Aids to recall / recognition: identikit and identity parades
Goldstein and Chance (1971) tested subjects for their recall of photos of women’s faces, snowflakes, and inkblots. Participants were shown 14 photos from each set, and tested immediately afterwards, and then again after 48 hours. The results showed that accuracy of recall for faces was good — 71%, as against 48% for inkblots, and 33% for snowflakes. Facial recall and recognition are, however, quite complex aspects of memory. For instance, most of us can easily recognise a familiar face, but if we were asked to describe that face in detail would experience some difficulty. Moreover, while we may often recognise a face as familiar we may not be able to pinpoint why, or recall the person’s name.
Recognition of faces is poor if the face changes between the encoding and the presentation of possibilities. This could be a change in glasses, hairstyle, beard or even expression. Patterson and Baddeley (1977) compared the accuracy of recall of a face that had changed in some way between the first and second time the participants saw it. For example, a change of wig reduces the accuracy of recall to 50%, while a change of wig and beard leads to only 30% accurate recall. These figures compare to 70% accuracy when the face does not change.
The accuracy of face recall is also influenced by: length of time since the face was seen, the seriousness of crime, the number of previous encounters with the suspect, and witness characteristics (like attention to detail).
Commentary
The most confusing method for recall is to show ‘mugshots’, then have an identity parade. This leads to confusion between recognition and episodic memory. In other words, the witness may recognize a face in the identity parade, but not know if it comes from the original event or from the ‘mugshots’.
Face recognition for familiar faces
The recognition of familiar faces and the naming of the person can be seen to have three stages.
1. The face is compared with a set of stored descriptions called ‘face recognition units’ (FRU), and this produces a feeling of familiarity or not.
2. The memory store is activated to recall facts about the person.
3. The retrieval of the name of the person is then made.
The research has tended to concentrate on whether the stored descriptions are based on individual features or the whole face. One possibility is a representation of how each face differs from the average prototype, and thus distinctiveness is the key (Bruce 1988).
Using computer images, Haigh (1986) found that the shape of the head was most important, followed by the eyes and mouth in full-face recognition, while the nose is more important in profile recognition.
In real life these problems associated with facial recall can have serious consequences. A vivid example concerns an Australian psychologist, Donald Thompson, whose research attracted some media attention, and he appeared in several TV programmes to discuss his findings. A few weeks later, however, he was picked up by the police, put in a line-up, and identified by a woman who claimed he had raped her. It subsequently emerged that the time of the rape had coincided with Thompson’s participation in a live TV discussion, providing him with an excellent alibi. The woman had been attacked while this programme was on air, and although she correctly identified Thompson’s face, having emotionally associated it with the event, she wrongly assigned it to the rapist. This can be regarded as unconscious transference of memory traces.
Media coverage of crime can play a part in this type of unconscious transference. For instance, Ainsworth (1995) showed participants a local news story about a series of sexual assaults which included two photos, one a photofit of the suspect and the other a photograph of a ‘good Samaritan’ who had helped one of the victims. A week later the participants were asked to select the suspect from a selection of six photos. One group was shown a selection containing the photofit of the suspect, a second group was shown a selection containing the photo of the ‘good Samaritan’, while the third group was shown a random selection. The first group picked the suspect on almost 40% of occasions, while the second group incorrectly selected the face of the ‘good Samaritan’ on 50% of occasions. The chances of selecting the suspect or the ‘good Samaritan’ were roughly equal, suggesting that unconscious transference was playing a part.
Labelling a face can also influence how people recall facial features. Shepherd, Davies and Ellis (1978) asked participants to view a photo of a male face for 30 seconds and then asked them to construct a photofit. Half of them were told the face belonged to the brave captain of a lifeboat, while the other half was told the face was that of a mass murderer. The subsequent photofit pictures were quite different, with the face of the alleged mass murderer judged to be more cruel, unpleasant and unintelligent.
