Abstract: PURPOSE: Traditional visual search tasks use 2D targets in 2D scenes, with experimental displays perpendicular to the line of sight. However, in real life, we search 3D objects in 3D environments, with non-perpendicular viewing angles. We investigate how 2D searches and 3D searches differ in terms of accuracy and efficiency (response time as a function of set size) s.
METHODS: 3D Lego blocks of similar sizes but different shapes were used as search targets and distracters. In Experiment 1, targets and distracters were 2D frontoparallel projections of the 3D blocks, which were placed in a concentric arrangement with no occlusion between objects (similar to a classic visual search task display). In Experiment 2, the search array was displayed with a perspective projection (we took pictures of 3D blocks arranged on a flat surface from a 30° angle pointing down). In both experiments, the set size was 3, 6, 9, or 12 and the search target was present in the mid of the search array in half trials.
RESULTS: When searching for 2D parallel projections of 3D shapes (Experiment 1), the presence of target throughout a trial increased accuracy, but did not affect efficiency. Regardless of this, the slope between response time and set size in target-absent trials was twice as much as that in target-present trials, align with the results of previous studies. When searching in a more realistic scene with perspective projection and occlusions between objects (Experiment 2), accuracy was higher but efficiency was similar to performance in Experiment 1. The presence of search targets during search increased both accuracy and response time.
CONCLUSIONS: aSearching in realistic 3D scenes and searching in 2D representations were likely to involve different mechanisms. Although 3D scenes were more complex, search performance was better. This might be because searching for 3D objects in 3D scenes was a more natural task that contained more information and bore more experiences.

Key words: visual search in real life, spatial configuration, perspective change