Summary: Why do some dreams feel like cinematic masterpieces while others are flat and chaotic? A large-scale study analyzed the semantic structure of more than 3,700 dream reports and reveals how personality, daily cognition, and even global events shape dreaming.
Using advanced natural language processing (NLP), researchers show that dreams are not random neurological noise. They represent a systematic reworking of waking life—blending fragments of routine settings with imagination—to produce immersive, often surreal experiences influenced by individual traits and shared circumstances such as the COVID-19 pandemic.
Key Facts
- Reinterpretation rather than replay: Dreams seldom replay waking events exactly. Instead, the brain recombines elements from work, school, and healthcare contexts into novel, immersive scenes.
- Mind-wandering predicts dream structure: People who frequently daydream tend to report more fragmented, rapidly shifting dream narratives.
- Belief shapes vividness: Individuals who consider dreams meaningful experience richer, more perceptually vivid dream content.
- Societal events leave traces: Dreams recorded during the 2020 lockdowns contained themes of constraint and heightened emotion, which faded as people adapted.
- AI can scale dream analysis: NLP models captured semantic and emotional patterns in dreams with accuracy comparable to human experts, enabling scalable studies of subconscious content.
Source: IMT
What makes some dreams vivid and immersive while others feel obscure?
A study led by researchers at the IMT School for Advanced Studies Lucca offers a clearer answer. Published in Communications Psychology, the research examined over 3,700 reports of dreams and waking experiences collected from adult volunteers aged 18 to 70. Across a two-week recording period, participants logged daily experiences while researchers collected detailed measures of sleep, cognition, personality, and psychological style.
Applying modern NLP techniques to this large, multimodal dataset, the team quantified the semantic structure of dream narratives. Their analysis indicates that dreams reflect a structured blend of personal traits and life experiences: stable characteristics such as an individual’s tendency to mind-wander, their interest in dreams, and subjective sleep quality all influence how a dream unfolds. External events—most notably the social impact of the COVID-19 pandemic—also leave identifiable signatures in dream content.
When comparing words and themes used to describe waking life versus dreams, the researchers found consistent transformation processes. Rather than straightforwardly replaying daily scenes, the mind reorganizes fragments—an office, a classroom, or a clinic—into scenes with altered perspectives, unfamiliar landscapes, and blended contexts. Dreams tend to emphasize perceptual features: vivid visuo-spatial detail, multiple interacting characters, and bizarre or unexpected events that depart from the ordinary coherence of waking thought.
Individual differences matter. People who report frequent mind-wandering commonly experience dreams that are more discontinuous and rapidly shifting. Conversely, those who attach importance to dreams—who report higher “dream interest”—tend to have more immersive, perceptually rich dream experiences. This suggests that attitudes and cognitive styles shape the brain’s allocation of processing resources during sleep.
The study also investigated societal influences by comparing a separate dataset gathered during the first COVID-19 lockdown with later reports. Lockdown-era dreams contained more references to limitations and displayed stronger emotional intensity, mirroring broader social stressors. Over subsequent months and years, those pandemic-specific patterns lessened, indicating that dream content adapts as people adjust psychologically to changing circumstances.
A notable methodological advance is the validated use of AI. Large language models and NLP pipelines were used to evaluate hypothesis-driven semantic dimensions and to identify lexical domains in a data-driven way. The models reliably detected emotional and thematic features that human raters would recognize, demonstrating that computational tools can rapidly and reproducibly analyze thousands of dream reports—opening new possibilities for research into memory, consciousness, and mental health.
Funding: Supported by the BIAL Foundation (grant #091/2020) and the TweakDreams ERC Starting Grant (#948891). The research was carried out at IMT School for Advanced Studies Lucca in collaboration with Sapienza University of Rome and the University of Camerino.
Key Questions Answered:
A: Not necessarily. The study finds that work often appears as a fragment that the dreaming brain recombines with other elements. If your workplace appears unchanged in a dream, it could suggest less of the creative, adaptive reprocessing that usually reshapes experiences during sleep.
A: Your level of interest in dreams matters. People who value and pay attention to their dreams often report richer, more immersive dream content; attitude toward dreaming correlates with perceptual vividness.
A: Yes—AI models analyze the semantic relationships between ideas, not just word frequency. In this study, NLP identified emotional and thematic patterns in dreams that matched expert human judgments, and it can do so at scale.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by editorial staff.
About this AI and dream research news
Author: Chiara Palmerini
Source: IMT School for Advanced Studies Lucca
Contact: Chiara Palmerini – IMT
Image: Image credited to Neuroscience News
Original Research: Open access. “Individual traits and experiences predict the content of dreams” by Valentina Elce, Giorgia Bontempi, Serena Scarpelli, Bianca Pedreschi, Pietro Pietrini, Luigi De Gennaro, Michele Bellesi, Giulio Bernardi & Giacomo Handjaras. Communications Psychology. DOI: 10.1038/s44271-026-00447-2
Abstract
Individual traits and experiences predict the content of dreams
Dreams are universal but highly personal. While past memories and concerns influence dreaming, how these influences change over time and how stable individual traits shape dream content has been unclear. This study quantified dream semantics across a large dataset of 3,366 dream and waking reports from adults collected between 2020 and 2024, alongside demographic, cognitive, psychometric, and sleep measures.
Combining language model–assisted assessment of predefined semantic dimensions with a data-driven lexical approach, the study found that dreams shift away from self-referential, thought-centered narration toward perceptual experiences dominated by visuo-spatial detail, multiple characters, and bizarre events. Stable traits—attitude toward dreaming, propensity to mind-wander, and subjective sleep quality—selectively shaped dream content. Data from the first COVID-19 lockdown showed increased references to limitations and heightened emotional intensity; these effects diminished over time. Together, the results indicate that both enduring personal traits and incidental experiences jointly influence the semantic structure of dreams.