REM sleep looks like a single, uniform state on a sleep report. Behind that label, though, the brain alternates between two very different modes of operation: one punctuated by bursts of rapid eye movements, the other quiet and steady. A new multicohort study, published in the journal Sleep, shows that this inner structure of REM sleep erodes with age in a way that the standard sleep summary completely misses, and that the erosion is steeper in women.
Two states hiding inside one stage
Phasic REM sleep is the portion of the stage where transient events fire, most visibly the rapid eye movements that give the stage its name, along with muscle twitches. Tonic REM is the rest: the same desynchronized EEG and muscle atonia, but without those bursts.
Phasic periods show stronger delta, theta, and gamma power on scalp EEG, while tonic periods carry more alpha and beta activity. They respond to the outside world differently, too. Sound thresholds are lower during tonic REM, and the brain shows more reactivity to external stimuli, which is why tonic REM has been described as a sentinel state, a period of heightened readiness to the environment. Phasic REM, by contrast, has been tied to internal processing, including the reprocessing of emotional memories, and REM density, the count of rapid eye movements in each minute of REM sleep, is consistently elevated in depression and post-traumatic stress disorder.
What 627 healthy brains showed
To map how this microstructure changes across the lifespan, the researchers pooled retrospective polysomnographic records from five databases, covering 627 healthy adults aged 18 to 80, roughly half of them women. Eye movements were detected automatically from electrooculography, with participant-specific amplitude thresholds, and hierarchical models separated age, sex, and their interaction.
At the macro level, the results matched what sleep medicine already knew: total REM duration and REM percentage both declined with age, and women, on average, had slightly more REM than men.
The microstructure told a different story. Phasic REM duration fell with age, the proportion of REM spent in phasic mode shrank, the number of phasic periods dropped, and both the total count of eye movements and REM density declined. The quiet tonic component, meanwhile, held steady. The result is a shift in the balance of REM sleep itself: less of the eye-movement-rich mode, more of the quiet mode.
The study also resolves a long-standing contradiction in the literature. Earlier work had often described REM density as stable across the lifespan, a stability some researchers linked to preserved cognitive function. Here, in a far larger sample, REM density clearly decreased with age, consistent with the minority of studies that had reported a decline.
A sex difference hidden at the macro level
The most striking result concerns sex. Men and women did not differ in phasic REM duration itself, and their macrostructural aging trajectories looked similar. But the age-related declines in phasic REM percentage and in REM density were significantly steeper in women. In other words, two people of different sexes can show identical REM duration and percentage on a sleep report while their underlying REM microstructure ages at different rates.
The researchers note that menopause-related changes in estrogen and progesterone are a plausible contributor, given how strongly those hormones shape sleep physiology. They also point out that women report worse subjective sleep quality overall, despite objectively better macrostructural sleep in many studies, a long-standing puzzle that microstructure may help explain.
The two modes age in opposite directions
Spectral analysis on a subset of 446 participants added a second layer. The frequency content of phasic and tonic REM diverged with age in opposite ways. The phasic mode showed a gradual slowdown in its EEG signature: delta power climbed while theta power fell, a pattern the authors call spectral slowing. Tonic REM went the other way, gaining alpha and beta power with age, without any slowing.
That dissociation matters because increased delta activity in REM has been linked in prior work to impaired memory consolidation and to markers of pathological aging, including Alzheimer’s disease neuropathology. The authors stress that this slowing is specific to the phasic mode and is not a general property of REM sleep aging.
Why it matters
Standard sleep staging throws away the phasic-tonic distinction, and this study suggests that the discarded information carries real signal. Because phasic REM declines visibly with age in healthy adults, deviations from the expected trajectory could serve as a reference for spotting accelerated change in psychiatric and neurodegenerative conditions. REM sleep is already known to be disrupted in depression, PTSD, and Parkinson’s disease, where REM sleep behavior disorder can appear in the prodromal phase, years before motor symptoms. Microstructural markers that shift before the macro level does could offer an earlier window on disease.
The study’s authors frame their work as a first step toward using REM microstructure as a candidate biomarker, and they caution that the field needs cycle-by-cycle analyses and validation in patient populations before that becomes clinical reality.
Limits
The study is retrospective, pooling records collected with different protocols across five sites, and some variables, such as antidepressant use, could not be systematically controlled. Participants with mild sleep apnea (AHI between 5 and 15) were not excluded, so residual sleep-disordered breathing cannot be ruled out. Eye-movement detection was automated, and the spectral analyses relied on a single central derivation, C3, which was the one electrode present in every database. The authors confirmed their phasic-tonic power differences survived correction for EOG contamination, and eye-movement amplitude itself remained stable with age, which argues against the delta increase being an artifact.
Bottom line
REM sleep is not one thing, and it does not age as one thing. The eye-movement-rich phase shrinks over the adult lifespan, faster in women, while the quiet phase persists, and the two diverge spectrally in opposite directions. The findings make the case that the next generation of sleep biomarkers may live in the REM microstructure.
Source
Bertin C, Jeanne R, Bourgin J, Leoncini-Janin E, Wunderlin M, Züst MA, Hot P, Eichenlaub JB. Age and sex-related changes in the human REM sleep microstructure. Sleep. 2026 Jul 31:zsag209. doi:10.1093/sleep/zsag209. PMID 42536402.