The Deployment of Visual Attention: Two Surprises
Abstract
The visual system is not capable of processing of all aspects of a scene in parallel. While some visual information can be extracted from all locations at once, other processes, including object recognition, are severely limited in their capacity. Selective attention is used to limit the operation of these limited-capacity processes to one (or, perhaps, a few) objects at a time. Searching for a target in a scene, therefore, requires deployment of attention from one candidate target to the next until the target is found or the search is abandoned. Common sense suggests that distractor objects that have been rejected as targets are marked in some fashion to prevent redeployment of attention to non-target items. Introspection suggests that sustained attention to a scene builds up a perception of that scene in which more and more objects are simultaneously recognized. Neither common-sense nor introspection are correct in this case. Evidence suggests that covert attention is deployed at random among candidate targets without regard to the prior history of the search. Rejected distractors are not marked during a search. Prior to the arrival of attention, visual features are loosely bundled into objects. Attention is required to bind features into a recognizable object. For an object to be recognized, there must be a link between a visual representation and a representation in memory. Our data suggest that only one such link can be maintained at one moment in time. Hence, counter to introspection, only one object is recognized at one time. These surprising limits on our abilities may be based on a trade off speed for apparent efficiency.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2000
- Accession Number
- ADP010550
Entities
People
- Jeremy M. Wolfe
Organizations
- Brigham and Women's Hospital