A dream is a stretch of mental activity that unfolds during sleep, typically experienced as a sequence of images, emotions, and sensations that feel real in the moment but are generated entirely by the brain. Researchers sometimes use the broader term “sleep mentation” to capture everything from vivid, story-like narratives to fleeting, fragmentary thoughts that surface during the night. That distinction between rich dreamlike content and simpler thought-like content turns out to matter more than most people realize, and so does the question of when during the night these experiences arise.
More Than Just Movies in Your Head
The casual image of dreaming as watching a film behind closed eyelids captures only the most dramatic version of what the sleeping brain produces. Sleep researchers distinguish at least two broad categories of nocturnal mental activity. Dreamlike mentation is characterized by strong visual imagery, emotional intensity, bizarre or impossible scenarios, and a narrative structure that resembles a story. Thought-like mentation, by contrast, involves more verbal, literal content with much weaker imagery.1PubMed Central. The neural correlates of dreaming When someone says “I had a dream last night,” they almost always mean the first kind. But both count as dreaming in the scientific sense, and both reveal something about how the brain organizes experience while you sleep.
The reason this matters is that how you define “dream” shapes what you find when you study it. Early research focused almost exclusively on vivid, narrative dreams, which made dreaming seem like a phenomenon tied to one specific sleep stage. Broadening the definition to include any conscious experience during sleep has opened up a much richer picture.
Dreams and Sleep Stages
For decades, dreams were considered a product of REM (rapid eye movement) sleep, the stage in which your eyes dart beneath closed lids and your brain is nearly as electrically active as when you are awake. REM sleep does produce the most elaborate, emotionally charged, narrative dreams. But researchers collecting reports after awakenings from non-REM stages, including the lighter stage known as N2, have found that people report mental content then too. The difference is structural: dream reports gathered after REM awakenings tend to have greater connectedness and complexity than those gathered after N2 awakenings, meaning the words and images in REM dreams recur over longer stretches and form more coherent storylines.2PubMed Central. Structural differences between REM and non-REM dream reports assessed by graph analysis
So the popular idea that you only dream during REM is wrong, but the intuition behind it is not entirely off base. The most vivid, story-like dreams cluster in REM periods, and those periods grow longer as the night progresses, which is why you are more likely to wake from a memorable dream in the early morning. Non-REM mentation tends to be shorter, more fragmented, and closer to idle thought than to a full-blown narrative.
The Hypnagogic State
There is an often-overlooked window where dream-like experiences begin before you are even fully asleep. The hypnagogic state is the transitional zone between wakefulness and sleep, and it can include sensory perceptions that feel surprisingly vivid: flashes of faces, geometric patterns, snippets of conversation, or the sudden sensation of falling.3PubMed Central. The hypnagogic state: A brief update These experiences are brief, usually lasting seconds to a couple of minutes, and they tend to lack the narrative structure of a full dream. Many people experience them regularly without even recognizing them as a form of dreaming.
There is a corresponding state on the other side of sleep, called the hypnopompic state, which occurs as you transition from sleep back to wakefulness. Both are interesting precisely because they sit at the boundary of consciousness, blending features of waking perception with the hallucinated imagery that characterizes sleep.
What Happens in the Brain During a Dream
One of the most important findings in recent dream research involves a region researchers call the posterior cortical “hot zone,” spanning the medial and lateral occipital lobe up through the precuneus and posterior cingulate gyrus. When people report having a dream experience, this region shows decreased low-frequency electrical activity compared to when they report no experience during sleep. The finding holds whether the person was in REM or non-REM sleep, and regardless of whether they could recall the specific content of the dream.4PubMed Central. The neural correlates of dreaming
This is a meaningful result because it suggests dreaming is not just an artifact of REM physiology. Instead, it appears tied to a specific neural signature in posterior cortex that can emerge across different sleep stages. In practical terms, researchers monitoring this brain activity in a sleeping person can predict with reasonable accuracy whether the person will report a dream experience when woken up.
Why We Dream
The question of why the brain generates dream experiences at all has produced several competing theories, none of which fully explains the phenomenon. Three stand out for having the most empirical traction.
The threat simulation theory proposes that dreaming evolved as a biological defense mechanism. By simulating threatening events during sleep, the brain rehearses how to perceive and avoid danger without any real-world risk. Studies of traumatized children have given this idea support: severely traumatized children reported more dreams, and their dreams contained more threatening events, as if the system had been upregulated by genuine danger.5PubMed. The threat simulation theory of the evolutionary function of dreaming: Evidence from dreams of traumatized children However, tests using recurrent dreams in adults found mixed results. While about two-thirds of recurrent dreams did contain at least one threat and dreamers tended to take reasonable defensive actions, fewer than 15% of the dreams depicted realistic situations critical to physical survival or reproduction.6PubMed. Evolutionary function of dreams: A test of the threat simulation theory in recurrent dreams The theory explains some features of dreaming well but struggles with the many dreams that are mundane, pleasant, or entirely surreal.
