We allowed the ideal observer only two choices (left or right) instead of the three choices (direct, left, or right) provided to human observers. This allowed us to specify the correct response for each trial, which in turn allowed us to vary the visibility of the stimulus according to the ideal observer’s responses. The availability of low-level visual information will constrain human observers’ sensitivity to the direction of gaze, whether two or three response alternatives are allowed. Hence, the current analysis allows inferences about the role of low-level visual information in the effects of facial expression on judgments of eye contact.