Abstract
Interpretation of eye-tracking data rests on the assumption that observed fixation patterns are mainly stimulus and task dependent. Given this assumption, one can draw conclusions regarding where the most diagnostic information is for a given perceptual task (e.g. face recognition, scene identification). If the assumption is true, then the fixation location at stimulus onset should not largely influence subsequent fixation patterns. However, we demonstrate that start location very strongly affects subsequent fixation patterns. Participants viewed upright and inverted faces and were told that they would be required to recognize the faces later in the experiment. We imposed five different start locations relative to the faces: above, below, right, left, and center (tip of nose). We found a distinct pattern of fixations for each start location that extended through at least the first five fixations. In particular, there was a strong fixation bias towards the side of the face opposite the start location. For the center start location, we found the classic result of more fixations to the eyes, particularly the right. The initial saccade from the center start location was delayed relative to the other start locations, suggesting that participants were already sampling information from the face. These general fixation patterns held regardless of face orientation. However, the difference in fixation patterns between upright and inverted faces was dependent on start location. While the central start location produced the classic result (more fixations to eyes for upright and to mouth and nose for inverted), the relative preference for eyes in upright over inverted was dependent on the start location. We conclude general biases in saccadic programming as well as stimulus information influence eye-movements during face perception. Eye-tracking allow us to tease these influences apart only if both factors are carefully controlled and analyzed.
NIMH Intramural program, United States - Israel Binational Science Foundation.