Visual Memory Distortions Paint a Picture of the Past That Never Was 

Consistencies in visual memory • Mental caricaturing • Recalling what should have been there

Rich Uncle Pennybags winking through his monocle forever sticks in the minds of Monopoly players. Er, strike that—the board game’s iconic mascot (also known as Monopoly Man) has never worn a monocle. So why did 50% of participants in one study draw him wearing one?

People consistently make the same false-memory error with Monopoly Man, despite the majority of visual experience being the canonical image, according to Deepasri Prasad and Wilma A. Bainbridge, who study vision and memory.  

A false memory of a famous image is popularly known as the visual Mandela effect (VME). It’s a subtype of the “Mandela effect,” a controversial term coined by researcher Fiona Broome after she found that many people incorrectly remembered Nelson Mandela’s death in the 1980s. (The South African anti-apartheid activist and president actually died in 2013.) 

The visual Mandela effect can occur for several reasons. Some images may invoke schema-based memory distortions. People may recall Monopoly Man wearing a monocle because they associate the accessory with wealthy 19th century capitalists, Prasad and Bainbridge suggest. People may also view noncanonical versions of these images online, causing them to confuse inaccurate fan art with the original. 

The VME cannot be universally explained by a single account, the researchers wrote in an article for Psychological Science. “Instead, perhaps different images cause a VME for different reasons—some related to schema, some related to visual experience, and some related to something entirely different about the images themselves.”  

Prasad and Bainbridge put each of these explanations to the test through a series of four studies on how people form false memories of pop-culture images and logos. They conducted their research when Prasad, a doctoral student at Dartmouth College, was a research assistant in Bainbridge’s lab at the University of Chicago.  

In their first study, they showed 100 online participants both correct (canonical) and manipulated versions of popular images—including Monopoly Man, Curious George, the Pokémon character Pikachu, the Volkswagen and Fruit of the Loom logos, Where’s Waldo, and C-3PO from Star Wars—and asked them to identify the correct version. Participants were significantly more likely to select the altered versions of the images than the canonical versions.   

The researchers repeated this experiment with a group of 60 adults using computer-based mouse tracking, which requires participants to move their mouse over occluded images to uncover a small section of the image. In each trial, participants were shown a blurred version of the canonical image, then instructed to inspect the image using their mouse before indicating which of two images they had seen, the canonical image or a related VME image. This revealed that participants looked at images the same way regardless of whether they chose the correct image during the recall phase. This suggests that the visual Mandela effect is not the result of attentional or perceptional differences in how people process these images, Prasad and Bainbridge wrote. 

A group of 50 participants also produced similar errors when asked to draw these icons from memory, regardless of whether they reported seeing the VME images before. 

“The fact that VME errors can occur during short-term recall, despite limited familiarity with the image, could suggest that there is something intrinsic to these stimuli that encourages these errors,” Prasad and Bainbridge wrote. 

Furthermore, when the researchers scraped Google for 100 real-world images of the seven icons for which a VME effect was found in their previous studies, most of the results depicted the canonical image without VME features. This wasn’t the case for all the images, however. C-3PO, for example, was shown to have two golden legs instead of his canonical silver one in 24% of scraped images, and 51% of images didn’t show his legs at all. VME features were also present in 28% of Where’s Waldo images and 18% of Volkswagen images, but in less than 5% of Pikachu images and in none of the images of Curious George, Monopoly Man, or the Fruit of the Loom logo. 

This suggests that the visual Mandela effect may arise in part due to previous exposure to inaccurate images but can also cause people to misremember images in ways that they are unlikely to have encountered in the real world, Prasad and Bainbridge wrote (Prasad & Bainbridge, 2022). 

Consistencies in visual memory 

Though the gaps in humans’ visual memory can lead to perplexing results, they also tend to be consistent. Images of people tend to be more memorable than landscape scenes, for example, while novel objects stand out more in memory than familiar ones, University of Pennsylvania researchers Nicole C. Rust and Barnes G. L. Jannuzi wrote in an article published in Current Directions in Psychological Science. 

“What’s more surprising is that across differences in culture and individual experience, a strong tendency exists for us all to find the same pictures memorable and forgettable,” Rust said in an interview. “This makes sense when you realize that memorability is linked to how our visual systems operate.”  

Research with rhesus macaque monkeys suggests that image memorability may be influenced by activity in the inferotemporal cortex (IT), an area of the brain associated with object recognition. Similar objects activate shared clusters of neurons in the IT, and image memorability is reflected by the strength of the IT response, Rust and Jannuzi explained. 

Researchers are also using deep neural networks (DNNs) to further explore findings generated by experimental research with humans and other animals, they added.  

“The development of these types of models holds tremendous promise for understanding the specific processing that happens in each visual brain area as well as how that processing contributes to visual behavior,” they wrote. 

DNNs are referred to as “deep” because of the numerous layers of networks within these models that allow them to sequentially process information in a way that is intended to mimic the human brain. Each DNN has its own unique “architecture,” a term for the order in which these layers are connected, and the layers may be optimized to function differently via various training methods and data sets. In the case of DNNs designed to reflect visual memory, the DNN’s architecture may consist of convolutional layers that process each part of an image separately and fully connected layers that process the entire image. Through supervised training, researchers can then teach DNNs to associate different images with particular labels, allowing them to “recognize” those images. 

