KNW 2333 Eyewitness Testimony

"The fallibility of human recall is not the problem. The problem is the lack of awareness of this fallibility"
--Ian Rowland, magician

Why is This Important?

As part of KNW 2333, you were part of an experiment which tests your ability to recall details of a strange event which happens without notice. The results of this exercise always show that most people cannot accurately recall many details. Their descriptions are neither detailed nor accurate. So why is this important?

In case you haven't kept up with such things, a large and growing number of convictions have been reversed by new DNA evidence. According to the Innocence Project, 330 convictions have been reversed in this way (the number is still growing). According to the Innocence Project, eighteen (18) of the people were under death sentence! Over 75% of these wrongful convictions were the result of mistaken eyewitness identification. The Northwestern University School of Law published a detailed analysis of exonerations from 1989 through 2003. This was published in The Journal of Criminal Law and Criminology.

We should note that at least 18 of the exonerations have occurred in Dallas County, making it the leader in the country. There are two primary reasons for this.

  • Dallas County has preserved the physical evidence.
  • The District Attorney is driving the case reviews.
Both are required.

Our justice system is not supposed to send innocent people to prison, but a mass of evidence shows that this has happened many times. The really bad part is that, given the long time involved, it is essentially certain that innocent people have been executed for crimes they did not commit. That's why this is important!


On October 12, 13, 14 2008 the Dallas Morning News ran a significant series titled MISTAKEN IDENTITIES dealing precisely with the problem of wrongful convictions resulting from eyewitness errors. The series also covers prosecutorial conduct, flawed police procedures along with the well-known unreliability of eyewitness identification.

Background of The Problem

It seems to us that there are two problems.

  1. The well-known fallibility of eyewitnesses.
  2. Some very unscientific thinking.

As KNW 2333 is about critical and scientific thinking, it seems appropriate for us to take a look at this. Be assured that lawyers all over the country are doing so also. A lot of papers, articles and court decisions are addressing the issue; we will cite a number of these.

The problems with eyewitness identifications are well-known. Considerable research has been done in this area. A little web searching led to LOTS of pointers to court briefs, law journal articles, books, etc. This subject is no secret within the legal community. The problems have, in fact, been known for a century. Two authorities - Munsterberg (psychology professor) and Wigmore (law professor) - wrote books 100 and 70 years ago. For some VERY good collections of research, see Loftus, Cutler and Penrod, and Wall (see references below).

Now for a few details. Wigmore, in Part II, goes into great detail on the problems of perception and, in Part III, on recollection. It makes fascinating reading. He explores perception via sight, hearing, taste, smell and touch. The perceptual errors possible are quite amazing. "There has always been a quarrel about the objectivity and reliability of sense perception." (p338)

Regarding Testimony Identifying Persons (p427), Wigmore notes that "Some of the most tragic miscarriages of Justice have been due to testimonial errors in this field." That could have been written this year instead of 1937. The investigator often "... cannot determine whether the identification is erroneous or not." Further, "... the danger increases in the proportion of B's actual likeness to A, while really a different individual." The vagueness of human features and observer's inability to recall fine details will cause false recognition. Seems we are learning this all over again.

Wigmore provides an examination of the psychic processes. For example, suppose that the real perpetrator of a crime has properties a,b,c,d,e, and f. What the specific properties are is not important, save that they be visible. Now suppose that the police pick up a suspect who has properties a,b,c,d,m and p. The two differ only in two lesser properties. Prof. Wigmore says that there is a very high risk that the witness will wrongly identify the suspect as the perpetrator and will not be aware of the error. As noted above, the investigator will not be able to detect the error either.

Figure 7.1 in Cutler and Penrod (p100) further illustrates this problem. This graphic shows four men of similar appearance but different hairstyles. The hairstyles make distinguishing them easy. If you cover the hair with a cap, however, telling them apart becomes extremely difficult. A witness trying to pick out one in this case would likely have a 75% probability of making an error.

A report known as the Devlin Committee Report, produced in the U.K., notes that "cases of mistaken identification constituted by far the greatest cause of actual or possible wrong convictions." That is also true here. The Devlin Committee was convened by the House of Commons because of two exonerations of persons convicted by eyewitness errors.

In the U.K., an article in the Criminal Law Review (Williams, see ref.) reviewing the Devlin Report noted that

"The general characteristic of these episodes was a naive belief by the police, judges and juries in the accuracy of visual identification, even by a single witness. It was in vain that the defendant gave evidence of an alibi; such a defence, even if supported by witnesses, generally had no power to countervail a witness's identification."

The Committee stated the problem clearly in paragraph 1.24:

"A witness says that he recognizes the man, and that is that or almost that. There is no story to be dissected, just a simple assertion to be accepted or rejected. If a witness thinks that he has a good memory for faces when in fact he has a poor one, there is no way of detecting the failing."

Section 1.25 of the report notes that:

"We have not as a result of our deliberations discovered any special method by which mistakes in identification can be detected. Modern research, which we consider in paragraphs 4.12-14 below, confirms the general risk of error in memory and observation, but does not as yet offer any convincing means of differentiating the good witness from the bad."

It is worthwhile to read the entirety of section 1.25 (p. 7). Paragraph 4.25 summarizes the problem nicely. Also see paragraph 8.1.

It seems that a major problem lies in jurors' general assumption that the identification is correct when they in fact may have no way to know if that is actually the case. Perceptual and/or recollection errors and illusions may be present and be absolutely undetectable. At trial, the witness may put on a stellar performance, convincing everyone that the defendant is the awful person who committed the crime, but any perceptual or recollection errors are completely invisible and the jury, if aware of the issue at all, cannot make any evaluation of possible errors that is any better than tossing a coin.

Have a look at these references.

The Kansas Supreme Court made a comment (to be repeated later on), which makes the rather pungent statement that

"The problem, of course, is that juries usually attach great weight to eyewitness identification while others involved in a trial and other disciplines have documented that such identification is often unreliable."

This means that everyone but the jurors are aware of how unreliable eyewitness identification is.

Regarding unscientific thinking, Peter Neufeld, writing in the New England Law Review (35:639 2001) writes, on page 643, that

"In fact, one of the fundamental problems is that a lot of us who practice criminal defense decided to go to law school in the first place and become lawyers because we were petrified of science." He adds "Once we got our licenses to practice law, we did not want to deal with science. Prosecutors also do not want to deal with science."

Egads! In this scientific and technological age we have lawyers who do not want to deal with science! In those few sentences Mr. Neufeld has put his finger on a major problem. But what do we do about it? There's a lot of inertia in the legal system; change will be slow.

There are other causes for wrongful convictions beside flawed eyewitness identification. In an op-ed piece, John Terzano of the Justice Project noted "Indeed, it is no longer a mystery why wrongful convictions occur: Study after study has shown faulty eyewitness identification, false confessions, bad lawyering on both sides (egregiously incompetent defense counsel and just as egregious prosecutorial misconduct), snitch and accomplice testimony, and faulty forensics are some of the key common causes of wrongful convictions in both capital and noncapital cases."

How Long Has This Problem Been Known?

Just to cite two of the references, the brief notes that John Wigmore and Edwin Borchard (see references below) wrote about the problem. The brief also notes that "However, the early academic literature on the topic was largely ignored by the legal community." Nothing happened until Patrick Wall produced Eye-Witness identification in Criminal Cases in 1965 (see reference below).

The earliest reference we found was to On The Witness Stand by psychology Professor Hugo Munsterberg of Harvard. The book was written in 1908! The reference said that Munsterberg pointed out the problem but, given the lack of science a century ago, could offer no solutions. Reading the relevant parts of the book confirmed this. John Wigmore wrote a marvelous piece for the Illinois Law Review in 1909 in response to Munsterberg's book. The piece is set as a trial in which Munsterberg is cross-examined. It is very entertaining, having such characters as Judge Solomon Wiseman and a lawyer named Simplicimus Tyro. Wigmore's thesis in the piece is that, while the science of psychology knew a great deal about eyewitness testimony, it could offer nothing of use to courts in evaluating such testimony.

Research scientist (psychology) Robyn Dawes, in his book House of Cards, dissects the current clinical practice of psychology with quite powerful results. His research (and that of many other investigators) shows that psychology still cannot offer any tools which can help in a specific case.

The Consequences

One consequence of the problem is obvious - innocent people spend years in prison for crimes they did not commit. In our opinion, the proper number of innocent people to send to prison is zero. Any more than that indicates a problem.

Another consequence is a fallout from the first - lawsuits. After being deprived of years of their lives, these innocent people are filing lawsuits against the entity that sent them to prison. These lawsuits can get VERY expensive for the cities, counties, etc. In Texas, wrongly convicted people are entitled to $80,000 per year of confinement, $100,000 for time on Death Row.

  • (p.13) City of Boston pays $3,200,000 to innocent man. Really disgusting police behavior
  • Marlboro MA man gets $13,600,000 award. City legal costs about $1,000,000.
  • State of Connecticut awards $5,000,000 to innocent man.
  • State of Maryland pays $250,000 to innocent man.
  • Illinois case results in $36 million settlement, the largest yet recorded.
  • In Part 2 of MISTAKEN IDENTITIES, the Dallas Morning News reports that Billy Wayne Miller received $500,000 in compensation from the State of Texas.
  • Texas Monthly magazine (Nov 2008 p. 153) reports that Richard Danziger sued the City of Austin and won $10.5 million. He had spent 11 years in prison.
  • The above Texas monthly story also reports that Billy James Smith has filed a lawsuit against the Dallas Police Dept. He spent 19.5 years in prison.
  • Dallas Morning News (25 March 2011) reports that the City of Irving paid a $700,000 settlement to a wrongly convicted man. Police misconduct in the lineup procedure is alleged. Including legal fees, the settlement cost Irving about $1,200,000. The problem was eyewitness error.
  • Dallas Morning News (12 May 2011) reports that a man convicted of two rapes in 1984 had been completely cleared by DNA evidence. He spent 27 years in prison. The false conviction was the result of eyewitness error; both victims testified and identified the man. Both were mistaken.
  • Dallas Morning News (6 April 2012) reports 3 men to be exonerated in a 1994 purse snatching robbery. Bad police procedure (a "showup" identification), prosecutorial misconduct (withholding evidence) and eyewitness error were the reasons for the wrongful conviction.
  • Dallas Morning News (28 April 2012) reports that 2 men were to be cleared (declared innocent) in a 1983 rape/assault case. The report noted that the victim "believed with all her heart that 'these were the men that raped me'. But she was wrong." Two more wrongful convictions due to eyewitness error.
  • (11 Feb 2013) Fox News reports that Randolph Arledge was exonerated after serving 29 years in prison for a murder did not commit. Modern DNA testing cleared him. A lying witness contributed to the false conviction.
  • (25 Jul 2014) Dallas Morning News reports that DNA analysis had cleared a man sentenced to prison for 12 years for a 1990 rape he did not commit. This the result of DA Craig Watkins' policy of reviewing old cases and performing DNA tests if evidence is available and testing was not done earlier. The cause of the wrongful conviction was eyewitness (victim) error. The man had been released from prison in 2002. The annoying part of this case is that the DNA test that exonerated him also pointed to the real villain. Unfortunately the statute of limitations had expired and he cannot be prosecuted. The real rapist got away with it.
  • (20 Mar 2015) In Ohio, Ricky Jackson was released from prison after 39 years served for a murder he did not commit. The conviction was based on the testimony of a 12-year old "eyewitness" who said he saw the crime. Years later the witness recanted, admitting that he had not actually seen the crime. The state of Ohio will pay Jackson over $1,000,000 in restitution.

The View from the Law

We are indebted to the late Prof. Dan Shuman of the SMU Law School faculty, who taught evidence law. His consultation was quite valuable.

The situation for the law is complex. Lawyers have a very big picture to learn. The criminal cases that get to trial are the difficult ones. Many prosecutions end in confessions and/or plea bargains. Others can't proceed because of lack of evidence. The toughest are those where there is no evidence other than one or more eyewitnesses.

We could note that the scientific and legal views of evidence are different.

  • Scientific: The evidence must be unimpeachable. You must always try to shoot down your hypothesis; find any holes in it. Eyewitness assertions and anecdotes are useless. Reliable measurements (or the equivalent) are required. Standards are very high in order to avoid fooling yourself. A scientist will do something a lawyer might not do - try to impeach their own evidence.
  • Legal: Useful evidence is determined by what can be allowed in court. Rules of evidence hold that relevant evidence can be used unless some other specific rule prohibits it. Eyewitness testimony, including identification, is widely accepted. Juries often consider it to be infallible.

Keep in mind what the situation of law is: the court cannot go out and rerun the crime in the way a scientist would replicate an experiment. This is not possible - crimes are one-offs. The court has to deal with these unique situations at times often significantly removed from the event.

Here's one interesting note. Agent Paul Ahner of the U.S. Secret Service told us that the Secret Service will not file a case with the U.S. Attorney if all they have is eyewitness testimony. They require something physical like a paper trail or other forensic evidence.


To help analyze all of this, we can formulate some hypotheses about our legal processes. We'll look at them in light of any evidence that might show that any of them are wrong and will cite some of that evidence.

  • The current safeguards work reliably to protect the innocent. As of mid- 2011 Dallas County has seen 32 exonerations in which people convicted of crimes were released because DNA evidence indicated that they could not be guilty. The safeguards clearly did not work in these cases. These wrongful convictions resulted mostly from mistaken eyewitnesses.
  • A judge can reliably decide whether safeguards are adequate. The previously mentioned exonerations show that any safeguards, if indeed any were used, were not adequate. Wrongful convictions resulted.
  • A jury can reliably detect an eyewitness error and protect the innocent. The exonerations prove that those juries could not detect the errors. Cutler and Penrod devote Chapter 12 to this subject, and their conclusions are not encouraging. Given that (according to Wigmore and others) the investigator cannot detect the eyewitness error, just how is a jury supposed to do it?
  • Proper cross-examination can expose an eyewitness error. Cross-examination, even if used properly, was ineffective during those 15 trials. The errors were not uncovered and wrongful convictions resulted.
  • A witness who is quite confident of an identification is always correct. Loftus (p101): "To be cautious, one should not take high confidence as any absolute guarantee of anything." That's the result of a lot of research. This hypothesis is false. Cutler and Penrod confirm this.
  • A trained person will not make an identification error. See Loftus (pp163-170). In tests, even trained police officers can make mistakes. Experiments in training officers to improve identification accuracy produced no real improvement.
  • A lie by a witness will always be detected. Prof. Shuman notes that witness lying is still somewhat of a problem. The real challenge is detecting the lie. Without additional evidence at hand, a clever lie can pass undetected.
  • We (the people) know how many people have been falsely convicted because of eyewitness errors. This cannot be true. If juries cannot detect the errors, the safeguards are not foolproof and cross-examination cannot reliably reveal the errors, then we cannot know how many innocent people are in prison at this time. The eyewitness errors were undetectable. All we can know is a minimum rate - the number of those who have been exonerated so far.
  • There are no known means for reliably detecting eyewitness error. This sounds extreme, but all that is necessary to refute it is a demonstration of some method that will reliably (with certainty) detect essentially all eyewitness errors. Keep in mind that Wigmore indicated that this hypothesis is almost certainly true. A very large body of research backs him up. The Devlin Report in the U.K. says the same thing. A law article by Glanville Williams in the Criminal Law Review (July 1976 p410) notes that
    "As the Committee succinctly says, the special difficulty with identification evidence is that its reliability is exceptionally difficult to assess. Ordinarily, a witness can be tested by considering whether his story is probable and coherent, and by observing his demeanour. But the evidence of identification rests on a single piece of observation; there is no story. And the witness may be wholly convinced of the correctness of his observation and yet be wrong."

What Can Be Done?

A number of references indicated that procedural reforms in the area of witness identification can help. It turns out that lineup and photospread presentations have been done in a very unscientific manner. Biases creep in all over. Here's a rough outline of what might be done.

  • The presentation must be double-blind.
    • The witness must be given no hints of any kind.
    • The officer presenting the suspects/photos cannot know who the suspect is.
    • The photos must be presented in several groups with the suspect appearing in only one group.
    • The witness should be told that the suspect may or may not be in the group presented. This removes pressure to make an identification even if wrong.
    • Someone who is NOT present during the witness' viewing will evaluate the choice.
  • Multiple witnesses must not be allowed to converse with each other or to compare notes or choices.
  • Photos/persons must be presented sequentially, not together. This will prevent "comparison shopping" by the witness.
  • All people presented must have similar physical/racial characteristics.
  • The photos should be presented in at least 2 separate groups of 4 to 6. When a group is presented, the witness must be told that the suspect may or may not be in that group. This will reduce pressure to identify "somebody."
  • Witness seeing a suspect's photo before the presentation is prejudicial. Photos must not be released to the press before lineups are conducted.
  • The entire identification process must be video-recorded with audio.
Fortunately, recent news stories indicate that many jurisdictions around the country are, in fact, modifying their identification procedures in line with these scientific principles. This change should significantly reduce the number of mistaken eyewitness identifications. As an example, the Dallas Morning News (16 Feb 2009, p3A) reports on efforts in the state legislature aimed at reforming processes in Texas. Most of the story is about DNA samples, but there is a note about an eyewitness error and police procedures. In the Timothy Cole case, rape victim Michelle Mallin picked Cole out of a photo lineup of six photos. "All were were standard jail mug shots except for Cole's color Polaroid." (DMN) Good grief!! It was patently obvious which photo was the person that police had arrested. It would be hard to do a better job of telling the witness which one to pick.

Courts have called flawed procedures "impermissibly suggestive."

Note: on 29 October 2007 the American Bar Association released a multi-state study of state death penalty systems. The study identified problems such as discrimination, procedural difficulties, inadequate defense and inadequate review processes. Stephen F. Hanlon, chair of the project noted

"After carefully studying the way states across the spectrum handle executions, it has become crystal clear that the process is deeply flawed."

In the large list of problem areas presented in the press release, this one is of interest here:

"States are not establishing policies or requiring prosecutor's offices to establish policies on exercise of prosecutorial discretion, or on evaluating cases that rely on evidence such as testimony of jailhouse snitches, or on eyewitness identification or confessions, considered as less reliable evidence." (boldface theirs, italics ours)

Prof. Shuman suggested scientific investigation into what conditions might "contaminate" an eyewitness. Does the time of day of the lineup presentation make any difference? Should the witness discuss the event with anyone else? What happens if the witness has seen the suspect's picture in the newspaper? Some investigation into the psychology of such situations might help. In fact, investigations of this type have been done. Cutler and Penrod plus Loftus (see below) detail large numbers of studies in these areas.

Some of the answers follow.

  • Does the time of day of the lineup presentation make any difference? We have not found any reference to this factor yet. It may be insignificant.
  • Should the witness discuss the event with anyone else? No. In particular, the witness should not discuss the event with any other witness.
  • What happens if the witness has seen the suspect's picture in the newspaper? This is terribly prejudicial. The newspaper photo can contaminate the witnesses memory and produce a false identification.

The courts do try to reduce the rate of error. On the other hand, there is a responsibility to get the "bad guys" off the streets. Law enforcement faces the difficulty that changing identification to a scientific model will almost certainly result in a large increase in failure to identify; the witness is not able to pick out the suspect. This poses a problem for the prosecutor, who has no case without the witness identification (positing that there is no other evidence available).

A recent report (June 2010) commissioned by the Supreme Court of New Jersey addresses the issue quite forthrightly.

There are three possible outcomes where eyewitness testimony is the only evidence available.

  • Getting it right (correct conviction or acquittal)
  • Type I error (false positive) - convicting an innocent person
  • Type II error (false negative) - actually guilty person escapes

We must remember that the prosecutor is under some public pressure to get the "bad guys" put away. They don't like Type II errors! It appears that the law knows that the error rate is non-zero. Type I errors are going to occur. Type II errors will occur when the witness fails to identify a really guilty person (these witnesses will obviously not be appearing in any trial) or when the jury rejects the eyewitness testimony and acquits when the defendant is really guilty.

A Kansas Supreme Court case (86,959) points to a very interesting reference.

"Astonished by the pardons of two individuals who had been independently convicted on the basis of erroneous eyewitness identifications, the British Home Secretary appointed a committee to investigate this area of criminal law and police procedure. The committee, chaired by Lord Devlin, recommended that the trial judge should be required by statute (1) to direct the jury that it is not safe to convict upon eyewitness evidence unless the circumstances of the identification are exceptional or the eyewitness evidence is supported by substantial evidence of another sort, (2) to indicate to the jury the circumstances, if any, which they might regard as exceptional and the evidence, if any, which they might regard as supporting the identification and (3) if a trial judge is not able to indicate either such circumstances or such evidence, to direct the jury to return a verdict of not guilty."

See reference below to "Evidence of Identification."

Farther along in that opinion, after discussing several models for handling eyewitness identification, the court notes that

"The problem, of course, is that juries usually attach great weight to eyewitness identification while others involved in a trial and other disciplines have documented that such identification is often unreliable."

Two Questions for the Jury

It appears that a jury presented with an uncorroborated eyewitness identification is being asked to evaluate TWO questions.

  1. Does the witness believe that the defendant committed the crime?
  2. Is the witnesses identification actually correct?
Herein, we believe, lies the problem. The answer to the first question is always YES. A witness who did not believe the defendant did it would not be testifying; a friend who is a Dallas County prosecutor confirms this. This question therefore requires no critical evaluation; the answer is a given. It's the second question that causes the problems.

Recent experience in Dallas County (any many other jurisdictions) proves that the answer to the second question can be NO while the answer to the first is YES. There is no necessary connection between the two questions. An eyewitness can be convinced and wrong at the same time. The story of Jennifer Thompson is instructive. She was absolutely certain that a certain black man raped her. She positively identified him in court, certain beyond any possible doubt. The man was sentenced to life. Thompson remained certain until 11 years later, when DNA evidence not only proved that she was wrong but also pointed to the real culprit. She had been absolutely positive and absolutely wrong. There's more on the Jennifer Thompson case. In this bit, a quote from the Associated Press, investigator Gauldin says that the system worked. An innocent man had been freed. We respectfully disagree. The system did not "work" - it failed badly. An innocent man spent 11 years in prison for a crime he did not commit.

As of late 2008, Jennifer Thompson and the man she sent to prison are jointly writing a book about the whole affair.

Eyewitness, Part 1 - Ronald Cotton and Jennifer Thompson on 60 Minutes.

Eyewitness, Part 2 - Ronald Cotton and Jennifer Thompson on 60 Minutes.

The difficulty we have with the second question is that ALL references so far (including Wigmore) indicate that there are no known means for detecting eyewitness errors in the absence of other reliable evidence. Mike Ware, head of the Dallas County DA's Conviction Integrity unit is quoted (DMN 30 Dec 07) as saying "how did it get to that point, where the victim is picking out the wrong person in the courtroom?" If Mr. Ware would study the scientific research on this subject he would find out how we got here. The problems with eyewitness identification are not news anywhere.

This leads to a conclusion: we are asking our juries to evaluate unsubstantiated claims while having no reliable tools for doing so. They cannot reliably do it correctly; the Dallas County experience proves that. The witness's sincerity of belief and performance on the stand are of absolutely no help in answering the second question; research has shown that. We are asking too much of our juries.

Chapter 12 of Cutler and Penrod deals with jurors' ability to detect eyewitness errors. Please note that it appears to us that judges and lawyers assume that juries can reliably do this. Note the summary conclusion.

"In sum, there are a variety of reasons to be concerned about jury decisionmaking in eyewitness identification cases.

  1. Jurors tend to overbelieve eyewitnesses.
  2. Jurors apparently have difficulty reliably differentiating accurate from inaccurate eyewitnesses.
  3. Jurors are not adequately sensitive to aspects of witnessing and identification conditions that are arguably better predictors of witness accuracy than is witness confidence.
  4. A major source of juror unreliability is their reliance on witness confidence -- which:
    • Is a dubious indicator of eyewitness accuracy even when measured at the time an identification is made and under relatively "pristine" laboratory conditions.
    • Appears to be high malleable and influenced by post-identification factors such as repeated questioning, briefings in anticipation of cross-examination, and feedback about the behavior of other witnesses. These factors do not increase witness accuracy, and are therefore likely to further reduce any relation between witness confidence and accuracy, and are therefore likely to further reduce the ability of jurors to distinguish accurate from inaccurate eyewitnesses."

That finding provides a lot of reasons to doubt eyewitness identification. Also - if you think the foregoing is depressing, look at what followed.

Chapter 13 in Cutler and Penrod deals at length with juries' sensitivity to factors influencing eyewitness accuracy. Take note of a primary conclusion.

"More comprehensive research has been conducted in jury knowledge about the factors that affect identification accuracy and their decision processes in eyewitness cases. Taken together, the survey studies, the prediction studies, and the mock-juror experiments converge on the conclusion that jurors are generally insensitive to factors that influence eyewitness identification accuracy, often rely on factors (such as recall of peripheral details) that are not diagnostic of witness accuracy, and rely heavily on one factor, eyewitness confidence, that possesses only modest value as an indicator of witness accuracy. The implications of this conclusion are profound. Even if attorneys were given the opportunity to gather the information necessary for effective cross-examination and even if they knew what questions to ask in eliciting eyewitness identification at trial, the effectiveness of cross-examination as a safeguard is still questionable in light of the lack of juror sensitivity to factors that are known to be diagnostic of eyewitness reliability."

See Chapter 6 of Cutler and Penrod for great detail on factors affecting witness accuracy.

Does anyone still believe that the "safeguards" work reliably?


One thing to emphasize - a false eyewitness identification does NOT mean that the witness was lying. The witness is convinced that the identification is correct. Psychological research confirms this. We have not yet seen notes about deliberate witness fraud, so we infer that such deliberate lies are infrequent. Munsterberg notes "...the witness who lies offers no psychological interest for the student of illusions." Prof. Shuman concurred that witnesses deliberately lying is something of a problem, but its magnitude cannot be known. This makes perfect sense - you can't know whether a witness is lying unless you have independent means of detecting the lie.

What emerges from all this is the idea that a witness' degree of certainty does not necessarily have any connection to the actual truth of the identification. It is possible to be completely convinced, quite certain, and dead wrong! This principle applies in many other areas as well.

In addition, it appears to us that making reforms would be highly beneficial to everybody.

  • False imprisonment would be avoided; no innocents sent to prison.
  • Judicial resources would be conserved.
  • Embarrassment of the judicial system would be avoided and public confidence bolstered.
  • Cities and counties could avoid large damage awards (multi-million $)

We, as citizens, have a real interest in having our legal system "get it right" while minimizing errors of both types.

16 April 2008. The Dallas Morning News reports that a 16th man has been exonerated by DNA evidence. A man named McGowan spent nearly 23 years in prison for a 1985 rape/burglary he didn't commit. He was convicted solely on the basis of a mistaken eyewitness. The problem was a badly flawed photo lineup procedure. There is one encouraging note in the article.

"Dallas police are exploring a pilot project in which lineup photos would be viewed one at a time and administered by an officer who does not know whom police suspect as the perpetrator. Mr. Ware said all photo lineups should embrace this 'double-blind' procedure, as well as making photos as uniform as possible" (DMN 16 Apr 08)
This procedure is exactly what has been recommended many times. It will, almost certainly, result in a lot higher rate of witnesses failing to identify the suspect. Serial presentation of photos prevents the witness from comparing photos and picking the one that seems to look most like the perpetrator. This "comparison shopping" is extremely dangerous.

16 September 2008. The Dallas Morning News (page 1 section B) reports that another man has been cleared by DNA evidence. A man named Lindsey is on the way to freedom after 26 years in prison. This brings the Dallas County total to 19. An article on the front page of that paper quotes defense attorney Richard Franklin as saying that "All the science points to the fact that eyewitness testimony is no good." This is in a comment on the DA's review of deat penalty cases that were based on eyewitness testimony.

19 January 2009. Progress. The Dallas Morning News (page 1 section B) reports that The Dallas Police Department has dropped out of a study and is proceeding to overhaul the way it performs photo lineups. The sequential blind procedure being adopted is exactly what many references suggest. In this method, an officer who does not know who the suspect is presents the photos to the witness serially (one at a time). This implements what the police department was studying in April of 2008 (see note and DMN reference above).

We would actually recommend tightening the controls even more.

  • Use THREE groups of six photos, with the suspect appearing in only one group. The witness would not know which group the suspect was in.
  • Require that all photos be of the same quality. If the suspect's photo is a color photo, ALL fillers' photos must be color also.
  • Digitally edit all photos to remove all backgrounds and clothing, so that only the face is left for the witness to see. No hats of any kind allowed.

We're Playing a Game

After a LOT of study, we reach the conclusion that prosecutions based solely on uncorroborated eyewitness identifications are very much like playing a casino game, with the significant differences being:

  1. In the casino game we are not gambling with someone's future - or life.
  2. In the casino game we can know the odds.

Another Conclusion

There is one more approach to this. Consider the following possible situations.

  • Witness fails to identify suspect
  • Guilty suspect erroneously acquitted in trial
  • Wrongful conviction sends innocent person to prison.
The significant common factor here is that, in all 3 cases, the guilty party gets away with it. Given that, we can evaluate the cases on other grounds. What other effects might distinguish the cases?
  • Witness fails to identify suspect. The perpetrator gets away with the crime but authorities can always go after him if more evidence appears.
  • Guilty suspect erroneously acquitted in trial. The suspect is completely free; he cannot be prosecuted again, even if ironclad evidence appears. Maybe that case should not have gone to trial.
  • Wrongful conviction sends innocent person to prison. The perpetrator gets away with the crime but can be prosecuted if more evidence appears. The disaster here is that the innocent victim is sent to prison. Also - the investigation usually stops. In one case the real perpetrator raped two more women before he was finally caught by DNA evidence. In a Texas case a man (David Pope) was freed after 15 years in prison; the DNA evidence actually identified the real perpetrator, but the statute of limitations had run out. This happened again. The April 6, 2012 DMN (and other newspapers) carried a story about three men sent to prison 17 years ago for a purse-snatching in Dallas. Beginning in 2002, a University of Houston law student began working on the case. The end result was exposure of bad police procedure, mistaken eyewitnesses and prosecutorial misconduct (withholding evidence). The men were declared as innocent of a crime they did not commit. The ultimate insult was that the student was actually to trace the actual perpetrators of the crime. The four still alive finally admitted to it. As in the case above, the statute of limitations had run out and the real criminals cannot be prosecuted. These two cases illustrate one of the worst side effects of eyewitness errors - the real criminals get away with the crime.
If we are going to make a value judgment regarding this, we might want to favor the situations which involve the least collateral damage. In those terms, failure to identify is least damaging; the guilty party can still be prosecuted if found and no innocent person's life is ruined.

Some Questions

After lots of reading about the situation, we have several questions that we would like to find answers to. Some progress is being made

  1. Why do science and law take such opposite views of eyewitness testimony? This is due to very different circumstances. Science insists on replicable experiments or observations, while such are impossible in court. The court must try to reconstruct something that happened in the past, relying on the adversary system to keep things straight. The standards of evidence are different.
  2. Given that the problems of eyewitnesses have been known for at least 100 years, why are we still getting false convictions because of those same problems? It could be that a lot of the Type I errors are due to flawed procedures or "bad lawyering on both sides." These problems are amenable to solution.
  3. In this 21st century scientific and technological age, why must we still have lawyers who don't want to deal with science? (See Neufeld above) We don't know - any suggestions?
  4. Given that the witness always believes the defendant did it, how is a jury supposed to reliably evaluate the claim? What tools are available?
  5. What is the legal profession's tolerance for Type I errors (false convictions)?
  6. Why is prosecution based solely on eyewitness identification allowed at all?

Some References

These are a selection of references concerning eyewitness identification. There are more out there - LOTS more. Note that the books by Munsterberg, Borchard and Wigmore are old; used copies of Munsterberg are available, but Wigmore will be hard to find. Borchard is available but pricey. Wall is available. All are out of print. They might be found in law libraries. The Wigmore piece in the Illinois Law Review is not available on-line; it will be found in law libraries. By the way, it is as entertaining as it is informative. The Criminal Law Review (UK) will likely be found only in law libraries. Most of the other books should be available.