Why Does Eyewitness Memory Fail? Recall Rebuilds Rather Than Replays

Encoding is selective and the conditions at a crime are close to the worst possible. Attention is narrow under stress, and weapon focus describes the documented tendency for a witness confronted with a weapon to fixate on it and remember the face less well. Exposure is brief, lighting is frequently poor, distance is greater than witnesses estimate and duration is consistently overestimated. Stress impairs rather than sharpens memory for detail, contrary to the widespread belief that traumatic events are burned in, with studies of soldiers undergoing intense interrogation training finding that a large proportion subsequently failed to identify their interrogator. Cross-race identification is measurably less accurate than same-race identification, a robust finding replicated for decades. None of these problems is visible to the witness, whose experience is simply of having seen something, which is why the resulting confidence bears little relation to accuracy at this stage.

The memory continues changing between the event and the courtroom, and the processes are well documented:

• •Post-event information, since details learned afterwards from other witnesses, news reports or questioning can be incorporated into the memory as if they had been observed, which Elizabeth Loftus demonstrated across decades of experiments
• •Question wording, since asking how fast cars were going when they smashed rather than hit produces higher speed estimates and more reports of broken glass that was never present
• •Unconscious transference, where a face seen in another context is recalled as the perpetrator
• •Repeated recall, which strengthens whatever version was retrieved rather than the original, so each telling can drift further
• •Confirming feedback, since telling a witness they identified the suspect substantially inflates their later confidence and their recollection of how good a view they had
• •The confidence and accuracy relationship, which is weak once contaminating feedback has occurred and considerably stronger at the moment of an initial well-conducted identification, which is why recording confidence immediately matters

Line-up design determines how many innocent people are identified. If the officer administering the procedure knows who the suspect is, subtle cues influence the witness, which is why double-blind administration is recommended. If the other people in the line-up do not match the witness's own description, the suspect stands out and the procedure becomes a test of who most resembles the description rather than a recognition test. If the witness is not told that the perpetrator may not be present, they tend to choose the closest match relative to the others rather than making an absolute judgement. Showing photographs one at a time rather than all together has been argued to reduce that relative comparison, though the evidence is more contested than it once appeared. Repeating a procedure with the same witness and the same suspect contaminates it. These are all fixable design problems, and the reforms that address them, including blind administration, proper fillers, explicit instructions and immediate confidence statements, are recommended by scientific bodies and unevenly adopted.

Analyses of exonerations established through DNA evidence have consistently found mistaken eyewitness identification as the leading contributing factor, present in a substantial majority of cases in which an innocent person was convicted and later cleared. Those cases are informative because DNA provided an independent check that is usually unavailable, which implies the underlying error rate in cases without biological evidence is not lower. The response has included courts requiring expert testimony on memory to be admitted, revised jury instructions explaining the limits of identification evidence, and a landmark New Jersey decision in 2011 that comprehensively reviewed the scientific literature and restructured how identification evidence is handled in that state. Interview technique has also changed, with the cognitive interview, which avoids leading questions and asks witnesses to recall freely in varied orders and perspectives, producing more accurate detail than conventional questioning and being adopted by police forces in several countries.

Memory reconstructs rather than replays, so it can be altered by post-event information, question wording and feedback, and confidence rises with confirmation rather than with accuracy. Stress narrows attention and impairs detail, weapon focus reduces face memory, and cross-race identification is measurably less accurate. Line-ups fail when the administrator knows the suspect, when fillers do not match the description and when witnesses are not told the perpetrator may be absent. Mistaken identification is the leading factor in DNA exonerations.