Denise Cullen
How is it possible to think you’re sure about something, only to learn that your memory’s let you down, and you were wrong all along? False memories can be so convincing that we never think to question their veracity. Denise Cullen investigates this odd, and little-understood, phenomenon.
Imagine learning about a famous person’s death, watching footage of the funeral, and listening to the eulogies – then, decades later, finding out that this person had been alive all along.
This was the scenario confronting Fiona Broome in 2009 when she shared her memory online, then subsequently learnt that Nelson Mandela was still alive.
Broome, a paranormal researcher, had a distinct memory of the human rights activist and Nobel Peace Prize winner dying in prison in the 1980s.
“I thought I remembered it clearly, complete with news clips of his funeral, the mourning in [South Africa], some riots in cities and the heartfelt speech by his widow,” she wrote on her website, in a post since removed.
As history records, Mandela died aged 95 – a free man and revered former South African president – in 2013.
“Recall is a more active and effortful process than mere recognition.”
Broome would have been willing to chalk it up to a glitch in her memory. But after discovering that many others shared the same memory, she decided it was instead a glitch in the matrix – a sign consistent with the many-worlds theory of quantum physics that there was a parallel universe in which Mandela had, indeed, died in prison in the 1980s.
Since then, many other examples of what’s become known as the Mandela Effect – or shared false memories – have emerged.
Common examples include that Rich Uncle Pennybags – aka the Monopoly Man – wears a monocle (he doesn’t), that Pokémon character Pikachu has a black-tipped tail (it’s yellow) and that there’s a hyphen in KitKat (there isn’t).
Geographically, some folks swear that there are 51 or 52 states in the United States (there are 50) or that New Zealand is located north-east of Australia (it’s south-east).
Cinematic examples include the Evil Queen in Snow White and the Seven Dwarfs saying “Mirror, mirror on the wall” (it’s actually “Magic mirror on the wall”). And who can forget the chilling moment in The Silence of the Lambs when Hannibal Lecter first meets Agent Starling and says, “Hello Clarice”? Thing is, it never happened.
Misremembering the finer details related to board game mascots, fictional characters or logos might sound inconsequential. Yet the Mandela Effect has spawned a fertile field of psychological research seeking to uncover why people develop false memories – and why, when they do, they are along much the same lines.
Wilma Bainbridge, who works in the Department of Psychology at the University of Chicago, has been interested in the workings of human memory since she and others discovered that people are surprisingly consistent in what they remember, forget and make false memories about.
In 2011, Phillip Isola and some of his colleagues at the Massachusetts Institute of Technology (MIT) identified that memorability was a stable property of an image shared across different viewers.
Presenting at the annual Conference on Computer Vision and Pattern Recognition (CVPR), they built one of the first computer vision systems that sought to predict the memorability of different images.
They also provided some of the first glimmers that low-level visual attributes of an image – such as its bright colours, or distinctive edges – cannot alone account for its memorability. Similarly, aesthetics (visual appeal), interest (how likely people are to be drawn to or interact with an image) or saliency (the area which draws people’s eye focus first) are insufficient to unlock the keys to memorability.
[There is a] tendency for people to consistently misremember characters or logos from popular culture – things that were, in fact, designed to be memorable.
While completing her PhD at MIT, Bainbridge, Isola and MIT colleague Aude Oliva drew on a 10,168-image database of facial photographs to see if the same intrinsic memorability was found in human faces.
Their research, published in the Journal of Experimental Psychology, found that some faces were consistently remembered or forgotten – and that this couldn’t be fully explained by attractiveness or other perceived character traits such as ‘trustworthy’ or ‘boring’.
Bainbridge says it was Isola’s paper in 2011 and hers in 2013 that launched the burgeoning field of memorability. Since then, 845 scientific papers have cited the two papers.
Currently on maternity leave after having twin girls, Bainbridge told me via email that she was originally inspired to probe the visual Mandela Effect because of how pervasive discussions were online about people having the same false memories. But no memory research had then investigated this intriguing phenomenon.
In a recent article in Psychological Science, Bainbridge and her colleague at The University of Chicago, Deepasri Prasad, explored the visual Mandela Effect for the first time.
This is the tendency for people to consistently misremember characters or logos from popular culture – things that were, in fact, designed to be memorable.
Over a series of experiments – using icons such as the Monopoly Man, Pikachu, Curious George, the Volkswagen logo and Waldo from Where’s Waldo – they provided the first experimental confirmation that the visual Mandela Effect exists. (Where’s Waldo? is known as Where’s Wally? in Australia. The discrepancy isn’t an example of the Mandela Effect. It arose because publishers believed ‘Waldo’ would better resonate with North Americans.)
In the first experiment, they presented 100 adults with images of 22 characters, 16 brand logos and two symbols, and made two altered images of each.
“Even though we’ve all lived different lives, there are some pictures that most people remember and some pictures that most people forget,”
For instance, they modified Curious George by adding a thin tail in one image and a bushy tail in the other.
Research participants viewed all three images and had to choose the correct one.
The results indicated that seven out of the 40 images elicited shared – and specific – false memories.
In the second experiment, they used eye-tracking methods to see if there were differences in the way participants looked at the images they correctly identified, versus those they got incorrect.
“We found no attentional or visual differences that drive this phenomenon,” Prasad and Bainbridge wrote.
