Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Behavioural neuroscience

The circuit of fear

Do you find it difficult to overcome an irrational fear? Blame it on the specific neural circuits hardwired in the brain that control fear recognition, and fear renewal even when fear has seemingly been overcome.

Learning to predict danger allows animals to defend themselves against harm and is crucial for survival. The neural mechanisms that subserve these functions are evolutionarily old, and their dysfunction is thought to underlie a host of anxiety disorders in humans, including post-traumatic stress and panic disorder1. Two papers in this issue, by Herry et al.2 (page 600) and Likhtik et al.3 (page 642), pinpoint the neural circuits that mediate the bidirectional transition between a defensive behaviour — fear — and the default exploratory behaviour.

To study the neural mechanisms that mediate fear responses, fear conditioning is widely used. In a typical procedure, animals are exposed to a normally harmless stimulus (such as a sound or light) before a brief exposure to an aversive stimulus — typically a foot shock. A few such rounds of pairing the harmless (conditioned) stimulus with the aversive (unconditioned) stimulus create an association between them in the animals' minds. So, on subsequent exposure to only the conditioned stimulus, they remember the association and express behavioural and physiological responses that correspond to fear in humans.

It is important to study how fear is first learned. However, the pertinent question for clinicians is how fear can be eliminated or reduced. Extinction of fear occurs when the association between the conditioned and the unconditioned stimuli is broken by repeated presentations of the conditioned stimulus only4. But does the association get completely erased from the memory? The answer is no. Although the conditioned stimulus eventually fails to elicit fear responses, much, if not all, of the original learned fear survives extinction5. So when the extinguished conditioned stimulus is tested either during, or shortly after, exposure to a dangerous context again, the conditioned fear is renewed spontaneously. Extinction therefore involves new learning, and its activation by situational cues inhibits the expression of fear responses to the conditioned stimulus.

Three brain regions implicated in fear conditioning and extinction are the amygdala, the medial prefrontal cortex and the hippocampus. The amygdala is the central structure involved in both fear conditioning and extinction1. It is divided into two main compartments: the basolateral amygdala, which receives sensory information about the conditioned and the unconditioned stimuli; and the central nucleus, which receives information processed in the basolateral amygdala. Neuronal projections from the central nucleus to the hypothalamus and brainstem then evoke the behavioural and physiological responses that constitute fear6 (Fig. 1a, overleaf).

Figure 1: Neural circuits that mediate fear.

a, During fear conditioning, convergence of inputs from the conditioned stimulus (CS) and unconditioned stimulus (US) to fear neurons in the basolateral amygdala (BLA) leads to potentiation of the conditioned input and activation by the CS of neurons in the central nucleus (CEA) that initiates physiological and behavioural responses characteristic of fear. b, During extinction, inputs from the medial prefrontal cortex (PFC) activate neurons in the intercalated cell masses (ICMs) — either directly or through activation of extinction neurons in the basolateral amygdala — which then inhibit the activity of fear output neurons in the CEA. c, During fear renewal, inputs from the hippocampus, which evaluates the current context, activate inhibitory interneurons in the basolateral amygdala that silence extinction neurons, thus restoring fear responses.

The learning of fear involves the co-occurrence of weak (conditioned stimulus) and strong (unconditioned) inputs to single neurons and leads to enhancement of neuronal activity known as excitatory postsynaptic potentials, which are elicited by the conditioned stimulus and mediated by NMDA glutamate receptors. This synaptic plasticity results in the conditioned stimulus evoking an increased response in basolateral-amygdala neurons that project to fear-output circuits in the central nucleus. Extinction results from NMDA-receptor-mediated changes in the amygdala7 that block the effects of this long-term potentiation, possibly by activating local neurons that secrete the inhibitory neurotransmitter GABA. The medial prefrontal cortex mediates consolidation of extinction, as well as the subsequent activation of these GABA-secreting neurons by the conditioned stimulus. The hippocampus probably regulates the situational control over this activation, and so over extinction. But how fear-conditioning and extinction memories are stored, and how they are translated into behavioural outcomes, is poorly understood.

Herry et al.2 set out to examine the role of the basolateral amygdala in fear learning and extinction in mice. In agreement with previous work8,9, they found that the pairing of conditioned and unconditioned stimuli results in enhanced firing of a population of basolateral-amygdala neurons — fear neurons — when animals are subsequently exposed to the conditioned stimulus alone. After extinction, these neurons no longer fire in response to the conditioned stimulus. But, intriguingly, another population of neurons — extinction neurons — emerged, which selectively respond to the conditioned stimulus undergoing extinction. Moreover, moving the mice from the extinction context to the conditioning context not only renews fear responses but also leads to reactivation of fear neurons and deactivation of the extinction neurons.

The authors show that the change from fear-neuron firing to extinction-neuron firing precedes the behavioural shift from a fear to a no-fear response, suggesting that activities of these neurons control the observed behaviour. In addition, infusion of the GABA-receptor agonist muscimol into the basolateral amygdala to inactivate neurons there impaired both extinction learning and fear renewal, indicating that these neurons' activity is crucial for the inhibition that underlies extinction.

