Visual crowding: a fundamental limit on conscious perception and object recognition

Abstract

Crowding, the inability to recognize objects in clutter, sets a fundamental limit on conscious visual perception and object recognition throughout most of the visual field. Despite how widespread and essential it is to object recognition, reading and visually guided action, a solid operational definition of what crowding is has only recently become clear. The goal of this review is to provide a broad-based synthesis of the most recent findings in this area, to define what crowding is and is not, and to set the stage for future work that will extend our understanding of crowding well beyond low-level vision. Here we define six diagnostic criteria for what counts as crowding, and further describe factors that both escape and break crowding. All of these lead to the conclusion that crowding occurs at multiple stages in the visual hierarchy.

Figure 1

Visual crowding—the deleterious effect of clutter on peripheral object recognition—is ubiquitous in natural scenes. A. It seriously impacts virtually all everyday tasks including reading, driving, and interacting with the environment. For example, fixating the bull’s-eye, near the construction zone, note that it is difficult or impossible to recognize the child on the left side of the road, simply because of the presence of the nearby signs. The child on the right, on the other hand, is relatively easier to recognize. B. While fixating the crosses, identifying the middle shape, letter, or line orientation—or even the number of tilted lines—is difficult or impossible on the bottom half of the panel. Crowding impairs the ability to recognize and scrutinize objects, but it does not make them disappear; one can see that some thing is present in panel (A), but it is difficult to identify the thing as a child as opposed to another sign. Crowding defines the spatial resolution of conscious object recognition throughout most of the visual field.

Figure 2

The critical spacing of crowding and Bouma’s proportionality constant (b). A. Fixating the crosses along the bottom, notice that the target orientation (central Gabor patch in each column) is easier to recognize on the right. B. Performance accuracy increases as the target-flanker separation increases. Bouma’s constant, b, may be defined as the target-flank separation (as a ratio of target eccentricity) that results in criterion performance (shown by the dashed line). Although the analytic methods and criteria used to compute b vary from study to study, it generally corresponds to the point at which performance begins to drop as flankers are advanced toward the target.

Figure 3

What escapes crowding? Although crowding limits conscious access to object identities in the periphery, there is much information that gets through or escapes crowding. A-B. Adaptation to low-level features. Adapting to the orientation (A) or motion (B) of a crowded pattern that is unidentifiable nevertheless causes strong local orientation and motion aftereffects, respectively. C. While fixating the central cross, notice that the array of Gabor patches on the right appears tilted more clockwise, while the array on the left appears relatively more tilted counterclockwise. In fact, the only difference between the two arrays is the single central Gabor patch; all the flankers are identical. Crowded objects can be unidentifiable, but nevertheless contribute their features to the ensemble or texture.

Figure 4

Crowding can be modulated or released depending on the nature of the flankers, and the target-flanker relationship. A. Reversing the contrast polarity of the target (left panel) reduces crowding compared to the right panel. B. Flankers of the same size as the target crowd more effectively (more crowding in right panel). C. Perceptual grouping of the flankers on the left reduces crowding (more crowding in the right panel). D-E. Object-centered, holisitic crowding. Faces crowd each other, and upright faces are more effective flankers than inverted ones. Crowding therefore occurs not just between features (Gabor patches, letters, but also between holistic representations of faces).

Figure I (Box 1)

Bouma’s constant (reflecting the critical spacing between target and distractors) varies depending on the similarity of the targets and flankers, and their complexity (top panel), the type of stimuli used (middle panel), and the attentional requirements of the task (bottom panel). That critical spacing can vary systematically suggests that Bouma’s rule is a rule-of-thumb, not a hard-and-fast law. Differences in methods and analytic approaches to calculating (b) cannot explain these graphs. The graphs show a within-study modulation of crowding (b), depending on similarity, complexity, stimulus type, and attention.