We’ve all seen cases in the news where police are seeking a suspect in a crime but have no photograph, only a witness or two to describe the individual they saw committing the crime.   In such cases it is not uncommon for a facial composite sketch to be created, intended to depict the likeness of the suspect based on the memory of a eyewitness or eyewitnesses.

Such composites have been seen in cases like the Oklahoma City Bombing, The Unabomber, The Green River Killer, The 1971 airline hijacking by Dan “DB” Cooper, “Hagamannen” or “The Hang Man,” A serial killer in Sweden.   Facial composites have become a common technique in investigations where a subject of unknown identity has been seen by one or more persons and are commonly used in the United States, Australia, Western Europe and elsewhere.

Yet despite the proliferation of facial composites as an investigation tool, there has been a continuing debate about the reliability of facial composite images and whether the problems which they introduce to an investigation may outweigh their benefits.   Some studies have suggested that the entire practice may be failing as an investigative tool.

Some famous cases of composite sketches compared with the person who was finally convicted for the crimes in question demonstrate that facial composites can be quite unreliable.  While some of these examples do have some resemblance to the perpetrator, whether or not they would be recognized by those who knew or had seen the suspect is more questionable.

Facial composts may come from one of three basic sources:   An individual witness, multiple witnesses, who are independently interviewed and whose information is evaluated to create a final image or from captured images such as CCTV frames.  In general, multiple witness descriptions are considered more reliable than single witness descriptions.

As security cameras become more and more common, they are being turned to as a means of producing a likeness of a suspect.   In many cases, the image of a subjects face may be grainy or may not be completely visible in any single frame.   In such circumstances, several frames, combined with some artistic interpretation may be used to produce a more recognizable face.

While this may seem like a reasonable way of improving the accuracy of a security camera image, recent studies have called this into question. A study from the The British Psychological Society found that facial composites generated from two or more partial facial images, such as those that might be captured by a CCTV camera result in a high percentage of errors in identifying the subject in question.

Standard Methods for Creating a Facial Composite:

1.   Sketch artist – the witness describes features to an artist who draws the face of the subject the witness remembers.  Several drafts may be made before one is arrived at that the witness is satisfied with.   The witness may comment on the sketches such as “that looks more like him, but his nose was more narrow and his chin was more prominent.”

2.   The original “Identi-Kit” -  This method uses a series of transparent overlays of facial features that can be combined to produce a face.   A standard kit contains a number of basic facial shapes as well as overlays for hair styles and features.  The image produced by using overlays can either be used directly or may be the basis for a sketch which may further refine the image.

3.  Software Systems – Several software systems exist which replicate the functionality of the original overlay systems.   Software can also be used to render a more realistic face in full color.

Potential Problems Created:

There are certainly cases where facial composite images have helped in identifying a subject, but there are a number of potential problems which can be created through the use of standard facial composite generating methods.   These may include any of the following:

1. Contamination of the memory of a wittiness – Although it may seem counter-intuitive, it is possible for repeatedly thinking and discussing an event to lead to a condition known as ‘false memories’ which can be very compelling and difficult to distinguish from real memories or events.   This is especially true when the memory of an event is hazy or incomplete.   A subject may be unsure if he or she saw a certain feature on an individual, but after repeated questioning and discussion may come to believe that the feature was definitely present.This can also lead to more difficulty identifying the actual individual in the future, such as in a photo lineup.   In circumstances where a number of choices are presented, an individual may become increasingly confused.   After seeing a number of different noses, for example, it can be come difficult to remember which one looks most like the person that the wittiness has seen.
2. Introduction of bias – one of the problems with any observations recalled from memory is that they are subject to human bias.  The bias effect can become even more acute in circumstances where the data is incomplete or the event was quick or traumatic.  The bias may be especially strong if the wittiness believes certain things about the individual they saw.   For example, if they believe that the person may have been Hispanic, they will be more likely to recall Hispanic features when recalling a face.  There also can be bias based on the nature and emotions of the event which tend to color the perception of features.   If the event is violent in nature, the person may remember a sinister looking face with deep set eyes, strong features and a slanted brow.    Features can also become exaggerated, such that a slim person may be recalled as being very scrawny and a person with a fair complexion may be remembered as extremely pale.
3. The bias of the sketch artist or investigator may come into play – This is especially true where there is an existing suspect who is being investigated.   It has been suggested that in the case of the Unabomber the sketch may have been influenced by the investigators to make it look more like a person of interest who was being investigated at the time. With something as subjective as a composite image, it’s easy to see how bias and suggestion could come into play, even if it is unintentional.
4. An inaccurate sketch can hinder an investigation severely.   It can divert attention from the actual perpetrator or lead to investigating false leads.   It is even possible that it can contaminate the memory of other witnesses, such as in a series of crimes by the same individual.   There have even been circumstances where wrongful convictions have been linked to facial composites.   All and all, an inaccurate sketch can lead investigators on a wild goose chase, searching for a person who does not exist due to confidence in the sketch.

