Australasian Science: Australia's authority on science since 1938

Faces in the Crowd

Credit: kritchanut

Credit: kritchanut

Can naturally gifted face-recognisers improve the detection of lawbreakers?

While many important justice, border and security procedures depend on accurate face recognition, prior studies have demonstrated that experienced and trained personnel may not be any better at the task than inexperienced untrained volunteers, with the performance of both groups marked by errors in matching individuals to identity documents or to previously viewed photographs. A new study in PLoS One (www.tinyurl.com/znssamk) has suggested that one solution to this problem would be to identify and recruit people to these roles who are naturally gifted at face recognition, and who demonstrate fewer errors on these tasks.

Previous studies have demonstrated significant individual differences in people’s ability to recognise faces. Most of us recognise familiar faces quickly and effortlessly, even when the image in a picture or video is of poor quality or viewed from an unusual angle. For unfamiliar faces, however, it is common that performance on standardised tests is marked by high error rates, with failures to match an identity card to its bearer and to correctly reject a mismatch.

One classic study reported that participants achieved only 70% accuracy at matching good quality, full-face photographs to individuals captured on video footage taken the same day. In another study, conducted with supermarket cashiers who were asked to match identity cards to subjects, the participants failed to correctly verify whether the bearer was the owner in 64% of cases when the card was presented by someone resembling the owner. When the card was presented by someone who did not resemble the photo on the card, an error rate as high as 34% was still observed.

But experience and training do not seem to enhance face recognition abilities. In one 2014 study, researchers found that trained passport control officers did not do better than untrained lay persons at matching people to photographs.

The PLoS One study by Anna Bobak and colleagues at Bournemouth and Stirling universities examined whether those who perform well on standardised face recognition and memory tests also excel when required to identify or match individuals on videos similar to those available through closed-circuit television systems. The participants had individually contacted the research team as they believed they were superior to their peers at recognising and recalling faces. This was confirmed using the Cambridge Face Memory Test (www.tinyurl.com/plvax9p), a standardised measure of memory for faces, with all participants performing markedly above the population mean.

The experimenters sought to discover whether the advanced skills demonstrated on standardised tests would also be evident on tasks requiring examination of video footage and stills, where faces may be presented at diverse or unusual angles and with different lighting.

The first task required participants to determine whether a face captured at an angle in a still frame taken from video footage was present in an array of ten photographs of same-sex faces depicted front-on. The target person was present in half of the 80 trials.

In the second task, participants were presented with front-on photographs of 20 faces and asked to identify these in 40 brief video clips. One of the “wanted persons” appeared in half of these clips.

The expert participants performed significantly better on both tasks than controls. On the face-matching task, the experts demonstrated an error rate of 8%, substantially lower than the 40% error rate of controls, and also reported a greater degree of confidence in their judgements. When scanning video clips for targets, the experts also performed at a higher level than controls, although the difficulty of the task produced a high error rate for both groups.

The study’s findings support the long-held assumption that some individuals are demonstrably superior to others at specific perceptual tasks.

The new discovery is that the advantage observed on standardised tests appears to generalise to more practical, “ecologically-valid” tasks, similar to those required by justice, border or security personnel. That this advantage was observed in self-nominated lay “experts” suggests that the process of selecting the right personnel to perform important identification roles may be more useful than any subsequent training.

A further finding of the study was that individual variation was observed between face matching and face memory tasks. If replicated, this may suggest that some experts may be particularly advantaged in matching faces to photos, such as at passport control, while others may be better at spotting faces in a crowd, as required for security or policing among large groups of people.

A/Prof Tim Hannan is Head of the School of Psychology at Charles Sturt University, and the Past President of the Australian Psychological Society.