Abstract
When performing face discrimination tasks (i.e. identification, gender discrimination etc.) using multiple saccades, humans use idiosyncratic fixation strategies (Kanan et al. 2015; Mehoudar et al. 2014; Peterson and Eckstein, 2012; Walker-Smith et al. 1977). Here we analyze observers' consistent use of their preferred fixation sequence across trials during face identification and investigate its relationship to perceptual performance. Methods: Fifty five observers completed a 1 of 10 face (15 deg) identification task in white luminance noise. Observers made free eye movements in varying numbers of blocks of 125 trials each (125 - 1500 trials total) with a face viewing time of 1500ms. Each of the 10 faces were divided into 10 areas of interest in order to assign a number (1-10) to each fixation position. We analyzed the similarity of fixation sequences across trials corresponding to specific faces within a subject using only the first 3 fixations from each trial. We used the Needleman-Wunsch algorithm (Scanmatch, Cristino et al. 2010) to obtain a similarity score for fixation sequences between trials. Results: Fixation sequences that included foveating the eyes as well as an area that is centered between the eyes and nose were the most common and represented 12% of all trials across observers. Consistent use of individual-specific fixation areas varied across observers from 8% to 50% of all trials. Critically, we show that observers that consistently use more similar fixation sequences achieve higher performance in the task (r = 0.4; p< .01). Conclusion: Our results extend previous findings on variations across individuals in fixation strategies to faces. The present findings suggest that consistent use of a preferred idiosyncratic fixation sequence also varies greatly across individuals and that observers with greater consistency achieve higher perceptual performance.
Meeting abstract presented at VSS 2016