Photofits
were first introduced in the UK in 1970 and involve trained operators selecting
alternatives for specific facial features to match a witness’s verbal
description until they have produced a type likeness. There are several ways in which researchers
have attempted to improve on this method. One is the use of computers to
produce from the witness’s description a good pictorial likeness of the
suspect, known as an E-FIT. However, the original E-FIT still built up a face
on a feature-by-feature basis: as the witness gave a description of, say, the
eyes, the computer chose the best set of eyes to fit this description, and so
on. The witness did not see the face until it was completed. Researchers from
the Face Processing Research Group have recently pioneered a new system of
compiling an image of criminal suspects based on the idea that an improved
image would be produced by allowing the witnesses to watch the face build up.
It is very important that the features are seen as part of the whole face, so a
cartoon outline and features are used to replace those elements that have yet
to be described. In addition to this, if possible, the evidence from several
witnesses is used to create a more accurate image of the suspect’s face.
While these methods are widely used there has been little in the way of their evaluation, nor has there been much recognition of the possibility that producing such a likeness may interfere with a witness’s original memory.
When witnesses experience problems recalling faces they may be presented with a ‘mugshot file’ which is a collection of photos of people already known to the police, on the assumption that these photos may act as a prompt. One of the problems associated with mugshot files, however is that the witness may feel under pressure to make a selection even though they do not feel particularly confident. This was certainly the case with the child participants in a study by Peters (1987). Another problem is that the witness is being asked to view a static representation of a suspect whom they saw live.
The same problems can occur when a witness is presented with a photospread or a show-up. A show-up is the presentation of a photo of one suspect for the witness to identify, and is the least satisfactory form of identification, though as Ainsworth (2000) points out, one form of show-up is used consistently in courtrooms without question. This occurs when witnesses are asked to point out the perpetrator in court, but it presumes automatically that the right person is in the dock, since the witness is highly unlikely to choose someone in the public gallery. With show-ups witnesses may be led to believe that the police are very confident about the suspect since this is the only one being presented. Yarmey (1992), however, was able to show that only 57% of witnesses in a show-up were able to correctly identify a person with whom they had interacted two minutes earlier. A photospread is more effective than a show-up, as it involves the presentation of at least 12 faces all resembling the suspect. Much, however, will depend on the fair selection of foils alongside the suspect, and the witness may still experience pressure to make a selection, possibly being unprepared to admit that the suspect is not present.
Photospreads would tend to be used in situations where the police have a suspect in mind, but need a witness to confirm the identification. A typical procedure would involve a witness viewing an array of perhaps twelve photographs and being asked if they recognised anyone in the array. If the witness indicated that they did recognise anyone then they should be asked to say which photograph they recognise and from where. Photospreads have a number of advantages over the procedures discussed so far. For example, they are preferable to mugshot inspections in that they allow the witness to peruse a much smaller number of photographs before making a decision. They are also fairer than show-ups because the witness is asked to make a decision about an array of possible alternatives rather than just about one face.
However, photospreads are not without their own problems. For example, a witness who has seen a ‘live’ perpetrator may have difficulty in relating that image to a static photograph. Further, the photograph itself may not be a particularly good likeness, or may have been taken when the accused’s appearance was somewhat different. Many studies carried out in the psychology laboratory suggest that people can be quite good at identifying a face seen previously (Cutler et al., 1994). However, many of these studies lack ecological validity and their results are perhaps not easily related to real life studies. For example, most lab-based studies involve showing subject-witnesses a photograph and then later asking them to identify the same photograph from an array. In such cases, however, the target person’s appearance is identical in both viewings, making recognition somewhat easy. Furthermore, most studies of this kind will use an array which does contain the photograph shown previously. In real criminal investigations it is possible that on some occasions the police will have the wrong suspect in mind and will thus show the witness a photospread which does not contain the real perpetrator.
