An economics professor made the audacious suggestion a few years ago that the medieval practice of trial by ordeal “worked” because innocent people were more likely than guilty ones to pick up pieces of hot iron or plunge their bare hands into boiling oil. The practice depended on suspects having unwavering faith that God would save the truly innocent from third-degree burns. And the priests in charge could temper the heat, just in case God wasn’t paying attention.

Likewise, less brutal means of psychological manipulation in U.S. criminal investigations may “work” to elicit confessions, but there’s a growing scientific case that currently acceptable tactics aren’t conducive to discovering the truth.

The recent mass distribution of videotaped interrogations, such as the ones featured in the Netflix documentary series “Making a Murderer,” reveal that confessions aren’t always spontaneous acts of contrition, motivated by the need for a clear conscience. The interrogation process can still be something of an ordeal, with confessions coaxed by combinations of psychological pressure, deception and sheer exhaustion.

The exhaustion factor interested psychologist Elizabeth Loftus, a memory expert known for demonstrating the unreliability of eyewitness testimony. She wondered whether it mattered that many interrogations are held between midnight and 8 a.m., or that in some cases, the process can drag on for many hours.


Teaming up with researchers at the Michigan State University sleep laboratory, she and colleagues set up an experiment to see if sleep deprivation might make subjects more prone to making false confessions. Their results were published Monday in the Proceedings of the National Academy of Sciences.


The study used 88 student volunteers who were asked to complete questionnaires and take tests on a computer, and were explicitly warned that pressing the escape key would destroy the researchers’ valuable data. Told the study had something to do with sleep deprivation and memory, they were asked to report back a week later to spend the night in the lab.

Half of them were allowed to sleep; the others were asked to stay up all night. Those in the all-nighter group were monitored and could do homework, watch TV, or play games with lab personnel. The next morning, all subjects were told someone had observed them press the escape key the week before.

Of the sleep-deprived, a full half – 22 – confessed on a first round of questioning, and eight more confessed on the second. Fewer of the sleepers did so: just eight on the first round and nine on the second.

Unfortunately, the researchers didn’t build in a mechanism to reveal how many people, if any, really did press the escape key. They said it’s unlikely anyone would press it by accident given its far corner location. But they had to rely on faith and the assumption that few subjects would press the forbidden key on purpose.


If the study indeed measured false confessions, it may help explain why the phenomenon is so common. False confessions are thought to account for 15 to 25 percent of wrongful convictions. So damning is a confession that in some cases, even DNA isn’t enough to reverse a conviction.


Take the case of Juan Rivera, who was convicted of the 1992 rape and killing of an 11-year-old girl in Chicago. Rivera initially denied having anything to do with the crime, but over a protracted interrogation, he signed a confession. Here’s how a New York Times Magazine story described the interrogation:

What followed was 24 hours of near constant interrogation, and around 11:30 on the morning of Oct. 30, after banging his head on a cell wall, pulling out a clump of his hair and being handcuffed behind his back and placed in leg shackles, Rivera finally provided investigators with a detailed confession.

Rivera later explained that he “blacked out” and didn’t remember signing the confession.

When a lab was finally able to analyze DNA found in the victim’s body, it didn’t match Rivera’s. But so damning was the confession that in a second trial, the prosecutor persuaded the jury that the 11-year-old victim must have had sex with someone else shortly before she was raped and killed. Rivera was convicted again, though he was subsequently exonerated in 2012.


There was no such exculpatory evidence in the case of Brendan Dassey, one of the two suspects questioned in “Making a Murderer.” There’s widespread disagreement over whether Dassey assisted in the murder for which he’s now serving time. But that’s beside the point, said law professor Adam Benforado, author of the book “Unfair: The New Science of Criminal Injustice.”


Benforado said he’s recently watched the documentary’s confession scenes, during which Dassey, with a reported IQ of around 70, comes across as stressed and confused. He’s led to think that signing the confession will allow him to go home. “I don’t know whether Brendan for sure didn’t do it,” said Benforado. “But I know the techniques being employed by the police are bad techniques.”

Until he became a law professor, he said, he didn’t realize how much deception police are allowed to use – whether it’s telling suspects they’ve been incriminated by nonexistent fingerprints or eyewitness testimony, or falsely claiming that they have failed polygraph tests.

Imagine you’re a suspect and you’re told there’s evidence against you, Benforado said. If you confess, you’ll get a two- year sentence, but if you fight it, you might end up with 25. When faced with this choice in his class, he said, many of his law students would confess. When presented that way, falsely confessing isn’t an irrational choice. Some of us would do it on a full night’s sleep.

Faye Flam writes about science, mathematics and medicine. She has been a staff writer for Science magazine and a columnist for the Philadelphia Inquirer. She is author of “The Score: How the Quest for Sex has Shaped the Modern Man.”

Only subscribers are eligible to post comments. Please subscribe or login first for digital access. Here’s why.

Use the form below to reset your password. When you've submitted your account email, we will send an email with a reset code.