A team of researchers from the Universities of Fribourg and Verona have studied two unconscious – and potentially conflicting – biases in our interpretation of visual information. On the one hand, the tendency to amplify differences between similar images and, on the other hand, the tendency to underestimate them. The first bias leads us to overestimate differences between two similar images, the second to consider them to be more similar than they are. Through experiments with volunteers, the team has worked out that the first bias is associated with the processing of visual signals by the brain, while the second occurs later, at the stage of decision-making.
These two perception biases are known to specialists, but both were thought to be related to the brain's processing of visual data. Through a series of eight experiments with volunteers, Dr David Pascucci of the University of Fribourg and his colleagues have shown that these steps, while both being unconscious, involve different and successive phases of perception. The tendency to amplify differences is indeed associated with visual processing, while minimizing them occurs at the decision-making level, which is closer to consciousness.
We might believe that our eyes deliver a high-fidelity replica of the surrounding world, yet our experience of vision is constantly influenced by what we have seen before. Many optical illusions are based on our tendency to exaggerate visual differences. For example, staring at a red dot on a white background for long enough will produce a blue-green ghost spot on a completely white sheet. This affects not only colors, but also positions and motions. After staring at a waterfall for example, we will have the illusion that the landscape flowing upwards.
But researchers have also found that we tend to minimize visual differences. This explains why we fail to notice the numerous "continuity errors” in movies - vehicles that change position, cups that disappear or change hands.
In practice, the researchers showed volunteers a series of images on screen displaying grooves with different orientations, then asked them to estimate the orientation of the grooves of some of them. They measured how the previous images that they had seen influenced the volunteers' estimates. In other words, how the recent past changed the perception of the present.
"Every perception experience is influenced by these two opposing forces," explains David Pascucci. “We have shown that the two perception biases are associated with different stages in the interpretation of visual information. The tendency to exaggerate differences is close to vision, while the tendency to minimize them is closer to consciousness, at the decision-making level."
A better understanding of these perception biases may have important applications, for instance in medical image analysis or air traffic control, where important decisions must be based on comparing images.