Caitlin House, via Flickr
Caitlin House, via Flickr

Why Do Babies Twitch in Their Sleep? (Adorable Video Edition)

When I brought my puppy home last August, I knew he would be fun to play with. I had no idea how entertaining he would be when asleep. He dozed constantly, and more often than not, his whole body — legs, tail, lips, eyes, ears — would twitch. This isn’t a quirk of canines. Sleep twitching happens to “literally every mammal that has been looked at”, says Mark Blumberg, a psychology professor at the University of Iowa. Dogs, cats, rats, ferrets, sheep, squirrels — they all twitch. Even whales twitch their flippers. “I have YouTube videos of a guy who recorded his girlfriend’s toes when they twitched,” Blumberg says.

Speaking of videos…(the first one is Blumberg’s dog, Katy):

I undoubtedly spent too much time in the past couple of days doing YouTube searches for twitching babies. What’s funny about many of these videos is the commentary of those behind the camera. They tend to say one of two things: “OMG, look at that spaz!” or, “Awww, he’s dreaming.” And that’s how sleep researchers have traditionally thought of twitches, too, according to Blumberg.

“The sleep field really started off in many ways as an offshoot of Freudian psychoanalysis and the study of dreams,” Blumberg says. “People see these movements and they think, ‘Oh, Fido is chasing rabbits in his dreams.’ But it turns out that that’s almost certainly not the case.” In an engaging new review in Current Biology, Blumberg argues that these sleep twitches actually have an indispensible purpose: to teach a newborn what all of its limbs and muscles can do, and how to use them in concert to interact with the big, wide world.

The first big study to propose this idea was published more than 40 years ago in Science. Howard Roffwarg, then director of the Sleep Laboratory at the New York State Psychiatric Institute, described the behaviors and brain-wave patterns of newborn human babies as they sleep. He noted that a newborn spends “one-third of its entire existence” in a REM state, with intense brain activity and continuous muscle contractions.

“Grimaces, whimpers, smiles, twitches of the face and extremities are interspersed with gross shifts of position of the limbs,” Roffwarg wrote. “There are frequent 10- to 15-second episodes of tonic, athetoid writhing of the torso, limbs, and digits.” Since newborns can barely see, the idea that these spasms are useless byproducts of their dreams is unlikely, he added. What if, instead, twitches play a key role in the development of the nervous system?

That paper has been cited more than 1,000 times, but it took awhile to percolate*. A decade ago, two big Nature papers reinvigorated the idea that sleep twitches are important, Blumberg says. In the first, Swedish scientists reported that in young rats, spontaneous muscle twitches during sleep help program the cells in the spinal cord to carry out the withdrawal reflex. In the second paper, a French group showed that sleep twitches in newborn rats trigger patterned bursts of neuronal firing that are known to be important for motor coordination. Blumberg’s own experiments have found similar things; last year, for example, he reported that newborn rats twitch their whiskers frequently during sleep, and that these twitches drive certain bursts of activity in several brain regions.

Blumberg uses nifty high-speed video to precisely track the jerky movements of baby rats as they fall asleep. He doesn’t use any anesthesia in these experiments, so I asked him how he manages to get the animals to fall asleep on command. “The hard part is keeping them awake!” he said. Turns out newborn rats cycle from asleep to awake every 10 to 30 seconds. The cycle: They wake up, stretch, yawn, kick, and lift their head around. After about five seconds, they suddenly go limp, with no movement other than breathing. Then individual twitches begin — a limb here, tail there. “Then it starts to build, and almost starts to get this real powerful look of epilepsy to it,” Blumberg says. Then they wake up and it starts all over again.

Here are a couple of Blumberg’s videos. The one on the left shows the movements in real time; the one on the right is in slow-motion:

To the naked eye, the movements seem random. But Blumberg’s experiments have shown that the flailing is actually quite ordered in space and time. For example, when an animal brings its right elbow in toward its shoulder, there’s a high probability that the left elbow will immediately follow in the same pattern. Similarly, on the same limb, if the shoulder moves in toward the body, there’s a high probability that the elbow would then flex. Blumberg suspects that these predictable couplings are building blocks that help the developing motor system learn more complex behaviors.

“The brain is trying to understand, what are my limbs, how many do I have, and how many joints, and muscles, and how do they all move together?” he says. Once these simple commands are learned, he continues, the brain can use them to learn more complex sequences. “So that later, you can fire off a command somewhere in your mind, and generate a whole series of joint movements that would bring a bottle to your mouth, or make it possible to step.”

Nobody knows for sure how to read this code — that is, how any particular pattern of twitches leads to a specific complex behavior. But Blumberg says the future of figuring this out is with robots. Researchers can design computer simulations of simple neural networks, program in some random muscle contractions, and see what kinds of circuit patterns emerge. For example, roboticists Hugo Gravato Marques and Fumiya Iida of the Swiss Federal Institute of Technology Zurich (who co-authored the new review) have used such simulations to show that twitches help form the spinal cord’s withdrawal reflex — a neat confirmation of the earlier Nature paper. In the future, the robots will get more sophisticated, modeling twitches in multiple joints and multiple limbs, Blumberg says. “These feedback loops all have to be integrated and mapped, and it’s a very difficult thing to study in an animal.”

I asked Blumberg how the rest of the sleep field has responded to these ideas. He said he had just been to a big sleep conference in Baltimore and that, for the most part, sleep researchers still aren’t giving much thought to development. “I can’t tell you how many people have theories about sleep, and they all want to have a grand theory of sleep.” Some people think sleep is for memory consolidation, others that it’s for pruning synapses, or conserving energy, or even just limiting the time we have to make stupid decisions and put ourselves in danger.

But for Blumberg, the question, What is sleep for? is just as silly as, What is being awake for? Just as being awake is a good state for eating, drinking, and reproducing, perhaps sleep just happens to be the best time for consolidating memories, saving energy, and learning motor patterns. “What we have to come up with is the reason why sleep is so conducive for all of those things.”


Blumberg wrote a comprehensive historical review of the idea in Frontiers in Neurology 

You can read about other scientists using biological principles to build robots in my 2011 feature in New Scientist