Sleep Talking: Is It Connected to What You Were Dreaming?

01The one study that put a number on it

In December 1970, four researchers at the Journal of Nervous and Mental Disease published something nobody has really improved on since. Arkin, Toth, Baker and Hastey paid 28 chronic sleep talkers to sleep in a laboratory, recorded what they said, woke them shortly afterward, and asked what had been going through their minds. Then they handed both the utterance and the report to judges and asked a simple question: do these two things match?

Across 166 sleep speeches, the answer depended almost entirely on the stage of sleep the person was in. Concordance in varying degrees was discernible in 79.2 percent of REM period speech and report pairs, 45.8 percent of stage II pairs, and 21.1 percent of stage III to IV pairs, which is the deep sleep of the early night.

Read that as a gradient rather than a verdict. In REM sleep, where dreaming is at its most vivid and story shaped, most of what people said lined up with what they said they had been experiencing. In the deepest non-REM sleep, most of it did not. And the phrase "in varying degrees" is carrying real weight: some of those matches were a word landing in the same scene, not a line of dialogue lifted whole out of a dream.

The limits are worth stating plainly because they cut both ways. Twenty eight people is a small sample, and 1970 was a long time ago. Every one of them was a chronic sleep talker, recruited because they did this often, which is not the same as the person who did it twice this year. They were woken deliberately and questioned immediately, which is the best possible condition for remembering a dream, and nothing like your bedroom. And judges rating whether two texts resemble each other is a softer instrument than a blood test.

None of that empties the finding out. It means holding it the way the researchers themselves did. Alfonsi and colleagues, reviewing the whole field in Sleep Medicine Reviews in 2019, put the position carefully: verbal utterances, reflecting the ongoing dream content, may represent the unique possibility to access the dreamlike mental experience directly. May represent. In the same abstract they note that despite decades of progress on what the mind does during sleep, the literature on sleep talking specifically remains dated and fragmentary. The people who study this for a living are saying the evidence is thin. That is the state of the answer, and pretending otherwise would be the dishonest version.

02Why the stage you were in decides the answer

Sleep is not one thing. Across a night you cycle between non-REM sleep, which is deepest in the first couple of hours and shallower later, and REM sleep, which is brief early on and gets longer and longer toward morning. Most of the dreaming you would recognize as a dream, with a plot and people and a place, happens in REM.

Sleep talking is the odd one out among night behaviors because it does not belong to either half. The American Academy of Sleep Medicine's own diagnostic text is blunt about it: the essential feature is talking, with varying degrees of comprehensibility, during sleep, and sleep talking may occur during REM or NREM sleep. Cleveland Clinic says the same thing to a general audience. Behavioral sleep medicine specialist Michelle Drerup told them there is some evidence that at times, sleep talking may be related to dreams, but this is not always the case, and that although most dreams happen in REM, sleep talking can happen during any stage.

That single fact is why the answer has to be a gradient. If your words arrived out of REM, they had a dream available to come from, and the machinery that normally keeps your muscles quiet during REM had let a little through. If they arrived out of deep non-REM, there was much less of a dream there to leak, and what you produced was closer to a piece of speech machinery firing during a partial arousal than to a line of dialogue.

Here is the one thing this page is not about, said once so it is clear. Sleepwalking and night terrors come out of deep non-REM sleep in the first third of the night and are not the sleeper acting out a dream, while acting out dreams while asleep is a REM event and is a different situation with different implications. Sleep talking can accompany either of them, and it can also happen entirely on its own, which is the common case. If you also want the difference between nightmares, night terrors and ordinary bad dreams, that is laid out elsewhere on the site and there is no need to repeat it here.

Those figures come from Arkin and colleagues in 1970, from 166 recorded speeches by 28 chronic sleep talkers who were woken and questioned in a laboratory. Concordance was judged in varying degrees rather than word for word, and no comparable study has been run at scale since.

03How to tell which kind of night yours was

Nobody is going to wire you up, so this is inference rather than measurement. But you have four ordinary pieces of information and, taken together, they narrow it usefully.

What time it happened. REM sleep is scarce in the first hours of the night and abundant in the last. An episode at five in the morning had far more REM available to it than one at half past eleven. Nobody has measured that directly for sleep talking, so it is two established things laid side by side, and it earns a lean rather than a conclusion.