A more recent proposal, the overfitted brain hypothesis, borrows a concept from machine learning. The idea is that during waking hours, the brain gets very good at processing the specific stimuli it encounters every day, but risks becoming too narrowly tuned to that daily input. Dreams, with their bizarre, distorted, out-of-distribution imagery, function as a kind of corrupted training data that helps the brain maintain its ability to generalize and respond flexibly to novel situations.7PubMed Central. The overfitted brain: Dreams evolved to assist generalization Under this framework, the weirdness of dreams is not a bug but the entire point: strange, impossible content forces neural networks to stay robust rather than narrowly specialized.
Dreams and Emotional Processing
A third major function attributed to dreaming involves the regulation of emotions. REM sleep is known to play a role in processing emotionally significant waking experiences, and neuroimaging studies have shown that the brain regions active during dreaming overlap substantially with those that regulate emotions during wakefulness.8PubMed Central. The Functional Role of Dreaming in Emotional Processes The suggestion is that dreaming may help defuse the emotional charge of difficult experiences, particularly traumatic ones, by reprocessing them in a safe neurochemical environment where stress hormones are suppressed.
A 2024 study provided some of the strongest evidence yet for this idea. People who recalled dreams showed reduced emotional reactivity to negative images the next day compared to people who did not recall dreams. More positive dream content was associated with even greater reductions in next-day emotional responses. The researchers also found that dreaming was linked to a kind of memory trade-off: emotional memories were maintained at the expense of neutral ones, suggesting that the dreaming brain actively prioritizes and reorganizes what to keep and what to let go.9Scientific Reports. Evidence of an active role of dreaming in emotional memory processing shows that we dream to forget The study’s title captures the paradox neatly: we may dream partly in order to forget.
How Children’s Dreams Differ
If dreams depend on the brain’s capacity for mental imagery and narrative construction, you would expect them to change as the brain matures. That is exactly what the evidence shows. A cross-sectional study of children aged five through eight found that until about age seven, dream imagery was more static than dynamic, and children typically experienced themselves as passive observers rather than active participants. Narrative structure remained simple until around age eight. Interestingly, dream report rates were predicted by visuospatial skills rather than verbal abilities, with performance on spatial tasks being the single best predictor of how often a child reported dreaming.10International Journal of Behavioral Development. REM Dreaming and Cognitive Skills at Ages 5-8: A Cross-sectional Study
Later research with slightly younger children, ages four through eight, found a somewhat more active picture. About 70% of preschoolers reported representing themselves in an active role, and the vast majority of their dream narratives involved kinematic (movement-based) content. Their dreams featured more human characters than animal characters, and social interactions and emotions were common.11PubMed Central. Content analysis of 4 to 8 year-old children’s dream reports The disagreement between studies likely reflects differences in methodology and how dream reports were collected, but the broader point holds: children’s dreams become more complex, narrative, and self-involved as their cognitive abilities grow.
What Blind People Dream About
The dreams of people who are congenitally blind offer a natural experiment in how sensory experience shapes dream content. People blind from birth have considerably less visual content in their dreams, but they show an increase in auditory, tactile, olfactory, and gustatory sensations compared to sighted people.12PubMed Central. Do congenitally blind people have visual dreams? Their dreams are clearly rich and immersive, just built from a different palette of sensory information.
What is more surprising is that some studies have documented reports of visual imagery even in congenitally blind individuals, despite these people never having processed visual input through their eyes. A 2023 review confirmed both the increased non-visual sensory content and the occasional presence of visual-like imagery, raising questions about what neural mechanisms could produce such experiences in people whose visual cortex was never driven by retinal input.13PubMed Central. Mental Imagery in Dreams of Congenitally Blind People One possibility is that the visual cortex, when freed from its usual job, gets repurposed during sleep in ways that generate something resembling sight. This remains an open question.
Lucid Dreaming
Most dreams are experienced without any awareness that they are dreams. The ordinary dreamer accepts impossible scenarios at face value because the brain regions responsible for critical self-reflection are largely offline. Lucid dreaming is the exception: the dreamer becomes aware that they are dreaming while still asleep. Despite having been physiologically verified for decades, the neurobiology of lucid dreaming remains incompletely understood. Preliminary neuroimaging data suggests involvement of prefrontal and parietal regions, which aligns with the idea that lucid dreaming involves a partial reactivation of the same executive circuits that support self-awareness during waking life.14PubMed Central. The cognitive neuroscience of lucid dreaming
Lucid dreaming is relatively rare as a spontaneous occurrence, though some people experience it frequently. Researchers have explored various induction techniques, from reality-testing habits during the day to targeted brain stimulation during sleep, with mixed results. The phenomenon is interesting not just for its novelty but because it provides a rare window into consciousness: a state where the dreamer can signal researchers (through pre-arranged eye movements) while still physiologically asleep, allowing scientists to time-lock brain measurements to reported dream events.
