Peer commentary
Article Type: Editorial From: Journal of Forensic Practice, Volume 17, Issue 4.
Peter J.B. Hancock
Professor Peter J.B. Hancock is based at the Department of Psychology, University of Stirling, Stirling, UK.
A recent survey that I conducted of papers on the psychology of facial composite construction revealed 50 published in the last ten years, so this collection represents more than a 10 percent increase in the total. What might a practitioner wish to know? The first aim of a facial composite system is to help a witness or victim of crime to create an image of the person she or he saw. There is much psychology here, in helping the witness to remember without leading them; how best to interview them; how best to present potential images. The second aim is for the face image to be recognised by someone who knows or has also seen the perpetrator. It turns out that there is much that psychology can contribute here, too, for example, in how the image is presented. Papers in this collection address each stage of the process.
It starts with a methodology paper, by Fodarella et al., which describes the main composite systems currently in use in the UK and abroad and the procedures for using them, including the cognitive interview. Marsh et al., then demonstrate the fragility of human encoding of visual information, such as the appearance of a face. It is not possible to ignore an incorrect verbal description of the face, even if instructed to do so.
There are three papers looking at the effects of how images are presented to the witness during composite construction. There is much evidence to show that we perceive faces as a whole, such that changing one aspect of it can produce apparent changes in the appearance of the whole face. In some composite systems, features are selected in isolation, while in others they are selected in the context of a whole face. In two experiments, Skelton, Frowd and Speers confirm the prediction that construction in the context of a whole face works better. Their second study used a day’s delay between seeing the target face and creating the composite. The recognition rate was around 14 percent, better than has often been found for feature-based composite systems,but distinctly worse than modern holistic systems such as EFIT-V and EvoFIT.
Ness et al., report the first evaluation of a three-quarter view version of PRO-fit, a feature-based composite system. There is some evidence that a three-quarter view of a face is good for recognition, since it affords more information about the face than a purely frontal view. It is also possible that a witness did not see the target head-on. The results show that while a three-quarter view composite on its own was no better recognised, presenting both images did improve recognition over either alone. It suggests that a proper 3D composite system might be a valuable way forward.
Davis et al., provide a valuable confirmation of two techniques to improve the recognisability of composites after they have been made. The first is combining different composites, either from different witnesses or multiple attempts by the same witness. The second is stretching the composite vertically. Both improved recognition rates substantially, though it is still unclear why stretching an image works.
Finally, Frowd et al., look back over ten years of research on facial composites, pooling data across many studies in order to detect consistent effects, such as differences between composite systems and changes to the interview. This form of analysis is important, as individual studies often produce noisy data. The cost of creating and evaluating composites places a practical limit on participant numbers and there can be large differences in the quality of individual composites, which can distort results. Pooling data from over 1,000 participants, as here, provides a reliable insight into the factors that matter.
