Abstract
What should a theory of selective attention look like? Almost all current theories treat selective attention as a unitary process that operates according to one set of principles to achieve a single computational goal. In contrast, we will argue that selective attention should be treated as a general type of process that is embedded within many different cognitive subsystems and is used within different subsystems under different conditions. Specifically, attention will operate within a given subsystem when that subsystem faces an overload of inputs and different tasks will therefore elicit the use of attention within different subsystems. In addition, because representational formats, processing algorithms, and computational goals vary across cognitive subsystems, the details of how selective attention operates will vary across subsystems. Moreover, because different cognitive subsystems may sometimes need to focus on separate sources of information at a given moment in time, attention may sometimes operate independently within different subsystems. Thus, theories of attention must recognize that attention is a set of partially independent processes that are embedded within specific cognitive subsystems and operate in a manner that reflects the properties of those subsystems. To provide empirical support for this embedded process metatheory of attention, we will review recent evidence that: (a) attention operates within different cognitive subsystems depending on the stimuli and task; (b) attention has different properties depending on the subsystem in which it is operating; and (c) attention can be simultaneously focused on different sources of information within different cognitive subsystems.