Does the presentation of multiple facial composites improve suspect identification?

Abstract

Previous research that has evaluated the accuracy of facial composites has reported low identification rates. Two studies are reported here that consider whether showing more than one composite of the same suspect might improve the rate of identification. Sixteen participant-witnesses saw one of two staged events, each involving a different unfamiliar target. Each participant-witness worked with a police operator to construct a composite of the target they had seen. One, four or eight composites depicting the same target were then shown to individuals familiar with the target. Overall, the results showed that presenting more than one composite increased the rate of identification. In addition, the results of Study 2 suggest that if the police must select just one composite from a number produced by witnesses, then a promising method might be to choose the one which bears most similarity to the other composites in the set.A facial composite is an image, constructed by an eyewitness working with a police operator, which represents a 'type-likeness' of the perpetrator and is regarded as a 'visual statement'. Rather than act as identification evidence per se, composites are employed by the police to generate possible suspects. These suspects can then be investigated further and charged or eliminated from the enquiry, based on other evidence. Many composites are shown to the public in the hope that someone familiar with the perpetrator will identify the person depicted and provide the police with a name and hence a lead.Empirical research has typically found that the composite images produced by mock witnesses after seeing a target face (either live or in a photograph) bear little resemblance to the target. The resulting images may be of poor quality for a number of different reasons including: limitations of the composite system; the skill of the composite operator; and the difficult task that the witnesses undertake in recalling and verbally describing the target's face.Earlier systems, such as Identikit and PhotoFIT, were shown to have serious limitations (e.g. Ellis, Davies, & Shepherd, 1978;Laughery & Fowler, 1980), which were attributed to limited feature databases and the use of a feature-based, piecemeal construction technique

Are two views better than one? Investigating three-quarter view facial composites

“…In general however, these results are consistent with previous research (e.g. Brace et al, 2006;Bruce et al, 2002) by indicating that presenting more information improves the identification of composites. They are also important from an applied perspective.…”

“…The results also suggest that just presenting more information does not facilitate increased performance at test. Instead, these results provide supporting evidence for the presentation of different types of information, as reported by Brace et al, (2006) and Bruce et al, (2002) The results from Experiment 1 revealed that a three-quarter-view performed as well as a full-face view when all views were presented at study. This is line with face recognition research (e.g.…”

“…As two different views of the same face can look very different and often more different than two different people, presenting both composites together may serve to increase identification rates in a manner that has been observed in previous research using composites from multiple witnesses and systems (e.g. Brace et al, 2006;Bruce et al, 2002;Ness et al, 2003). Stage 1 examines this for composites that have been constructed by the same participant-witness.…”

“…Recent research has found that presenting varied information from multiple witnesses does increase identification rates. In particular, showing multiple composites (Brace et al, 2006) or combining composites from four different witnesses increased the number of correct identifications above the level observed for a single composite (Bruce et al, 2002). Similarly, combining composites from the same witness using two different composite systems also increased identification significantly (Ness et al, 2003).…”

“…For example, researchers have examined morphing composites from single and multiple witnesses (e.g. Brace et al, 2006;Bruce et al, 2002), caricaturing (Frowd et al, 2007) and by manipulating expression (Mcintyre et al, unpublished). Although this research has had some success in improving recognition rates, no research to date has examined the impact of viewpoint on both the construction of a facial composite and the subsequent identification of the image.…”