Can Outside Stimuli Sneak Into a Dream
If you have ever heard your alarm clock become a ringing telephone in a dream, you have experienced stimulus incorporation. A 2024 systematic review cataloged the ways external sensory stimulation during sleep can alter dream content and found two broad categories. Incorporation occurs when the stimulus shows up as an identifiable element in the dream, such as a spritz of water on the face becoming rain. Modulation is subtler: the stimulus changes general features of the dream, like its emotional tone or complexity, without appearing directly.15PubMed Central. Influencing dreams through sensory stimulation: A systematic review
The brain does not simply pipe external signals into a dream unfiltered. It weaves them into the ongoing narrative in a way that preserves the dream’s internal logic, however bizarre that logic may be. This is part of what makes dreams so interesting from a consciousness perspective: the sleeping brain is still monitoring the external world at some level, but it integrates that information on its own terms, fitting reality into the dream rather than interrupting it.
Do Other Animals Dream
You cannot ask a rat about its dreams, but you can look at what its brain does during sleep. Research has shown that during REM sleep, rats replay the same sequences of hippocampal neuron firing that occurred while they were navigating a maze earlier that day, and these replays unfold at roughly the same timescale as the original experience.16PubMed. Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep This replay is not random. External cues played during sleep can bias which memories are replayed: an auditory tone associated with a particular spatial task increases hippocampal replay of that specific memory.17PubMed Central. Biasing the content of hippocampal replay during sleep
Whether this constitutes dreaming in the subjective sense is unanswerable with current methods. What it does establish is that the neural machinery for replaying and reorganizing waking experience during sleep is not unique to humans. If dreaming is, at its core, the brain running through and recombining recent experience, then the foundations for it exist across a wide range of species with similar brain architecture.
When Dreams Break Through Into Movement
During normal REM sleep, your voluntary muscles are essentially paralyzed, a mechanism called muscle atonia that prevents you from physically acting out your dreams. When this mechanism fails, the result is dream enactment behavior, which can range from mild twitching and talking in your sleep to full-blown physical activity like punching, kicking, or leaping out of bed. The process likely begins with emotionally charged dream content that occasionally overwhelms the normal motor suppression.18PubMed Central. Dream enactment behavior: review for the clinician
In its clinical form, REM sleep behavior disorder (RBD), this breakdown of atonia becomes chronic and can lead to injuries for the sleeper or their bed partner. RBD has drawn considerable attention because it is now recognized as an early marker for neurodegenerative diseases. Dream content itself may have clinical and prognostic value in conditions ranging from schizophrenia and borderline personality disorder to temporal lobe epilepsy and dementia.19PubMed. The clinical use of dream content in modern psychiatry
The dreams of people with schizophrenia illustrate how dream content can mirror waking cognitive differences. Compared to controls, patients with schizophrenia report more nightmares but similar overall dream recall. Their dream narratives tend to be shorter, and they rate their own dreams as less bizarre than external judges do, despite a similar actual density of bizarre elements.20PubMed. Dream content in chronically-treated persons with schizophrenia This disconnect between perceived and actual bizarreness echoes the broader cognitive disruptions seen in schizophrenia during waking life, suggesting that the same processes shaping waking perception also shape the dream world.
Decoding Dreams From Brain Scans
One of the more remarkable developments in dream science is the attempt to decode dream content from brain activity. In a landmark 2013 study, researchers trained machine-learning models on fMRI data collected while participants viewed images while awake, then applied those models to brain activity recorded during sleep onset. The models were able to predict the visual content of the participants’ dreams at above-chance levels based purely on neural signals.21PubMed. Neural decoding of visual imagery during sleep
Follow-up work showed that the connection between dreamed objects and brain activation patterns operates across multiple levels of visual processing. Decoded features from the dream fMRI data correlated positively with the visual features associated with the dreamed object categories, and these correlations were strongest at mid- to high-level layers of a deep neural network model of visual processing, suggesting that dreaming engages abstract category representations rather than low-level pixel-like patterns.22Frontiers in Computational Neuroscience. Hierarchical Neural Representation of Dreamed Objects Revealed by Brain Decoding with Deep Neural Network Features More recently, researchers have begun exploring whether these decoded signals could be used to reconstruct dream imagery as video, though that work remains at an early, proof-of-concept stage.23arXiv. Making Your Dreams A Reality: Decoding the Dreams into a Coherent Video Story from fMRI Signals
How Dream Research Actually Works
Studying something that only exists in a sleeping person’s subjective experience poses obvious methodological challenges. The standard approach is the serial awakening paradigm, in which researchers wake participants at predetermined intervals throughout the night and immediately collect verbal reports of any mental content. A typical protocol might involve a dozen awakenings spread across early, middle, and late periods of sleep, covering all sleep stages.24PubMed Central. The memory sources of dreams: serial awakenings across sleep stages and time of night This approach is time-efficient and allows both within-person and between-person comparisons, which is valuable given how much dreaming varies from individual to individual and night to night.25PubMed Central. Assessing sleep consciousness within subjects using a serial awakening paradigm
The method has inherent limitations. Awakening someone disrupts the very thing you are trying to study. Reports depend on the person’s ability to remember and articulate an experience that was fleeting and often incoherent. And the lab environment itself changes how people sleep. Still, serial awakenings combined with EEG monitoring and, increasingly, neuroimaging have produced the bulk of what we know about when dreams occur, what they contain, and how the brain generates them. The field is limited not by lack of interest but by the fundamental difficulty of studying an experience that vanishes the moment the subject opens their eyes.