Tuning of the Visual Perception of Meaning to Natural Image Statistics in Childhood

North Dakota State University Fargo, Fargo ND

Investigators

Abstract

Seeing and recognizing faces is an important part of social interaction and most people are excellent at quickly finding faces in the world around them. In fact, the tools that the visual system uses to detect faces also lead human observers to sometimes see faces in patterns and objects that are not faces at all. Human observers may see a face in the clouds, or in a rock formation, for example, or the windows and doors of a building may look like a smiling face to a human observer. Accidental face sightings like this are an important clue about what the visual system needs to see in a pattern to decide that a picture is face-like enough to count as a face. Seeing these kinds of faces also depends on the ability to combine shapes and textures from different parts of a picture to make a larger whole. This project investigates how both of these things change during childhood, using the detection of accidental (or “pareidolic”) faces in noisy texture images to see how children’s tendency to see faces in random patterns changes as they get older. The results advance understanding of how children’s increasing experience with their visual world affects their visual sense of what makes something look like a face. In many different ways, vision is tuned to the patterns of color and light that human observers most often in the natural world. Generally, the visual system is most sensitive to patterns that are like the ones seen in real scenes and less effective when presented with patterns that differ from this. This sensitivity to the lawful structure of the natural world has not typically been investigated in terms of how and why human observers see patterns in noise. The tendency to see faces in randomness is a particularly useful way to determine how children’s emerging understanding of natural image statistics between the ages of 5-12 years old supports the detection of complex patterns like faces and objects in clutter. To test the hypothesis that face pareidolia (detecting faces in non-face images) becomes tuned to the statistics of the visual world in childhood, the researchers ask children to find faces in noise that has been manipulated to have different statistical properties. Some of these closely reflect the structure of the real world, while others differ substantially. The results of this work establish important and novel connections between extended development of the visual system in concordance with natural image structure and the manner in which that development supports complex inferences about faces and objects in cluttered scenes. This project is jointly funded by Developmental Science, the Established Program to Stimulate Competitive Research (EPSCoR), Perception, Action and Cognition (SBE/BCS), and the Science of Broadening Participation (SBE). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.