Toy trumpets and cartoon lasers from old video games: This is the sound of zebra finch song. The birds weave a noisy rhythmic chatter of beeps and clicks and trills into complex melodies for courting lifelong mates and coordinating social time. Young birds learn tunes much like babies learn to talk—by mimicking adults and then practicing and refining.

They probably also learn by singing in their dreams.

REM sleep, the kind associated with vivid dreaming, is common not just across the mammalian kingdom, but has also been found in reptiles, cuttlefish, and birds. And during that deep phase of sleep, many species of birds, zebra finches included, silently sing their songs. Not only does activity in the parts of the birds’ brain dedicated to song follow a similar pattern when the birds are asleep as when they’re awake, this replay actually extends in at least some cases to the birds’ vocal muscles.

A couple of years ago, researchers from the University of Buenos Aires found a way to track muscle activity in a vocal organ called the syrinx of a bird known as the great kiskadee, a species of flycatcher. They then created synthetic songs from the data they collected and published their results in a journal called Chaos. “The possibility of entering the mind of a dreaming bird—listening to how that dream sounds—is a temptation impossible to resist,” study author Gabriel Mindlin, who specializes in bird song mechanics, told Newsweek.  

Read more: “The Dreams of a Bumblebee in Autumn”

Now, a team of scientists from University of California, San Francisco has tried to decode the dream songs of zebra finches in another way: using AI and neural recording technology. In ongoing work that hasn’t yet been peer reviewed or published, a team led by Walter Gonzalez, a neuroscientist at the Kavli Institute for Fundamental Neuroscience of UCSF, has found that activity patterns in the song centers of the brains of dreaming zebra finches actually matched the patterns they sang during the day only 5 percent or less of the time, suggesting that the actual notes in their dream melodies were different.

The variation is surprising because the songs male zebra finches sing are simple, unique, and repetitive: When awake, they sing their tunes with little variability. The findings so far suggest that learning through dreaming may be more about experimentation than straight rehearsal or passive memory storage, says Gonzalez.

The researchers will later play the notes decoded from the dream song back to the birds and record what is happening in their brains. Zebra finches only recognize their own songs, so their song circuits won’t become active if they hear an unfamiliar tune.

Gonzalez and his team also want to see if they can change the birds’ behavior by tweaking the neurons associated with song learning when the birds are asleep. They plan to teach the birds two songs when they’re awake, and then silence the neurons associated with one of the songs while they’re off in dreamland, to see what this does to singing patterns.

“Can you change the way a bird learns to sing by playing with their dreams?” Gonzalez asks in a statement. Ultimately, they hope their research will help us understand larger truths about sleep and dreaming and why it evolved.

Dreaming, it seems, arose long before humans came on the scene. So whatever function these nocturnal reveries serve may predate us, too.

Lead image by Tasnuva Elahi; with images by natian and chinnarach / Adobe Stock