Eyewitness testimony is commonly regarded as very high quality evidence. But recent research has shown there are many ways memories of events can become contaminated. This article is based on a presentation to the NZ Skeptics conference in Wellington, 27 September 2009.

In 2003, a woman was tragically attacked and raped after leaving a bar in Christchurch. She remembered her assailant as a man with “rat- like” features. Later, she chose the police suspect from a photographic lineup, indicating that she was “90 percent sure” that he was her assailant. This identification became the central piece of evidence that convicted Aaron Farmer. But, in June 2007, Mr Farmer was exonerated after DNA proved that he could not have been the rapist – he had spent almost three years in prison.

Unfortunately, Mr Farmer’s case is not an isolated incident. Decades of legal and psychological research have shown that eyewitness identification error is the leading cause of wrongful conviction. Recently the former High Court judge, Sir Thomas Thorp, published an extensive review of legal research on miscarriages of justice. In that paper, he estimated that there are at least 20 innocent people in New Zealand prisons, and he emphasised eyewitness error as a leading cause of convictions. This conclusion fits neatly with exoneration data from the Innocence Project, based in New York. Since 1992, the Innocence Project has exonerated over 250 wrongfully convicted people, over 75 percent of whom were identified by at least one eyewitness.

How can human memory be so fragile as to lead a witness to choose an innocent person from a lineup? Over 30 years of research has shed light on this question. Ultimately, this research has shown that memory can go wrong in several ways. The best way to understand these errors is to think of memory as a three-stage process:
[1] encoding,
[2] retention, and
[3] recall.

At the encoding stage, information is perceived and transferred from the environment, through our senses. These perceptual processes allow us to lay down memory traces. Next, those traces are retained for a period of time. Of course this retention stage can last for anywhere between seconds and years, until finally we recall that information from memory. It is important to know that any one of these three stages can go awry.


Encoding depends heavily on our ability to pay attention to information in the environment. However, our attentional systems are limited. We can only pay attention to a few things at once. Anything that does not receive the requisite amount of attention does not have the chance to make it through the encoding phase of memory.

Furthermore, many variables, such as stress, can limit our attentional processes even more. As a result, witnesses will often not pay attention to details that could be forensically relevant. For example, a witness under stress may pay particular attention to the weapon being brandished by the offender, rather than paying attention to his facial details. If this is the case, those facial details may never be stored in memory, and if information is not stored, it cannot be recalled later.


The information that makes it into memory can be distorted easily. Perhaps the best known psychological science research in this field is the misinformation effect pioneered by Elizabeth Loftus. This research shows that a simple suggestion can change witnesses’ memories. In a typical misinformation experiment, there are three stages.

First, participants watch a simulated crime, such as a man stealing a maths book from a bookstore. After a delay, participants are exposed to post- event information (PEI), which is usually a narrative describing the simulated crime. For some participants, the PEI is accurate but generic (eg, “the man stole a book”), and for others the details are misleading (eg, “the man stole a science book”).

Finally, participants are questioned to determine their memory’s accuracy for the event. These participants are often specifically told to ignore everything they read in the narrative and only rely on what they saw during the event. Typically, those participants who read misleading details during the PEI have less accurate memories than those who read generic information.

This research shows the ease with which a person’s memory can be changed. Decades of research have shown that people can come to remember having seen a crime when in fact they have seen an innocuous event. Using this paradigm people can even come to remember having seen an innocuous event, when in fact they have seen a crime. Witnesses can often be exposed to misleading details from co- witnesses, suggestive interviewing techniques or sometimes, media reports of the crime. Any of these sources can lead witnesses to remember details that did not happen.


Psychological science has also shown that the way we test witnesses can also affect their memories for what they have seen. Some of the most prolific research in this field has examined the way that we test witnesses’ memories for offenders’ faces using the lineup technique. Photographic lineups are the most common method of testing eyewitness recall for offenders.

Usually, a lineup depicts a police suspect surrounded by known innocent people – known as distracters. A witness chooses a person from a lineup in the same way that a person chooses an option from multiple- choice question. When people choose the correct answer from a multiple- choice question it is considered evidence that they recognised the correct answer by relying on memory; and when witnesses choose the suspect from a montage, it is considered evidence that they recognised the suspect from the crime scene.

However, people do not always rely on their memory in either multiple- choice questions or lineups. A multiple-choice question can be biased towards the correct answer, as in this example:

What is the capital of Burundi?

Most people cannot rely solely on their memory to answer this question. Now consider these choices:

(a) Paris;
(b) Sydney;
(c) Wellington;
(d) Bujumbura.

You probably chose the correct answer (d), not because you had a memory for Burundi’s capital, but because you used a process of elimination to choose that answer. Similarly, a lineup is sometimes constructed so that witnesses do not need to rely on their memory for the offender; instead, they use a process of elimination – the suspect becomes the Bujumbura of the lineup.

Lineup bias

The danger arises when the wrong person is suspected of a crime and then included in a biased lineup. Research shows that witnesses will often choose from a lineup, even when the actual offender is not present. If the lineup has been constructed in a biased way (like the multiple choice question above), witnesses are even more likely to choose from the lineup. It is misidentifications like these that often lead to wrongful convictions.

Taken together, this research shows that witnesses’ memories are susceptible to several sources of error. As such, we need to ensure that we collect and test witnesses’ memories with scientifically valid interview and lineup techniques. Scientific recommendations regarding best practice procedures for witness evidence have been available for several decades, but few jurisdictions worldwide have taken them up. This lack of recognition for scientific validation is surprising given the relatively fast uptake of forensic science methods, such as DNA testing.

As a result, the best way to think of witness memory evidence is like biological evidence at a crime scene. If we were unlucky enough to stumble across a bloody crime scene, most people would be careful not to contaminate the scene by trampling through the blood spatter patterns, or handling any evidence. Similarly, we should treat witness memory with the same caution. When a witness has been exposed to a crime, we should not contaminate their memories with suggestive questioning and biased lineups. Instead, we should collect and preserve their memories with scenically valid techniques. Only then can we hope to reduce the increasing number of wrongful convictions caused by erroneous witness evidence.

Recommended Posts