Why do we dream?

The answer begins with an open question

The US National Institute of Neurological Disorders and Stroke says the exact purpose of dreaming remains unknown. Its public booklet describes emotional processing as a possibility, rather than an established explanation for every scene. That distinction matters: uncertainty about purpose does not make dreams inaccessible to research, and a promising hypothesis is not a complete account of an individual dream.

Three different meanings of why

It helps to choose the question before choosing an explanation. “How was this experience produced?” asks about a process. “Does having this experience do something useful?” asks about a function. “What does this scene mean to me?” asks about a personal response. Those questions can be interesting together, but one answer cannot automatically stand in for the others.

Consider an invented example: someone dreams of an old classroom and later thinks about a teacher they liked. The classroom is the reported content. Remembering the teacher is a present association. Deciding to write a thank-you note would be a separate choice. None of these descriptions, on its own, establishes why the brain produced that particular classroom.

A memory experiment gives a testable clue

Wamsley and Stickgold studied a virtual maze learned before sleep. Of 39 college students who completed the study, 21 entered the sleep group; 17 remained in its performance analysis after exclusions. Maze-related dreaming accompanied greater overnight improvement, but those participants started with poorer performance. After accounting for baseline completion time, the association was no longer statistically significant. This limits the claim of a learning benefit.

The result concerns a particular task and selected sample. Participants were not randomly assigned to dream about the maze. It supports a link between reported content and learning, not proof that consciously experiencing the dream caused improvement. The dream descriptions also differed from literal replays of the task.

An emotion hypothesis needs equally careful wording

Sterpenich and colleagues combined two studies. In one, 18 participants provided reports during laboratory awakenings with high-density EEG. Fear in reports was associated with activity in the insula and midcingulate cortex. In another, 89 participants completed dream records and an imaging task while awake. More reported dream fear was associated with lower responses to aversive images in several regions and greater medial prefrontal activity.

These observations are consistent with an emotion-regulation hypothesis. They do not establish that deliberately having frightening dreams improves wellbeing, or that persistent nightmares are necessary therapy. A group association with a laboratory response is narrower than a demonstrated benefit for a person.

Separate sleep, experience and later performance

Here is an original reasoning exercise. Imagine that a group sleeps after practising a puzzle, remembers puzzle dreams, and performs well the next day. Three things have been observed together: sleep, a report and performance. To claim that the experienced story itself produced the improvement, an experiment would need a way to distinguish that contribution from the other processes and differences between people.

Now imagine that someone recalls no puzzle dream but also improves. That does not automatically disprove every dream hypothesis; the hypothesis must first specify whether a remembered report is required. Equally, a favourable result cannot be credited to a dream merely because one was recalled. These are questions about the design of evidence, not instructions to interrupt sleep or test yourself.

Competing explanations can fit the same scene

Memory: a learning experiment asks whether reported task-related dream content accompanies later performance. Even a relationship would not establish that dreaming caused the improvement; the adjusted result in the experiment above illustrates that limit.

Emotion: a study can compare reported dream emotion with measured waking responses. The fear study above supports a relationship worth investigating, not a conclusion that each frightening dream is beneficial.

Process and purpose: identifying activity during sleep is different from showing what the conscious dream experience does. A proposed function needs a prediction that distinguishes it from alternatives, rather than an explanation that can fit any story afterward.

For an original example, a dream about rehearsing a presentation could resemble a recent task, contain emotion and still fail to show why dreaming exists. Ask what observation would count against each explanation. Personal resonance alone cannot settle the scientific comparison.

Personal significance is a separate choice

You can value a dream without assigning it a biological job. An imagined reunion may be worth remembering because you enjoyed it; a surprising landscape may suggest a drawing. These are optional ways to respond, not empirical claims about what dreaming evolved to accomplish.

Keep the description, your association and any intended action separate. For example: “I saw an old classroom”; “it reminded me of encouragement”; “I might contact a teacher if I still want to tomorrow.” This format leaves room for the association to change. It also avoids giving the dream authority over someone else’s intentions, a health decision or an expected future event.

Animal sleep offers evidence, but no spoken dream report

Bendor and Wilson trained rats on an auditory-spatial task and recorded hippocampal activity. During subsequent sleep, task-related sounds biased which spatial pattern was reactivated. The experiment measured neural replay, not the animal's subjective experience.

This distinction is important when a headline says researchers know what an animal dreamed. A measured pattern linked with earlier behaviour gives evidence about brain activity. It does not provide the animal's description of a scene, feeling or storyline. Inferring an experience and receiving a first-person report are different routes to knowledge; neither should be silently substituted for the other.

An observation of a sleeping pet is further removed from this controlled experiment. It cannot identify the exact remembered event behind a movement. The study helps investigate memory-related sleep processes while leaving the character of animal experience open.

Use the next guide for your actual question

For the sleep stages in which reports are collected, continue to REM and non-REM dreaming. For the learning experiments, use Learning, memory and dreaming. For research on reported emotion, use Stress, emotion and dream reports. These guides address different pieces of the scientific question.

For reflection on your own scene, Using a dream dictionary explains how to work with personal associations without treating them as established scientific meanings. You do not need to choose a theory before finding an experience interesting. A scientific question can end with uncertainty, and a personal note can remain a description.

Sources & their limits

These sources support the statements indicated above. Their authors do not endorse this website or provide an assessment of your experience.

1. NINDS — Brain Basics: Understanding Sleep (official booklet, 2025)

   Official overview: dream purpose remains unresolved.

   Cited in this guide · 2 places

   • The answer begins with an open question
   • Has science discovered one reason for all dreams?
2. Wamsley & Stickgold (2019) — Dreaming of a learning task is associated with enhanced memory consolidation: Replication in an overnight sleep study

   Specific maze-task association; no causal test of experiencing a dream.

   Cited in this guide · 3 places

   • A memory experiment gives a testable clue
   • Competing explanations can fit the same scene
   • Does dreaming about studying prove I learned better?
3. Sterpenich et al. (2020) — Fear in dreams and in wakefulness: Evidence for day/night affective homeostasis

   Dream-fear and waking-response associations, not a treatment trial.

   Cited in this guide · 3 places

   • An emotion hypothesis needs equally careful wording
   • Competing explanations can fit the same scene
   • Does fear in a dream mean it was good for me?
4. Biasing the content of hippocampal replay during sleep

   Original rat experiment: auditory-spatial task, hippocampal replay results and online methods; neural reactivation, not subjective dream reports.

   Cited in this guide · 1 place

   • Animal sleep offers evidence, but no spoken dream report

Put the context beside an experience

Continue through the science library