Japanese scientists figure out how to read ‘visual contents’ of dreams with brain scans

A team of Japanese scientists have revealed they can scan your brain and tell what you are dreaming about at least 60% of the time.

In a study published Thursday in the journal Science, the team from Japan revealed how they were able to figure out what people were dreaming about.

“In this study, we tried to read out what we called the visual contents of dreaming from brain activity during sleep which is measured by functional MRI (fMRI),” said team scientist Yukiyasu Kamitani in an interview for Science’s podcast.

To do this, the scientists hooked subjects up fMRIs and then tracked their brain activity when they slept, and when the brain activity became the most active and indicative of visual stimuli, the scientists woke the subjects up and asked them to explain what they saw.

“Dreaming often accompanies vivid visual experience, and we assumed that visual content represented disparity between common brain activity patterns between dreaming and perception during wakefulness,” Kamitani explained to Science. ”By using context specific activity patterns associated with stimulus perception, we are able to reveal dream contents from brain activity during sleep.”

This wasn’t something that they just did a few times. Each subject was woken up over and over and over again.

When hooked up to an fMRI “the subject verbally reported what they saw [in their dream], and then went to sleep again. We repeated this procedure to get at least 200 visual reports from each subject,” Kamitani said.

That data was then correlated and organized so that it could be parsed out in a meaningful way.

“Since verbal reports were not well structured, we performed text analysis on the reports…. We group them into about 20 basic visual categories,” Kamitani said on the podcast. “So each verbal report can now be represented by a vector with elements indicating the presence or absence of each visual category.

“We collected images from the Internet depicting each of the basic visual categories, and presented them to the subjects, while measuring with the [fMRI], and that was used for the training of the decoder.”

The decoder that Kamitani is talking about is a database built from brain scans when a person is awake, that data is then mixed and correlated with the “dream” reports to create a predicative matrix.

That matrix was then used on fMRI scans of new sleepers to predict what they were dreaming about, which it could do about six out of ten times.

The reason it the computer was able to do this was because the brain’s reaction to images seen by the eyes was very similar to those experienced in dreams.

“There is a similarity amongst the subjects, so from that result, we could pick up some basic dream content and then we can build a model from those base contents, and they may apply to other people,” study co-author Masako Tamaki told Live Science.