“From the sky, I saw something like a bronze statue, a big bronze statue,” the drowsy man told neuroscientists as he lay inside an MRI machine that moments earlier had been recording his brain activity as he dozed. “The bronze statue existed on a small hill. Below the hill, there were houses, streets, and trees in an ordinary way.”

This description of a just-interrupted dream, along with nearly 500 more descriptions gathered from the man and two other research subjects in a similar fashion, is the basis for this report by Horikawa and colleagues. Using these descriptions of dreams, the authors have, for the first time, successfully predicted images seen in sleep based exclusively on MRI scans of brain activity. A neural decoding approach was used in which machine-learning models predicted the contents of visual imagery during the sleep-onset period. They were able to predict what subjects had seen with 60% accuracy, which is higher than can be attributed to chance.

Rather than waiting more than an hour for subjects to enter rapid eye movement (REM) sleep, the researchers took advantage of the frequent hallucinations that occur during the onset of sleep, called stage 1. The authors of the report suggest that this technique can be applied in clinical sleep research and eventually to help people who experience bad dreams.

Horikawa T, Tamaki M, Miyawaki Y, Kamitani Y: Neural decoding of visual imagery during sleep. Science 340(6132): 639-642 (2013).


This entry was posted in Uncategorized. Bookmark the permalink.

Comments are closed.