Whether a dream was in your hands when you woke. In Arkin's laboratory the sleeper was roused right after speaking, which is why anything was recoverable at all. If your partner nudged you mid sentence and you surfaced with a scene still attached, you are the only person who can check the match, and you should trust the check more than any general rule. If you woke hours later with nothing, that absence tells you very little, because most dreams are gone within minutes either way. Our page on how to remember your dreams covers the recall side of that.

Whether it was language or noise. This is the most deflating and most useful clue. When Arnulf and colleagues analyzed 883 speech episodes recorded on video polysomnography from 232 adults, 59 percent of those episodes were nonverbal: mumbles, shouts, whispers and laughs. Only the remaining minority contained the 3,349 understandable words in the whole dataset. Most sleep speech never reaches the level of a sentence.

Whether you have any awareness of it. The diagnostic text notes that the sleep talker is rarely aware of their sleep talking, which is why the person who has to tell you about it is always somebody else. That gap is completely ordinary and signals nothing at all.

04The question underneath the question

Almost nobody who searches this actually wants a lecture on sleep architecture. They want to know whether the words counted. Whether the name meant something. Whether the person who heard it now knows something about them that they had not chosen to say.

The clearest statement on that comes from the sleep medicine classification itself, and it is worth quoting because it is the sentence you came for: the content of sleep talking has not been shown to reflect actual prior waking behavior or memories, although substantial syntactical and semantic structure reminiscent of waking speech is often preserved. Read those two halves together, because they are the whole answer. Your sleeping brain can produce something that sounds like a real sentence, with grammar and timing and tone. That does not make the sentence a report about your life.

There is a direct experiment on this, and it is quietly devastating to the idea of sleep speech as a leak. Uguccioni and colleagues gave people a list of 16 words and a story of a few hundred words to learn before bed, then recorded them overnight. Eleven of the patients spoke during REM sleep. A single one of them uttered a sentence judged to be semantically, but not literally, related to the text they had learned. Researchers deliberately loaded specific material into people's heads a few hours before they talked in their sleep, and it came back out once, sideways, in one person out of eleven. That is what a leak actually looks like, and it is a trickle.

The same paper carries a number that puts the whole thing in scale. Those eleven speakers pronounced a median of 20 words each, which the authors calculated as 0.0003 percent of sleep spent with spoken language. Whatever you said, you said very little of it.

And the content itself, when researchers looked at it in bulk, is not confessional in the least. In the Arnulf dataset the single most frequent word was "No". Negations made up 21.4 percent of clauses, questions appeared in 26 percent of episodes, and 9.7 percent of clauses contained profanity. The authors' summary of what sleep talking mostly is: a familiar, tense conversation with inaudible others, suggestive of conflicts. It is an argument with somebody who is not in the room. That is a very different thing from a secret being told.

WebMD puts the folk belief and its answer side by side and lands in the same place: scientists still are not sure if such chatter is linked to dreams, and the talking can occur in any stage of sleep. Cleveland Clinic opens its own article by naming the folklore directly, that pop culture might have you believe people are more prone to sharing their deepest, darkest secrets when they talk in their sleep.

So, if what you are carrying around today is the fear that you were tested and failed, put it down. There is no evidence base for sleep speech as a truth serum, a lie detector, or a window onto anything you were hiding. Nobody has ever demonstrated that it works that way, and the people best placed to demonstrate it have been at it for more than fifty years.

05Whether you can ask a sleep talker anything

This is the follow up question everybody has, usually because somebody in the house has already tried it. WebMD's own advice is deadpan about the result: go ahead, try asking a question while they are sleeping, and do not be surprised if you get a single syllable answer. Arnulf's team found that apparent turn taking in these conversations respects the usual gaps of waking language, which is exactly why it feels like a conversation from the outside. The rhythm is right. The content is not addressed to you.

Real two way communication with a sleeping person has been done, and the fact that it took a laboratory to manage it tells you how far ordinary sleep talk is from an interview. In 2021, four independent research groups got answers to questions out of people during verified REM sleep. Konkoly and colleagues report that across 36 participants, correctly answered questions came back on 29 occasions, from 6 of the individuals tested, and the answers arrived as deliberate eye movements and facial muscle contractions rather than speech. It required lucid dreamers who knew they were dreaming, electrodes, and prearranged signals.

