By Paul Muchowski, Ph.D.
The Washington Post recently told the story of Larry Mitchell, a 51-year-old factory worker in Kokomo, Indiana, who smoked cannabis regularly for years and, without quite noticing, stopped dreaming. When he quit last summer, his dreams returned in force: vivid, emotionally rich scenes that included long conversations with a beloved cousin who had died. He told the Post he became excited to go to sleep. He does not plan to go back.
Mitchell's experience is not an oddity. It is a window into one of the most instructive questions in sleep science: what happens when a drug makes people feel that they sleep better while the instruments that measure sleep suggest otherwise? According to a National Sleep Foundation population survey published in Sleep Health, approximately 9 percent of American adults currently use cannabis to help them sleep. That is tens of millions of people making a nightly pharmacological decision, and most of them have never been told what that decision may cost.
A mismatch worth taking seriously
The Post's reporting highlights a pattern that sleep researcher Nicole Bowles of Oregon Health Sciences University (OHSU) has observed across roughly a decade of work: people who use cannabis consistently report that their sleep improves, yet when their sleep is measured objectively in a laboratory, most studies find more fragmentation and more time awake after sleep onset, not less.
To understand why both things can be true, it helps to distinguish the different outcomes that sleep science measures. Sleep onset is how quickly you fall asleep. Sleep continuity is how often and how long you wake during the night. Subjective sleep quality is how rested you report feeling. Sleep architecture is something different: the objectively measured proportion of the night spent in each sleep stage, including deep slow-wave sleep (deep sleep) and rapid eye movement (REM) sleep, the stage most closely associated with vivid dreaming. Architecture can only be assessed with objective tools such as polysomnography, the laboratory technique that records brain waves, eye movements, and muscle activity throughout the night. A drug can improve how a night feels while degrading what the night is made of.
What THC does to REM sleep
The evidence on tetrahydrocannabinol (THC), the intoxicating constituent of cannabis, deserves an honest accounting rather than a slogan. The clearest signals of REM suppression come from older laboratory studies that used high doses of THC. A 2025 systematic review and meta-analysis in Sleep Medicine Reviews examined the polysomnographic literature and concluded that cannabis administration does not consistently alter most sleep parameters, and that while early high-dose trials reported REM reductions, modern studies at lower therapeutic doses have produced mixed results. The evidence base remains thin, and anyone who tells you the question is settled is overstating it.
What is difficult to explain away, however, is what happens when regular users stop. A systematic review of human studies in Substance Abuse documented that sleep disturbance is among the most consistent features of cannabis withdrawal, and strange or vivid dreams are a validated withdrawal symptom. This is the REM rebound that Larry Mitchell experienced: dreams returning not gently but all at once. A rebound of that magnitude implies that something was being held down. In my judgment as a neuroscientist, the withdrawal phenomenon is the most persuasive evidence that chronic THC exposure meaningfully alters REM sleep in regular users, even as the acute dosing literature remains mixed.
The mechanism is well characterized. THC acts on cannabinoid type 1 (CB1) receptors, the most abundant class of receptor in the brain's endocannabinoid signaling system. A positron emission tomography study published in Molecular Psychiatry demonstrated that chronic daily cannabis smokers show downregulation of cortical CB1 receptors, meaning the brain reduces receptor availability in response to constant stimulation. The same study found that receptor density returned to normal after approximately four weeks of monitored abstinence. This explains both tolerance and the rebound: a dialed-down system cannot regulate sleep normally when the drug is suddenly removed, and it takes weeks to rebalance. People who do not know the rebound is temporary often misinterpret a week of intense nightmares as evidence that they need the drug, which is precisely how the cycle sustains itself.
The stakes of REM sleep are not trivial. Researchers link REM to memory consolidation and emotional regulation, and the long-term consequences of suppressing it remain an open question. Cannabis is well documented as causing temporary memory loss and reducing working memory, which may be attributed in part to its effects on REM sleep. A 2025 study in JAMA Neurology of more than six million adults in Ontario found that individuals who required emergency department care or hospitalization due to cannabis had a 72 percent higher risk of a new dementia diagnosis within five years than the general population. That finding applies to a specific population with severe use, it is an association rather than proof of causation, and it says nothing reliable about moderate use. I cite it not to alarm but because the truth is we do not yet know where the boundaries of risk lie.
Cannabidiol is a different molecule with a different profile
Here the science requires a distinction that popular discussion of "cannabis and sleep" routinely collapses. Cannabidiol (CBD) is a non-intoxicating constituent of the cannabis plant, and it does not behave like THC at the CB1 receptor.
The polysomnographic evidence reflects that difference. A small randomized, double-blind, placebo-controlled crossover study published in Frontiers in Pharmacology administered 300 mg of CBD to healthy volunteers and recorded a full night of polysomnography. CBD produced no significant alterations in the sleep cycle, in contrast to the architecture disruption associated with benzodiazepines and other conventional sedatives, and in contrast to the REM effects attributed to THC in the older high-dose literature. Where THC's signature is suppression of a sleep stage, CBD's signature in that controlled setting was the absence of interference. More recently, a novel formulation of CBD (300 mg) and terpenes (8 mg) increased deep and REM sleep in participants with severe insomnia, as measured by a wrist-worn sleep tracking device.
There is additional clinical evidence suggesting benefit, and it must be represented at its actual strength. A retrospective case series of 72 adults published in The Permanente Journal found that sleep scores improved within the first month in 66.7 percent of patients receiving CBD alongside usual psychiatric care, with anxiety scores decreasing in 79.2 percent. These were subjective measures from a chart review, not polysomnography from a randomized trial, and the sleep improvements fluctuated over time. The fair summary is that CBD does not appear to disrupt sleep architecture the way THC can, early data suggest it may support subjective sleep quality, plausibly through its documented calming effects, and rigorous placebo-controlled trials with objective sleep measurement are what the field needs more of.
What this means for your nights
If you are among the millions using THC-dominant cannabis to sleep, the research suggests you are making a trade you may not have known about: faster-feeling nights purchased, at least in regular high-potency use, at the potential cost of REM sleep and the dreaming that comes with it. If you stop, expect roughly one to three difficult weeks of rebound insomnia and intense dreams, and know that the disruption is temporary because the receptors recover.
And if you take one scientific point from Larry Mitchell's story, let it be this: how a night feels versus what happens to your objective sleep architecture are two different readouts. Any sleep aid, whether pharmaceutical or botanical, should be judged on both. The molecules THC and CBD in the cannabis plant are not interchangeable, and the difference between a compound that suppresses a sleep stage and one that leaves the night's architecture intact, or even boosts it, is not a marketing detail.
Paul Muchowski, Ph.D., is a neuroscientist and former professor at the University of California, San Francisco.