2.4.3. Apparatus and Stimuli

Stimuli were presented on a Dell monitor. Participants saw 44 photographs of varying indoor and outdoor scenes. Images were taken from a publicly available database (Shuffle database, Large Change Images) [57], http://search.bwh.harvard.edu/new/Shuffle_Images.html. The experimental routine was programmed in Python using Psychopy [58,59]. The participants’ gaze behavior was recorded using a Pupil Labs eye tracker, using 120 Hz binocular gaze tracking and 60 fps world camera recordings [60].

The eye tracker was calibrated at the beginning of the experiment. All calibrations were done using the 9-point calibration routine implemented by Pupil Labs. Gaze points were mapped to the screen via the screen marker solution implemented by Pupil Labs. For that purpose, the monitor was defined as a surface based on 10 markers attached to the edge of the screen. To validate the calibration, participants were asked to fixate on a fixation dot presented at the center of the screen at the beginning of each trial (Figure 2A). The pupil labs eye tracker offers an accuracy of up to 0.6∘ and a precision of 0.2∘. If online-computed deviations between the recorded gaze position and the fixation dot exceeded 50 px (corresponding to a viewing angle of 1.15∘), the eye-tracker was recalibrated.

Images presented during a trial categorized into easy, medium and difficult with respect to the search task prior to the experiment. Ratings were done by three experimenters independently resulting in 75% of all ratings to be 100% consistent, whereas for the remaining 25% (eleven images) of ratings deviated by one (e.g., easy, easy, and medium). To resolve these inconsistent cases, the median of the ratings was chosen as a label (e.g., easy). The image dataset was split in half, assuring an equal distribution of difficulty among the two. Half of the images were used for the TC condition (n=22) while the other half was used for the TUC condition (n=22). Within each condition, images were shown in a randomized order, with no image shown twice to the same observer.