In your every waking moment, whether you mean to or not, you are absorbing new experiences, and changing your brain. Specifically, some of your neurons become more strongly connected. The sites where they meet, known as synapses, become larger and more numerous, and an electrical signal in one of the neurons more easily triggers a signal in another. This is how we learn and store memories, in the changing strengths of our synapses.

But there’s a limit to this process.  

It takes a lot of energy to maintain these connections, and since we only have so much energy to spare, we can’t keep strengthening our synapses indefinitely. Nor would we want to. If our synapses kept getting stronger, our neurons would eventually become twitchy and hyperactive, leading to seizures or epilepsy. “In theory, the system could get to total saturation—all synapses would be strong and couldn’t get any stronger, and you wouldn’t be able to encode any more information” says Richard Huganir from Johns Hopkins University.

Brains have a way of avoiding this disaster. Neurons can scale down their synapses en masse, weakening them proportionally so that their relative strengths are the same but their absolute strengths go down. If one synapse was previously stronger than another, it would stay that way, although both would become weaker.

In 2003, Chiara Cirelli from the University of Wisconsin-Madison theorized that this mass downscaling happens specifically while we sleep. In fact, she argued, it might be one of the reasons that sleep exists at all—to provide a quiet time when our brains can effectively renormalize our synapses, ready for another day of learning. That may partly explain why sleep is so universal among animals, and why our mental abilities take a hit after a sleepless night. Sleep is the price we pay for the ability to learn, and it’s non-negotiable.

Now, working independently, Cirelli and Huganir have both found support for this idea.

Cirelli’s team collected the brains of mice that were either awake or asleep, and used a powerful microscope to measure the size of almost 7,000 synapses. “It took four years, with around six people working manually,” she says. Chief among them was Luisa de Vivo, who found that, on average, synapses seem to shrink during sleep. Specifically, the contact area where the two neurons meet is 18 to 20 percent smaller in sleeping animals, compared to waking ones—a clear sign that they had weakened. And as predicted, they shrank by roughly the same proportion, maintaining their relative strengths.

While Cirelli’s team was focusing on the physical size of sleeping synapses, Huganir’s group looked at their chemistry. Specifically, he looked at receptor proteins—little docking stations, found at synapses, which allow neurons to receive chemical messages from their neighbors. The levels of these receptors, and particularly a class known as AMPA receptors, are a good indicator of a synapse’s strength. And that strength, as team member Graham Diering showed, falls during sleep.

He isolated large numbers of synapses from the brains of mice, both asleep and awake, and measured the levels of thousands of proteins. He also tagged some receptors with fluorescent molecules, and used a microscope to track them in the brains of living rodents. Both techniques revealed that many receptors, AMPA ones included, move away from the synapse while the mice snooze. “The composition of the synapse is really totally changing from wake to sleep,” says Huganir.

But #notallsynapses. Based on her data, De Vivo calculated that around 20 percent of the synapses—the largest ones, and therefore the strongest—were unaffected by this downscaling. And it’s not clear what that means. “Our interpretation is that the largest synapses are those that have been there for a long time,” says Cirelli. “They may be the repositories of very strong memories—like the name of your mum—which you’re very unlikely to forget, even if you’re sleep-deprived. The majority of synapses, which are scaled down, may be the ones more engaged in what happened recently. If they’re not linked to anything relevant over a few days, they disappear. That’d be my guess.”

The next step, she says, is to follow the very same synapses over time, to see how they change physically and chemically as an animal learns new tasks, falls asleep, and wakes up. If a synapse is involved in recent learning, it is spared the nocturnal downscaling that affects the majority of its peers?

It’s possible that synapses might be both weakened and strengthened during sleep, says Amita Sehgal from the University of Pennsylvania. “In adult animals exposed to considerable sensory stimulation during their waking hours, down-scaling may dominate,” she says. “But in other conditions, such as early life,” strengthening may be more important.

There are other theories about the function of sleep. One says that sleep provides the brain with a chance to clean itself out, removing harmful toxins and repairing worn out cells. Another suggests that sleep gives the brain a chance to consolidate memories that were created during the day. The idea of sleep as a synaptic-rescaling period doesn’t contradict any of these other explanations—and might actually add to them.

“In recent years, there’s been evidence that sleep is important not just for remembering relevant stuff, but also for forgetting irrelevant things,” says Cirelli. Perhaps the mass-downscaling of synapses is part of that ‘smart forgetting.’ “If you think about how many details there are in your life, you’d better forget!”