Errors in Eyewitness Identification Procedures
Eyewitness testimony may be questioned on three scientific grounds. First, visibility conditions may be poor - low light, poor weather, etc. Second, many research studies report that even under good visibility, humans are poor at facial identification. Third, the procedures used to obtain the identification may be biased. This article addresses the requirements of a proper and unbiased eyewitness identification procedure rather than the ability of individuals to perform face recognition.
Jurors treat eyewitness identification as compelling evidence in both civil and criminal trials. The strength of eyewitness testimony is demonstrated by a study (cited in Loftus and Doyle, 1992) that recorded verdicts in a mock trial. Two separate sets of the jurors heard evidence differing only by the presence or absence of an eyewitness. With no eyewitness, only 18% of jurors gave guilty verdicts. Addition of an eyewitness identification increased the proportion of guilty verdicts to 72%. Moreover, even when the identification was impeached, the guilty rate was still 68%. Several other studies have similarly found that juries tend to base their decision on a confident eyewitness identification even when other factors (such as poor visibility or bias) question its validity.
Although jurors rely heavily on eyewitness identification, there is overwhelming evidence that eyewitness identification is highly fallible and that eyewitness confidence is a poor guide to accuracy. Here are just a few examples:
A recent study (Wells, et al, 1998) examined the first 40 cases where DNA exonerated wrongfully convicted people. In 90% of the cases, mistaken eyewitness identification played a major role. In one case, 5 separate witnesses identified the defendant.
Huff (1987) studied 500 wrongful convictions and concluded that mistaken eyewitness identification occurred in 60%. This is an amazingly high number since eyewitness identification is an important factor in only 5% of all trials (Loh, 1981).
Cutler and Penrod (1995) examined eyewitness identification accuracy from controlled studies performed in "natural setting." In the typical study, a person enters a convenience store and performs some memorable action (such as paying in pennies) to ensure drawing the clerk's attention. Later the clerk views a photospread and identifies the "customer." The percentage of correct identification ranged from 34-48% and the percentage of false identification is 34-38%. It is hard to know how far to generalize such studies, but they suggest that eyewitnesses are almost as likely to wrong as to be correct when identifying strangers. Moreover, these results occurred under highly favorable circumstances: extended duration, good lighting, clear visibility, and no "weapons focus."
Why is mistaken identity so common? One reason is poor encoding at time of initial perception. This could be due to poor visibility (bad lighting, brief duration, long distance, etc.) or to the tricks played by human perception. A second reason is faulty memory. Memory has several quirks which affect reliability, including 1) low resolution (a remembered face is not as clear as one actually viewed), 2) the tendency for memories to be constructed so that missing information is supplied from expectations/biases or from an external source (TV, newspaper, other witnesses, the police, etc) or from other memories and 3) systematic perceptual distortions in memory (small sizes grow and large sizes shrink, color are remembered as brighter, etc.)
A third reason for error lies in the procedures used during photo-identifications and lineups. Below, I briefly review the mistakes commonly found in identification procedures.
Error Sources in Lineups and Photo-identifications:
From a scientific view, valid finding of fact requires that the outcome is not contaminated by confounding variables (those other than the ones whose effects are explicitly being measured) or biased toward desired results. Moreover, each observation should be independent and unrelated to any other. Photospreads and lineups generally fail to meet these criteria because they offer many opportunities for a biased result.
1. The witness is often not told explicitly that the culprit's picture might not be among the alternatives.
Eyewitness research has repeatedly found that identification is a relative, not an absolute judgment. That is, the witness does not simply compare each picture to memory, making a series of independent yes-no decisions. Instead, the eyewitness looks at all the pictures and then picks the one most likely to be the culprit.
There are several consequences of this decision strategy. One is that the witness is highly likely to make a "false alarm," pick a picture even if the criminal is not in the array, even when unsure. First, an eyewitness probably starts with the assumption that the culprit must be among the alternatives. Why else would the police bother with the photospread/ lineup? Then, likelihood of false identification increases when the police put pressure on the witness to make an identification. Moreover, anything which causes a witness to expect that the culprit is present in the array (e. g., police say "we think we have our man") will increase false alarm rate.
Many studies show that explicitly telling the witness that the culprit may not be in the array greatly reduces false identifications while have little effect on correct ones. The result is much higher overall accuracy. Therefore, the identification examiner should always inform the eyewitness that the culprit might not be present. In fact, the person calling the eyewitness to setup the lineup/photo-identification should also say that the culprit might not be present.
2. The "distractors" are poorly chosen.
Since the eyewitness chooses the "best picture" relative to the others, it is important that the suspect not stand out from the "distractors" (other photos or people) due to different height, weight, coloring, coloring, clothes, behavior, etc. In photospreads, there are numerous ways that one picture can be subtly different: lighting, color tone, brightness, sharpness, viewing angle, background, location of face in the frame, etc.
