MRI scans of a sleeping person’s brain can help predict what’s seen in the land of Nod.
Researchers in Japan report in the journal Science on Thursday that they could predict images with 60 percent accuracy. First, they watched three subjects as they fell asleep and woke them in early sleep stages to ask them what they saw in their dreams. From those reports, the researchers built a database that grouped together the things the subjects saw in their dreams (e.g. “house” and “hotel” were grouped into a ”building” category). Later, when the volunteers were awake, the researchers showed them photos that matched these categories while recording their brain activity with MRI.
The researchers next built a computer model that linked patterns of brain activity to different types of images. This model was able to predict what kinds of images the subjects saw in their dreams in subsequent naps at a rate slightly higher than would be expected by chance.