STRATEGIES AND STRUCTURE IN SELECTIVE ATTENTION TASKS

If the ability of observers to respond to objects on the basis of a single attribute reveals how objects are represented perceptually, experimental results concerning this ability should converge.; Two experimental tasks are described. In unconditional classification, the observer responds to one object according to its level of an attribute. In simultaneous comparison, the observer responds to a two-object display according to whether its objects share a level of that attribute. Because stimulus displays differ between the tasks, the perceptually represented objects may differ--two objects presented for comparison may be perceived as one object.; Experiments 1 and 2 evaluated this conjecture. Subjects attended selectively to an attribute in unconditional classification. In simultaneous comparison, random variation over objects of the irrelevant attribute interfered with performance, but poorer performance was restricted to one kind of display. Manipulated display properties influenced whether two objects were perceived as a single object. The results suggested that some two-object displays are always perceived as one object, some as two objects, and some as one object or two objects, depending on the conditions.; Experiments 3-5 evaluated the invariance of perceptual representations of displays that contain two nominal objects. Despite its use to assess selective attention, the comparison task requires that attention be divided over two sources of information. If objects are defined empirically by simultaneous availability of information about their attributes, divided attention should be easy to the attributes of one object but difficult to the attributes of two objects. Responding to an attribute of one object in a display represented as two objects should be as easy as to an attribute of a display represented as one object. Classification performance using rules that either required responses to be based on both nominal objects or allowed responses to be based on one nominal object confirmed these hypotheses.