Bouncing Babies Betray Awareness of Others' Beliefs

Dec 29 2010 Published by under Behavioral Neuro, Neuroscience

...if only I could have found synonyms for "awareness" and "adult" or "other person" that started with life is eternally unfulfilled now.

Today's post on babies is dedicated to Glendon Mellow of the Flying Trilobite, who just yesterday welcomed a bouncing TriloBoy! Congratulations to Glendon!!! That's going to be one amazingly awesome family. 🙂


So let's talk about the ability to recognize other people's beliefs, and other people's point of view. It's a skill that most people think other people lack entirely (especially around the holidays, it seems). But it turns out that all humans can understand what other people believe. And not just adult humans...BABIES.

(Though I still think the planet would be a better place if we ALL had one of these. Fire at will.) Kovacs et al. "The Social Sense: Susceptibility to Others’ Beliefs in Human Infants and Adults" Science, 2010.

Now this paper isn't EXACTLY about point of view, but it is about knowing the beliefs of others. Humans HAVE to be able to know, to some extent, the mental states of other humans. If everyone really DID walk around in the world seeing only their point of view on everything, society would be completely unable to function. I'm not just talking high politics here, I'm talking things like collective hunting, or, for example, playing soccer. If you're playing soccer, you've got the ball and you're headed toward the goal, you need to be able to infer that your team mates are somewhere nearby and ready to take over the ball if you need to pass. You need to infer that the opposing team has the goal of stopping you. We generally take these things for granted, but we go through our lives every day inferring other people's beliefs and points of view.

And this is complicated. It involves holding your own mental state, and ANOTHER's mental state, in your head at the same time, and often reconciling the two when it seems that they conflict. This is part of what is called a "Theory of Mind", which is our ability to know our own beliefs, and to understand that other people have beliefs which may differ from yours (whether or not your belief or the other guy's belief is RIGHT is another question entirely).

Many people believe that the Theory of Mind develops around 4 years old. The test for this goes like this.

1) Show a 5 year old and a 3 year old another child with a toy.
2) Have the other child put the toy in a box and walk away.
3) Move the toy from the box to a bag on the other side of the room.

Now ask the kids where, when the other child comes back, she should search for her toy. The 5 year old will usually say the kid will look in the box, where the child first left the toy, as they will not realize it has been moved. A 3 year old will say that the kid should look for the toy in its present location in the bag, rather than in the box, because the 3 year old (presumably) doesn't understand that the kid now out of the room doesn't know the toy has been moved.

The 5 year old will then call the 3 year old a stoopidhead. Fighting will probably ensue.

But the question becomes: is this test accurate, and does a child being able to 'pass' it prove theory of mind develops around 4? After all, a younger child could "get it" and still get the test wrong due to issues with problem solving, or being unable to inhibit their own belief from taking precedence, an inability to deal with conflicting beliefs, rather than simply knowing the other person has a different belief. Some studies have shown that, in fact, this test isn't good enough, that younger children could pass the test, if the test itself were simpler.

These authors thus made up a test, testing whether people understood the difference between their belief and the belief of another agent. And they started with adults playing a game.


Above you can see the game. Goes like this. You've got your little gnome, and you've got a ball, and you've got a block. You start with the ball visible. Then four different things can happen.

1) The ball hides behind the block and the gnome sees it happen.
2) The ball leaves the scene and the gnome sees it happen.
3) The GNOME leaves and the ball stays hidden
4) The gnome leaves and the ball goes AWAY.

Then the gnome comes back. Either the ball is there or it isn't. The object of the task was to click your mouse as soon as you detected the ball. But here's the trick.

The thing they were looking for here was reaction time. The idea is this: If the person knows the ball isn't there, he'll click slower. If the person knows the ball isn't there and the GNOME knows the ball isn't there, he'll click slower. That's baseline. The reaction time difference they were looking for was when the ball WAS there, and the gnome either did or did not see the ball stay. The idea is that you'll be fast when you know the ball is there, but you'll be FASTER when you know the ball is there and you ALSO know the gnome thinks it's there.


And it actually kind of worked. You can see the longest reaction time (second bar down) was when neither the person nor the gnome believed the ball was there (what's called "true belief"). The ball isn't there. Similarly, the people had the shortest reaction time when the person and the gnome both believed the ball was there (top bar), another true belief. The reaction times were a little longer when the gnome's belief was at odds with what the humans believed (the bottom two bars). The error bars don't look great but they say it's significant. This suggests that people not only know what the gnome's belief IS, they are aware of it and process it unconsciously, and this affects their reaction time. Not only that, this worked even when the gnome didn't come back at the end. When they had two true beliefs instead of one, participants reacted faster. When they substituted a pile of boxes for the gnome, there were no differences in reaction time, showing that it requires another "agent" like a gnome, not just a pile of stuff.

