Stimuli were constructed from the same set of line drawings as in
Experiment 1.
The original line drawings were manipulated in two different ways, one version removing contours at and around junctions, called the
junctions-removed condition, and the other version removing contiguous contour sections in-between junctions, called the
middle-removed condition. To this end, locations of junctions were determined as points where two contours crossed, points where the angle between adjacent contour elements was less than 130°, and endpoints of contours. We then marked breakpoints at 25% and at 75% of the distance between two junctions. This allowed us to remove junctions by removing, from all contours, the sections between 0 and 25% and between 75% and 100% of the distance between two adjacent junctions and to remove middle parts by erasing the sections between 25% and 75% of the distance (see
Figure 6, top right for an illustration). Note that both manipulated versions are complements of each other, containing half of the total contour content, only having overlap at the end points of the middle segments and junction segments. An example of these manipulations on an office scene is shown in
Figure 6. These manipulations change the distributions of contour features. For example, the number of pixels is halved, so for each feature, the distribution is scaled. Additionally, because the contours were separated into smaller pieces, the length distributions are shifted. A comparison of the distributions of contour features for the intact and the manipulated image are in
Figure 7. There is no figure for how the junctions were changed, because either they were maintained (and thus identical to those in the intact scene) or completely removed. By design, the length distributions are shifted to the left (shorter) because the contours were separated into smaller pieces. Also, with fewer pixels than in the intact scenes, the overall distributions are smaller. The changes between the distributions are similar for all scene categories.
Unlike the stimuli in
Experiment 1, these stimuli are not randomly shifted in any way. Thus, the stimulus generation is deterministic, and the images will not differ between two participants who are shown the same scene exemplar in the same experimental condition.