Being fooled (or not) by Hyperrealistic masks: When “Five blimps” can reveal the Dark Knight

In October 2010, an elderly white man boarded an Air Canada flight bound from Hong Kong to Vancouver. During the flight, this passenger visited the bathroom and emerged an Asian man in his early 20s. No, this wasn’t an episode of Scooby Doo, this was an actual case where a hyperrealistic mask was used to assume a new identity. The deception was so successful that the mask went unnoticed even through multiple identity checkpoints in the airport on the way to board the flight.

While a typical mask simply hides the wearer’s identity, a hyperrealistic mask allows the wearer to assume a new identity and, if used to commit a crime, completely throws off the search for a suspect. In one case, a white man committed multiple robberies wearing a black male mask. Unfortunately, an innocent man who looked like the mask was arrested for the crimes instead. The similarity was so striking that the innocent suspect’s own mother recognized her son’s face from surveillance photos from the robberies.

In another case, three black men disguised as white men successfully stole over $200,000 from a check-cashing store. The only reason they got caught was because one of the thieves decided to write an unexpected endorsement to the mask makers, a special effects company called CFX, after the robbery. “I’m sending you this message to say I’m extremely pleased by CFX work on the mask. The realism of the mask is unbelievable,” he wrote.

While anecdotal evidence like this suggests that the realism of masks can indeed be unbelievable, to date no research has examined people’s perception of hyper-realistic masks and the factors that affect our ability to detect them. A recent article by Jet Sanders and colleagues in the Psychonomic Society’s new journal CRPI (Cognitive Research: Principles and Implications) fills this gap.

In one experiment, the researchers showed participants 20 photographs of faces, one of which was a face wearing a hyper-realistic mask. To encourage deep processing of the faces, participants were asked to rate how trustworthy, dominant, and attractive each face was.

Participants were then asked a series of increasingly-probing questions to see whether they had noticed the mask. First, they were asked an open-ended question – “What did you think of the faces you saw?” – to capture any spontaneous identifications of the mask. Next, they were prompted “Did you notice anything unusual about any of the faces?” to capture vaguer suspicions.

Not one of the 60 participants reported detecting a mask.

Participants were then explicitly asked whether they thought any of the faces were masks. Participants were told that some of the participants were in a Mask group (where at least one of the photos had a mask), and some were in the No Mask group (where none of the photos had a mask), and asked which group they thought they were in.

Participants largely didn’t think they had seen a mask – only one-fifth (21.7%) thought they were in the Mask group.

Everyone was then told they were in the Mask group, presented again with all the faces they had rated, and were asked to identify which of the faces were masks from the array.

You can try this out yourself on the image below – can you spot the mask?

Answer: It’s face number 9.

Participants were pretty good at spotting the mask when prompted – 70% were able to identify it. However, they thought some of the real faces were masks, too. In fact, most people (78%) identified at least one real face as a mask!

Perhaps people’s inability to spontaneously detect masks in this experiment was due to seeing the faces in static pictures and not in the flesh. Would mask detection improve in real life?

To address this question, the researchers recruited passers-by on university campuses to evaluate faces, including ones wearing masks, in real life. This experiment also varied the distance at which faces were seen and the race of the faces.

A research confederate wore either a high-realism mask, a low-realism mask, or no mask, and appeared near the participant (distance of 5 m) or far from the participant (distance of 20 m). To compare people’s ability to identify own-race and other-race masks, people in the U.K. and Japan were tested using high-realism masks that were either Asian or white.

The figure below shows a low-realism mask (left), a high-realism mask (middle), and a real face (right).

Even when tested in real life, rates of detecting high-realism masks were still very low. Only two of 160 participants identified seeing the mask in response to the open-ended question and five in response to the prompted suspicion question. Notably, all of these participants saw faces at a near distance. No-one spontaneously identified the mask at a far distance.

When explicitly asked whether they had seen a mask or a real face, less than half of participants who saw the high-realism mask (42.5%) thought they had seen a mask.

It wasn’t that people were blind to masks. In the low-realism mask condition, around half of the participants identified seeing a mask in response to the open-ended question and most did so in response to the prompted suspicion question. When explicitly asked, nearly everyone reported having seen a mask.

The researchers also found an effect of face race, but it was not specific to masks. People were more likely to think they had seen a mask in the other-race condition, both for the high-realism masks and the real faces.

These findings bolster anecdotal evidence that hyper-realistic masks can easily fool viewers. Spontaneous mask identification in a real-life context was rare, and there was no evidence that people could spontaneously detect masks in still images, like those that might be used to identify criminals from security camera footage.

What strategies might help people wearing masks to avoid detection? One easy strategy suggested by the current findings is to keep a distance – spontaneous mask detection was only successful at a close distance. Another strategy could be to wear masks that do not draw onlookers’ attention. There’s evidence that people look longer at attractive than unattractive faces and can’t help but pay attention to attractive faces. Perhaps wearing an unattractive mask could minimize attention drawn to the wearer and lower the likelihood of getting caught.

On the flip side, what strategies might improve onlookers’ mask detection? Movement and speech might help. Hyperrealistic masks look good at rest, but they don’t fully transmit subtle movements of the face, lessening their realism. This constraint is also important for speech production. The masks fully cover wearers’ lips, and make it difficult to produce sounds that require lips to touch each other (like “b”, “p”, or “m”) or teeth to touch the lower lip (like “f” or “v”).

Perhaps the best clue that someone is wearing a mask has nothing to do with appearance after all. Asking the person to pronounce a simple phrase like “five blimps” might do more to reveal their true face.

Psychonomics article highlighted in this post:

Sanders, J. G., Ueda, Y., Minemoto, K., Noyes, E., Yoshikawa, S., & Jenkins, R. (2017). Hyper-realistic face masks: a new challenge in person identification. Cognitive Research: Principles and Implications, 2, 43. DOI: 10.1186/s41235-017-0079-y.

 

The Psychonomic Society (Society) is providing information in the Featured Content section of its website as a benefit and service in furtherance of the Society’s nonprofit and tax-exempt status. The Society does not exert editorial control over such materials, and any opinions expressed in the Featured Content articles are solely those of the individual authors and do not necessarily reflect the opinions or policies of the Society. The Society does not guarantee the accuracy of the content contained in the Featured Content portion of the website and specifically disclaims any and all liability for any claims or damages that result from reliance on such content by third parties.

You may also like