It would be naïve to presume that on all such occasions the witness will simply tell the investigators that the perpetrator is not present and that they should start looking elsewhere. The pressures on a witness, especially one who has been a victim, may encourage him/her to guess or to point to the photograph which is most like the face of the real perpetrator. Such ‘guesses’ will not, however, be distributed randomly across the twelve photographs presented. In all likelihood the police will have plumped for a particular suspect because they match the verbal description given by the witness originally. If one face in the photo array matches all the elements provided by the witness, it is quite likely that the witness, even if unsure, will choose that photograph. But, of course, this does not necessarily prove that the face which is picked out is that of the real perpetrator — it may simply be that it matches the description more closely than do the others.
Although the regulations governing the conduct of identifications would not allow investigators to use foils whose appearance was completely different from that of the suspect, there will inevitably be some foils which can be rejected immediately because some aspect of their appearance is at odds with that of the real perpetrator. Furthermore, the fact that the witness can view all the photographs at once makes the procedure similar to a multiple choice examination question. The witness may go over the array a number of times and gradually eliminate the less likely foils until they are left with perhaps just one or two from which to choose. In such situations, an identification would not necessarily prove that the face picked out was that of the real perpetrator, but instead may simply signify that it was the one whose appearance was most like them. Laboratory studies which have used a ‘target absent’ array show that people do on occasion pick out an innocent person, rather than admit that they are unable to spot the face seen earlier. This can be a particular problem with younger children (Davies et al, 1988) who may feel pressured into picking someone out even when they are unsure.
The authorities are aware of some of the potential problems of photospread identifications and in many cases will try to prevent mistakes from occurring. For example, it is now routine practice for witnesses to be told in advance that the person whom they saw ‘may or may not be in the photo array’ and that witnesses should only identify someone if they are certain. However, such warnings may be ineffectual when a victim-witness is highly motivated and wants desperately to see his/her attacker brought to justice. Regulations also dictate that when a photospread is used it should be made available to the court so that jurors and defence counsel can see for themselves whether the procedure was fair.
Photospreads can have some advantages over the traditional live identification parade. For example, photospreads may be considered more suitable for young witnesses or for those who are particularly nervous. For such individuals, attending a live ID parade might be particularly nerve-wracking and they may find the mere physical presence of a perpetrator intimidating. In such situations, the real perpetrator may be more likely to be identified if witnesses view a photo-array than if they attend a live ID parade.
Identification parades
(line-ups)
Perhaps the best know type of formal identification process is the identification (or ID) parade. In such procedures a suspect stands among a number of innocent foils and a witness is asked to inspect the parade and then to decide whether the perpetrator is present. The technique is in some ways similar to that of the photospread and as such is prone to the same types of errors. Thus, a witness may feel under pressure to pick someone out even if they are unsure; the appearance of the foils may allow the witness to eliminate some immediately, etc. As with photospreads, there are regulations which govern the fair conduct of ID parades, including the need to inform a witness that the perpetrator may or may not be present and that they should make an identification only if they are sure. It is also normal for the accused’s legal representative to be present at the parade and for the whole procedure to be video-recorded. However, guidelines and advice can never guarantee that mistakes will not be made. As has been noted recently:
Such guidelines may have eliminated some of the more questionable tactics used in the past, but as some recent cases of wrongful conviction have shown, there is still room for bias and error to creep in.
(Ainsworth, 1998: 92)
There are a number of distinctive aspects of live ID parades which may contribute to possible errors. One obvious consideration is the number of foils who should be put into a parade in order to make the procedure fair. Rules concerning this vary from country to country, though in Britain the minimum acceptable number is generally seven. Research carried out in the USA suggests that the minimum acceptable number to ensure fairness is five foils plus the suspect (Wells et al., 1994). However, the actual number of foils on the parade is perhaps less important than their similarity to the suspect. As with photospreads, foils should only be included if their appearance is broadly similar to the description given by the witness. Any foils which can be immediately rejected (because their appearance does not match the description) may just as well not be there at all, as their rejection will be almost automatic. Brigham and Pfeiffer (1994) suggest that the fairness of a line-up should be judged not with reference to its actual size, but rather to its functional size. Thus an eight-person parade containing three foils which can be immediately rejected might be considered unfair while one which contains five appropriate foils may be considered fair, when in effect both sizes of parade are equally fair, as each have a functional size of five foils.