In the third experiment, the researchers scraped the top 100 Google Image results for each of the seven images to see if previous exposure to non-canonical (incorrect) versions might explain it. But they concluded that there was “no single unifying account for how prior perceptual experiences could cause these visual false memories – which had previously elicited the visual Mandela Effect – to occur”.
The fourth experiment involved having participants draw the images, given that recall is a more active and effortful process than mere recognition.
Some participants viewed the canonical (correct) images prior to being required to reproduce them, while others, who’d flagged that they were already familiar with the images, did not.
One-fifth of all images drawn by the former group, and about half of those drawn by the latter group, showed characteristic Mandela-Effect-type errors. For example, the Monopoly Man frequently appeared with a monocle, while Waldo was often depicted sans cane.
The common production of such errors during both short- and long-term recall suggests there’s something intrinsic to these images that leads to people generating the same sorts of fallacies – but Bainbridge says that researchers are only just beginning to probe what that might be.
Her laboratory is concerned with broader questions about why some images are intrinsically memorable.
“Even though we’ve all lived different lives, there are some pictures that most people remember and some pictures that most people forget,” she explains.
Interestingly, when people view an image, high-level visual and memory areas in their brains show a sensitivity to its memorability – regardless of whether they consciously remember seeing it or not.
Recent work shows that the images people remember or forget can even be used to identify early signs of Alzheimer’s disease.
Several functional magnetic resonance imaging (fMRI) studies, including one conducted by Bainbridge and her colleague Jesse Rissman of the University of California Los Angeles and published in Scientific Reports, have demonstrated distinctive brain activation patterns (neural signatures) when memorable images are viewed.
These processes take place outside conscious awareness, suggesting they occur automatically.
Humans aren’t alone in this, with research led by Nicole Rust at the University of Pennsylvania and published in eLife in 2019 identifying similar patterns in rhesus monkeys who completed visual memory tasks.
In a 2022 paper published in Computational Brain & Behavior, Bainbridge and her then University of Chicago master’s student Coen Needell wrote that they had developed a deep learning neural network that can predict people’s memories.
“We’ve recently developed a web tool called ResMem using deep learning artificial intelligence where you can upload an image and it will tell you the per cent chance someone will remember that image,” Bainbridge says. “Anyone can try it out with their own photos.”
Recent work shows that the images people remember or forget can even be used to identify early signs of Alzheimer’s disease.
Research published by Bainbridge and colleagues in Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring in 2019 found that a small, specific set of images reliably differentiated people with mild cognitive impairment (MCI) or subjective cognitive decline (SCD) from healthy controls.
Using data drawn from the DZNE-Longitudinal Cognitive Impairment and Dementia Study (DELCODE), an observational, longitudinal memory clinic–based study across 10 sites in Germany, Bainbridge and colleagues analysed the memory performance of 394 individuals.
Each participant viewed a randomly selected subset of 88 photographs from a total pool of 835.
The performance of 193 healthy controls was compared to 136 participants with SCD – elderly individuals who self-report a decline in cognitive abilities but don’t yet meet clinical thresholds – and 65 participants with MCI: elderly individuals who show early clinical signs of cognitive decline, but are not yet at the level of Alzheimer’s disease.
(Bainbridge notes that Alzheimer’s disease is more severe than MCI, which is more severe than SCD; however, it is possible to have MCI or SCD and never end up developing Alzheimer’s disease.)
The researchers found that there was a lot of overlap in what the different groups remembered and forgot.
However, there was a small subset of images that were highly memorable to healthy controls, but highly forgettable to those with mild cognitive impairment or subjective cognitive decline.
A subset of as few as 18.3 images could distinguish between the two groups.
In this way, the intrinsic memorability of images might ultimately pave the way towards quicker, easier and more reliable diagnostic tests of precursors to Alzheimer’s disease.
The study of false memories also has weighty implications for criminal defence, given that some people might be wrongfully identified as suspects just because their faces cause false memories more easily.
Though this research is not the focus of Bainbridge’s laboratory, work in this area is continuing, with the promise of some yet-to-be-published data suggesting that these more diagnostic images also better tap into the underlying brain pathology in those with MCI.
“We’re now interested in creating a neural network tool that can predict your chance of making a false memory to an image – and then, theoretically, you could make images that cause lots of false memories,” Bainbridge explains. “These next steps are still in very early stages, though, and sadly, we don’t really have anything yet [on what features may prompt false memories],” she says. One goal of the research is to make the neural network tool available to any scientist who wants to study what makes something cause false memories.
Bainbridge’s research on memorability has potential applications for further research as well as education, which may be enriched, for example, with textbook images or infographics that are more likely to stick in students’ minds. The findings are also likely to enhance clinical practice, given that memory problems are the most common cognitive deficits in dementia.
Bainbridge says those experiencing dementia typically benefit as a result of specially designed environments or tools to aid their memory – for example, memorable cues to help them remember to take essential medication.
The study of false memories also has weighty implications for criminal defence, given that some people might be wrongfully identified as suspects just because their faces cause false memories more easily.
“You’d want to make sure to control for that when choosing a line up,” Bainbridge says.
“It’s pretty amazing to think about how our brains can build up vivid memories of images that don’t really exist and that we’ve never seen before.”