The official board game of my house is One Night: Ultimate Werewolf, and whenever we play, the unspoken (or, more often, spoken) assumption is that my roommate Adam is always the werewolf. To be fair, he is often the werewolf. And he has a habit of saying “I’m the werewolf,” right at the beginning of the game, essentially short-circuiting everyone’s thought processes, because the point of the game is to lie, and to find the liars. Admitting upfront to being a werewolf just does not compute. Unless, of course, he’s lying. But what if he isn’t?

If you’ve ever played Mafia at camp, this is a similar sort of game. Everyone is assigned a role, and is either on Team Villager or Team Werewolf. There’s one quick round of play that leaves no one sure who anyone else is, or even who they themselves are (though the werewolves know who each other are, for cahooting purposes).

Then you just… discuss. And through discussion, try to find out who the werewolves are and kill them. The werewolves, of course, are lying through their fangs, trying not to be killed, and it behooves the villagers to lie also, to try to squeeze information out of the others.

You would think that all this deception and confusion would lead to chaos, which the wolves would turn to their advantage, but my experience is that it’s actually really difficult to win the game as a werewolf. Perhaps this is because, as a recent study by Nadav Klein and Nicholas Epley of the University of Chicago Booth School of Business shows, people are better at detecting lies when they discuss things as a group.

People are generally not amazing at distinguishing between truth and lies. A meta-analysis from 2006 found that people were a little better at telling when someone was being truthful—61 percent of the time, on average. They were able to correctly identify lies only 47 percent of the time.

If you know how to look, there are physical and vocal cues that can signal a lie, writes the psychologist Paul Ekman, who has studied deception. Ekman developed tools, which are used by U.S. intelligence organizations, to help people better read facial “microexpressions” and thus detect emotions (and lies). Not everyone is very good at reading a face, though, and there is some scientific debate as to the usefulness of these expressions.

It’s also questionable whether it helps if you know the liar personally. A couple of studies have found no real difference in people’s ability to detect lies coming from friends versus lies coming from strangers. The werewolf could be your best friend, and you’d never know.

So you discuss. Does it mean something that my roommate Gabi claimed to be a villager and then went to the kitchen to get a glass of water? Was she trying to keep us from contradicting her or was she just thirsty?

According to the study, published in the Proceedings of the National Academy of Sciences, we are much more likely to figure it out through discussion than if we just go around the table, let everyone make a statement, and then vote on who we think is the werewolf. Because the advantage seems to come from talking it out, not just from averaging everyone’s individual guess together.

Participants in the study judged video clips of people speaking, and tried to determine whether they were lying or telling the truth. In some cases, they were allowed to discuss as a group and make a joint decision, in other cases, researchers had several people complete the task at once, and turn in their individual answers without collaborating.

The “real groups,” as the study calls them, did consistently better than the groups that were really just a collection of individual answers. This means there’s something about the interplay of ideas, the back-and-forth of exchange, that leads to more accurate judgments.

“Groups were not simply maximizing the small amounts of accuracy contained among individual members but were instead creating a unique type of accuracy altogether,” the study reads.

This is somewhat surprising, given how many factors are at play in group dynamics. What if someone is charismatic and wrong, and brings everyone around to their point of view through sheer force of personality? What about the herd mentality? The study didn’t address how group discussion led to better accuracy, only that it did. Neither does it say what exactly they talked about before coming to a decision, but when we play Werewolf, reasons for being considered a liar can be anything from talking too much, or too little, believing someone else’s story too readily, being too skeptical, or just “acting weird.” When you think about it, it’s remarkable that this sort of collective guesswork ever leads to the right answer.

Future research on the exact mechanisms at play here could have real implications—not just for board games, but for jury deliberations, where separating truth from fiction relies on a group discussion.