Despite their proximity to the crime and the suspect, government informants can sometimes get it wrong. WGBH's Morning Edition Host Joe Mathieu spoke with Northeastern University law professor and WGBH legal analyst Daniel Medwed about how the information government informants give can occasionally lead to wrongful convictions. The transcript below has been edited for clarity.

Joe Mathieu: What exactly does the term [government informant] mean?

Daniel Medwed: Well, an informant is someone who, as the term suggests, provides information in exchange for a benefit. There are two basic types: there's a confidential informant, who's typically out in the street, often embedded in a community where there's some criminal activity; and there are jailhouse informants, who are currently incarcerated, and often provide information about cellmates or blockmates. Now the type of benefit they receive can range considerably, from, at the low end, money or something else to, at the high end, leniency on a pending serious charge. I think the most problematic type [is] jailhouse informants.

Mathieu: What's the problem? I suspect that prisoners talk about their cases with their cellmates or blockmates and they know stuff.

Medwed: Well, in theory I think that's right. But in practice I'm pretty skeptical, and here's why. For one thing, prisoners tend to know the drill. They know there are sharks swimming in that ocean, looking for information that they can parlay into favorable treatment for their own situation. So for that reason inmates are a lot more tight-lipped in my experience than they appear to be in TV dramas. For another thing, jailhouse informants have this incredibly powerful incentive to fabricate. What would you say to get out of prison early? What would you say to get transferred to a facility closer to your family? For those reasons, about one in five documented DNA exonerations — folks who've been freed through post-conviction DNA testing — their cases involved jailhouse informant testimony at trial.

Mathieu: Let's assume there's a fundamental credibility problem for jailhouse informants. Won't jurors be able to figure that out on the stand? Are they told about the benefits the government has promised to the witness, or did they not hear that?

Medwed: This is another example where theory deviates from practice. In theory, the government is supposed to turn over all so-called impeachment information — information that would discredit a government witness, including any incentive to lie. But in practice, what prosecutors do to avoid disclosing these arrangements, is they sometimes enter loose "wink and nod" agreements with the witness, where it's not an ironclad promise, so they claim that they don't have to turn it over. And then the witness on the stand, in good faith [and] without necessarily committing perjury can say, "I wasn't promised anything." So in theory, yes, but in practice not always.

Mathieu: That's pretty tough. What can be done about it, if anything?

Medwed: Well, on the one hand, we could get rid of informants altogether, but I don't think police and prosecutors want to do that. It's a pretty powerful tool in their investigative tool shed. But on the other hand, we can come up with a bunch of targeted reforms. We could have pre-trial reliability hearings, where judges evaluate the testimony of the jailhouse informant outside the presence of the jury [to] figure out whether it's credible. Second, we could have what are called corroboration requirements — we could only allow jailhouse informants to testify if what they have to say is independently reinforced or corroborated by other evidence. And lastly, we've talked about this throughout the summer, we could have jury instructions to inform jurors about the hazards of jailhouse informants.

Mathieu: You've worked around this a long time. What's the most extreme offer you've heard given to an informant like this?

Medwed: Well actually, it comes from one of the cases I handled when I was running an Innocence Project in New York. Somebody testified against my client. He was a prisoner, and he testified against my client, who was also incarcerated for an unrelated crime at the time. And the benefit that that person received was a transfer to a prison closer to his wife to facilitate conjugal visits. So for that reason he made up a murder charge against my client.