Doris I. Braun¹ (), Karl R. Gegenfurtner¹; ¹Giessen University Dept. of Psychology

How important are object colors and contours for the guidance of eye movements? We selected high-quality digital reproductions of 20 still-life paintings of fruit and flower arrangements from the 17th-18th century from a recent database (Van Zuijlen et al., 2021). The “eidolon factory” (Koenderink et al., 2017) was used to disrupt object recognition while maintaining local image structure. To investigate the influence of color and contour information on visual exploration we presented each still life image on a computer monitor in four versions: as original in color and grayscale and as Eidolon in color and grayscale. 26 young participants first had to indicate the most interesting part of each image after it was briefly flashed for 120 ms. In a second experiment, they were asked to view each image for 5 s and then to rate how much they liked it on a 7-point scale. During the image exploration eye movements were recorded with an EyeLink 1000. For each participant and image we analyzed the number and amplitude of saccades, the position and duration of fixations and the extent of image exploration (spread). Eye movements reflected content and structure of each image. When objects were defined by contours as in both versions of the originals, significantly more saccades were made, saccade amplitudes and fixation durations were significantly shorter, and spreads were larger. Also inter-subject agreement was significantly higher when contour information was present. This was even further improved by color. For both Eidolon versions, impaired object recognition prolonged fixation durations and increased saccade amplitudes. For eidolon images, color information had a major effect on which image regions were preferentially fixated. We conclude that contour and color information results in a more thorough exploration of the paintings with shorter fixations, smaller saccades, a larger spread and higher intersubject agreement.

Acknowledgements: Supported by ERC Advanced Grant Color 3.0 (Grant no. 884116) and Deutsche Forschungsgemeinschaft SFB/TRR 135 A1,C2 (project number 222641018)