Although researchers are still exploring how these complex models work, DNN lower-level convolutional layers have been found to respond to images in a way that is similar to early visual processing in the V1 and V4 regions of human visual cortex. Likewise, higher-level fully connected layers produce activity similar to later image processing in other parts of the human visual cortex, as well as monkey IT, Rust and Jannuzi wrote. 

These findings suggest that visual DNNs not only reach the same conclusions as humans and other animals, but that they likely recognize images in a similar way, the researchers concluded (Rust & Jannuzi, 2022). 

Mental caricaturing 

A potential explanation for common visual memory distortions is that these cognitive biases may make images more memorable by amplifying their distinct features, according to research reported recently in Psychological Science. Johns Hopkins University scientists Zekun Sun, Subin Han, and Chaz Firestone found this mental caricaturing appears to occur with everything from simple geometric shapes to faces and other objects.

“Though often used intentionally for political or comedic effect, human cognition sometimes engages in a caricaturing process of its own, encoding and even misremembering stimuli in exaggerated form,” Sun, Han, and Firestone wrote. “Our minds encode the world around us in ways that emphasize what is distinctive, even when the relevant stimuli have no particular significance.”

The researchers explored this effect through a series of six experiments in which 700 participants were tasked with reproducing novel shapes. In their first pair of experiments, Sun, Han, and Firestone found that when participants were asked to recall shapes by adjusting a slider to alter an existing shape, they consistently produced shapes that were more exaggerated and “spikier” than the original. Subsequent experiments amplified the effects, even without explicit demand for participants to remember or distinguish multiple objects at the same time, preexisting schematic associations or knowledge, or other contextual factors that might tend to encourage cognitive biases.

“These results thus provide initial evidence for the hypothesis that the mind encodes and stores even novel contextless shapes as more informationally dense than they really are, caricaturing even some of the most basic stimuli we encounter,” Sun, Han, and Firestone wrote. 

Though a single instance of mental caricaturing may have a relatively subtle influence on memory distortion, the effect can become magnified as we repeatedly retrieve these images from memory or share these images with other people, the researchers added. For example, when participants were tasked with reproducing shapes on the basis of previous participants’ reproductions, the researchers found that shapes quickly became more jagged and even sprouted new appendages. 

“Considering that daily life often involves repeatedly remembering and recalling various objects, the biases we observed here may well accumulate in real-world contexts as well,” the researchers concluded (Sun et al., 2024). 

Recalling what should have been there 

Perceived distance is another property that can cause us to misremember a scene. When we perceive scenes as closer to us, we are more likely to engage in boundary extension, leading us to create visual memories that extend beyond our actual field of vision, cognitive psychologist Alon Hafri and his colleagues at Johns Hopkins reported in Psychological Science. 

“How you remember a visual scene depends on its spatial scale—whether you see it as close up or farther away,” Hafri said in an interview. “For instance, if you see a model train up close, your mind ‘fills in’ more of the surrounding scene than it would for a similar scene viewed from a distance, like a real train on a distant track.”  

In a series of experiments on the subject, Hafri and his team tested participants’ memories of outdoor scenes by showing them realistic images of faraway scenes as well as altered versions of these images that had been made to look like artificial miniatures. For the altered versions, the researchers used blurring effects that simulate the shallow depth of field that arises from viewing scenes at close range, Hafri explained.

Participants were shown two identical images of an outdoor scene. The participants were more likely to report the second image as a closer view than the first image when they viewed a “miniaturized” scene than an unedited distant scene, even though the second image was always exactly the same as the first. This suggests that when images appear closer to the viewer, they are more likely to engage in boundary extension, causing them to recall more of the scene than was actually visible to them, the researchers explained. 

This effect did not persist in control experiments in which the researchers rotated or obscured the blurring effect so that it did not alter the perceived distance of the image, Hafri said. It was found to generalize, however, when the researchers “sphere-ized” an image by editing it so that it appeared to be wrapped around a sphere, causing the center of the scene to appear closer to the viewer. 

Taken together, these findings suggest that boundary-extending memory distortions are a response to the viewer’s perceived distance from a scene (Hafri et al., 2022). 

“The current results broaden the view of what kinds of contents the mind may add to visual memories and under what conditions it may do so,” Hafri and his colleagues wrote. “In our case, we have found that beyond scene content itself, visual memories are biased by the spatial contexts in which they are formed.” 

These and other findings suggest that while visual memory distortions may be common, at least our memories are reliably unreliable. 

“Memory doesn’t function like a camera, capturing the world exactly as it is,” Hafri said. “However, it’s not simply a less detailed or ‘noisy’ version of reality either. Our memories are distorted and biased in specific ways.”  

Back to top

Feedback on this article? Email [email protected] or login to comment.

References 

Hafri, A., Wadhwa, S., & Bonner, M. F. (2022). Perceived distance alters memory for scene boundaries. Psychological Science, 33(12), 2040–2058. https://doi.org/10.1177/09567976221093575  

Prasad, D., & Bainbridge, W. A. (2022). The visual Mandela effect as evidence for shared and specific false memories across people. Psychological Science, 33(12), 1971–1988. https://doi.org/10.1177/09567976221108944  

Rust, N. C., & Jannuzi, B. G. L. (2022). Identifying objects and remembering images: Insights from deep neural networks. Current Directions in Psychological Science, 31(4), 316–323. https://doi.org/10.1177/09637214221083663  

Sun, Z., Han, S., & Firestone, C. (2024). Caricaturing shapes in visual memory. Psychological Science. https://doi.org/10.1177/09567976231225091