Many wrongful convictions hinge on a misidentified culprit, a new report suggests. The study indicates that in our efforts to make sure that good guys don't get locked up, we may be letting more bad guys go.

The researchers hope their work will help create strong links between social science data and public policy. "To the extent that social science research has a useful role in shaping policy decisions," study researcher Steven Clark, of the University of California, Riverside, said in a statement. "Social scientists must do for policymakers what they do best and what policymakers cannot do for themselves: conduct careful studies, and provide a clear and complete analysis of the empirical data."

The study and rebuttals by other groups were published in the May issue of the journal Perspectives on Psychological Science.

In recent years, researchers and policymakers have called for specific reforms to eyewitness identification procedures that would help to reduce the rate of false identification. These reforms affect various aspects of identification procedure, including how lineups are constructed, what witnesses are told prior to the lineup, and how police officers should behave during the procedure.

Such procedural reforms are often viewed as having 'no cost' — they are thought to reduce the false identification rate without affecting the correct identification rate. The researchers argue that 'no cost' view may not actually be true.

After extensive review of the existing data, Clark says that existing data suggest that when we choose to enact reforms that are designed to reduce false identifications, we may also reduce the number of correct identifications at the same time.

But this tradeoff does not tell the whole story. Other researchers responded to Clark's argument with their own data.

Research led by Nancy Steblay, of Augsburg College, indicates that while reform procedures may reduce the number of 'hits,' they do so by minimizing the influence of suggestive and coercive practices, such as biased instructions and cues from lineup administrators.

A second group, led by Elizabeth Loftus, of University of California, Irvine, agrees with Steblay that these so-called 'lost' hits aren't actually relevant, because hits that result from suggestive practices are not legitimate identifications. They argue that eyewitness identification evidence should be based solely on the independent memory of the witness, not on the results of suggestive or coercive procedures.

A group led by John Wixted, of University of California, San Diego argues that there is a scientifically valid way to compare witness-identification procedures. If we identify the procedures that reliably differentiate between innocent and guilty suspects over time and across different situations, we will be able to determine which techniques are diagnostically superior to others.

Until we have such comprehensive data, Larry Laudan of the University of Texas said in his rebuttal, the best way to protect innocent defendants is by clearly communicating the fact that eyewitness identifications, regardless of their format, are fallible.

The researchers say we now have enough empirical data to be able to inform jurors about the error profiles of various eyewitness identification procedures. Sharing this information, he argues, is more important than trying to arrive at "the one unique and definitive format for conducting identifications."