When new scientific evidence casts doubt on convictions, the justice system offers no easy path to freedom—even when it’s the prosecutors doing the asking. The story of Russell Maze, put in prison for life after a jury found that he’d killed his infant son, puts a spotlight on this pressing problem at the heart of post-conviction relief. Pamela Colloff’s latest feature was co-published by The New York Times Magazine:

[Sunny] Eaton rose and walked across the courtroom to face the judge. Gripping the lectern, her face rigid with concentration, she tried to find the right words. “Our office receives hundreds of applications for review per year,” she began. “Out of those hundreds, we take on less than 5%. And of that 5%, sometimes we have to ask experts to review the information in the case.” She continued: “We’ve had experts look at cases and tell us, ‘No, you got this right—this was trauma, this was abuse.’ And we turn down those cases. But sometimes, your honor, a case is different.”

She spoke quickly, as if by racing forward, she could prevent the judge from interrupting her. “Over the last two years, this unit has analyzed every detail of this case,” she said. “We’ve read every record. Every line of testimony. We’ve consulted expert after expert. And we did not just rely on the petitioner’s experts. We got baby Alex his own independent experts, including the chief medical examiner for Knox and Anderson county, who more typically testifies for the state. Including a local practitioner trained at Vanderbilt, who we trust with our babies every single day. Including the former chief pathologist for Vanderbilt University. And one by one, expert after expert, told us this was not abuse—”

Dozier leaned forward in his high-backed chair. He wanted to know about the doctor who had diagnosed Alex with shaken baby syndrome, Starling, and whether she had been consulted. “But she wasn’t?” he asked sharply.

Eaton was startled by the question because it showed a fundamental misunderstanding of the work that the conviction-review unit did. Her duty was not to double-down on the state’s original trial theory but rather to investigate whether there was new evidence to consider, and whether that evidence was consequential enough that it should change the outcome of the case. Just as she did not ask the original prosecutors to evaluate the soundness of the conviction, so she did not ask Starling to review the accuracy of her diagnosis. Eaton had sought out physicians who did not have a record to defend.