Marathon Police Interrogations May Send Innocent People To Prison

Police Interrogations Past Six Hours May Multiply False Confession Risk

A police interrogation that runs six hours could carry roughly ten times the estimated risk of ending in a wrongful conviction built on a false confession, compared with routine questioning that lasts an hour or two. That is the finding of a new statistical analysis in the Journal of Criminal Justice. In the researchers’ simulations, estimated risk keeps climbing, reaching about 25 percent by the 12-hour mark.

Those numbers push back on a widely discussed claim published earlier this year by researchers Scott Mourtgos and Ian Adams, who concluded that legally permitted interrogation tactics are associated with only about 1 percent of wrongful convictions that rest on false confessions. A figure that reassuring could shape police training, how courts weigh confessions, and reform efforts.

A team led by Thomas Bryan Smith, a criminal justice researcher at the University of Mississippi, rebuilt that analysis from scratch. Their verdict was not that the 1 percent figure was made up. It fits short, routine questioning, they say, but it hides what happens once interrogations drag on.

A 1 Percent False Confession Risk Figure Rested on a Built-In Assumption

Hard data on these cases is scarce, so Mourtgos and Adams used a statistical shortcut that doctors rely on to read test results. A positive result means little without knowing how common a disease is and how often healthy people also test positive. Here, the disease stands in for a false confession that ended in a wrongful conviction, and the positive test stands in for a problematic interrogation tactic.

Smith and his coauthors say the catch is that nobody knows how often such tactics show up in each kind of case. The earlier analysis handled that gap by treating every possibility as equally likely, which quietly assumed tactics are just as common in false confession cases as in ordinary ones. Build in that assumption and the answer lands near the starting number almost automatically. The Mississippi team ran the same model with its own data, got close to 2 percent, and saw the same pattern. Yet earlier research shows false confession cases more often involve accusatory tactics and long questioning. The authors stress the problem is a quirk of the model’s design, not a belief held by Mourtgos and Adams.

Estimated False Confession Risk Climbs From 1 Percent to 10 Percent by Hour Six

To sidestep the guessing, the team used something earlier studies had actually measured: how long interrogations last. Routine questioning usually lasts between 20 minutes and an hour, and averages roughly an hour and a half across eight studies. For false confession cases, the researchers picked a deliberately cautious average of about three hours, drawn from vulnerable groups such as people with serious mental illness. One earlier study of proven false confessors found much longer sessions, averaging 16 hours, with some running past a full day.

Simulating cases at each length produced a steep climb. At an hour and a half, estimated risk sat near 1 percent. By three hours, it was roughly 2.7 percent, and past six hours it topped 10 percent. At 12 hours, it reached about 25 percent. The model projected 50 percent at nearly 34 hours, though sessions that long almost never showed up.

Here, risk means the model’s estimated chance that a case of a given length turns out to be a wrongful conviction resting on a false confession, not the odds that any one suspect will confess falsely.

Long sessions were rare, with only about 179 of every 10,000 simulated interrogations expected to reach the higher-risk range. Rarity does not make them unimportant, the authors argue, since a low rate in routine cases says little about the odds once questioning runs that long. In lab experiments, innocent people held out longer than guilty people before signing a confession, which may help explain why marathon sessions matter.

Authors Warn Even Their Own Estimates Are Too Early to Guide Policy

Smith and his coauthors stop short of claiming they found the true number. Length is only one difference between interrogations that go wrong and those that do not. Others include the tactics used and a suspect’s age and mental state. Officers also choose approaches based on how strong the evidence is and how vulnerable a suspect seems, so tactics and risk are tangled together. Simulations cannot show cause and effect, either. Smith said in a University of Mississippi news release that such simulations are “too preliminary to be used to inform decisions in individual cases.”

Many European countries and Canada already use interviewing methods built around gathering information rather than confronting suspects, the authors note, and experiments find these approaches produce more true confessions and fewer false ones. Even a 1 percent rate, they add, is a problem in a system that imprisons 1.9 million people.

Smith and his colleagues make a clear case that interrogation length can change the estimated risk dramatically, and that police departments, courts, and lawmakers leaning on a single reassuring figure may be missing where the greatest danger lies: in interrogations that stretch well beyond the typical hour or two.

---

Disclaimer: This article reports on a single statistical simulation study and is provided for general information only. It is not legal advice, and the study’s estimates should not be used to judge any individual case or interrogation.

---