Unconscious processing of real-life scenes revealed by eye movement dynamics

When viewing scenes, the eyes are consistently attracted to meaningful and visually salient areas. Does this occur also when scenes are rendered invisible? This is unknown, as free viewing of invisible scenes has never been documented. In two preregistered experiments, fifty-two participants freely viewed scenes presented consciously or unconsciously using Continuous Flash Suppression. Despite being unaware of the scenes, participants gazed more, but not longer, on objects, and were attracted to both semantically and visually salient areas. Comparison of gaze patterns to a Convolutional Neural Network (CNN) mimicking the human visual ventral pathway revealed that unconscious gaze patterns correlated only with layers corresponding with early visual processes. Our results provide first evidence for meaningful gazing on invisible scenes, while highlighting the limitations of such processing.