This story was co-published with the Los Angeles Times.

California Gov. Jerry Brown is considering granting clemency to Shirley Ree Smith, a grandmother convicted in 1997 of shaking to death her 7-week-old grandson, Etzel Glass. Sentenced to 15 years to life in prison, Smith insists she's innocent.

Prosecutors built their case against Smith almost entirely on the findings of forensic pathologists at the Los Angeles County Department of Coroner. During an autopsy, doctors discovered a small amount of bleeding on the infant's brain and in his optic nerves. Based on this bleeding, the forensic pathologists concluded that somebody had violently shaken Etzel's body, killing him.

They ruled the death a homicide. Others, however, aren't so sure.

Etzel's body exhibited few of the signs typically associated with fatal head injuries. One doctor testifying on Smith's behalf dismissed the notion that Etzel was murdered as "fantasy." In a 2006 opinion, the U.S. 9th Circuit Court of Appeals expressed a similar skepticism, saying there was "simply no demonstrable support for shaking as the cause of death."

Whether or not Smith is ultimately pardoned, her story provides a window into the increasingly rancorous scientific debate about shaken baby syndrome, a once-widely accepted theory that violent jostling can cause fatal head injuries in infants. Based on studies dating back to the 1960s, many doctors came to believe that a signature trio of symptoms — bleeding and swelling of the brain, and hemorrhaging of the retinas — provided conclusive proof that someone had shaken a child to death.

But now a growing number of experts have doubts about the diagnosis. Official reviews in Canada and Britain have uncovered cases in which people were wrongly convicted based on the shaken baby theory.

As a journalist for ProPublica, I've spent more than a year scrutinizing the inner workings of the nation's system of coroner and medical examiner offices responsible for probing sudden and suspicious fatalities. In the course of that research, my colleagues and I analyzed roughly two dozen instances in which people were wrongly accused of killing babies or small children.

Questionable autopsy findings played a central role in each of these cases. Some experts, like Dr. Michael Laposata, the chief pathologist at Vanderbilt University Medical Center, think there are more innocent people still serving time in prison. "I don't think it's a handful," Laposata told me last year. "I think it's far more."

When children die unexpectedly, the authorities often look for unnatural causes, and rightly so; in recent decades our society has become far more vigilant about detecting and prosecuting child abuse, an admirable accomplishment.

But child deaths can pose special problems for forensic pathologists. When babies and small children die, the clues can be quite subtle. Oftentimes, doctors are hunting for microscopic indicators that a child has suffered head trauma or has been asphyxiated. Such cases require a high degree of expertise.

A key concern is that forensic pathologists may confuse the symptoms of a natural ailment for a sign of abuse. That could well be what happened with baby Etzel. There are dozens of afflictions that cause bleeding and bruising and can easily be mistaken for child abuse. One of the leading textbooks on child abuse now includes two chapters on these mimics, which range from certain forms of cancer to sickle cell anemia to trauma suffered during the birthing process.

Unfortunately, many forensic pathologists aren't prepared to deal with the complexity of child death cases. According to a 2009 report by the National Academy of Sciences, only a third of the coroner and medical examiner offices in the U.S. had the equipment to do a microscopic tissue analysis. The report painted a dismal portrait of the profession, citing poor funding, a lack of decent facilities and a severe shortage of qualified doctors.

Reality, it turns out, bears very little resemblance to "CSI."

My colleagues and I surveyed the nation's 69 busiest coroner and medical examiner offices. More than 1 in 5 of the physicians working in these morgues — including the chief medical examiner of Washington, D.C. — were not board certified in forensic pathology, the branch of medicine focused on the mechanics of death. Experts say such certification ensures that doctors have at least a basic grasp of the science, and should be a requirement for anyone performing autopsies in a possible homicide case.

Located at the intersection of science and the law, coroners and medical examiners are subject to little regulation in most states and, as the National Academy pointed out, there are no national standards for the field. By contrast, nursing homes, hospitals, dialysis centers, clinical laboratories and many other medical facilities receive oversight at both the state and federal level, and can be fined if they fail to operate properly.

So as Gov. Brown considers Smith's request to commute her sentence, he might consider another question, as well:

Are there any other criminal convictions he should be taking a close look at?