Which is to say: what you get from a sleep talker at two in the morning is not testimony and cannot be treated as consent, agreement, or an answer to anything. If you live with someone who does this, the kind and correct thing is to stop interrogating them, however entertaining the results.

06Why that sentence and not another one

There is a reasonable middle position here that gets lost between "it means nothing" and "it means everything". Sleep speech is not random noise. It is produced by a brain that spent all day in your life, using your vocabulary, in your voice, with your preoccupations somewhere in the background. When a name or a worry or a phrase from work turns up, chance is not really the explanation.

What it is not is evidence about the thing itself. A colleague's name arriving out of your mouth at three in the morning tells you that the name is available to your brain. It does not tell you what you think about them, what you want, or what you have done. The distance between those two statements is the entire subject.

The one measured association worth knowing sits on the sleep side rather than the meaning side. Camaioni and colleagues followed 29 sleep talkers and 30 controls aged 18 to 35 for a week, recording vocalizations and collecting dream reports and sleep logs. The sleep talkers reported significantly worse sleep quality, and the number of vocalizations correlated positively with sleep fragmentation and negatively with the emotional load of their dreams. More broken sleep, more talking, and if anything, flatter dreams rather than more intense ones. That is a correlation in 29 young adults over seven days, so hold it loosely. But it points somewhere useful, and it is the opposite of the assumption most people arrive with, which is that a big night of talking means a big night of dreaming.

07The other reading people reach for

Search for this and you will find, sitting right beside the medical questions, a steady stream of people asking what talking in your sleep means spiritually. It comes up constantly, it comes up unprompted, and it is usually somebody's second guess once the clinical answer has failed to satisfy them.

That reading is old, it is widespread across a great many traditions, and it is not the same kind of claim as the ones on the rest of this page. Sleep research can tell you which stage produced a sound and how often the words matched a report. It has nothing to say about whether the night means something, and it never did. If your framework holds that the sleeping mind is open in ways the waking one is not, nothing measured here contradicts that, and nothing measured here supports it either. Those are two separate conversations and this page is only qualified to have one of them.

08Why it started, and why your sister does it too

Two thirds of people have done this. Bjorvatn and colleagues interviewed 1,000 randomly selected Norwegian adults by telephone and found a lifetime prevalence of sleep talking of 66.8 percent and a current prevalence of 17.7 percent, meaning within the past three months. The authors flag their own limits, a low response rate and single questions per parasomnia, and the sleep medicine classification quotes essentially the same figures. Whatever else is true, you are not unusual.

It also runs in families, and the evidence for that is stronger than most people expect. Hublin and colleagues put the question to the Finnish Twin Cohort and got responses from 1,298 identical and 2,419 fraternal twin pairs aged 33 to 60. Genetic influence accounted for roughly half the liability in childhood sleep talking and a little over a third to a half in adults, and childhood and adult sleep talking were highly correlated, which is to say it tends to be a persistent trait rather than a phase. If your mother did it too, that is most of the reason.

As for why now, the useful frame is the one Dr Drerup offers. Since most parasomnias are thought to be a mixed state between wakefulness and sleep, she says, sleep talking may be more likely to happen when sleep patterns are disrupted or disturbed. That fits the Camaioni finding about fragmentation, and it makes the list of usual suspects predictable rather than sinister: jet lag, short sleep, stress, fever and illness, alcohol, and untreated sleep apnea, which fragments sleep all night without the sleeper knowing.

That last row is on the table because partners mix it up constantly. The classification text explicitly separates catathrenia, a groaning sound made while exhaling, from sleep talking, and notes it tends to happen nightly rather than coming and going.

09What is worth doing about it

Honestly, for most people, very little. WebMD is direct that there is no known way to reduce sleep talking, and the practical advice everywhere reduces to the same handful of things: a consistent schedule, enough hours, less alcohol, caffeine well clear of bedtime, and a cool dark room. Those are worth doing because the underlying picture is disrupted sleep, not because there is a treatment for talking.

