You close your eyes. Hours later, you surface from a world that felt real: a childhood home you have not seen in decades, a conversation with someone long gone, a staircase that leads nowhere. Why that scene? Why that night?
A landmark study published in Communications Psychology offers the most detailed answer yet. Dreams are not random neural static. They are shaped by your personality, your cognitive traits, and your real-world experiences in systematic, measurable ways. The researchers analyzed nearly 2,000 dream reports using artificial intelligence and found that the content of a dream is a fingerprint of the person who dreamed it.
What They Found
The study, led by Giacomo Handjaras and colleagues at the IMT School for Advanced Studies Lucca and Sapienza University of Rome, enrolled 287 participants aged 18 to 69. Over two weeks, each person kept a diary of both their dreams and their waking experiences, yielding 2,038 dream reports and 1,679 waking reports. The scale is what sets this work apart from previous dream research, which typically relied on small samples.
To analyze this mountain of text, the team turned to natural language processing. Three large language models (LLaMA 3, ChatGPT-4, and ChatGPT-4 Turbo) rated each report across 16 predefined semantic dimensions, including visual perception, bizarreness, social interaction, and emotional intensity. Separately, a statistical technique called Non-Negative Matrix Factorization extracted 32 lexical domains from the raw vocabulary of the reports. Human raters validated the AI judgments, and the agreement was robust: all correlation coefficients exceeded 0.60, with a mean of 0.65.
Dreams versus Waking: Two Different Worlds
The first major finding is that dreams and waking thoughts are qualitatively different. The differences are large by statistical standards.
| Dimension | Effect Size (Cohen’s d) | Direction |
|—|—|—|
| Visual perception | 1.52 | Stronger in dreams |
| Spatial perception | 1.40 | Stronger in dreams |
| Bizarreness | 2.85 | Much stronger in dreams |
| Social interaction | 1.40 | Stronger in dreams |
| Self-referential thought | -1.77 | Stronger in waking |
| Agency (sense of control) | -1.58 | Stronger in waking |
| Time anchoring | -1.20 | Stronger in waking |
A bizarreness effect of 2.85 standard deviations is enormous. It means that the strangeness of dreams is not a subtle difference from waking cognition. It is a defining feature, as central to dreaming as sight is to vision. Participants described dreams as having an immersive, spectator-like quality where the dreamer experiences events rather than directing them.
Lexically, dreams are enriched for animals, nature, locations, objects, food, and thriller elements. Waking reports, by contrast, are anchored in self-reflection, agency, and a firm sense of time.
Personality Shapes the Story
Here the study breaks genuinely new ground. The researchers correlated dream content with 15 individual difference measures, including personality traits, cognitive abilities, sleep quality, and chronotype (morning versus evening preference). The results show that who you are predicts what you dream.
People with a positive attitude toward dreaming reported dreams with heightened arousal, bizarreness, spatial detail, and visual richness. Those with a high propensity for mind-wandering experienced more bizarre dreams with more shifts in setting, as though a wandering mind during the day carries its restlessness into the night.
Poor subjective sleep quality was linked to more bizarre dreams and a higher number of object references. Higher visuo-spatial memory scores predicted more object references as well, suggesting that people who think in terms of spatial layouts and objects carry that cognitive style into their dreams.
Evening chronotypes showed a different pattern: they reported more communication content (talking, messaging, conversing) in their waking lives than in their dreams, while morning types showed the opposite distribution. Younger participants included more job-related details in their dreams.
Interestingly, some measures produced no meaningful link. Trait anxiety, the vividness of mental imagery, and verbal memory did not significantly predict dream content, suggesting that certain psychological dimensions leave less of a trace on the nightly narrative.
The Pandemic Left Its Signature
The study also captured a natural experiment. Some of the dream data were collected during the COVID-19 lockdown in April and May 2020. During that period, dreams contained significantly more references to limitations, confinement, and emotional intensity. Over the following years, those themes gradually faded as psychological adjustment occurred. The finding is striking evidence that dreams track not only stable personality traits but also the brain’s long-term adaptation to collective stress.
AI as a Dream Interpreter
A practical implication of this work is methodological. The study demonstrates that large-scale computational dream research is not only feasible but reliable. Where earlier dream studies relied on human coders scoring a handful of reports by hand, AI can now process thousands of narratives with consistency approaching human judgment. This opens the door to studies that were previously impractical: tracking dream content over months or years, comparing clinical populations, and probing the links between dreaming and mental health at scale.
The research was funded by the BIAL Foundation and the TweakDreams ERC Starting Grant.
Why It Matters
Dreams have long been considered a window into the unconscious mind. This study suggests they are also a window into the waking mind: your cognitive habits, your emotional history, your sleep quality, and even your age all leave a measurable signature in the stories you generate at night. For researchers studying consciousness, memory consolidation, and emotional processing, the finding that dreams are systematically patterned rather than random opens new avenues for investigation.
For clinicians, the prospect is tantalizing. If dreams reliably reflect psychological states, they could become a tool for monitoring mental health, tracking recovery from trauma, or detecting early signs of conditions like depression or post-traumatic stress disorder.
Limits
The study has important caveats. The findings are correlational, meaning the researchers cannot prove that a given trait causes a specific dream feature. Dream reports are self-reported and written after waking, which introduces memory biases. And the sample, though large for a dream study, is predominantly Western and educated, leaving open the question of how culture shapes the dream landscape.
The Bottom Line
You are not just the hero of your dreams. You are the author. Your personality, your cognitive style, your sleep quality, and even the collective events you have lived through all leave their mark on the scenes your mind constructs each night. The science of dreaming is entering a new era, one where AI helps us decode the stories we tell ourselves in the dark.
Source
Elce V, Bontempi G, Scarpelli S, Pedreschi B, Pietrini P, De Gennaro L, Bellesi M, Bernardi G, Handjaras G. Individual traits and experiences predict the content of dreams. Communications Psychology. 2026;4:69. DOI: 10.1038/s44271-026-00447-2.