A dream EEG and mentation database

Read in full here

A new open database connects brain activity with what people dream about

This article presents the DREAM database, a large shared resource that brings together sleep-brain recordings and dream reports from many different research labs. It was created to solve a long-standing problem in dream research: most studies are small and use different methods, making results hard to compare.

What did they do?

The authors gathered and standardised data from 20 separate studies, building a single dataset with 505 people and 2,643 awakenings, each paired with a dream report (or confirmation that no dream was recalled). They cleaned and aligned all the data so it could be analysed together. Examples include:

• EEG and MEG recordings taken just before awakening, all converted to the same format.

• REM and NREM awakenings, where participants said whether they had dreamed or not.

• Simple, consistent questions about dream length, vividness, emotions, and sensory details.

• Sleep stages re-scored using the same rules so the datasets could be merged.

What did they find?

Across the combined dataset, the authors found that patterns in brain activity can help predict whether someone is dreaming — not only during REM sleep, but also during lighter stages of NREM sleep. They also showed expected patterns: dream reports were rarer after deep sleep, and the EEG signals before awakening matched well-known sleep signatures.

Overall, the work shows that bringing many small studies together makes it possible to answer questions that no individual study could. The authors note that the database should help future research into how specific dream features connect to brain activity, and how dreaming works across different sleep stages.

Wong, W., Herzog, R., Andrade, K.C. et al. A dream EEG and mentation database. Nat Commun 16, 7495 (2025). https://doi.org/10.1038/s41467-025-61945-1