Abstract
Introduction: Many studies have shown a normal human preference for fixating the eyes during a variety of face-related tasks (Barton et al., 2006). Previously, we showed using ideal observer analysis that the eye region also contains an impressively large and dense concentration of identification-relevant visual information (Peterson et al., 2007). Furthermore, human face processing preferentially uses the eye region information when making an identification (Peterson et al., 2008). There remains the question of how optimal these eye movement strategies are given the face's distribution of information and the eccentricity-dependent processing properties of the retina and visual cortex. Here, we present evidence for observer-specific face identification eye movement strategies driven by individual differences in the fixation location that leads to maximal recognition performance. Methods: 10 cropped grayscale male Caucasian faces were embedded in white Gaussian noise. In each of 3 sections conducted in an eye-tracker the observer identified a randomly selected noisy face. In the first two sections, observers began each trial fixating around the edge of the screen. The stimulus was shown for either 1500ms (Section 1) or 500ms (Section 2), during which eye movements were allowed. Section 3 required the observer to maintain fixation on one of four randomly sampled locations along the midline of the face (between eyes, nose tip, center of mouth, chin tip) during a 500ms display. Results: In Sections 1 and 2, fixation patterns were highly observer-specific, with average landing points ranging from the eyes to just below the nose tip. Results from Section 3 showed that observer performance as a function of maintained fixation was also variable. Regression analysis reveals a very strong correlation between individuals‘ maximally performing fixation point and their preferred landing point. Conclusion: Human eye movement strategy for quick face identification beneficially incorporates a representation of the observer's fixation-dependent task ability.