Skip to main content

In the face of Crime


August 11, 2006

By Dawn McQuiston-Surrett

News coverage of wrongful convictions ought to suggest even to a casual observer that the U.S. judicial system is fallible.

My research focuses on the application of psychological science to issues relevant to the legal system, and has specifically been concerned with evaluating aspects of police procedure in the collection of eyewitness evidence.

In a recent book, “Actual Innocence: When Justice Goes Wrong and How to Make it Right,” the authors document the exonerations of many wrongfully convicted individuals. The book was written by associates of The Innocence Project — a non-profit legal clinic and criminal justice resource center that has helped exonerate at least 160 people through post-conviction DNA testing.

A major conclusion in the book is that an overwhelming proportion of wrongful convictions, more than 80 percent, are the result of mistaken eyewitness identifications. The Innocence Project reports that inadequate procedures used by police and others in the justice system are largely responsible for facilitating mistaken identifications.

The construction and distribution of a facial composite image of someone who has committed a crime is a procedure that can be central to the beginning of a criminal investigation. Forensic artists sometimes work with an eyewitness to develop a sketch of the culprit’s face, and a sketch can often result in a good match to the culprit.

However, recent research has raised concerns about the usefulness of computerized composite systems. According a study in Psychology, Crime, and Law (McQuiston-Surrett, Topp, & Malpass), the reliance on forensic artists has largely been replaced by computerized composite software programs that offer on-the-spot access to a large database of facial features. The programs are easy to use and inexpensive.

Using these systems, witnesses select photographic elements from a kit that represent individual facial features, and then assemble them into what they believe to be their recollection of the culprit’s face. Along with researchers at the University of Stirling, Scotland, I have spent the last several years studying the use of computer facial composite systems. Among the questions explored: How well can eyewitnesses recall specific facial information about a person they have seen commit a crime in order to reconstruct the culprit’s face? And how useful to American and British police forces are these computer programs in accurately depicting a criminal?

Our research, recently published in "Legal and Criminological Psychology" and "Forensic Psychology and Law," found there is cause for concern.

The problems associated with computerized composites appear to be due to both system design and witness limitations. The software programs only provide a limited range of facial features, skin tones and textures, and accessories from which to choose. Options for creating more ethnic features and skin tones are all but absent. There is a limited ability to manipulate and modify facial features. The general structure of the facial feature catalogs force witnesses to view up to hundreds of individual features in sets of anywhere between 6 and 30 at a time. And the images often do not appear lifelike.

More importantly, people have difficulties when they are asked to recall a face feature-by-feature, because when we process a face we not only encode individual features but also the relationships between them (i.e., holistic processing). It appears that facial composite tasks were initially designed based on strategies that made sense to law enforcement pragmatically rather than based on theoretically-driven investigations regarding facial processing in memory.

A recent string of sexual assaults in the Phoenix area committed by an admitted serial rapist (see preceding page) highlights another potential problem with the use of computer-generated facial composites: composite production involving multiple eyewitnesses. These particular crimes resulted in at least two different facial composites of the alleged culprit. Both of the composites were released to the public. Neither, it was later discovered, resembled the man police arrested.

A recent Arizona Republic article titled “Chandler is drawn to sketches: Computer-generated images lack accuracy, some experts say” (August 31, 2005) sheds light on some of the problems researchers have uncovered. According to Sgt. Scott Yarbrough, “They’re supposed to look like real people, but you’re taking lip number 42 and putting it with eye 150.” His reaction to the composite process is on target, and the Chandler police have begun to lean more toward the use of traditional sketch artistry.

Until radical improvements are made to the composite construction process, I would advise police to rely on experienced forensic sketch artists, whose work is notably superior, in spite of the extra time and costs required for their construction.

In the meantime, however, facial composites created using computerized programs are being used as evidence — generally incriminating — in trials nationwide. But how do lay people (i.e., jurors) evaluate the computerized renderings and incorporate that evidence into their assessments of witnesses and defendants?

I am currently examining how jurors weigh such information, and under what conditions facial composites lessen or strengthen the perception of a defendant’s alleged culpability.

Dawn McQuiston-Surrett is a professor in the Department of Social and Behavioral Sciences. She received her Ph.D. in applied experimental psychology from the University of Texas at El Paso in 2003. Her research is in legal psychology.