I’ve always thought of REM sleep, the dreaming stage, as the part of the night I could protect but never push. You skip the late glass of wine, keep the phone out of the bedroom, get up at the same time on Saturday, and hope. So when ScienceDaily ran a headline this week about a wearable ultrasound patch that “boosts REM sleep without drugs or surgery,” getting people into REM 43 minutes faster, I got excited for about a paragraph. Then one question wouldn’t leave me alone: faster than which night?

FROM THE DOCUMENT
a fixed order was adopted due to the unknown long-term effects of FUS
The paper's stated reason for never putting the ultrasound night first. Source: Nature Communications: Tang et al., NEUSLeeP ultrasound patch for REM sleep enhancement (2026)

The answer is in the methods of the Nature Communications paper behind the headline. Every volunteer slept the sham night first and the stimulation night second, on consecutive nights, in a lab bed, wired up with the patch and a full set of sleep electrodes. The authors explain the choice in one line: “a fixed order was adopted due to the unknown long-term effects of FUS.” FUS is focused ultrasound, and the caution makes sense. The trouble is what a strange bed does to you on night one.

Sleep researchers call it the first-night effect. A meta-analysis in Sleep Medicine pooled 53 studies covering 1,422 healthy people recorded on consecutive nights and found that on the first night, REM arrived later and there was less of it than on the second. Those are the two measures this patch is now famous for moving. A study that always puts the sham night first hands the second night a head start, whether or not the device does anything.

None of that makes the device dull. NEUSLeeP comes from Huiliang “Evan” Wang’s lab at the University of Texas at Austin. It’s a soft patch that sticks beside the temple, sends focused sound through the thin bone there toward the left subthalamic nucleus, deep in the brain, and records brain waves through hydrogel electrodes on the same patch. On the stimulation night, the sound came in five-minute blocks, 30 seconds on and 30 seconds off, every 90 minutes, five times over the night. Lead author Kai Wing “Kevin” Tang told the university’s news office that it was “the first time we’ve been able to noninvasively target deep brain regions involved in REM sleep.”

The target is what hooked me. I had the subthalamic nucleus filed under movement, the spot where surgeons implant deep brain stimulation electrodes for Parkinson’s, not anywhere near dreaming. The authors’ case is that people with those implants have reported better sleep, and that the nucleus is wired into brainstem hubs, including the pedunculopontine nucleus, that help switch REM on. In earlier studies, implanted stimulation there raised the REM share of the night by 1.3 to 6.2 percentage points and cut the wait for REM by about 15 minutes. So how does a sticker on your skin nudge a knot of neurons sitting several centimeters past bone, in someone who’s asleep? If sound can do what an operation does without opening the skull, that would change what sleep medicine can offer people.

Of the 28 volunteers, 26 produced usable recordings; two were dropped for poor signal quality. Across those 26, the paper reports that REM rose from 16.3 percent of the night to 20.9 percent, a gain of 4.6 percentage points, or from 63.8 minutes to 79.7. The wait for the first REM period fell from 177 minutes on the sham night to 134 minutes on the ultrasound night (p = 0.0098). Sleep efficiency and the number of arousals barely moved. Those REM gains land inside or above what implanted electrodes managed, which is exactly why the fixed order is so maddening.

The authors don’t hide it. Their limitations section names the non-counterbalanced order, says the study “was single-blinded and not double-blinded,” and flags possible placebo, expectancy and order effects, along with the lack of an active control such as stimulating a region that has nothing to do with sleep. The sleep scorer is described as blinded to condition, and the trial registry lists masking as none. All of that is in the paper. None of it made the headline.

I went looking for the detail that would let me wave the result away, and the data didn’t hand it over cleanly. The meta-analysis found that first nights also drag down sleep efficiency, yet here it sat at 80.6 percent on the sham night and 81.4 percent with stimulation. It also found the first-night effect weakest in people aged 20 to 29, and the healthy volunteers averaged 22.7 years old. Both details make a pure strange-bed artifact harder to argue. Neither rules it out, because in this design the ultrasound and the second night are the same variable, and no amount of analysis can pull them apart afterward.


The people most likely to buy a sleep patch got less of the spotlight. Eleven analyzed volunteers had Pittsburgh Sleep Quality Index scores of 6 to 10, poor sleepers the registry describes as a subclinical sleep-disturbance group, all between 19 and 38 like everyone else. Their REM share climbed from 17.9 to 23.2 percent. Their average wait for REM went from 176 to 146 minutes, a 30-minute drop with spread so wide that chance can’t be excluded (p = 0.21). The 43-minute figure is the pooled average across all 26, and eleven people can’t tell you much about whether poor sleepers respond differently. The morning rise in heart rate variability is a different matter: it showed up in the healthy sleepers and not in this group, and on that measure the paper’s own interaction test does suggest the groups split.

The trial registration describes a smarter machine than the one in the methods. It lists the study as non-randomized with no masking, describes night two as ultrasound “EEG-guided and timed to REM,” and promises “closed-loop stimulation time-locked to REM using EEG.” The published methods lay out a fixed schedule instead, one block every 90 minutes regardless of sleep stage, even though the paper’s own discussion describes stimulation as “synchronized with stage detection.” REM latency, the number in every headline, was registered as a secondary outcome; the primary outcomes were total REM time, REM percentage and a brain-imaging signal from the targeted nucleus. The record was first submitted on July 31, 2025, sixteen days after the study’s listed start. The registry pictures a patch that reads your sleep and answers it. The night that was actually tested ran on a clock.

My read is that the money shaped the framing more than the data did. The lead funding line is DARPA’s REM REST program, which exists “to demonstrate new technologies that enhance the means by which REM sleep supports healthy adaptation to stress and trauma” and cites poor sleep before deployment as predictive of a nearly fivefold jump in PTSD likelihood among service members. That mission fits how a two-night sleep study arrived wrapped in talk of PTSD and depression, with Vincent Mysliwiec telling ScienceDaily that the technology “could help millions of people get the restorative sleep they need.” The same writeup concedes that the researchers “have not yet established whether NEUSLeeP can improve symptoms of mental health disorders.” Four authors, Tang and Wang among them, declare a filed patent application, and the team is working with UT’s commercialization arm, Discovery to Impact, to bring the patch to market.

I want this to work. A way to steer REM without a pill or a surgeon is exactly what sleep science should be chasing, and aiming sound through the temple at a target that deep is clever engineering. The cleaner test isn’t exotic: an adaptation night before anything counts, a randomized and counterbalanced order so half the volunteers get ultrasound before sham, a sham nobody can tell apart from the real thing, and enough people to see a difference. The team says larger trials are planned, including in PTSD, depression and chronic insomnia.

If they run it that way, I’d sign up to sleep in their lab. Until then, I’m not sticking a patent-pending ultrasound array to my temple, and I’m keeping my Saturday alarm.

Sources

  1. Nature Communications: Tang et al., NEUSLeeP ultrasound patch for REM sleep enhancement (2026)
  2. ClinicalTrials.gov: NCT07190287, NEUSleeP wearable EEG-tFUS REM modulation study
  3. Sleep Medicine: Ding et al., A meta-analysis of the first-night effect in healthy individuals for the full age spectrum (2022)
  4. DARPA: REM REST, Rapid Eye Movement Restoration and Enhancement of Sleep-derived Trauma-adaptation
  5. UT Austin News: Get Better Sleep With Ultrasound Patch That Boosts REM Rest (2026)
  6. ScienceDaily: New wearable ultrasound patch boosts REM sleep without drugs or surgery (2026)