The “Whole Face” Approach:

An alternative to the traditional approaches for creating facial composts is the more recent “whole face” concept.   This method attempts to better utilize the basic nature of how humans remember another face.  In general, facial reconition is not based on the recognition of individual features but rather of the entire face as a whole. This can make it difficult for individuals to visualize a single feature or to describe the features in a subjective way.

Using this approach, subjects are not asked to describe individual features but instead simply pick from a lineup of faces, choosing the one which more resembles the individual they remember. The faces are generic in nature, although they may be chosen to match the approximate age, race and build of the individual they remember.   Initially the faces picked will not have all that much resemblance to the face a person remembers, but through numerous iterations, the final image becomes more and more accurate.

It is believed that this approach may not only generate better facial composites, but may also be less prone to creating false memories, confusing the witness or otherwise hindering identification.   Evofit is a new piece of software which utilizes the whole-face approach and may be able to address these problems and revolutionize the concept of facial composite generation.

However, it’s important to note that while the theory behind whole-face methods of composite generation appears to be valid, it has yet to see wide adoption and thus empirical evidence of the value of such methods remains scarce.   The approach also can also introduce new problems, such as producing a completely inaccurate final composite if the initial population of generic faces is poorly chosen for the individual.   This could reintroduce the problem of memory confusion for the witness.

Arguments for And Against:

The whole face approach may also cause other prejudices in human perception to come through into the final image and it certainly does not change the fact that a person’s perception could be tainted well before they begin the process of generating the composite.   In the end, the limits of human perception and memory mean that any approach to creating an image exclusively from memory will tend to be error prone and will never reach the point of one hundred percent reliability.

There is good reason to suspect that facial composites may be overused in general.   Several studies have found the use of composites to be problematic on many levels.   However, there are some instances where the images clearly have proven their usefulness.   It is clear that an accurate composite can indeed generate valuable leads, but the difficulty arises in determining how reliable a given composite image is.   In cases where there is more than one wittiness has clearly seen a suspect, there is the possibility of at least comparing for consistency, but this is a luxury that is not always available.

There definitely are circumstances where an accurate sketch has helped to identify a person or has jarred the memory of another witness who saw the subject.

Besides helping to identify a suspect, it has also been contended that a suspect may be made uneasy by seeing his (or her) likeness being posted or showing up in the press.   This might be useful in some circumstances, although making a suspect uneasy is not necessarily useful in and of itself.  Even if the composite does not look exactly like a suspect, it’s possible that someone might tend to see it as resembling themselves.   However it could also backfire. If the sketch does not look very much like the suspect at all, it’s possible that it could give them a feeling of impunity.

In the end, the most useful and reliable method of identifying a perpetrator from witnesses alone is a line up, either of live individuals or of photos. Given the importance of this, it is important to consider whether creating a facial composite at all may be more trouble than it is worth, since doing so could contaminate the memory of the wittiness when and if a suspect lineup is available.  Other useful methods of identifying potential suspects come from less subjective, even if more general descriptions, such as hair color, approximate height, distinguishing marks or other simple and easy to remember factors.

In circumstances where an unknown person has been seen by a wittiness, investigators may do well to consider whether or not a composite is a worthwhile investigative tool or whether it should be omitted entirely due to the problems and confusion it may cause.   Sometimes a picture is worth far less than a few words, such as “A tall, thin Caucasian male with dark hair, no facial hair and a small tattoo of some kind on his left ankle.”

---

This entry was posted on Wednesday, February 4th, 2009 at 8:20 pm and is filed under Bad Science, Good Science, Humor, Misc. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

---

View blog reactions