Listen to this story



How Your Brain Learns to Fear

And how it’s different from a cheap scare

Emotion is a dirty word in neuroscience. A science of the brain needs to measure, test, explain. How do you measure relief? Test ambivalence? Explain fear? Actually, we can, and we have. We know how fear is learned.

Fear is not the schlocky, cheap jump-scares of films and games. That bit in Resident Evil 2 where your character bends down to pick up a scrap of paper from the interview-room floor? And that thing, that pink, muscular assemblage of claws and teeth, erupts from the two-way mirror behind you and makes you drop the controller, whimpering? That’s not fear. That’s startle.

Scaring the crap out of you — we can measure that with ease. Sudden movements, loud noises, erupting glass make you jump, heart racing, senses alert. Play an unexpected loud noise to a rat, and, just like you, it jumps six inches into the air, heart racing, and freezes when it lands. Poised for action. We know the rat is startled because we can see that it’s startled.

Fear is not jumping six inches into the air. Fear is dread, the anticipation that something bad is about to happen, that something bad must happen, any moment now. How do you measure dread?

We can measure fear by training rats to dread something innocuous. We can explain fear by finding where in the brain that innocuous thing is represented after the training but wasn’t represented before. Where the representation of fear now resides.

Such fear conditioning is disarmingly simple. You occasionally and briefly play a little sound. This, by itself, causes no response except a modicum of interest from the rat: In what is an otherwise extraordinarily dull, featureless, gray box, a sound turning on and off is entertainment gold, right up there with Taylor Swift. So rats pay attention to the sound. After a while, you follow up the sound with a brief, mild electric shock. Just a little one—brief, maybe a moment or two. But it’s enough to frighten the rat. To make it freeze in place, defensive, wary, wondering where that small but sudden pain came from.

Repeat that sound-then-shock pairing a few times. Then play the sound by itself. The rat freezes anyway. It stops, anticipating the coming shock. For a while, every time you play the sound, the rat freezes. The rat has learned to dread the sound, to fear what it portends.

And where does this happen? The amygdala. Cut out the amygdala and the rats do not fear. They never learn to dread the sound; they never freeze when it’s played. But knowing the where does not explain the how. The how is that somehow, somewhere, the brain links the sound to the shock.

We use sound to create fear because we already know a lot about how sound is represented in the brain. We know in detail the route sound takes from the ear, down through the brainstem, up through the thalamus, and on to the cortex. We also know in detail the pathway pain takes, from the sensory nerves running up the spinal cord, up, up to the thalamus. So we can trace the sound’s route and work out where it might meet the pain’s pathway.

They meet in the amygdala. The bits of sound thalamus and bits of pain thalamus are separate, but they send their outputs to the same part of the amygdala. Often to the same neurons. An amygdala neuron activated by sound is often also activated by pain. Now we have our “how” in sight.

You do your first pairing of sound-then-shock. A neuron in the amygdala is first activated by the sound, then activated by the pain. Its two blips of activity in a row represent the link between the sound and pain. But at first, the blip of activity to the sound is small, weak. It has no consequence on the rest of the brain. Repeatedly following the sound activity with the much larger, stronger pain activity teaches the neuron to increase its attention to the sound. The input carrying the sound to the neuron gets stronger and stronger with each pairing, until that sound input is strong enough to activate the neuron just as much as the pain input. Until the neuron’s output for the sound and the pain are the same.

The amygdala’s output goes directly to the brainstem circuits that make rats freeze. Now that the rat has learned to dread the sound, what happens? You play the sound. It triggers a cascade of activity through the sound route in the brain until it arrives at the amygdala. Here, it now strongly, fiercely activates the attentive neuron. The brainstem gets this signal, which looks just like the pain signal, and makes the rat freeze.

While this is a beautifully simple idea, inevitably it is not quite the whole story. (You suspected that, did you not? Since when has anything about the brain been simple?) The amygdala has more than one neuron. Many more. Very recent work has revealed what happens to many neurons at the same time during fear conditioning.

When we look at many neurons in the amygdala, we see some neurons that increase activity during the pairings of sound-then-shock. We see those neurons that learn to pay attention to the sound. But some decrease their activity, as though they were paying less attention to the sound. And here’s the weird bit: Whether a neuron increases or decreases its activity, it does not have to respond to the pain at all. So how do these neurons learn to link the sound to the pain?

