Christmas Day is a time for opening presents and finally ending the suspense of what one is receiving this year, but chances are some of us may have already guessed what’s under the wrapping—perhaps by strategically shaking the boxes for clues about its contents. According to a November paper published in the Proceedings of the National Academy of Sciences, if someone happened to see you shaking a wrapped gift, they would be able to tell from those motions what you were trying to learn by doing so.
“There are few things more delightful than seeing a child’s eyes light up as they pick up a present and wonder what might be inside,” said co-author Chaz Firestone of Johns Hopkins University, who studies how vision and thought interact. “What our work shows is that your mind is able to track the information they are seeking. Just as they might be able to tell what’s inside the box by shaking it around, you can tell what they are trying to figure out when they shake it.” Christmas presents are “the perfect real-life example of our experiment.”
According to Firestone et al., there is a large scientific literature devoted to studying how people represent and interpret basic actions like walking, reaching, lifting, eating, chasing, or following. It’s a vital ability that helps us anticipate the behavior of others. These are all examples of pragmatic actions with a specific aim, whether it be retrieving an object or moving from one place to the next. Other kinds of actions might be communication-oriented, such as waving, pointing, or assuming an aggressive (or friendly) posture.
The JHU study focused on so-called “epistemic” actions, in which one is seeking information: dipping a toe into the bathtub to see how hot is, for example, testing a door to see if it is locked, or shaking a wrapped box to glean information about what might be inside—like a child trying to guess whether a wrapped Christmas present contains Lego blocks or a teddy bear. “Epistemic actions pervade our lives, and recognizing them does, too,” the authors wrote, citing the ability to tell that a “meandering” campus visitor needs directions, or that someone rifling through shallow drawers is probably looking for keys or similar small objects.
For the first experiment, 16 players were asked to shake opaque boxes. In the first round, they tried to guess the number of objects inside the box (in this case, whether there were five or 15 US nickels). In the second, they tried to guess the shape of a geometric solid inside the box (either a sphere or a cube). All the players scored perfectly in both rounds—an expected outcome, given the simplicity of the task. The videos of those rounds were then placed online and 100 different study participants (“observers”) were asked to watch two videos of the same player and determine which video was from the first “guess the number” round and which was from the second “guess the shape” round. Almost all the observers guessed correctly.
This was intriguing evidence that the observers could indeed infer the goal of the shaking (what the game players were trying to learn) simply by interpreting their motions. But the researchers wondered to what extent the success of the observers relied on the game players’ success at guessing either the number or shape of objects. So they tweaked the box-shaking game to produce more player error. This time, the videotaped players were asked to determine first whether the box held 9, 12, or 16 nickels, and second, whether the box contained a sphere, cylinder, or cube. Only four out of 18 players guessed correctly. But the success rate of 100 new observers who watched the videos remained the same.
Firestone et al. ran three more variations on the basic experiment to refine their results. With each iteration, most of the players performed shaking motions that were different depending on whether the round involved numbers or shapes, and most of the observers (500 in total) successfully inferred what the players were trying to learn by watching those shaking motions. “When you think about all the mental calculations someone must make to understand what someone else is trying to learn, it’s a remarkably complicated process,” said Firestone. “But our findings show it’s something people do easily.”
DOI: PNAS, 2023. 10.1073/pnas.2303162120 (About DOIs).