People who constructed the identification procedure will likely say that the distractors were similar to the suspect, but they seldom present any objective evidence to support their assertion. The only real way to be sure is to test "naïve observers," people not present at the crime with the same alternatives. In a photospread, for example, naïve observers would view the same pictures and then make a choice. In a fair test, they should pick pictures at random, since they cannot use memory to select. However, if there is something innately suggestive or distinctive about a suspect's picture, it may be chosen at a rate above chance. Such a result would seriously question the photospreads validity. Attorney's who have any doubts about the fairness of distractors in a photospread (and this is hard to know a priori because even subtle differences can be critical), should have an experimental psychologist design and conduct an unbiased test.
It would be more difficult to retest a lineup, since
the distractors (people this time) may not be
available or wearing the same clothes. In addition, there is no guarantee that
they will behave the same way as during the ID. It should be remembered, however,
that a lineup following a photo-identification is not an independent event. If
someone identifies a suspect in the photospread, the
witness will almost certainly identify the same person in the lineup - for
consistency sake. Who would want to appear a fool by picking a different person
from photo and from a live group? Moreover, the lineup would be, at best, not a
comparison of people vs. memory but rather of people vs
previously seen photographs (and previous composite or identi-kit
sketches if there were any). In fact, eyewitnesses have strong tendency to stay
with initial identifications even when they are later proved in correct (e. g.,
3. The person conducting the photospread/lineup knew who the suspect was.
There are two reasons that neither the person conducting the lineup/photo-identification nor anyone else in the room should know who the suspect is. First, there is a possibility that he/she will intentionally or unintentionally signal this expectation. The signal need not be blatant ("look at no. 3's picture again!") as even subtle changes in body posture can be enough to tip off the witness. For example, a slight lean forward while the eyewitness views a picture can be enough to draw big red circle around it.
The tendency to signal expectations is so pervasive that drug and other important scientific studies are rejected without a "double blind" procedure, one where neither the subject nor the experimenter knows that the expected outcome. Similarly, courts now generally require that surveys conducted to support litigation in intellectual property cases be performed by questioners who have no knowledge of the desired outcome or even of the issues in dispute. It is ironic that criminal courts, where there can be much more at stake, freely permit introduction of such potentially biased evidence as identifications conducted without double blind procedures.
There is a corollary to the necessity of double blind procedures: the witness must be told that the examiner has no idea who the suspect is. Otherwise, the eyewitness might look for a sign of confirmation, real or imagined. Some personality types constantly seek approval from authority figures, such as the police. They are likely to seek affirmation in feedback from the examiner.
Second, the examiner can easily influence witness confidence after the choice. If the examiner says "good" or "um hmm," after the choice, the eyewitness will feel more confident and likely later express a stronger belief in his/her accuracy. This can be crucial because juries look at not just the identification, but also at the witness's certainty. In fact, one study found that witness confidence is about the only aspect of an identification that jurors consider (Cutler, et al, 1990). This is probably one of the reasons the correlation between eyewitness confidence and accuracy is low (Bothwell, et. al, 1987)
Although eyewitness identification is highly fallible, it still carries great weight with jurors. There are some situations where identification is more likely accurate. For example, if the suspect is someone previously known to the victim, then high accuracy is more probable. When it comes to strangers, however, identifications are frequently in error. Lastly, eyewitness confidence provides only modest assurance that the identification is correct.
Above, I have briefly discussed some of the causes, focusing on procedures in photo-identifications and lineups. Since the law does not require that suspects be represented at the time of a lineup or photo-identification, an attorney who takes a case afterward (or even before) the identification procedure can have no compelling method for evaluating fairness. The best chance of challenging the identification may be scientific evidence obtained by testing the photospread with naïve observers.
· Bothwell, R. R., Deffenbacher, K. and Brigham, J. (1987) Correlation of eyewitness accuracy and confidence: Optimality hypothesis revisited. Journal of Applied Psychology, 72, 691-695.
Brigham, J. and
Cutler, B. and Penrod, S. (1995)
· Cutler, B., Penrod, S., and Dexter, H. (1990) Juror sensitivity to eyewitness identification evidence. Law and Human Behavior, 12, 41-56.
· Huff, C. (1987) Wrongful conviction: Societal tolerance of injustice. Research in Social Problems and Public Policy, 4, 99-115.
Loftus, E. and Doyle, J. (1992) Eyewitness Testimony: Civil
and Criminal. The Michie Company:
Loh, W. D. (1981) Psycholegal
research: Past and present.
· Wells, G., Small, M., Penrod, S., Malpass, R., Fulero, SD., and Brimacombe, C. (1998) Eyewitness Identification Procedures: Recommendations for Lineups and Photospreads. Law and Human Behavior, 22, 603-647.