Can thousands of eyewitnesses get it wrong?

Yes, and a television experiment showed it. The psychologist Bob Buckhout, in a study titled "Nearly 2,000 Witnesses Can Be Wrong," had a TV network show 2,145 at-home viewers of a news broadcast a 13-second clip of a mock robbery: a man in a hat knocked a woman over and grabbed her purse, his face visible for only about 3.5 seconds. Viewers then saw a lineup of six numbered suspects and could call in to name the culprit. The real one was suspect number 2, yet he got about as many votes as two innocent suspects, numbers 1 and 5, and the single largest group, about a quarter of callers, said the culprit wasn't in the lineup at all. Even police precincts that called in named the wrong man.

Eyewitness testimony is the account a bystander or victim gives in court describing what they observed during an incident, and psychologists have been probing how reliable that account really is since the early 20th century. One early pioneer, Hugo Münsterberg, wrote a controversial 1908 book demonstrating the fallibility of eyewitness accounts, though it drew fierce criticism from legal circles at the time. His ideas did win over the public, and decades later DNA testing backed his warning: studies found that many DNA-based exonerations involved eyewitness evidence.

Memory distorts for specific, identifiable reasons. When a weapon is present during a crime, victims tend to focus on the weapon rather than the perpetrator's features, a pattern Elizabeth Loftus tested by showing people slides of someone either handing a cashier a check or pointing a gun at them; the person was correctly identified slightly less often in the gun condition. Even the wording of a question can plant a false detail: people asked how fast two cars were going when they "smashed" into each other estimated higher speeds than people asked about the same cars that merely "hit" each other.