Why eyes seem to follow viewers in some paintings

You’ve seen it in horror movies, or even in real-life at the local museum: a painting in which the eyes of the person portrayed seem to follow you around the room, no matter where you go.

People have described the effect as creepy or eerie, and some have thought it supernatural. But now researchers have demonstrated the very natural cause for this visual effect.

All it takes for the effect to work is to have the person in the painting, or photograph, look straight ahead, said James Todd, co-author of the study and a professor of psychology at Ohio State University. Our visual perception takes care of the rest.

“The core idea is simple: no matter what angle you look at a painting from, the painting itself doesn’t change. You’re looking at a flat surface. The pattern of light and dark remains the same,” Todd said.

“We found that our visual perception of a picture also remains largely unchanged as we look at it from different vantage points. If a person in a painting is looking straight out, it will always appear that way, regardless of the angle at which it is viewed.”

The study was conducted at the University of Utrecht in The Netherlands in collaboration with Jan Koenderink, Andrea van Doorn, and Astrid Kappers. Their results were published in a recent issue of the journal Perception.

While scientists have considered the reasons behind this visual effect for more than a century, advances in the field of perception now allow for better ways to study why it occurs, Todd said.

“Researchers have developed powerful techniques that allow us to measure the perception of complex shapes in a very precise way,” he said.

In this study, the authors viewed on a computer screen a picture of a medical mannequin human torso in a richly sculpted gilded frame, which appeared to be hanging on a brick wall. The wall and frame were shown in color, the torso in neutral gray.

In order to answer their questions, the researchers needed to determine how the apparent 3D structure of the object depicted in the picture was influenced by changes in the viewing direction. They were particularly interested whether points that appeared to be closest or farthest in depth relative to other neighboring points would remain the same when the picture was observed at different viewing angles. They also wanted to determine how the relative magnitude of the perceived depth in different regions of the picture would be affected when viewed at different angles.

In order to address these issues, the researchers did two types of tasks. In one, they moved a dot around on the computer screen to show which points on the torso appeared to be nearest and which appeared to be the furthest away. The researchers did this hundreds of times to find near points and far points on various parts of the torso.

In a second task, they used a gauge figure (a circle with a needle sticking out) that had to be placed on the torso so it looked to be flat against the surface (The needle had to appear like it was perpendicular to the surface of the torso). This allowed the researchers to determine how viewers perceived the 3D shape of the depicted object.

The researchers repeated this process for six different conditions, including sessions in which they looked straight at the monitor, and others in which they looked at it from an angle. Each researcher repeated these tasks three times for each of the six experimental conditions.

“These experiments took hours,” Todd said. “We made judgments at numerous probe points on each image, so that when all of the different conditions were completed we ended up making thousands of settings over the course of the experiment.

“From all that data we were able to mathematically construct a surface that is most consistent with the overall pattern of judgments in each condition.”

However, the different viewing conditions didn’t yield many different results.

“It turned out that that changes in viewing direction had remarkably little effect on the observers’ perceptions,” Todd said.

The only difference they found is that, when viewed from an angle, the torso looked “squashed” – in other words, it looked thinner to viewers. But the far points and near points, and the overall relief of the depicted object, remained proportionally the same.

The key is that the near points and far points of the picture remained the same no matter the angle the picture was viewed from, Todd said.

“When observing real surfaces in the natural environment the visual information that specifies near and far points varies when we change viewing direction,” he said.

Todd said people may be surprised by this phenomenon because of the unique perceptual aspects of viewing a picture. We perceive the object depicted in a painting as a surface in 3-dimensional space, but we also perceive that the painting itself is a 2-dimensional surface that is hanging on the wall.

“When we look at a picture, you have these two perceptions simultaneously, but it is difficult to make sense of that conceptually. That’s why this issue has fascinated people for hundreds of years.”

In fact, many researchers have continued to follow the theories of La Gournerie, a French researcher who proposed a mathematical analysis in 1859 of why eyes in a painting seem to follow viewers.

“One of the contributions of our study is that we showed that while La Gournerie had the basic idea right, his mathematical description was wrong,” Todd said. “We were able to use new methodologies to give a more correct mathematical analysis of what is going on.”