The part that actually needs saying is about the other person in the room, because sleep talking is one of the few sleep behaviors with an audience. The diagnostic text is clear about where the real problem sits: complications usually arise when sleep talking is very frequent or loud, or if the content is objectionable to others. In other words, the trouble is almost always the disturbance and the awkwardness, not the sleeper.

Three things follow from that, and they are worth being deliberate about.

Ask before you record. Somebody asleep cannot agree to being taped, and a recording of sleep speech is an unusually intimate object to make of a person without asking. If you are in a household where this is a running joke, make sure it is a joke both people are in on.

Do not treat the words as evidence. There is no version of this where a sentence spoken in stage two sleep is a fair thing to put to somebody over breakfast as though it revealed their position. The classification says the content has not been shown to reflect waking behavior or memories, and holding a sleeper to it anyway is unkind on top of being unsupported.

If it is wrecking someone's sleep, treat that as the actual problem. A partner who is awake at three every night because of noise has a real complaint, and it is a solvable one: earplugs, white noise, sometimes separate rooms for a stretch. That is a logistics question, not a verdict on anybody.

If, on the other hand, you find yourself genuinely curious about the dream side of this, the useful move is the gentle one. Keep a note by the bed and write down whatever is left when you wake, without trying to reverse engineer it from what you were told you said. Starting a dream journal is a better instrument than a phone recording, because it captures the part only you have access to. And if the dreams themselves have been unusually loud lately, why dreams get more vivid covers what actually drives that, which is a separate question from the talking.

10When it is worth mentioning to a doctor

On its own, sleep talking is not classified as a disorder at all. The American Academy of Sleep Medicine files it under isolated symptoms and normal variants, which is the category for things that show up on a sleep study and do not need treating. Cleveland Clinic notes that unlike other parasomnias, sleep talking usually carries little to no risk.

There are a few situations where a conversation is reasonable, and the line drawn by clinicians is consistent enough to quote. Dr Drerup: if your sleep talking occurs suddenly as an adult, or if it involves intense fear, screaming or violent actions, you should consider seeing a sleep specialist. WebMD adds the practical one, which is if the talking is badly disrupting your sleep or your roommate's. To that list, add loud snoring, gasping or witnessed pauses in breathing, and heavy daytime sleepiness, because those point at something treatable underneath.

One finding deserves its context rather than a headline. The Finnish twin study found psychiatric comorbidity roughly twice as common in adults who talked in their sleep frequently, and highest in those whose sleep talking began in adulthood. The authors state in the same breath that most cases of sleep talking are not associated with serious psychopathology. Both halves are the finding. It is a reason for a new, frequent, adult pattern to be worth mentioning at an appointment you were having anyway, and it is no reason for anybody to worry tonight.

What a doctor will do with it is unglamorous and reassuring. There is no test for sleep talking itself. If they order an overnight sleep study, it is to rule other things out, and as Dr Drerup describes it, that study records brain waves, heart rate and breathing along with arm and leg movements, and it is video recorded so behavior can be reviewed.

11Common questions about sleep talking and dreams

Does sleep talking mean I was dreaming?

Sometimes. Sleep talking can arise in either REM or non-REM sleep, and the connection to a dream is much stronger in REM. When Arkin and colleagues compared what sleep talkers said with what they reported having in mind moments later, the two matched in about 79 percent of REM pairs and about 21 percent of deep non-REM pairs. So it is a real possibility rather than a certainty, and the time of night is your best clue.

Do sleep talkers tell the truth?

There is no evidence for it, and the sleep medicine classification states that the content of sleep talking has not been shown to reflect actual prior waking behavior or memories. Sleep speech is mostly nonverbal sound, and the language that does appear reads in bulk as a tense argument with somebody who is not there. It is not a lie detector, a confession, or a statement anybody can be held to.

What do sleep talkers usually say?

Far less than people imagine, and much of it is not words. In the largest recorded set, 59 percent of episodes were mumbles, shouts, whispers or laughs. Among the actual language, the most common single word was "No", about a fifth of clauses were negations, a quarter of episodes contained a question, and roughly one clause in ten included profanity. It is closer to the sound of a disagreement than a diary entry.

