Abstract
It is long known that pupil size is modulated by luminance of stimuli, and arousal or cognitive effort of subjects. Recently, several groups suggested that higher-level perceptual interpretation also impacts pupil size, e.g., perceived brightness and lightness (Laeng & Endestad, 2012; Naber & Nakayama, 2013). Here we investigated pupillary responses when subjects were viewing scenes from three different top-level categories: indoor, man-made (outdoor), and natural. In Experiment 1, 10 subjects were showed 210 distinct grayscale images (70 from each category) automatically drawn from a large database, intermixed with their phase scrambled version. All images were preprocessed to have identical mean pixel value of 0.5. The subjects engaged in a visual search task for a local spatial distortion randomly embedded in the image, while binocular eye movement and pupil size were continuously recorded at 1000 Hz with EyeLink 1000. In Experiment 2, vertically inverted, as well as polarity (black-white) inverted versions of the original images were shown instead, to another group of participants. After initial contraction following stimulus onset, the pupil dilated as the subject kept on searching. Interestingly, pupil size was significantly larger for "indoor" scenes than "man-made", which in turn larger than "natural". Physical luminance, contrast, and task difficulty failed to explain the results, since "indoor" was actually associated with slightly higher mean luminance and contrast (luminance variance), and was less difficult (shorter response time and higher percentage correct) than "natural", which predicted the opposite results. Notably, the difference between "indoor" and "man-made" diminished in scrambled, vertically inverted, and polarity inverted conditions, although "natural" still resulted in smaller pupil size than other categories. We suggest that pupil size was larger when subjects were viewing indoor compared with outdoor scenes, partly because the latter might be subjectively perceived as brighter, a learnt regularity from everyday experiences.
Meeting abstract presented at VSS 2014