Slate is running a series this week on cases where the justice system got it wrong; somewhat spurred by the Illinois legislation abolishing the death penalty, partly just a good conversation.

The author, Brandon L. Garrett, is a bit pimping his new book but is also providing a good look at two of the most widely believed – but at the same time not entirely reliable – types of evidence on which many criminal cases rely. The first is eyewitness accounts and identification, the second is the confession of guilt.

Now that we know—with the benefit of the DNA tests—that Sterling is innocent, one wonders how an innocent man could have guessed at incredibly specific crime scene details? Sterling later explained it this way: “They just wore me down.” “I was just so tired.” “It’s like, ‘Come on, guys, I’m tired—what do you want me to do, just confess to it?””

In a pair of videos I link to very often, there’s a great answer to how someone “knows unreleased details” – the cops slip them to the accused in one form or another, or lead the accused into guessing until they have them “guessing right” on tape.

More interestingly to me, however – Garrett finally comes up with some hard numbers. I’ve chided the Innocence Project before about this, because they make a habit of releasing only their “number of innocent people freed” number, rather than giving us the chance to see the total number of cases they’ve examined. Will has said – and I agree – that even this may not be an exact figure, since IP only takes cases “likely to exonerate” on their early examination before proceeding all the way down the line, but it at least would give us something to work with.

Garrett, however, gives us a gem.

In 16 percent of the first 250 DNA exonerations, or 40 of the 250 cases I studied for my book, Convicting the Innocent, innocent defendants confessed to crimes they did not commit. (Additional DNA exonerees did not deliver confessions in custody, but they made incriminating statements or pleaded guilty to crimes they did not commit).

The false confessions pose a puzzle. All but two of the 40 DNA exonerees who falsely confessed were said to have confessed in detail.

Now, this is not perfect. His study is only on those cases that are proven false convictions. But we at least have a hard number here – 250 cases of proven innocent, 40 cases of false confession, 38 of which are said to be an “in depth” confession. And every one proven innocent almost-definitively by DNA evidence. This leads to at least a reasonable suspicion that confessions in districts across the nation are contaminated or even coerced by the cops – perhaps by cops who don’t know what they are doing, or perhaps by the type of behavior we commonly associate with not-so-honest cops who start and stop the recording on TV shows, only recording the parts of the interview they want to be available in court.