Can I ask someone questions while they are sleep talking?

You can, and you may get a syllable back, because the turn taking rhythm of sleep speech follows the ordinary gaps of conversation. What you will not get is a reliable answer. Genuine two way communication with a sleeping person has only been achieved in laboratories, with trained lucid dreamers responding through prearranged eye and facial signals during monitored REM sleep. Nothing said in sleep counts as agreement or consent.

Why did I say a specific person's name?

Because their name is in your head, which you already knew. Sleep speech uses your vocabulary and your preoccupations, so recognizable material turns up in it routinely. What has never been demonstrated is that the material carries its waking meaning with it. A name is not a message.

Is it normal to talk in your sleep as an adult?

Yes. Around two thirds of adults have done it at some point and roughly one in six has done it in the past three months. Twin research suggests it is a persistent, substantially heritable trait rather than something most people grow out of. What is worth raising with a clinician is a pattern that is new in adulthood and frequent, or one that comes with intense fear, screaming, or violent movement.

Can you stop sleep talking?

Not directly. There is no established treatment for sleep talking itself, and the honest advice is to improve the conditions around it: a regular schedule, enough sleep, less alcohol, no late caffeine, and getting any untreated snoring or breathing problem looked at. If it is disturbing somebody else's sleep, solving the noise is a fair and separate goal.

Should I record my partner talking in their sleep?

Ask them first, while they are awake. A recording made without asking is a strange thing to have of somebody, and it is worth remembering what it can and cannot do. It can be genuinely useful history to bring to a sleep appointment. It cannot tell either of you what the words meant.

If you came here mortified, the most useful thing on this page is probably the arithmetic: the people whose sleep speech has been measured said a median of twenty words a night, most of it not even language, and the field's own reviewers describe the research as dated and fragmentary. You were not tested. Whatever was funny at breakfast is allowed to stay funny. DreamTold treats dreams as a window into your own mind rather than a transcript anybody else gets to read, so if something from that night did stay with you, look up what keeps recurring in the dream dictionary and take it on its own terms. This guide is educational and meant for reflection, not medical advice. If sleep talking is new and frequent in adulthood, comes with intense fear, screaming or violent movement, or arrives alongside snoring, gasping or heavy daytime sleepiness, that is worth raising with a qualified clinician.

12Sources

Every claim here that comes from somewhere else, with the somewhere else, so you can read the original.

• American Academy of Sleep Medicine: ICSD-3-TR parasomnias text, sleep talking under isolated symptoms and normal variants
• Cleveland Clinic: Michelle Drerup, PsyD, on why people talk in their sleep
• National Institute of Neurological Disorders and Stroke: Brain basics, understanding sleep
• Journal of Nervous and Mental Disease: Arkin, Toth, Baker and Hastey, the degree of concordance between the content of sleep talking and mentation recalled in wakefulness, 1970
• PubMed: Arkin and colleagues, the degree of concordance between the content of sleep talking and mentation recalled in wakefulness, Journal of Nervous and Mental Disease, 1970
• PubMed: Arnulf and colleagues, what does the sleeping brain say? Syntax and semantics of sleep talking in healthy subjects and in parasomnia patients, Sleep, 2017
• PubMed: Alfonsi and colleagues, sleep talking, a viable access to mental processes during sleep, Sleep Medicine Reviews, 2019
• PubMed: Bjorvatn and colleagues, prevalence of different parasomnias in the general population, Sleep Medicine, 2010
• PubMed: Hublin and colleagues, sleeptalking in twins, epidemiology and psychiatric comorbidity, Behavior Genetics, 1998
• PubMed Central: Uguccioni and colleagues, sleep-related declarative memory consolidation and verbal replay during sleep talking in patients with REM sleep behavior disorder, PLoS One, 2013
• PubMed Central: Camaioni and colleagues, the influence of sleep talking on nocturnal sleep and sleep-dependent cognitive processes, Journal of Clinical Medicine, 2022
• PubMed Central: Konkoly and colleagues, real-time dialogue between experimenters and dreamers during REM sleep, Current Biology, 2021
• WebMD: Talking in your sleep, causes and treatments
• TikTok: the caption that names the morning after, on being caught talking in your sleep again