Now, how do we test this in BABIES?

Well, you can't tell a 7 month old baby to click a button, now can you. But what you can measure is their sense of SURPRISE, measured by how long they stare at final scene. If the ball is there and they know that it shouldn't be, they should show more surprise, and stare longer, than if they knew the ball was supposed to be there. For the babies, they simplified it down, and their baseline was when the baby and the gnome thought the ball was supposed to be there, and it WASN'T. This made the babies stare longer, detecting something wrong.


You can see the top bar, where the baby and the gnome both thought the ball was there and got fooled, the babies stare longer than on the bottom bar, where the baby and gnome know the ball isn't supposed to be there, and they were correct.

And then they did the crucial experiment, testing the babies on how long they stared when they knew the gnome would have a false belief, thinking the ball was behind the box when it wasn't.


The bottom bar is the normal comparison, where both the gnome and the baby know the ball isn't there. The top bar, however, shows the baby knowing the ball isn't there, and the gnome not knowing. And the babies STARED LONGER, which suggests that they had realized that the gnome thought the ball was there, and taken that into account.

This whole study suggests that not only adults, but BABIES as young as 7 months can tell when another thing (in this case a gnome) is an "agent", and can understand that that agent has beliefs which are different than their own. The authors believe that this means that humans have an innate "social sense" which allows us to have a theory of mind even when we are babies, realizing that other people have different beliefs than we do.

There is one problem with this study. In the adult group, they had a bunch of boxes stand in for the gnome during a whole test. When they did that, the people's reaction times reflected only their own beliefs, and served as proof that they had to have an observer that was thought to be an agent, rather than just a pile of stuff. But in the baby condition, they didn't do this. They DID substitute the boxes at the very end, rather than having the gnome come back in, and this didn't change the babies' responses, which they interpret as the babies' still computing the beliefs of the gnome. I'm not so sure about that, and I think they might have done a condition with boxes all the way through. But still, it's a really interesting study! And it highlights how incredibly important it is to humans to be able to understand other people's beliefs and points of view. Even if we know they're wrong. 🙂

Kovacs, A., Teglas, E., & Endress, A. (2010). The Social Sense: Susceptibility to Others' Beliefs in Human Infants and Adults Science, 330 (6012), 1830-1834 DOI: 10.1126/science.1190792

4 responses so far

  • AK says:

    I wonder how the babies would react with cartoon boxes, Teletubbies, and the like? (And what about cats, cartoon cats, etc.) Perhaps they could come up with a theory of innate (perceived) agency. I also wonder if it's an all-or-nothing thing, or a gradient depending on perceived likelihood of a potential agent having a point of view.

  • James Kornell says:

    If the question is, can babies infer intention, then a pretty simple experiment is telling: lean over a baby with hands outstretched ("I'm benign and I'm going to pick you up"), check the body language. Do it again with your arms crossed tightly over your chest. Check the body language.

    This seems more like an experiment showing babies can parse pictures (incredibly sophisticated), babies can identify particular collections of squiggly lines as representing human-like agents (incredibly sophisticated), and babies can project conjoined intention with frustrated action into extremely abstract representations (like static 2D images of a dynamic 3D world). Amazing, yes, but not for the reasons (I don't think) the experimenters claim.

    Sarah Blaffer Hrdy is awesome for exposition of what babies do.

  • mokawi says:

    AK: it's been done. Kris Onishi, who's probably the first to have babies pass the false belief test, used color boxes and a toy. I also heard of tests with puppets; although I'm not sure if there's anything about teletubbies in particular. In any case, there's been quite a few of these.

    Kornell: sometimes, jargon kills it all. It feels like you make an interesting point, but I can't say for sure what it is. Could you make it again?

    Is it me, or is 3½ years a bit long for knowledge to pass from practice to words?

    • Jim Kornell says:

      mokawi -- Sorry. For the second para, we look at pictures and understand their semantics. Other animals don't do this, because flat and static is utterly different from fully-dimensional. Babies infer a human narrative from a flat screen. I suppose the experimenters have to take that for granted, but babies are little geniuses. -- Jim

Leave a Reply