As reported by the New York Times:
False criminal confessions set in motion other errors by police, informants and witnesses, suggests a review of 240 overturned convictions.
In experiments, confessions have been shown to shape the interpretation of other evidence by polygraph examiners, fingerprint experts and witnesses, but real-world proof of the phenomenon has been hard to come by. To remedy that, researchers looked at cases overturned with the help of the non-profit group the Innocence Project, on the basis of DNA evidence, from 1992 through July 2009.
Fifty-nine of the cases involved false confessions — mostly by the suspect, sometimes by a supposed accomplice. The confession cases were significantly more likely than the others to have also involved other mistakes, such as incorrect findings by forensic labs (63% vs. 48%), or an informant whose tale did not reflect the truth (19% vs. 11%).
The researchers verified that most confessions happened very early in investigations, making it plausible that they clouded what happened later. Errant eyewitnesses were actually the most common glitch in the cases studied, appearing in three-quarters. But cases that involved eyewitness errors (and not confessions) were less likely to have multiple errors than cases that involved confessions (and not witness errors); in other words, eyewitness errors weren’t as contaminating as false confessions.