In 2011, researchers Jason Tangen, Sean Murphy, and Matthew Thompson at the University of Queensland discovered a striking visual illusion while preparing a set of face images for a study. As they were going quickly through the faces to check their spatial alignment, they started noticing that the faces appeared highly distorted, almost cartoonish. They then realized that these distortions were most pronounced when the faces were flashed about 4-5 times per second in peripheral vision.
The effect, which they termed the Flashed Face Distortion Effect (or FFDE), won second place at the 2012 Best Illusion of the Year Contest and was published as a research article in Perception that year. Below is a video example of the illusion:
To get the full effect, keep your eyes fixated on the central cross while the faces flash in your peripheral vision. If you pause the video at any point, you can confirm that the faces are not actually distorted at all.
In the 15 years since its publication, many researchers have attempted to explain its mechanisms, and in particular, the degree to which this illusion depends on high-level processes. For example, researchers have shown that upside-down faces (which are known to be processed less holistically than upright faces) show only result in a modest reduction of the effect, suggesting that the FFDE doesn’t necessarily require face-specific expertise. Similarly, another study found that the effect is not modulated by the race of the face or the observer, suggesting that the FFDE does not depend on having full expertise of the faces one is observing.
The Importance of Spatial Location
A new study published this month in Perception examined whether a low-level aspect of this illusion can explain its power. Specifically, Yong Hoon Chung and co-authors at Dartmouth University asked whether changing the spatial location of the flashed faces halfway through a presentation would alter the strength of the illusion. They reasoned that if the effect depends on low-level effects based on which part of the retina is being stimulated, then the strength of the FFDE should decrease whenever there is a change in the location of the faces.
Across three experiments, the researchers manipulated the position of faces across the visual field and the method by which participants reported the illusion’s strength. Across all experiments, the researchers found that perceived face distortions were substantially diminished the moment the stream of faces switched from one side of the screen to the other, or from the upper half of the screen to the lower half (or vice versa). Moreover, this decrease was observed regardless of whether participants responded continuously with a joystick (reporting the distortion magnitude throughout each experimental trial) or with a single button press at the end of each trial.
The findings by Chung and colleagues confirm that the FFD is highly sensitive to the location of faces in the visual field, suggesting a large portion of this effect can be attributed to low-level retinal processes. We know, for example, that perceptual phenomena such as color afterimages (e.g., seeing red following prolonged exposure to green) and motion aftereffects (e.g., seeing upward motion following prolonged exposure to downward motion) are also location-dependent, manifesting much more strongly when the test stimulus matches the location of the adapting stimulus. Now the FFDE can be added to the class of illusions that depend critically on retinal location.
Although there is still much to understand about the mechanisms of the FFDE, this new research seems to have identified one of the main factors.