Why REM sleep can look like it helps and hurts fear memory at the same time
Fear learning works like a smoke alarm. After a scare, the brain should remember what actually predicted danger, but it should not start shrieking at every harmless object in the room. REM sleep, the dreaming-heavy phase, seems to be involved in both sides of that balance.
On one side, REM sleep supports keeping the fear memory itself. In mice, a drug that restored REM sleep that had been suppressed after fear learning was linked to better consolidation of contextual fear memory, meaning the animals retained the learned fear association more strongly [4]. The same pattern appears in a companion report of that work [5]. In that sense, REM sleep looks like it helps recall of the specific fear memory.
On the other side, REM sleep also appears to restrain overgeneralization, the tendency to treat safe, similar situations as threatening. In a study of 126 people, losing REM sleep through late-night sleep deprivation produced fear overgeneralization comparable to total sleep deprivation, while normal sleep, especially the REM-rich phase, inhibited it [1]. So more REM sleep was associated with less spreading of fear, not more.
How the brain keeps fear from spreading too far
The human study offers a concrete mechanism. The percentage of REM sleep a person got was negatively associated with how much their fear generalized, and positively associated with activity in the dorsolateral prefrontal cortex during a generalization test shortly after sleep [1]. In plain terms, more REM sleep went with a more active control region and a tighter, less trigger-happy fear response.
Theta oscillations, rhythmic electrical waves in the frontal lobe, carried much of that effect: they mediated 44% of REM sleep's influence on fear generalization [1]. That number means roughly two-fifths of the statistical link between REM sleep and reduced overgeneralization ran through these frontal rhythms, not that theta is the whole story.
A separate line of work in mice points to a different control region, the infralimbic cortex, which is important for learning that a once-frightening cue is now safe. Most infralimbic neurons were strongly active during REM sleep, and suppressing them specifically during REM sleep in the hours after fear conditioning impaired later extinction memory, the learned safety response [3]. This suggests REM sleep may help install a brake on fear, though the study measured extinction learning rather than generalization itself.
What happens when REM sleep is lost, and what that means for anxiety
The clearest picture comes from sleep loss. Reducing REM sleep, whether by cutting the second half of the night or by total deprivation, made people generalize fear more broadly [1]. In rats, stressful fear conditioning and re-exposure to the fearful context reduced REM sleep in vulnerable animals, and blocking a small regulatory molecule called miR-221 in the basolateral amygdala, a fear-processing hub, lessened that REM sleep reduction [2]. That study did not test recall or generalization, but it shows the fear system and REM sleep regulate each other in both directions.
Put together, the findings are consistent with a protective role: REM sleep after a frightening event may help preserve the useful memory while keeping its alarm range narrow [1][3]. Disrupting REM sleep after trauma could therefore tilt the balance toward broad, hard-to-extinguish fear, which is the pattern seen in anxiety-related disorders [1][3]. This remains a proposed link rather than a demonstrated cause of those disorders.
Is the opposite-effect idea proven?
Not directly. The recall benefit comes from mouse studies of contextual fear memory and extinction [3][4][5], while the generalization benefit comes from a human experiment using risk ratings and skin conductance [1]. No single study in this evidence set measured both recall and generalization after the same REM manipulation, so the two effects have not been shown to trade off within one brain at one time.
The honest summary is that REM sleep appears to support retention of a fear memory and, separately, to limit how widely that fear spreads. Whether those are truly opposing forces, or two parts of one organizing process, is still an open question that would need a study measuring both outcomes together.
Sources used in this answer
REM sleep reduction leads to fear overgeneralization via negatively modulating prefrontal theta oscillations
In 126 people, reduced REM sleep through late-night sleep deprivation caused fear overgeneralization comparable to total sleep deprivation, and prefrontal theta oscillations mediated 44% of REM sleep's effect on generalization.
MicroRNAs in Basolateral Amygdala Associated with Stress and Fear Memories Regulate Rapid Eye Movement Sleep in Rats
In rats, fear conditioning and context re-exposure altered microRNAs in the basolateral amygdala and reduced REM sleep in vulnerable animals, and blocking miR-221 attenuated that REM sleep reduction.
Infralimbic activity during REM sleep facilitates fear extinction memory
In mice, infralimbic cortex neurons were strongly activated during REM sleep, and suppressing them specifically during REM sleep after fear conditioning impaired extinction memory consolidation.
Atypical hypnotic compound ML297 restores sleep architecture following emotionally-valenced learning, to promote memory consolidation and hippocampal network activation during recall
In mice, the compound ML297 restored REM sleep suppressed after contextual fear conditioning, and this restoration was associated with improved contextual fear memory consolidation and more activated hippocampal neurons during recall.
Atypical hypnotic compound ML297 restores sleep architecture immediately following emotionally valenced learning, to promote memory consolidation and hippocampal network activation during recall
In mice, ML297 reversed post-conditioning reductions in REM sleep and NREM spindle power, and renormalization of these sleep features was associated with improved contextual fear memory consolidation.