Research Seeks to Improve Eyewitness Accuracy Through National Science Foundation Grant

Julia Fishman

Assistant Professor of Criminal Justice Amanda Bergold is at the forefront of a project designed to find the best ways to construct lineups.

December 3, 2019—How does the level of similarity between photographs in a lineup influence accuracy by eyewitnesses?

That’s the question Marist College Assistant Professor of Criminal Justice Amanda Bergold seeks to answer through an ambitious research project that was recently funded by the National Science Foundation’s (NSF) Division of Social and Economic Sciences. Bergold is working with former colleague Paul Heaton of the University of Pennsylvania’s Carey School of Law. They were awarded a $174K NSF grant earlier this year.

“Eyewitness lineups are an essential tool for solving crimes,” said Bergold. During a typical lineup, a witness views six to eight photographs; lineups usually only include one suspect and all other images used are known innocents, called fillers. Past research suggests that lineups where the suspect stands out from the fillers (known as biased lineups) reduce identification accuracy. Lineups in which the suspect resembles the fillers too closely to be difficult to distinguish from them also reduce identification accuracy, making the task too difficult for otherwise reliable witnesses. “Cognitive literature on memory shows that too much similarity can actually impair performance on memory tasks,” said Bergold.

As a result, lineup administrators currently receive opaque guidelines: include photos that are not too similar, but not too different. “This poses a real challenge,” Bergold explained. “It’s very subjective. If research can establish clear guidelines for building a lineup, it would be an enormous help in this process.”

She and Heaton are hoping to uncover the sweet spot: the optimal level of similarity between photos to help predict accurate identification by a witness.

The work is timely. Several states have reformed or are seeking to reform how eyewitness testimony is discussed in the courts.

In a related line of inquiry, Bergold and Heaton are also assessing the role facial recognition can play in determining what photos are included in a lineup. “Law enforcement now has greater access to facial images than ever before,” noted Bergold. “How do the size of databases of available photos impact outcomes? Our research will examine that as well.”

Dr. Amanda Bergold speaking at the 2019 Social Justice Conference
Dr. Amanda Bergold speaking at the 2019 Social Justice Conference

Bergold’s work is underway. Using an experimental platform built upon an underlying database of over 570,000 potential filler photographs, she and Heaton have created a series of tasks for the study participants: view a stimulus photo, complete a distraction task, and then view a lineup. The study will then attempt to use algorithmic ratings of filler/suspect similarity in order to predict eyewitness accuracy. “We will be analyzing accuracy using a variety of different analysis techniques,” Bergold said. Ultimately, she and Heaton hope to test 2,000-3,000 subjects during the study period.

If an optimal range of similarity in a lineup does exist, the implications could be significant. Such an established range could improve the quality and accuracy of lineup identifications which would, in turn, reduce the incidence of wrongful convictions.

Bergold holds a BA in psychology and history from Williams College. She earned her PhD from John Jay College and the City University of New York Graduate Center.

The study will continue through 2021.

Asset Publisher