Technology with attitude

Dream Reading Research Predicts With 75–80% Accuracy Whether A Car, Computer Or Man Is In the Dream



Although our ultimate dream – to play around with a dream reading machine – has still not become a reality, scientists are a step closer to understand and reconstruct our dreams. A team of Japanese researchers used fMRI (functional neuroimaging) to scan the brains of three people as they slept, while also recording their brain waves using electroencephalography (EEG).As reported today, a Kyoto-based research team, led by Yukiyasu Kamitani of the ATR Computational Neuroscience Laboratories, can now predict certain features of dreams by looking at the brain activity of sleeping volunteers.

In 2008, Kamitani and his colleagues reported that they could decode brain activity associated with the earliest stages of visual processing to reconstruct images shown to participants. Now, they have found that activity in the higher order brain regions could accurately predict the content of the participants’ dreams.

This is how the research went on:

1. The sleeping volunteers were woken up whenever the researchers detected the pattern of brain waves associated with sleep onset.
2. The researchers asked the volunteers to tell what they had just dreamed about, and then asked them to go back to sleep.

This was done several times on different days, for each participant. Each volunteer reported having visual dreams six or seven times every hour, giving the researchers a total of around 200 dream reports. The researchers extracted keywords from the volunteers’ verbal reports, and picked 20 categories — such as ‘car’, ‘male’, ‘female’, and ‘computer’ — that appeared most frequently in their dream reports.

Kamitani and his team then selected photos representing each category and scanned the volunteers’ brains again while they viewed the photos. Then they compared these new brain activity patterns with those recorded just before the participants were woken up.

Yukiyasu Kamitani dream reading
Yukiyasu Kamitani and his team can predict certain features of dreams by using fMRI and EEG to monitor the brain activity of sleeping volunteers.

The researchers analysed activity in brain areas V1, V2 and V3, which are involved in the earliest stages of visual processing and encode basic features of visual scenes, such as contrast and the orientation of edges. They also looked at several other regions that are involved in higher order visual functions, such as object recognition.

“We built a model to predict whether each category of content was present in the dreams,” says Kamitani. “By analysing the brain activity during the nine seconds before we woke the subjects, we could predict whether a man is in the dream or not, for instance, with an accuracy of 75–80%.”

The findings, presented at the annual meeting of the Society for Neuroscience in New Orleans, Louisiana, earlier this week, suggest that dreaming and visual perception share similar neural representations in the higher order visual areas of the brain.

“This is an interesting and exciting piece of work, it suggests that dreaming involves some of the same higher level visual brain areas that are involved in visual imagery.” says neuroscientist Jack Gallant who also conducts a research on reconstructing videos from film viewers’ brain activity, which might one day lead to the invention of a device capable of visualizing dreams.

Kamitani and his colleagues are now trying to collect the same kind of data from the rapid eye movement (REM) stage of sleep, which is also associated with dreaming. “This is more challenging because we have to wait at least one hour before sleeping subjects reach that stage,” Kamitani says.