Abstract
In visual search, the more similar the target to the distractors, the less efficient the search; the less similar the target to the distractors, the more efficient the search. Very little has been done to understand how similarity arises in multi-dimensional stimuli. In previous work, we have studied how bi-dimensional dissimilarity arises when stimuli differ along two feature dimensions in efficient search (color and shape; shape and texture; color and texture). Here, we report our first evaluation of how visual complexity arises when the target differs from distractors along three feature dimensions (color, shape and texture). In Experiments 1-3, we first tested people’s search efficiency when the target differed from a set of homogeneous distractors either along color (a red target among green, pink or orange distractors), shape (an octagon target among triangle, square or house-shape distractors), and texture (a cross-textured target among dotted-, hashed- or solid-textured distractors). In Experiments 4-12, we fully crossed all possible combinations of the tested distractor colors, shapes and textures to define the multidimensional distractor stimuli (e.g., green-dotted triangles, pink-hashed squares). We used a model comparison approach to evaluate the extent to which efficiency in Experiments 1-3 could predict performance in Experiments 4-12. We used computational modeling to understand the mechanistic instantiation behind the competing models. Our results suggest that three-dimensional search efficiency (R-squared = 0.85 over 27 datapoints) as well as RT (R-squared = 0.86 over 108 datapoints) were best predicted by a model where all three dimensions are racing towards a rejection threshold simultaneously. The race occurs at each item location independently and results in substantial processing efficiency savings. From an information processing standpoint, this is the most efficient manner to discard unlikely targets. These findings suggest that search efficiency might be driven by different forces in uni-dimensional compared to multi-dimensional contexts.