A New False Memory Study Suggests People Can’t Tell What’s Real

A New False Memory Study Suggests People Can’t Tell What’s Real
Photo: Marcelo Hernandez, Getty Images

A new study is the latest to show how indistinguishable false memories are from real ones. People who watched someone else recount a supposed recollection were unable to tell whether the memory was false with any better accuracy than chance, the study found. Moreover, in many cases the people describing their own memories had been fooled into “remembering” something that never happened.

In 2015, memory researcher and psychologist Julia Shaw and her co-author published a study on false memories. Over the course of several interviews, they claimed to have convinced a substantial percentage of volunteers into recalling false childhood memories. In one especially frightening experiment, they reported that around 70 per cent of volunteers could be made to recall false memories of having committed a crime as preteens.

That memory trickery was done with the aid of the volunteers’ parents, who told the researchers about a true and highly emotional childhood experience that the volunteers had. The volunteers would later recall that memory for the researchers, who used the true memory to build a false recollection from scratch. They told volunteers that their parents had told them about the faux crime or another highly emotional but made-up incident and asked them to visualise that event as if it had happened. The volunteers did not know this “memory” had been invented. The made-up crimes ran from theft to assault to assault with a weapon.

Shaw’s latest study, published this week in Frontiers in Psychology, took some of the recorded accounts from the original experiment and showed them to groups of volunteers in two new experiments, totaling over 200 people. Shaw and her research assistants asked these volunteers to guess whether someone’s recounted memory was true or not. Eight sets of memories, a true and false one, were used, and the false memories involved either a crime or another highly emotional event. One of the true accounts told by someone in the original study, for instance, was about losing their family dog, while their false account involved being attacked by a vicious animal.

Across both experiments, volunteers appeared to do only slightly better than chance in guessing correctly that a memory was true, with roughly a 60 per cent accuracy rate. And they weren’t statistically better than chance when it came to sniffing out a false memory. In the first group, 57 per cent of volunteers guessed correctly for either kind of false memory, with a 55 per cent accuracy for recognising false memories of a crime.

“The main thing this research tells us is that false memories of highly emotional and criminal events look really convincing to other people,” Shaw, now an honorary research associate at University College London, told Gizmodo via email. This is significant because it could have serious implications for police, juries, and judges who rely on eyewitness testimony.

The second experiment added a twist to the design. Volunteers were divided into three groups. They were all asked to watch or listen to people describing a memory, but only one group had access to both video and audio. The other two saw either a video-only account, where the only thing they could use to identify a recollection as false or not was body language, or an audio-only account. As before, the video and audio group still performed no better than chance at detecting a false memory (53 per cent), as did the video-only group (45 per cent). But those listening only to audio recordings now did worse than chance, having a 32 per cent hit rate for false memories.

“What’s shocking is that people were better at classifying silent videos than when they had the audio account. In other words, they were better when they literally didn’t know what the memory being told was about at all than when they could hear it being told,” Shaw said. “I honestly don’t know why though—the data didn’t give me an answer for this. It seems that somehow in an audio-only context we actively rely on the wrong cues… but it’s unclear what those cues are.”

Shaw’s original findings in 2015, while considered the first experimental evidence that richly detailed false memories can be created with nothing more than suggestion, came under some scrutiny in 2018. Other researchers argued the criteria used for determining a false memory by Shaw was too broad. Their analysis concluded that half of the volunteers said to have a false memory may have instead developed a false belief that the event had happened, rather than a genuinely embedded memory of the event. Using their criteria, roughly a third or fewer of Shaw’s volunteers were considered to have a “true” false memory—a rate more in line with other studies of memory implantation.

Shaw rebutted the criticisms in a subsequent paper. She argued that it’s practically impossible to tell whether someone’s false recollection is earnestly remembered or merely believed, particularly in a real-world context where innocent people have been forced into false confessions of a crime, and that her criteria was intentionally different than others used in the past for that reason. The study’s findings have not been retracted, though an unrelated mathematical error was brought to her attention that same year and corrected (the error didn’t change any of the main conclusions). Still, for the new study, Shaw only used memories that were classified as false by her own and her critics’ criteria.

That research dispute might further highlight how difficult it is to know the truth behind any retold event, even among scientists who study memory. And Shaw’s new findings are especially worrying in a judicial context, given that juries sometimes only have audio evidence to rely on. Police and judges shouldn’t assume that they easily know how to tell apart a real memory from a fake one.

“We need to make sure that we use evidence-based tactics in memory interviews so we don’t contaminate memories or create fake ones… If the process through which a memory was recalled was leading, suggestive, or shows signs of coercion, it’s going to make for less reliable evidence,” Shaw said. “The only way to tell a true memory from a false one is to look at how and when the memory was first remembered—and even this can’t tell you for sure whether something is true, false, or a lie. But it’s a much better approach than just using your gut feeling to decide.”