Despite what watching Law & Order might have taught you, eyewitness accounts of crimes are far from trustworthy. Human memory is fallible, and in fact, eyewitnesses are notoriously bad at providing accurate evidence. Misidentification by eyewitnesses has played a role in 70 percent of convictions overturned by DNA evidence in the U.S., according to the New York-based Innocence Project. But a new study suggests one way to help eyewitnesses provide more accurate evidence: sleep.

Published in the journal PLOS ONE and spotted by Futurity, the study finds that witness testimony is more accurate during police lineups if the person is well-rested.

In the study, Michigan State University students watched a video of a staged crime in which a man planted a bomb on a rooftop. They were then asked to look through several black-and-white lineups of six people, some of which included the perpetrator from the video, and some of which only had look-alike fillers. The participants had to wait 12 hours between watching the video and picking out the criminal from the lineup, but not all of them spent it the same way. Some watched the video in the morning and looked at the lineup at night, while others watched the video at night and then looked at the lineup in the morning, after they had slept.

The researchers found that when participants slept between seeing the “crime” and being presented with potential suspects, they were less likely to pick someone out of lineups that only included fillers. In other words, they were better able to reject innocent lookalikes, noting correctly that the perpetrator they had seen wasn’t present at all. People who had not slept misidentified an innocent person as the perpetrator 66 percent of the time compared to 42 percent of the people who had slept. That doesn’t seem like a huge difference, but even one misidentification in real life can have serious consequences for innocent people accused of crimes. The results track with existing research on sleep that has found that it plays an important role in cognitive processes like learning and memory.

However, sleep didn’t make the participants any better at identifying the correct suspect when he was in the lineup. Both groups could pick him out with about 50 percent accuracy. Essentially, sleep isn’t going to solve crimes, but it could help ensure that the wrong person doesn’t go to prison.

Previously, a 2016 study in Psychological Science found that it’s relatively easy for police to nudge witnesses into picking their prime suspect from a lineup by making sure that they’re the only person in the group with a particular distinguishing feature, like a beard. These unfair lineups not only resulted in misidentification, but participants who chose from them were more confident about their (incorrect) choice than people who saw more fair lineups.

Taking a long nap isn’t the only way to improve lineup accuracy. The criminal justice advocates at the Innocence Project recommend several methods to make sure that witness testimony is more accurate, such as making sure that the people shown in the lineups all roughly match the description of suspect (beard and all), having the lineup administered by an officer who doesn’t know which person in the lineup is the police’s suspect, and making sure the witnesses know that the perpetrator might not be in the lineup at all, and that the investigation will continue even if they don’t pick anyone.

[h/t Futurity]