The Real Cause of Recurring Nightmares

Nightmares are caused by a failure in the brain's REM sleep fear-processing circuit. Every night during REM, the brain replays emotionally charged memories and gradually reduces the intensity attached to them — a process that runs silently on nearly everything the waking mind generates. When it works, a difficult experience feels slightly less charged the next morning than it did the night before. When it fails, the emotional load overwhelms the circuit before it can complete its work, sleep ruptures, and you wake with the images still vivid and your heart rate elevated.

That forced waking is the failure itself — the readout that the circuit hit a load it could not process. In trauma populations, the failure is driven by a specific neurochemical disruption during REM sleep. In the broader population, the cause is less settled, and recent research suggests nightmare disorder may not be one thing at all. But one Swiss psychiatrist, Carl Jung, predicted the core of it nearly a century ago, without instruments, working only from what his patients told him in his office.

Neuroscience spent most of the twentieth century treating that prediction as mysticism. What the research of the last two decades has produced is a mechanistic account of nightmare recurrence that maps onto Jung's clinical framework in ways researchers did not anticipate — and the most recent finding, from a 2022 Geneva sleep lab, looks like something a Jungian analyst would have designed if given a sleep headband.

What Makes a Nightmare a Nightmare

Sleep researchers draw a harder line between a bad dream and a nightmare than most people expect, and the distinction matters for the biology.

A bad dream produces fear, dread, grief, or anxiety while the dreamer stays asleep. Distressing, but below the threshold that would rupture sleep. A nightmare forces waking. The dreamer surfaces with heart rate elevated, images still vivid, the body in a mild fight-or-flight state. Two people could have nearly identical dream scenarios — one surfacing and one sleeping through it — and only one had a nightmare. What crosses the threshold is the brain's failure to contain the emotional load the imagery is generating, not the imagery itself.

Ross Levin and Tore Nielsen formalized this in a 2009 paper in Current Directions in Psychological Science. The waking-up is the event that requires explanation. The brain's regulatory circuit encountered something it could not manage, and the arousal overwhelmed the system that was supposed to be processing it (Levin & Nielsen, 2009). Why some people's circuits fail nightly while others sleep through comparably distressing content is the question that unlocks the mechanism.

It also turns out the answer involves not just sleep but how the fear-regulation system operates across the entire day.

The Mechanism: What REM Is Actually Doing

The Affective Network Dysfunction model, published by Tore Nielsen and Ross Levin in Sleep Medicine Reviews in 2007, starts from a specific claim: REM sleep is active emotional work. During it, the brain reactivates emotionally charged memories from recent experience and runs them through the fear circuits again. The amygdala activates. Threat-relevant material is rehearsed. But the neurochemical environment during REM is radically different from waking: norepinephrine, the primary stress neuromodulator, is nearly absent from the brain. The fear circuits are running, but the chemistry that would normally amplify distress is suppressed (Nielsen & Levin, 2007).

The result is a brain that can revisit frightening experiences without re-traumatizing itself. Over successive REM cycles across a single night, the emotional charge on a difficult memory decreases while the factual content stays intact. Matthew Walker at UC Berkeley described this as "overnight therapy" — and the description is technically accurate: the brain processes emotional experience during sleep in a way that reduces its affective weight by morning (Walker & van der Helm, 2009). The only evidence it worked is that yesterday feels slightly less charged than it did the night before. This is the same mechanism explored in Why Do We Dream? What Neuroscience Actually Found.

Nightmares are what happens when this circuit fails. The amygdala activates during REM, as expected. The medial prefrontal cortex, which normally holds the amygdala's response in check, cannot. Instead of the fear response being down-regulated across the night's REM cycles, it escalates. The emotional load exceeds what the circuit can contain. Sleep ruptures.

A 2019 study in SLEEP gave this failure a signature extending beyond sleep. Nightmare severity was inversely correlated with medial prefrontal cortex activity not only during sleep but during waking hours: people with frequent severe nightmares show measurably lower mPFC activity while awake, at rest, looking at emotionally neutral images. The brake is weaker all day. Nightmare disorder is a chronic feature of how the fear system is regulated, and sleep is simply where that feature becomes most disruptive.

When the Chemistry Never Arrives

About 2 to 5 percent of adults meet criteria for nightmare disorder. In PTSD populations, that figure runs between 50 and 70 percent, with some veteran cohorts higher. The gap points directly at the mechanism.

Walker and van der Helm documented the specific PTSD failure in their 2009 paper in Psychological Bulletin. In post-traumatic stress disorder, norepinephrine does not drop during REM sleep. The neurochemical suppression that makes fear extinction possible never arrives. The amygdala runs hot throughout the night. Instead of the fear response being attenuated across REM cycles, the brain reactivates the original fear response intact — the person wakes up, and the same memory carries the same charge the following night because the process that would have reduced it could not complete (Walker & van der Helm, 2009).

In PTSD, recurring nightmares are what the disorder does to the mechanism responsible for emotional maintenance during sleep. The circuit that should be running nightly therapy on the traumatic material cannot run. The material stays charged. The circuit fires on it again. Same result, every night, because the conditions for a different result never arrive.

