Interocularly merged face percepts eliminate binocular rivalry.

Faces are important visual objects for humans and other social animals. A complex network of specialized brain areas is involved in the recognition and interpretation of faces. This network needs to strike a balance between being sensitive enough to distinguish between different faces with similar features, and being tolerant of low-level visual changes so that a given face is stably perceived as a particular individual. Such stability may require feedback from higher brain regions down to the level where details are represented. Here, we describe a phenomenon in which interocular competition between face features is stabilized and eliminated when observers attend high-level face characteristics. Two different face images presented to the individual eyes do not cause the perceptual fluctuations that are typically observed in binocular rivalry. Instead, they merge into a stable percept of an intermediate face that combines features from both eyes' images. The stability of the intermediate face percept depends on the observer attending holistic face properties such as identity or gender. It disappears when observers explicitly attend facial features, suggesting a crucial role of top-down stabilizing feedback from high-level areas that represent holistic faces back to lower processing levels where detailed face features compete for conscious representation.