The pattern of activity across all neurons is the important bit: the pattern of how much each neuron is active. When the sound is first played, the neurons respond as a group with a particular pattern to that sound: Some are silent, some whisper, some are loud. When the shock arrives, the neurons respond as a group to the pain in a different pattern to that of the sound. Some respond the same way, but some previously silent neurons now shout, some shouters now whisper, and all other combinations besides.

But over the pairings of sound-then-shock, the pattern of activity created by the sound changes. With every sound-then-shock, the pattern created by the sound becomes more similar to the pattern created by the pain. So, then, when just the sound arrives, the pattern of activity across all neurons spells “pain,” as though the amygdala learns to literally represent the sound as “pain.” And the rat responds appropriately: Freeze!

You most likely have firsthand experience that tells you fear is real. But can it be forgotten? Yes. If you carry on playing the sound by itself, never following it with the shock, the rat will soon stop freezing. Its dread of the sound has extinguished, gone extinct.

Extinguishing fear is not a straight line. There is no smooth decay of dread. On the first day of playing just the sound, you will see the rat freezing for less and less time after each play of the sound. But then take a day off, and put the rat back in the box. Now the first freeze on this new day is much longer than the last freeze of the previous day. Some of the forgetting of fear has been forgotten. The fear has somehow regrown a little in the meantime.

While the fear is extinguishing, the neurons’ pattern of activity created by the sound is moving away from the pattern for pain. The neurons’ spelling of “pain” becomes increasingly fuzzy. But this move is not going back to their original pattern, their original spelling of “sound.” Instead, it is to an entirely new pattern of activity. The important bit being that whatever this new pattern means, it no longer means pain. And indeed, the more different the new pattern is from the pain pattern, the less the rats will freeze to the sound.

The extinguishing of fear is not unlearning. The brain has not returned to the way it was. Put your rat back in its box after a few weeks off and play that sound again. It will freeze. But that freezing very quickly wears off. Information is retained about both the fear and the loss of fear. After all, that fear may be useful in future. Events that once predicted imminent danger may do so again, so it makes sense that rat brains and our brains should not completely forget fear. They just put it on ice.

We know an extraordinary amount about how rats learn to fear. Does that help us? Does that help our acquired fears? A cautious yes. Phobias are an obvious target here. Not all phobias are newly learned fears, and not all are irrational. Heights can kill, spiders can be venomous, the dark could be full of danger to our hominid ancestors.

That said, our knowledge of how rats learn and forget fear is the basis for how we treat fear in humans. Just like the extinguishing of fear in your rat when you played the sound over and over again without consequence, so too does exposure therapy do the same for your fear. By repeatedly experiencing the innocuous but fearful thing (the spider, the kiwifruit, pictures of Pennywise the Clown <shudder>) without consequence, the idea is that your fear should extinguish.

We have seen why such therapy may not be a straight line to a cure. Dread does not decay evenly: It gets better, then maybe a little worse again, then better, then maybe a little worse, until finally the fear is extinct. And we have seen that if, many months or years later, a blast of that fear comes from nowhere, it does not mean the fear is back for good. It could just be the rational, sensible check by your brain that this fearful thing is no longer important.

But our stubborn cases, the fears of innocuous things that will not change, suggest something has gone wrong with the extinguishing of fear. The finger of suspicion here points not to amygdala but to a tiny part of the cortex (the medial prefrontal cortex, if you like that sort of thing). Remember: Extinguishing fear is not unlearning. Rather, the brain puts fear on ice; it suppresses the link between the innocuous thing and the dread. And it seems the amygdala needs permission from this tiny part of cortex to suppress the fear. This tiny part of cortex seems to provide the signal to say, “Okay, we can stop responding to this now.” So perhaps when this “okay” signal is missing, we cannot stop fearing.

We have much to discover in our search for learning to fear and forgetting to fear. We need to head up to the cortex and find out where this fear suppression happens and how it goes wrong. We need to find out how neurons can bridge that gap in time between the sound playing and the pain arriving. We need to find out how neurons that seem to know nothing about the pain change anyway. We need to find out so many more things — but that’s true of every area of neuroscience. And every new thing we find out will take us a tentative step further down the path of controlling our fears.

Theorist and neuroscientist. Writing at the intersection of neurons, data science, and AI. Unceasingly British.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store