Integrating Perceptual Organization and Attention: A New Model For Object-Based Attention

Abstract

Integrating Perceptual Organization and Attention: A New Model For Object-Based Attention Jerzy P. Jarmasz (jjarmasz@ccs.carleton.ca) Cognitive Science Program and Centre for Applied Cognitive Research, Carleton University, Ottawa, Canada, K1R 7W9 Abstract Recent research shows that, under certain conditions, visual attention is object-based. That is, attention preferentially selects objects in the visual field. These objects are processed, culminating in object recognition. On this formulation, the objects selected by attention are perceptual groups determined by the principles of perceptual organization of Gestalt psychology. These groups are formed independently of attentional processes and conceptual knowledge. This view is not consistent with available data about the visual system, which shows that perceptual organization is sensitive to conceptual information, depends on attentional processes, and infers representations that best explain the visual stimulus. Here, I propose a new account of visual attention that aims to correct these limitations of the Gestalt-based formulation. The nature of the object representations underlying perceptual and attentional mechanisms is discussed. It is proposed that attention and perception interact in an iterative process wherein constraints imposed both by the visual stimulus and an observer’s cognitive set determine the “objects” to which attention is allocated. Thus, visual attention is object based precisely because it is intricately involved in perceptual organization, and not because it selects the output of perceptual organization, as is generally claimed. Experimental results that support the claim that attention influences perceptual organization are reviewed. Finally, the implications for human factors research and the metaphysics of everyday objects are discussed. Introduction Vision is generally assumed to have the functions of identifying, locating, and directing action towards objects (Solso, 1996). It is also assumed that the visual system requires attentional mechanisms to limit the amount of sensory information it processes (Fernadez-Duque & Johnson, 1999). Thus, awareness of objects in the environment is supposed to result from a series of processing stages that select sensory information and then construct representations of objects by extracting regularities from the visual stimulus and matching them to patterns in memory (Palmer, 1999). It was first assumed that visual attention selects certain regions of the visual field, much the way a spotlight illuminates part of a stage and leaves the rest in the dark. Accordingly, this model is known as the spotlight model of visual attention (Fernandez-Duque & Johnson, 1999). On this model, attention is first directed to a region of the visual field, and only the information within that region is processed for object identification. This assumption was questioned when researchers observed that people respond to visual features that belong to a single object more quickly and accurately than when the features belong to two objects (Duncan, 1984; Treisman, Kahneman & Burkell, 1983). Subsequent research confirmed that it is usually easier to process information within a single object than across objects (Lavie & Driver, 1996). These findings have lead to the object-based model of visual attention (Duncan, 1984; Lavie & Driver, 1996). It is now generally recognized that the spotlight and the object-based models capture complementary aspects of visual attention (Driver & Baylis, It is undeniable that information can be processed more readily within one object than across many (Lavie & Driver, 1996; Driver & Baylis, 1998). However, the object-based explanation for this difference in processing efficiency is problematic. Cognitive psychologists generally distinguish between spatio-temporally bounded physical objects and the mental representations of these objects. Physical objects correspond to what philosophers call concrete particulars (Loux, 1998), and will subsequently be referred to as c- objects. Similarly, the mental representation of visual objects will be henceforth referred to as p-objects (for “phenomenological” objects). The generally accepted story about object perception is that the visual system constructs p-objects, which represent c-objects via various perceptual and cognitive processes. Researchers who accept the object- based model contend that attention selects “objects” for further processing. Which objects are these – p-objects or c- objects? P-objects are supposed to be the end product of visual processing (Solso, 1996), so attention must presumably be engaged prior to the construction of p- objects. However, the alternate claim that attention directly processes c-objects themselves instead of sensory input is nonsense. Most researchers assume the visual system first constructs low-level representations of c-objects, based on the physical properties of the stimulus. These representations are then elaborated into p-objects by higher- order visual and conceptual processes (Hoffman, 1998; Palmer, 1999). These low-level representations will be referred to as a-objects (for “attentional objects”). The object-based model can be restated thus: Visual attention selects a-objects, which are passed on to higher visual processes for elaboration into p-objects, which are representations of c-objects. Philosophers are actively studying the nature of c-objects (see Loux, 1998) and perceptual psychologists are