When a crime has been committed and the identity of the perpetrator is unknown, eyewitnesses are often asked to attempt to create a likeness of the face of the perpetrator. An eyewitness can do this by creating a facial composite, either through the assistance of a sketch artist or by using a mechanized composite system. However, facial composites tend to be poor representations of the intended face, even if it is a face that is very familiar to the composite creator. This is probably due to a mismatch between the way in which people encode faces and the way in which they attempt to recall faces when building a composite.
When facial composites were first introduced in the criminal justice system, eyewitnesses would work together with a sketch artist to create a likeness of the intended face. Today, law enforcement agencies typically use mechanized composite production systems, and computerized composite production systems are used more than twice as often as noncomputerized versions. The original mechanized composite production systems, such as the Identi-Kit and Photo-Fit, are composed of overlays of facial features (e.g., noses, eyes, chins, hair) that can be combined to create a face. Modern, computerized versions, such as E-fit, Mac-a-Mug, and FACES, consist of features that can be combined, and typically resized, in any order to create a face. Currently, however, composite production systems are being created that move away from producing a face at the feature level and, instead, focus on whole faces.
Many of the mechanized and computerized systems have attempted to increase the number of features available from which a composite creator may choose, the realism of the final product, and the user friendliness of the interface. FACES, for example, has more than 3,700 features, ranging from relatively prominent features such as hair, eyes, and lips to detailed features such as eye lines and mouth lines. The computerized systems result in a fairly realistic product and can be used after a minimal training session. However, even when people view a face that has been created with a composite system and attempt to re-create the face using the same system, thereby ensuring that all the features are available, they are still unable to create good likenesses of the intended face. Furthermore, composite producers themselves are poor judges of how well the composite that they have created matches the target face. Even if a person who creates a composite rates the composite’s similarity to the face that it is intended to represent, this rating is not predictive of how others rate the similarity of the composite to the target face.
Researchers have typically assessed people’s ability to create composites of faces through naming tasks, matching tasks, and similarity-rating tasks. Naming tasks show people a composite of someone who should be familiar to them (e.g., a famous person) and ask them to name the person the composite is designed to depict. Matching tasks have people choose which face the composite is designed to depict from a larger set of faces. Similarity-rating tasks have people rate the similarity between a composite and the face it is designed to depict. In general, facial composites tend to be poor likenesses of the faces that they are intended to represent, regardless of the composite production system and regardless of how the similarity of the composite to the intended face is assessed.
Although facial composites can be a helpful tool for law enforcement, they can potentially be problematic. This is because a composite that does not truly represent the perpetrator of a crime can lead the police to investigate innocent suspects who do resemble the composite. Additionally, creating a composite and viewing it can bias an eyewitness’s memory away from the original face toward the composite face. Recent research has shown the advantages of morphing (averaging at the pixel level) composites of the same target face that have been created by different people. But, at best, morphing of composites can only be used in multiple-witness cases, and although a morphed composite does tend to resemble the target face more than do individual composites, there is only a modest increase in similarity.
The main reason why composites do not tend to resemble the faces that they are designed to depict appears to stem from the difference between the way in which people naturally encode faces and the way in which creating a composite forces them to retrieve information about the face. People tend to encode faces through a holistic process, which enables them to be better at facial recognition than facial recall. Composite-production forces people to recall faces at a feature level, as they attempt to piece together a face while looking at many different variations of the same feature.
Newer, whole-face production systems that are still in very early, experimental phases attempt to correct for this disconnect between the encoding and retrieving phases in composite production. These systems start by generating a random set of faces; the user selects the face that best matches the user’s memory of the intended face. From that, a number of different algorithms are used that produce a set of faces that are variations of the initially selected face. The user again selects the face from this set that most closely resem-bles the intended face, and this process is repeated until the faces all resemble that target face equally well. Although the few comparisons to date of the whole-face systems with feature-level composite systems do not show the whole-face systems to be superior, they do present the composite creator with a retrieval task that is more similar to the encoding task than do the other systems. Consequently, these new systems may eventually prove to be a better tool for eyewitnesses to create a likeness of the perpetrator for the police and the public.
References:
- Davies, G. M., & Valentine, T. (2007). Facial composites: Forensic utility and psychological research. In R. C. L. Lindsay, D. F. Ross, J. D. Read, & M. P. Toglia (Eds.), Handbook of eyewitness psychology: Vol. 2. Memory for people (pp. 59-86). Mahwah, NJ: Lawrence Erlbaum.
- Frowd, C. D., Carson, D., Ness, H., Richardson, J., Morrison, L., & McLanaghan, S. (2005). A forensically valid comparison of facial composite systems. Psychology, Crime, & Law, 11, 33-52.
- Hasel, L. H., & Wells, G. L. (2006). Catching the bad guy: Morphing composite faces helps. Law and Human Behavior [Electronic version]. Retrieved June 25, 2015, from http://www.ncbi.nlm.nih.gov/pubmed/17053949
Return to the overview of Eyewitness Memory in Forensic Psychology.