One of the major battles in the fight for criminal justice reform relates to the reliability of eyewitnesses. For years, eyewitness accounts were treated as the gold standard for evidence in criminal cases. However, as science progressed we have learned a great deal about how eyewitness accounts can be misleading or even manipulated. This has led many supporters of criminal justice reform to argue that eyewitness identifications should be always be treated with skepticism. However, the issue appears to have been more nuanced than that. A 2015 paper authored by University of California San Diego psychology professor John T. Wixted argues that there is data to suggest a witness' confidence when initially making an identification can predict whether the identification is correct.

Critics of Wixted's approach have long claimed the initial confidence of a witness has little or no predictive ability as to whether the ID was correct. These scholars claim Wixted cherry picks data to support his conclusion; a claim Wixted denies. Wixted and his supporters believe there are a few steps that can be taken when considering witness confidence that will show the predictive accuracy of witness confidence.

Cutting the Filler

According to Wixted, the first step is to not consider so-called “filler” identifications. Many researchers consider witnesses that incorrectly identify one of the “filler” photographs in a lineup as the perpetrator. Traditionally, researchers have included in their analysis the 20% to 25% of lineups that have involved a witness picking a filler picture when calculating the accuracy of witness identifications at trial. However, as Wixted points out, none of these witnesses ever testify at trial. Witnesses who select a filler image are, for obvious reasons, not used by the state to identify a witness at trial, so it makes little sense to include them in the data. According to Wixted's theory, including these “filler” picks in the data as evidence of witnesses misidentifying a defendant at trial skews the results to imply eyewitnesses are less accurate than in reality. 

Avoiding Confidence Inflation

A point that both camps agree on is that the confidence of a witness at trial has no predictive value. While the belief is that a witness' initial reaction to a photograph may have predictive value, there are far too many other variables that could alter a witness' reaction in subsequent identifications.

“By the time the witness reaches trial, their confidence is likely to be very misleading,” said psychology lecturer Ruth Horry, “since confidence tends to inflate over time, as the witness comes to believe that their ID must have been correct—otherwise, why would the police be prosecuting the suspect?”

This sort of “confidence inflation,” the name given to a steady increase in the confidence a witness has in the accuracy of an identification, is a major factor in a number of wrongful convictions. Research conducted at the University of Virginia School of Law analyzed 161 cases in which a conviction based on a witness identification was later overturned due to DNA. In those cases, half of the trials involved a witness who was originally unsure about identifying the witness but who eventually expressed confidence in their decision at trial.