Murray Raskind at the University of Washington tested this account directly. His 2003 study in the American Journal of Psychiatry used prazosin, an alpha-1 adrenergic blocker that crosses the blood-brain barrier and reduces the brain's responsiveness to norepinephrine during sleep. In a twenty-week crossover trial with ten combat veterans with chronic PTSD and severe trauma nightmares, prazosin outperformed placebo on nightmare frequency, sleep disturbance, and overall PTSD severity. The logic was clean: block the norepinephrine, restore the neurochemical conditions for fear extinction, let the circuit complete (Raskind, 2003).

The Theory That Worked Until It Didn't

In 2018, a randomized controlled trial published in the New England Journal of Medicine enrolled 304 veterans across thirteen VA medical centers and found that prazosin did not outperform placebo on distressing dreams or sleep quality. The small-trial effect did not hold at scale.

The failure matters beyond the clinical question. If nightmare disorder were reliably produced by elevated norepinephrine disrupting REM fear extinction, blocking norepinephrine should work consistently. It does not. The AND model may describe the mechanism accurately for one subset of nightmare disorder, while other presentations run through different or differently weighted circuitry. Same surface output, possibly several different underlying failures. The nightmare is the output. The input is not always the same thing.

That mystery set the stage for the most interesting finding of the decade.

What Happened When Researchers Stopped Trying to Kill the Nightmare

In 2022, Lampros Perogamvros and colleagues at the University of Geneva published a study in Current Biology that approached nightmare disorder from a direction the pharmacological line of research had not considered: rather than interrupting the circuit generating the nightmare, what if you gave that circuit different material to work with?

Their approach built on imagery rehearsal therapy, a well-established treatment in which patients consciously rewrite their nightmare while awake, constructing a less threatening version of the scenario and rehearsing it deliberately. The Geneva innovation was pairing each rehearsal session with a specific neutral sound, then playing that sound through a wireless headband during REM sleep over the following two weeks.

The technique is called targeted memory reactivation. During REM sleep, the brain preferentially consolidates memories cued by stimuli present during the original learning episode. Play the sound during REM and the brain works preferentially on the memory connected to it. In this study, that memory was the rewritten, less threatening version of the nightmare. The brain's consolidation system was handed better material than the loop it had been running.

Of the 36 patients in the trial, those in the sound-cued condition had significantly fewer nightmares and more positive dream emotions at two weeks, with effect sizes above 1.0. At three months, the effect had grown rather than faded, reaching a Cohen's d of 1.45. The nightmare loop closed differently because the consolidation process was given different material.

The three-month trajectory is the data point that changes the frame. Suppression produces a declining effect over time. Perogamvros found the opposite. Something in the underlying circuit resolved. The nightmare did not return — not because it was blocked, but because the brain completed the work differently.

The Part Jung Already Knew

Imagery rehearsal therapy, the technique Perogamvros built on, was developed by Barry Krakow in the 1990s. Its core logic is this: engage with the nightmare's content while awake, understand what it is doing, and consciously give it a different resolution. Change your relationship to the material rather than trying to make it disappear.

Jungian analysts have been doing a version of this for a hundred years.

Jung's approach to recurring nightmares was to take the material seriously. The threatening figure is not to be fled from or suppressed. It carries something the waking ego has refused to face. In his framework, these figures represent shadow content — the rejected parts of the self that return as threat precisely because the dreamer has refused to meet them. The recurring nightmare is the psyche being insistent. Meet the material consciously, and the dream changes.

That is not a metaphor for what Perogamvros found. It is a description of the same functional event, arrived at through entirely different means. The nightmare is the brain's emotional processing circuit failing to complete its work on specific material. The circuit runs the same process on the same material and arrives at the same failure, every night, because the conditions for completion are not there. Change what the circuit is given, change the dreamer's orientation toward the material, and the circuit can finish.

Jung called this conscious reception. Neuroscience calls it targeted memory reactivation. The frameworks are operating at different levels of description, and the mechanism is not identical. But they are pointing at the same functional event: the nightmare resolves not by being killed but by being finished. Jung said this in different language nearly a century ago, working from patient sessions with no instruments and no sleep lab. The instruments eventually caught up. For more on this convergence, see Modern Neuroscience Proves Carl Jung Was Right About Dreams.

For the Jungian interpretation of threatening figures in recurring dreams, see Being Chased in a Dream.

What Is Still Open

The prazosin replication failure is the most important unresolved question in nightmare disorder research. It forces the category open. If the AND model were complete, the pharmacological approach should work reliably. It does not. The model likely describes one subtype accurately while others, presenting identically, require different approaches because they run through different mechanisms. Identifying those subtypes — and matching treatments to underlying mechanisms rather than surface presentations — is where current research is concentrated.

The Perogamvros finding raises a question the 2022 trial could not answer: does the content of the imagery rewrite matter, or does targeted memory reactivation work by attaching positive valence to the nightmare circuit regardless of what the new narrative says? The sound cue during REM may change the emotional tone of whatever the brain is processing, independent of the specific imagery the dreamer rehearsed. If so, the brain running the nightmare does not care about the plot. It cares about the charge.

That would mean the AND model had the observation right from the start: the nightmare is a failure of charge reduction, and the resolution comes from letting the circuit complete. What Jung understood, and what neuroscience has taken the better part of a century to describe mechanistically, is that completion does not come from suppression. It comes from engagement. The nightmare keeps running the same loop because it has not finished. The way to stop it is to help it finish.