Abstract
Visual search for orientation is often thought of in terms of the angular difference between the target and distractors and modelling typically use four reference orientations: 0°, 45°, 90° and 135°. However orientation is a dimension which wraps around after 180° and studies on orientation in visual search often run into problems with linear separability, a phenomenon in visual search where the presence of distractors that flank the target along a given dimension makes search for that target difficult. We investigated the limits of linear separability in orientation by systematically varying both the target orientation (0°, 15°, 45°, 75° and 90° from vertical) and the angular difference between target and distractors (7.5° to 75°). Displays were presented for one second and we measured participants' accuracy at detecting the presence of a gap on the target. Psychometric curves were fitted to the accuracy data, which showed the lowest thresholds for 0 and 90° targets, followed by 45°. The highest thresholds were found for 15° and 75°. However, the lapse rate for 45° targets was much greater than for the other target orientations. We suspect that participants are creating an on-line representation of 45° using their internal representations of vertical and horizontal, allowing for a higher threshold than 15° and 75°, which do not have accessible representations. However, the on-line nature of this representation results in slower processing times, leading to a higher lapse rate. Search for the 15° and 75° targets are likely treated as search for a slightly off 0° and 90° target respectively, leading to greatly increased thresholds.
Meeting abstract presented at VSS 2016