How Our Brains Play Angry Birds With The World

New research suggests we may use an "intuitive physics engine"—similar to those found in video games—to make rapid everyday inferences.

Every day our brains make lightning quick surveys of the physical world whether we realize it or not. We know how dishes would fall off the tray if the waitress doesn't carry it carefully. We know it's time for a child to come down from the tree if we see a branch bend a certain way. We know where to release the Angry Bird to beat the level we've been playing since the subway door closed. Most of the time, at least, we know what's coming.

Cognitive scientists have proposed any number of theories for how people perform these advanced evaluations so successfully, regardless of their grades in high school physics. Given the widespread affection for physics-related games, people seem to have some innate appreciation for the physical complexity of these concepts and moments. But the precise mental process we use remains a bit of a mystery.

A trio of MIT scientists has suggested a new theory: people possess an "intuitive physics engine" capable of running rich physical simulations on everything around us. These simulations are similar to the ones computers run during video games with two exceptions. They take into account uncertainty and they trade precision for speed; in technical terms, they're approximate and probabilistic.

"What happened in the past, what will happen—the way you form those judgments is based on these simulations supported by an intuitive physics engine," Peter Battaglia, lead author of a paper on the idea, recently published in Proceedings of the National Academy of Sciences, tells Co.Design. "It's a very powerful way of making these predictions."

The intuitive physics engine theory says that our brains develop three-dimensional geometric models of real-world scenes we wish to evaluate. These models consider static variables like the shape and spatial arrangement of objects in the scene, dynamic ones like motion and friction, and external factors like gravity. Using approximate rules of physics, our intuitive engines can simulate how these scenes will look a moment later, and suggest a probable outcome—as if our brains were playing Angry Birds with the world.

Battaglia and colleagues tested the theory in a series of experiments designed to gauge how we understand everyday scenes. One test was called "Will It Fall?" As the name suggests, study participants looked at about 60 different arrangements of building blocks and determined whether or not each would fall.

Image: Courtesy of Peter Battaglia

Other experiments tweaked the scene a bit (adding blocks with varying weights) and asked different questions (such as which direction the blocks will fall). The most complicated scene showed test participants a table with red and yellow blocks on top and asked them to imagine someone bumping into the table hard enough to knock them off. They had to decide whether red or yellow blocks were more likely to fall.

Image: Courtesy of Peter Battaglia

Across five of these psychophysical tests, people responded in a way that matched up well with the predictions made by an intuitive physics engine and much less well with other potential cognitive models. These rejected alternatives included a "ground truth" model, which runs a pure and precise physics simulation, as opposed to an approximate and probabilistic one. These ground truth models don't account for uncertainties like a gust of wind, or a minor misalignment.

Take responses to one "Will It Fall?" test that showed a stack of precariously balanced blocks (below). In fact, the blocks are arranged in a way that would not fall over, which a perfect ground truth physics model would recognize. But test participants believed the stack would topple over because their intuitive physics engine considered—and evidently accepted—the possibility of a slight but critical imbalance.

Image: Courtesy of Peter Battaglia

"It would take so many coincidences for the thing to stand up that basically you can say, 'I don't believe those things happened. I think this should fall over,'" Battaglia says.

The intuitive physics engine will have to pass subsequent experimental tests to hold up, but the recent research is a promising start. If it does stand as a solid explanation of physical intuition, Battaglia hopes educators will revise their teaching methods in response to the theory, perhaps by incorporating games into their curriculum. "It'd be great if you could play a video game and learn physics at the same time," he says.

Welcoming Angry Birds in the classroom—there's something our intuitive engines didn't see coming.

[Image: Flickr user Tomaž Štolfa]

Add New Comment


  • Peter Battaglia

    Simone, that's a good idea. We haven't yet tested people younger than 18, but we'd like to in the future. There is some very interesting evidence that people develop powerful intuitions about the physics of solid objects, fluids, and gravity as infants, even before they speak their first words (here's a great overview: I can't really say whether memories built from experience are more or less important than simulations, especially because their relative contributions are likely to vary between different activities and judgments. My suspicion is that our experiences play a crucial role, but in order to get so much out of them, and also to fill in the gaps where experience is missing, our brains use independent mechanisms such as simulation.

  • simone speziale

    Hi, very interesting results! But wouldn't you say that memories built from experience play an even more important role in making predictions, than the type of simulations you are suggesting? Was there any change in the answers reducing the age of the tested people to say less than 10, than less than 8, etc?

  • Peter Battaglia

    That's a great question Patrick, and I don't think we really know the full answer at this point. The "Intuitive Physics Engine" mechanism that we proposed here would be especially useful when a person needs to make predictions that are not very close to previously experienced events. That's because even though everyday scenes can be very different, they are often governed by the same underlying rules of physics, and so learning about one setting (eg. balls that collide with each other) can help you predict what will happen in another (eg. how towers of blocks will topple). There was another very interesting paper published earlier this year ( which proposed a way that people might use previous experiences when they judge the physical properties of objects, such as their mass. I think a great question for future research is whether, and how, mental simulators such as the Intuitive Physics Engine might cooperate with our previous experiences to help us make better physical judgments about our everyday surroundings.

  • patrick andrews

    One question which fascinates me is, given that each indvidual has a limited number of experiences, how can we say "this situation is close enough to some previous set of events to allow us to use it to predict the immediate future by rough analogy" ?

  • Chuck Mahoney

    Anyone who has thrown a ball accurately to a person catching it knows this. We don't have to run projectile motion analysis in our brains to figure out how much force we need, or at what angle we need to throw the ball for it to reach the target. This data is gained through experimentation and experience, saved in our brains, then recalled when needed. Similarly with judging how fast oncoming traffic is moving so we can pull on to the highway safely. It's actually pretty amazing.