The choice of foils is thus crucial in determining the ease or difficulty of an identification procedure (Lindsay, 1994). The police may wish to increase the likelihood of a witness picking out the suspect by choosing foils who are significantly different from the suspect. By contrast, defence counsel might prefer to have foils whose resemblance is very similar to the suspect, thus making it less likely that the witness will be able to identify the suspect. In practice, the police now have lists of foils whom they can call upon to attend an ID parade. Such people’s physical appearance is kept on record so that in theory it should be possible to call only those whose physical appearance resembles that of a suspect. This is a considerable improvement on past procedures in which it was not unheard of for police officers who were about to go off duty to be called onto a parade to act as foils. In many cases, the physical appearance, stance and behaviour of the ‘foils’ made it very easy for any witness to spot the one person on the parade who stood out from the rest!
Even if the foils’ physical appearance is broadly similar to that of the suspect, it may still be possible for a witness to identify who the real suspect is on the parade by studying participants’ behaviour carefully. For example, it is feasible that the real suspect may be extremely nervous while the foils will be much more relaxed. Sweating, trembling or even subtler non-verbal behaviour may send a signal to the witness as to who the ‘real’ suspect is. While a guilty suspect may well be nervous, it might also be the case that a genuinely innocent person will also show signs of tension. While many humans may be quite adept at spotting when an individual appears to be experiencing stress, they may be less accurate in their presumptions about the factors underlying the tension. Although laboratory-based research studies examining identification procedures have been helpful in many areas, they are unable to replicate the tension which a real suspect on a real ID parade might experience. For this reason, it is not easy to call upon reliable research evidence to establish the extent to which wrongly accused suspects are likely to be picked out as guilty.
A further worry concerning ID parades is that, as with photospreads, the witness can look at the whole parade at once and make a decision on that basis. In theory, a witness’s task is to study each face in turn and to decide whether or not that face belongs to the person who they saw commit the crime. However, in practice, the witness may actually narrow down the possibilities by rejecting those which are the most dissimilar to their memory of the accused and eventually select the one whose appearance is most similar to the image in their memory (Thomson, 1995).
This latter problem could be eliminated if witnesses were prevented from seeing the whole parade at once. An alternative would be to have the witness simply view each face in turn and to say whether the face was or was not that of the accused. The witness would not necessarily be told in advance how many faces there would be on the parade, thus possibly making their decision more reliable. When such procedures have been used, they appear to be successful in reducing significantly the number of false positives which normal identification parades produce. (In this case a ‘false positive’ would be where a witness incorrectly identifies a face on the parade as that of the suspect.) Despite the demonstrated advantages of this method over traditional ID parades (Cutler and Penrod, 1995; Thomson, 1995), psychologists have to date been unsuccessful in persuading the authorities to adopt such procedures.
Research in which participants see a staged incident and then try to select someone from an identity parade demonstrates that eyewitnesses frequently make mistakes. It is worrying to note that witnesses are more likely to pick someone in clothes similar to those worn by the culprit than select them on the basis of such characteristics as height and facial features (Sanders, 1984). Brigham & Malpass (1985) point out that criminals often change their appearance before an identity parade — with considerable success at evasion, so it would seem.
One problem with identity parades is that people assume that the suspect is present (why otherwise organise the line-up?) and will therefore identify the person who most closely resembles the person they saw, rather than no one. Kohnken et al. (1996) therefore recommend that witnesses should be told that the suspect may or may not be present and be urged not to guess.
All members should match the general description of the culprit as described previously by the eyewitness: the target figure should not in any way stand out from the others (Wells et al., 1993). They should all be dressed in similar clothes but not similar to those worn by the perpetrator at the scene of the crime (Yarmey et al., 1996). Wearing clothes that are different gives the police the opportunity to conduct a second test in which the witness will be asked to identify the clothes worn by the perpetrator (Vrij, 1998).
