have been communicated to the amygdala through projections from the insular cortex (Fig. 2c). The left lateralization of the activation is consistent with the common view that the left hemi- sphere is more involved in language processing95. However, brain imaging results cannot rule out involvement of the right amygdala, or indicate a critical role for the left amygdala in expression of fears learned through verbal instruction. Further support that the left amygdala mediates physiological expression of instructed fear learning was demonstrated in subjects with unilateral amygdala damage after a similar learning protocol. Those with damage to the left, but not right, amygdala showed an impaired expression of instructed fear. Instructed fear is dependent on awareness41, further indicating that learning based on abstract representations of contingencies may involve neural net- works partially different from those involved in fears acquired through classical conditioning and observation.
A model of social fear learning Social fear learning offers the opportunity to study transmission of biologically relevant information between individuals. Indeed, social learning at large may lie at the core of the forces that create and maintain culture31,96, which might then affect biological evolution96,97. Fear learning also provides insights into neurobiological mechanisms of social learning and thus may serve as a model for the intricate links between biological principles of learning and cultural evolution. Here we provide a framework for the relationship between neural mechan- isms underlying fear conditioning and two forms of social learning: observational and instructed fear. The model is centered on the amygdala, which is critical to physiological expression of learned fear, regardless of how learning is acquired.
As outlined earlier, in classical fear conditioning (Fig. 2a), informa- tion about the CS is communicated to the lateral nucleus of the amygdala by way of the sensory cortices and thalamus; this information converges with US input from the somatosensory cortex and thalamus. Through synaptic plasticity in the lateral nucleus, the CS-US associa- tion is formed. An additional, distributed cortical representation of the CS-US contingency is also acquired through the hippocampal memory system and may be expressed in regions associated with pain proces- sing, such as the ACC and insular cortex. In the presence of the CS, learned fear is expressed through projections from the lateral nucleus to the central nucleus, which in turn mediates autonomic expression. (Other means of expression may depend on other pathways8.) In addition, projections from the cortical representation of the CS-US contingency to the amygdala may contribute to autonomic expression of fear learning when there is subjective awareness of the CS-US contingency.
We propose that the mechanisms underlying learning through social observation (Fig. 2b) may be similar, with a few exceptions. First, the US in observational fear learning is the perceived fear expression of a conspecific and, as such, is conveyed to the lateral nucleus through the sensory cortices and perhaps the sensory thalamus. The representation of the strength of the US in the lateral nucleus may be modified by MPFC input related to perception and interpretation of the learning model’s mental state during the observed painful experience, as well as
a cortical representation of empathic pain through input from the ACC and insular cortex. We propose that, as in classical fear conditioning, the lateral nucleus is a site of plasticity underlying memory for the CS-US association, in addition to a distributed cortical representation of the CS-US association acquired through the hippocampal memory system. The output mechanism for observational fear learning does not differ from that for fear conditioning.
Fears that are acquired through verbal communication (Fig. 2c), we suggest, rely on a slightly different representation, given the symbolic nature of the learning. It is unlikely that abstract representa- tions of verbal threat are represented in subcortical structures, such as the amygdala. Although sensory information about the CS is conveyed to the lateral nucleus, we hypothesize that the association between the CS-US is only represented in a distributed cortical network. Further- more, this cortical representation is left-lateralized, reflecting the verbal nature of the US. We propose that memory for this cortical association depends on the hippocampal complex for acquisition, and that plasticity in the amygdala is not necessary. Nevertheless, autonomic expressions of instructed fears occur through communication of the cortical representation of the CS-US association and the potential for pain to the amygdala, perhaps by way of the insular cortex. As with other means of fear learning, we propose that the central nucleus mediates autonomic expression of instructed fear.
This proposed framework is simply our best guess of the processes underlying social learning of fear based on a limited literature, so a few caveats are appropriate. First, another brain region that may be involved is the striatum. Human brain imaging studies on fear learning, including those examining social fear learning42,93, report activation of the striatum98,99. Animal models of fear conditioning have not emphasized the striatum beyond its role in avoidance learning and active coping8, but this region, which is important in reinforcement learning100, may represent the CS-US association. Second, we have emphasized unidir- ectional projections in our model, but most of the regions we discuss have bidirectional connections with the amygdala. Third, this frame- work outlines how fear learning is first expressed after social and nonsocial means of acquisition. Once a CS is experienced and a fear reaction occurs, further learning may result, which could change the nature of the representation further. For instance, in instructed fear, co- occurrence of the CS and autonomic arousal may cause the CS to act as a secondary reinforcer, which projects its emotional salience to the lateral nucleus to facilitate an amygdala-dependent representation of the CS- threat association that was not present after initial verbal instruction. In this way, representation of verbally communicated fears may change over time and be experienced to be more similar to conditioned fears.
In spite of these caveats, the proposed framework represents a neural model that can begin to help us understand the complexity and subtlety of human fear learning in a social and cultural environment. This understanding may provide important knowledge about the under- lying socio-emotional impairments that are hallmarks of many psy- chological disorders, such as phobias and anxiety disorders, which are characterized by dysfunctional assignment of emotional value to certain stimuli and situations. Finally, a better understanding of the neural mechanisms supporting socially transmitted fears is essential to integrate our knowledge about the biological foundations of learning and cultural change to evolution at large.
ACKNOWLEDGMENTS We thank J. LeDoux for comments. This research was supported by the US National Institutes of Health MH62104 (to E.A.P.).
COMPETING INTERESTS STATEMENT The authors declare no competing financial interests.
SEPTEMBER 2007 NATURE NEUROSCIENCE