Abstract
Human observers sometimes falsely perceive patterns in random images as significant. This phenomenon of pareidolia is frequently reported for seeing faces in particular. On the other hand, the visual system of humans is thought to be hardwired and specially tuned to process faces due to the social significance of faces. If the visual system is highly sensitive to differentiate faces and non-faces, falsely perceiving a face in random images should be very rare. Previous neuroimaging studies have shown fovea preference for face representation (Levy et al., 2001; Malach, Levy, and Hasson, 2002). A recent study, using centrally presented noise patterns, further revealed face-related activation in the right fusiform face area when participants perceived face pareidolia versus letter pareidolia (Liu et al., 2014). However, it remains possible that sensitivity to discriminate faces and non-faces is poor in peripheral vision, leading to face pareidolia may appear to occur more frequently than other types of pareidolia. To test this hypothesis, we presented noise patterns randomly in the left or right visual field, and asked participants to report whether they saw a face or a house. In 80% of the trials, random noise patterns were shown. Whereas in only 10% of the trials a degraded face image was blended with a noise pattern and shown to participants; in the rest 10% of the trails a degraded house image was blended with a noise pattern and shown to participants. Surprisingly, for the random noise trials, participants reported seeing significantly more house pareidolia than face pareidolia. Taking response bias into account, participants were more sensitive for detecting faces than houses even in peripheral vision. No significant differences were found between the left and right hemifields. These results question the notion that human observers naturally tended to falsely “see” face patterns in random images.
Meeting abstract presented at VSS 2015