Fig 1.
The graph depicts the mean saliency values for the different regions of interest (head, body, areas of low and high saliency) for all social images according to the graph-based visual saliency algorithm [4].
The values depicted within the bars describe the mean number of ROIs per category and image across all social stimuli. Error bars represent standard errors of the mean.
Fig 2.
Example of a social stimulus as published in End & Gamer (2017) [13].
(A) Original scene. (B) Example of defined regions of interest for head (red), body (magenta), low saliency (yellow) and high saliency (green). (C) Overlay of a saliency map according to the Graph-Based Visual Saliency algorithm [4] with cool colors representing low salient regions and warm colors defining areas of high saliency. (D) Overlay of a fixation density map derived from fixation patterns of 31 participants who viewed the stimulus for 10s under free-viewing conditions. Image taken with permission from the Nencki Affective Picture System [24]. Please note that the stimulus shown here was not used in the current study and is only depicted to illustrate the current data analysis strategy.
Fig 3.
Reaction times (ms) as a function of task and stimulus category (A) and difficulty ratings for the different tasks (B). Error bars represent standard errors of the mean.
Fig 4.
Direct comparison of relative area normed fixation density for social (above) and non-social stimuli (below) on different regions of interest for two different time windows (before and after a given response) as a function of task.
The time window before the mouse click is referred to as “task relevant” (left panels), as it reflects possible top-down strategies on viewing behavior that were induced by the explicit tasks. The time window after completing the task is termed “task irrelevant” (right panels), as attentional allocation in this period should be less governed by task instructions. As there was no postulated task for the free-viewing condition, the task irrelevant time window does not contain any data for this modality. Viewing behavior for non-social stimuli was analyzed without social regions of interest, but allowed for comparing relative area normed fixation densities for areas of low and high saliency. Error bars indicate standard errors of the mean.
Fig 5.
(A) Relative area normed fixation densities as a function of ROI (head, body and background) and saliency (high vs. low) of these regions across different tasks. (B) Difference in relative area normed fixation densities between highly salient and less salient areas for each region of interest (head, body and background). Please note that the color coding of ROIs was adapted to the values of the highly salient regions. Error bars represent standard errors of the mean.
Fig 6.
Relative area normed fixation density of the first three fixations after stimulus onset (before response) on the different regions of interest for all tasks separately.
Top row: first fixations for social stimuli on four regions of interest (head, body, areas of low saliency and high saliency), bottom row: first fixations for non-social stimuli on two regions of interest (areas of low and high saliency). Error bars indicate standard errors of the mean.