To examine connectivity of fear neurons and extinction neurons, Herry et al. next stimulated the medial prefrontal cortex and the hippocampus. Their findings suggest that fear neurons receive inputs from the hippocampus and project to the medial prefrontal cortex. Extinction neurons, by contrast, are connected to the medial prefrontal cortex in both directions.

How does the extinction neurons' activity produce the shift from fear to safety? Between the basolateral amygdala and the central nucleus are clusters of inhibitory neurons known as the intercalated cell masses (ICMs). These neurons are innervated by input neurons from the basolateral amygdala, and their activity inhibits output neurons in the central amygdala by shunting dendritic excitatory inputs10. Neurons in the medial prefrontal cortex also densely innervate ICM neurons11. It has been suggested12,13, therefore, that activation of ICM neurons by these cortical neurons inhibits the amygdala's output during extinction, but conclusive evidence has been lacking.

Likhtik et al.3 used an innovative technique to selectively silence ICM neurons. Following fear conditioning and its extinction in rats, the authors delivered a toxin called saporin specifically to the animals' ICM neurons, to block protein synthesis and cause ICM cell death. They find that, with a decrease in ICM neurons, extinguished fear responses can once again be reactivated, suggesting that these neurons are required for the expression of learned extinction (Fig. 1).

These authors' work, together with that of Herry et al.2, clearly defines some of the neural circuits that mediate the extinction of conditioned fear. The amygdala, together with the hippocampus and the medial prefrontal cortex, uses situational information to evaluate an extinguished fear-conditioned stimulus and acts as a switching circuit that guides the appropriate behavioural response (Fig. 1). But, as with all interesting studies, these observations also raise questions.

The existence of fear and extinction neurons in the basolateral amygdala has been inferred from changes in the activity in this region after learning fear and extinction. These neurons are yet to be identified. Do extinction neurons activate ICM cells directly or through projections to the medial prefrontal cortex? Inhibitory neurons in the basolateral amygdala strongly inhibit the activity of local pyramidal neurons14.

Does inhibition of fear neurons during extinction result from activation of such local feedback interneurons in the basolateral amygdala? Herry and colleagues' observations suggest that long-term memories of both learned fear and extinction are not ultimately stored in the amygdala. So another question is how these learned behaviours are 'gated' after consolidation.

Answers to these questions will require exploration of the local circuitry of the amygdala, identifying fear and extinction neurons and uncovering the final targets of these cells. Neuronal receptors in these circuits — such as those targeted with saporin in Likhtik and colleagues' study — are likely to become targets for the development of specific treatments for many anxiety disorders.


  1. 1

    Davis, M. Annu. Rev. Neurosci. 15, 353–375 (1992).

    CAS  Article  Google Scholar 

  2. 2

    Herry, C. et al. Nature 454, 600–606 (2008).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Likhtik, E., Popa, D., Apergis-Schoute, J., Fidacaro, G. A. Jr & Paré, D. Nature 454, 642–645 (2008).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Pavlov, I. P. Conditioned Reflexes (Dover, 1927).

    Google Scholar 

  5. 5

    Rescorla, R. A. Annu. Rev. Neurosci. 11, 329–352 (1988).

    CAS  Article  Google Scholar 

  6. 6

    Sah, P., Faber, E. S. L., Lopez de Armentia, M. & Power, J. Physiol. Rev. 83, 803–834 (2003).

    CAS  Article  Google Scholar 

  7. 7

    Falls, W. A., Miserendino, M. J. & Davis, M. J. Neurosci. 12, 854–863 (1992).

    CAS  Article  Google Scholar 

  8. 8

    Quirk, G. J., Repa, J. C. & LeDoux, J. E. Neuron 15, 1029–1039 (1995).

    CAS  Article  Google Scholar 

  9. 9

    Collins, D. R. & Paré, D. Learn. Mem. 7, 97–103 (2000).

    CAS  Article  Google Scholar 

  10. 10

    Delaney, A. J. & Sah, P. J. Neurophysiol. 86, 717–723 (2001).

    CAS  Article  Google Scholar 

  11. 11

    McDonald, A. J., Mascagni, F. & Guo, L. Neuroscience 71, 55–75 (1996).

    CAS  Article  Google Scholar 

  12. 12

    Paré, D., Quirk, G. J. & Ledoux, J. E. J. Neurophysiol. 92, 1–9 (2004).

    Article  Google Scholar 

  13. 13

    Quirk, G. J., Likhtik, E., Pelletier, J. G. & Paré, D. J. Neurosci. 23, 8800–8807 (2003).

    CAS  Article  Google Scholar 

  14. 14

    Woodruff, A. R. & Sah, P. J. Neurophysiol. 98, 2956–2961 (2007).

    Article  Google Scholar 

Download references

Author information



Rights and permissions

Reprints and Permissions

About this article

Cite this article

Sah, P., Westbrook, R. The circuit of fear. Nature 454, 589–590 (2008).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing