In 1981, a Texas jury convicted Gary Graham of murder on the basis of testimony offered by a single eyewitness, Bernadine Skillern, who recognized Graham after having seen him fleetingly, in the dark, through a rain-splattered windshield.

Last month, after two decades of failed appeals, Graham was executed by lethal injection. The widespread protests and media coverage that accompanied this event reflected a growing concern among Americans about serious penalties–death in Graham’s case–that can result from convictions based solely on dubious eyewitness testimony.

Although we will never know whether Ms. Skillern was right or wrong in her identification of Mr. Graham, it is becoming increasingly clear that eyewitnesses can, like Ms. Skillern, be absolutely certain of their identification, and yet be absolutely incorrect. One such error was described on June 18, 2000 in the New York Times Op-Ed pages by Jennifer Thompson who described the wrenching experience of discovering that Ronald Cotton, a man whom she had confidently identified as the person who had brutally raped her, was shown by incontrovertible DNA evidence to be innocent. However, not everyone is swayed by the seriousness of such errors: Four days later, in the same editorial pages, Asa Hutchinson, a former federal prosecutor, expressed dismay that, with the validity of eyewitness testimony coming under increasing scrutiny, a cornerstone of convicting guilty defendants is in danger of crumbling.

Disparate though they might seem, these Thompson’s and Hutchinson’s views can be reconciled quite easily: In making a decision about whether to convict or acquit, juries should give a great deal of weight to testimony of eyewitnesses who are likely correct, but little weight to testimony of eyewitnesses who are likely incorrect. Although this sounds like a judicial version of "buy low, sell high", it is not. More than a century's worth of research on human perception and memory has provided a great deal of information that allows us to distinguish those circumstances under which an eyewitness's memory is believable versus those under which an eyewitness's memory should be gravely doubted.

To understand these circumstances, it is first important to picture the dramatic courtroom scene in which a witness like Jennifer Thompson swears to tell the truth, assumes the witness stand and, in response to the prosecutor's questions, expresses utter confidence in her identification of the defendant. As any defense attorney will attest, such testimony is devastating to the defendant's prospects of acquittal: Jurors almost always believe that an identification expressed with great confidence is accurate.

And why shouldn't they? Jurors are normal human beings with lifetimes of experience in monitoring their own memories and, in the process, discovering that strong memories–those that evoke a great deal of confidence–are generally accurate. The jurors' common sense dictates that they should attribute these same qualities to the witness whose memory they must evaluate.

In normal everyday experience, confidence-evoking memories are indeed usually accurate; if they weren't, our memories would be virtually useless as a guide to orienting our lives. However, there are certain circumstances, unusual in most of life, but common in legally relevant events such as crimes that, on the basis of careful scientific research, are understood to produce a witness who is highly confident, and thus highly believable–but dead wrong.

There are two necessary ingredients for such a confidence-evoking but incorrect memory, say of the appearance of a culprit in a crime. The first ingredient is that the circumstances of the crime prevent the witness from forming anything more than an incomplete or hazy memory of the culprit to begin with–for example, it is dark; the witness's view of the culprit is fleeting; the witness's attention is drawn away from the culprit's appearance by a threatening gun; or the witness is experiencing abject terror that her life is about to end.

The second ingredient is some form of biased identification procedure. There are many such biasing procedures which, unfortunately, occur with dreary regularity. Such procedures include a "showup" procedure in which the witnesses is shown a single suspect, under highly suggestive circumstances, and asked, "Is this the guy?"; a lineup procedure in which only the suspect fits the witness's description of the culprit; suggestive behavior on the part of police officers (sometimes subtle or even unconscious, but nonetheless effective) as to which lineup member is the suspect–again, the list goes on and on.

There are two consequences of such a biased identification procedure. The first is obvious: The defendant, whether guilty or innocent, is likely to be identified by the witness. The second consequence is more subtle, but also more insidious. Having identified the suspect, the witness is in a position to reconstruct her memory of the crime such that her originally hazy representation of the actual culprit's appearance is replaced by a much stronger representation of the suspect's appearance. Such reconstructed memories seem very real, and the more the witness rehearses such a memory–perhaps in anticipation of her trial testimony–the stronger and more confidence-evoking the memory becomes. It is on the basis this strong, but unwittingly reconstructed memory of the original event, that the witness will so confidently identify the defendant in the courtroom.

This account of confidence-evoking but dangerously dubious memories has been borne out again and again in scientific laboratories, and is also entirely consistent with the circumstances of real-life misidentifications. One astounding feature of Jennifer Thompson's experience as an eyewitness is that when she was eventually presented with the real rapist, Bobby Poole, she claimed–again with utter confidence and again entirely incorrectly–that she had never seen him before in her life. How could this be? Almost certainly, the answer is that Ms. Thompson's original identification of Ronald Cotton afforded her the opportunity to reconstruct her memory of her rape such that her memory of the rapist's appearance morphed from whatever it had originally been into the appearance of Mr. Cotton. To the degree that Bobby Poole does not resemble Mr. Cotton, it is entirely understandable that she wouldn't have recognized Mr. Poole. Any traces of Mr. Poole's appearance that had ever existed in Ms. Thompson's memory were eradicated when her memory of the rapist transformed into Mr. Cotton's face.

As a researcher in human perception and memory, I have been contacted hundreds of times by defense attorneys who believe that their client may have been misidentified. As I listen to the facts of a case, my goal is to determine whether the two ingredients that I described above–poor circumstances for forming a memory of the culprit to begin with, plus some form of biased identification procedure–are present. If they are present, then I tell the attorney that his or her case might benefit from having a memory expert testify at trial about how memory works and in particular about the circumstances under which a confident identification does not necessarily imply an accurate identification. If, on the other hand, these two ingredients are absent, I tell the defense attorney that the eyewitness is likely correct and that a memory expert probably couldn't testify about any facts of memory, relevant to the case, that are inconsistent with the jury's common sense.

Ex-prosecutor Asa Hutchinson is correct in asserting that a universal disbelief in eyewitness testimony would be a disaster for the orderly functioning of the judicial system. However, there is no need for universal disbelief. There is only a need for juries to be provided with scientifically based knowledge that would allow them to make an informed decision as to what weight they should give to the testimony of a highly confident eyewitness. Such knowledge on the part of Ronald Cotton's jury might have prevented the appalling miscarriage of justice that robbed him of a decade of his life.