Ever wondered why you sometimes have conversations in your sleep? These are the three leading explanations, according to biology.
getty
Every night, your brain performs a kind of controlled disappearing act. It dims your awareness, loosens your grip on the waking world and cycles through a choreography of sleep stages, each with its own distinct neurological signature. Most of the time, your body plays along. Your muscles go quiet. Your eyes dart behind closed lids. Your voice, for the most part, stays silent. Except when it doesn’t.
Sleep talking, formally called somniloquy, from the Latin somnus (sleep) and loqui (to speak), is one of the most common altered behaviors in human sleep. A large epidemiological study of 1,000 adults found that 66% had experienced it at some point in their lives, with around 17% having talked in their sleep within the past three months alone. It affects children even more: roughly half of young children are reported sleep talkers, a rate that generally declines with age.
What makes somniloquy fascinating and, to sleep researchers, genuinely useful, is that it doesn’t happen the same way twice. Its neuroscience depends entirely on which stage of sleep a person is in when it occurs.
During non-REM (NREM) sleep, particularly stages 2 and 3, the brain is oscillating in slow, rhythmic waves. These aren’t flat periods of nothingness. They alternate between inhibitory “down states” and excitatory “up states.”
Occasionally, an up state is strong enough, and localized enough to the speech motor regions, that it briefly activates the larynx, tongue and lips without triggering a full arousal. The person stays asleep and unaware. The brain has simply let slip a burst of motor output. This is what researchers sometimes call motor cortex leakage. The result is usually short and garbled: a word, a moan, an incomplete phrase.
REM sleep is a different story entirely. During this phase, the dreaming brain is remarkably active and almost indistinguishable from wakefulness on a brain scan. To prevent sleepers from acting out their dreams, a dedicated brainstem circuit produces motor atonia: a chemically induced full-body paralysis mediated by GABA and glycine receptors.
Research published in the Journal of Neuroscience confirmed that neurons in the ventromedial medulla are critical for maintaining this state, and that when they fail, the consequences are dramatic: animals (and humans) move, gesture and speak directly into their waking environment. In REM somniloquy, what you’re hearing is dream speech escaping into the physical world. The paralysis worked everywhere except the vocal apparatus.
Both mechanisms are, in essence, failures of suppression. The sleeping brain talks when it can’t quite hold it together.
In 2017, neurologist Isabelle Arnulf and her colleagues at the European Sleep Research Laboratories in Paris published what remains the most rigorous linguistic analysis of sleep speech to date, in the journal Sleep. They recorded 232 adults over multiple nights of polysomnographic monitoring, capturing 883 sleep speech episodes and 3,349 understandable words.
The most common word that people often uttered in their sleep, as they discovered, was “no.” Negations made up over 21% of all clauses, more prevalent in NREM than REM sleep. Interrogative phrases appeared in 26% of episodes. Subordinate clauses, grammatically complete and correctly structured, showed up in nearly 13%. And profanities featured in roughly 10% of all clauses, used at a rate substantially higher than the same speakers would use while awake.
What’s striking is not just the content, but the form. Sleep talking preserves the architecture of waking language. Speakers left pauses for imaginary conversational partners to respond, as if observing real turn-taking conventions. Broca’s area and Wernicke’s area, the canonical speech production and comprehension regions, appear to remain connected and active during sleep.
As Arnulf herself noted, “nocturnal language is negative, tense, more vulgar and addressed to somebody, not to oneself.” The sleeping brain, it turns out, is not mumbling at random. It’s usually arguing with someone.
Science doesn’t yet have a single, settled answer to why somniloquy evolved or persists, but there are several well-grounded frameworks, each illuminating a different piece of the picture. The memory consolidation hypothesis holds that sleep talking may be a byproduct, or even a signal, of the brain’s nightly memory processing.
A 2019 review published in Sleep Medicine Reviews proposed that verbal utterances during sleep reflect the activation of psycholinguistic circuits involved in replaying recently learned information, positioning somniloquy as a potential observable window into the consolidation of episodic and declarative memory.
In one striking case from a related study, a patient with REM sleep behavior disorder uttered a phrase during sleep that was semantically, though not literally, related to a passage they had studied before bed — an inadvertent verbal readout of the brain’s filing system.
The threat simulation theory, developed by neuroscientist Antti Revonsuo, offers a broader evolutionary framing. Revonsuo proposed that dreaming itself is an ancient biological defense mechanism, shaped by natural selection to rehearse threat perception and avoidance during sleep.
In a study testing this theory using 212 recurrent dream reports, 66% contained at least one threat and dreamers typically responded with defensive or evasive behavior. If the sleeping brain is running adversarial simulations (e.g., confrontations with predators, rivals, social dangers, etc.) then the negativity, refusals and verbal aggression of sleep speech start to make evolutionary sense. So the “no” isn’t random. It may be the most ancestrally rehearsed word in the human lexicon.
The motor breakthrough view is mechanistic rather than evolutionary: sleep talking simply happens when motor inhibition systems are imperfect. The brain is an extraordinarily complex biological system, and the circuitry responsible for silencing the body during sleep wasn’t optimized for flawless execution. Occasional leakage, especially in children, whose inhibitory systems are still maturing, is less a feature than a side effect of running such a complicated piece of hardware every night without maintenance windows.
These hypotheses are not mutually exclusive. The most defensible position is probably that sleep talking is overdetermined: it arises from incomplete motor suppression, while its content reflects the brain’s ongoing emotional and cognitive processing.
For the vast majority of people, sleep talking is harmless: nothing more than a curiosity, occasionally an embarrassment and largely self-resolving. It requires no treatment beyond addressing known triggers: sleep deprivation, fever, alcohol and emotional stress are among the most reliably documented.
That said, there are situations worth taking seriously. When sleep talking is frequent, loud and accompanied by physical movement like punching, kicking or jumping out of bed, it may signal REM sleep behavior disorder (RBD), a condition in which the atonia mechanism fails more comprehensively.
RBD is not merely an inconvenience: research has established it as a significant early marker of neurodegenerative disease. Studies tracking RBD patients over time have found that a substantial proportion go on to develop Parkinson’s disease or Lewy body dementia, often a decade or more after the sleep symptoms begin.
Sleep talking that occurs alongside night terrors, sleepwalking or breathing disruptions also warrants clinical attention. And if a bed partner is regularly losing sleep because of vocalization, that disruption itself, regardless of cause, is worth a conversation with a physician.
For most of us, though, the occasional murmured “no” in the night is nothing more than the brain doing what brains do.
Do you find it hard to fall asleep at night? Take my science-inspired Insomnia Test to know how severe your sleep difficulties might be.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。