Researchers at Italy's IMT School for Advanced Studies Lucca ran natural language processing over more than 3,700 reports of dreams and waking experiences collected from 287 adults, and concluded that the sleeping brain is neither a tape recorder nor a random noise generator. Instead, dreams appear to recombine memories, emotions, familiar places and imagined scenarios into new constructions.

The study, published in Communications Psychology, paired two weeks of dream and daily-experience diaries with data on participants' personalities, sleep quality, cognitive abilities and psychological traits. NLP was used to compare the semantic structure of waking and dream reports. Everyday settings such as workplaces, hospitals and classrooms turned up blended with unrelated locations, shifting perspectives or unfamiliar surroundings, and separate parts of a person's life merged into single scenes.

Dream content also tracked individual traits: habitual mind-wanderers reported dreams that hopped rapidly between scenes, while people who considered dreams personally meaningful described more vivid, immersive ones. A comparison with dream reports gathered by a Sapienza University of Rome team during COVID-19 lockdowns showed those dreams contained more references to confinement, barriers, restrictions and heightened emotion — the pandemic leaked straight into the dream corpus.

Lead author Valentina Elce framed the takeaway as dreams being a dynamic process shaped by identity and lived experience rather than a simple replay of the past. The NLP pipeline's assessments closely matched those of independent human reviewers, suggesting the method could spare future researchers the chore of hand-coding thousands of dream reports.

Caveats remain: the work depends on people remembering and describing dreams accurately, and it says nothing about why we dream at all. It also cannot explain why your subconscious keeps scheduling exams you never studied for.