The Road to Success: Necessary and Unnecessary Visual Features in Parallel Individuation

Each day, we interact with and make judgments about objects we see in the visual field. These interactions depend on the perceptual segmentation of figure from ground, and the subsequent processing of the segmented representations. In order to survive in our increasingly complex world, it is not enough to know what is object and what is background; we must be able to rapidly infer information about sets of objects, such as their identity, in what direction are they moving, and how many are there. Parallel individuation is the rapid selection of multiple targets for precise and rapid processing. One of the hallmarks of parallel individuation behavior is the ability to re- port the number of items in small sets with extreme speed and accuracy; this behavior is called ‘subitizing’. Previous research has suggested that subitizing, and other tasks de- pendent on parallel individuation, rely on targets that are spatially separate from each other: objects rather than object parts. In this thesis, we explored what visual features, like connectivity, interfere with or are necessary for parallel individuation to occur. We first demonstrated that both connected and unconnected targets can be subitized. In order to ensure the same neural mechanism was responsible for the behavior ob- served in both stimulus conditions, we used functional magnetic resonance imaging to compare the neural responses to connected and unconnected stimuli. We targeted two regions in the parietal lobe, the inferior and superior intraparietal sulci, which have been previously associated with individuation and identification, respectively. Activity in both regions was modulated by the numerosity of targets in the connected and un- connected conditions. While multi-voxel pattern analyses revealed that the two regions additionally held representations of number, only the inferior IPS could discriminate connected from unconnected stimuli. We concluded from these results that individuation in the inferior IPS does not depend on spatially separate targets, but rather can flexibly select a level from the object hierarchy of a scene within which to define figure and ground. We then investigated the role of other visual features in parallel individuation, working from a condition of failure to pinpoint visual characteristics that are necessary for subitizing to occur. Resolving line ownership, providing unique centers of mass, and removing bounding enclosure information all did not prove sufficient for subitizing to occur in concentrically arranged squares. Changing the arrangement of the squares in space however did demonstrate that subitizing occurs over overlap- ping targets, with no effect of amount of overlap. Manipulating the presentation time of overlapping targets showed that parallel individuation can operate over both unresolved and completed amodal representations of a scene. We proposed that successful parallel individuation is dependent upon the constraints of three stages of processing: segmentation, individuation, and task specific demands. We suggest that the individuation stage is dependent upon each target occupying a unique location in space, occurring outside the border of all other targets. Finally we discussed the generalization of this model to other tasks involving parallel individuation.Psycholog