Cutler & Penrod (1988) tested a rather different type of identity parade procedure, known as a sequential line-up, in which the individuals in the line-up are seen one at a time and the witness has to say ‘yes’ or no according to whether or not they recognise them. This makes it less likely that they will simply pick the person who, in the whole line-up, most closely resembles the culprit. Cutler & Penrod conducted experiments in which undergraduate participants watched a videotaped re-enactment of an armed robbery and later attempted to identify the robber from an identity parade. They found that participants were significantly more accurate in their identification on the sequential line-up than on the traditional one and were less likely to make a false identification.
Cutler & Penrod’s research also demonstrated that further improvement in accuracy of identification would be made if witnesses heard voice samples of the members of the identity parade, if the line-up members were shown from three-quarter poses and if witnesses were allowed to watch the line-up members walking in and out of the observation room (Cutler & Penrod, 1988).
A further improvement would accrue if witnesses were less nervous, since the more relaxed they are, the more accurate they are at picking suspects from identity parades (Wells et al., 1999). Anecdotal evidence indicates that some people find the procedure in an identity parade more stressful than the original incident, especially if the individuals in the line-up are not separated from them at all, even by a sheet of glass, as sometimes happens. People need to be told exactly what to expect, to be familiarised with all aspects of the procedure and would also be more relaxed if one-way mirrors were used.
As shown in the Wells & Bradfield (1998) research, witnesses should not be given feedback that confirms their identification as this artificially improves their confidence and distorts other judgements such as how good a view they had in the original incident. The identity parade and photospread should be administered by someone who is blind as to which person is the suspect so that the administrator cannot give the eyewitness feedback about the identification.
Finally, researchers are also investigating
whether running identity parades on video is better than doing them live. It
has been found that witnesses can be so overawed by the conventional live
line-up that they only look at each face for a few seconds. Using video, they
can pause and rewind as much as they like. In tests this led to a 100 per cent
correct identification. When the target was not present, 100 per cent of the
witnesses said that they did not recognise anyone in the parade.
We have already noted that perception can be influenced by prior beliefs and attitudes. Even support for different football clubs seems to affect perception, as Hastorf and Cantril (1954) demonstrated. They showed a film of a football game to fans from both sides and asked them to count instances of inappropriate behaviour and fouls. Guess the result? Each group of fans reported many more fouls from the opposing side.
Racial prejudice can very easily affect the way the world is viewed, as was vividly demonstrated by Duncan (1976) (see below). Chance and Goldstein (1996) have shown that identifying faces from a different ethnic group is more difficult than identifying faces from one’s own group, because of inappropriate and unfamiliar encoding strategies. The area of cross-race identification thus produces a particularly problematic issue in evaluating eyewitness testimony.
The difficulties in cognitive processing
faces of a different race are compounded by stereotypical attitudes held about
the race. Duncan (1976) was interested
in the effect of stereotypes on perception. White participants were asked to
view a video of a disagreement between two people which resulted in one pushing
the other. In fact four versions of the video were shown to different groups:
Tape 1 (White assailant/white victim)
Tape 2 (White assailant/black victim)
Tape 3 (Black assailant/white victim)
Tape 4 (Black assailant/black victim)
Participants were asked whether the behaviour
of the person who pushed the other could best be described as ‘playing around’
or as violent behaviour. When the perpetrator was black, 70% of participants
described his behaviour as violent, while only 13% of participants chose this
descriptor when the perpetrator was white. The attribution of violence was
highest for Tape 3, and lowest for Tape 1.
Peter B. Ainsworth, 2000, Psychology and
Crime, Longman, 0-582-41424-5
Kevin Brewer, 2000, Psychology and Crime,
Heinemann, 0-435-80653-X
Diana Dwyer, 2001, Angles on Criminal
Psychology, Nelson Thornes, 0-7487-5977-8
Julie Harrower, 2001, Psychology in
Practice – Crime, Hodder & Stoughton, 0-340-84497-3