Object and Gist Perception in a Dual Task Paradigm: Is Attention Important?

Object and Gist Perception in a Dual Task Paradigm: Is Attention Important? Maria Koushiou 1 (maria_koushiou@yahoo.com) Elena Constantinou 1 (elena_constantinou@yahoo.gr) Department of Psychology, University of Cyprus, P.O. Box 20537, 1678 Nicosia, Cyprus Abstract An experiment was conducted to investigate object and gist perception under conditions of inattention. Participants performed an attentionally demanding central task while responding to a secondary peripheral task involving object categorization or gist identification in natural scenes. Participants were unexpectedly more accurate on object than gist categorization, a finding attributed to a possible facilitatory effect of object figure saliency. This hypothesis was confirmed by a second experiment comparing the peripheral tasks under single-task conditions. A third dual-task experiment was conducted to compare object and gist perception when controlling for figure saliency by using the exact same stimuli for both peripheral tasks. No significant differences were found in the third experiment. Conflicting results and methodological issues are thus discussed. Keywords: object categorization; gist perception; dual-task methodology. Introduction Modern computational models of visual attention propose as a necessary stage in the processing of visual stimuli a pre- attentive level during which early visual features are computed in a bottom-up fashion in order to direct attention towards stimuli of interest (Itti & Koch, 2001). However the extent of processing that takes place at this pre-attentive stage is not quite clear, with empirical evidence often offering contradicting results. Phenomena like inattentional blindness and change blindness, provide evidence for the presence of significant pre-attentive perceptual limitations since humans fail to perceive highly visible objects (Mack, 2003) or to detect large changes in a visual scene (Simons & Ambinder, 2005) when their attention is allocated elsewhere. However, the use of priming methodology in inattentional blindness experiments, has shown that unconscious information prime subsequent behavior, indicating that visual information (e.g. the observer’s name) may be processed perceptually prior the engagement of attention (Mack, 2003). This empirical contradiction is also depicted in theory since early selection theories (Broadbent, 1958) support that only a rudimentary analysis of physical features occurs before attentional selection while late selection theories (Deutch & Deutch, 1963) claim that perception is achieved in an automatic and parallel fashion, with attentional selection intervening only after full perception of items is achieved. Lavie (1995) proposed a model in an attempt to resolve the early vs. late selection controversy. According to this model, the perceptual load of high-priority relevant stimuli determines whether selection is early or late. Experimentally, one way of testing the applicability of the above theories is provided by the dual-task paradigm. In this paradigm attention is drawn to a perceptually demanding central task while irrelevant to the central task stimuli are presented on the periphery. In dual task experiments researchers assess whether secondary tasks involving peripheral stimuli interfere with performance in the concurrent central task. Experiments using this procedure have shown that our perceptual system is subject to capacity limits when the stimulus load is increased (Pashler & Johnson, 1998). Efficient performance on each task requires perceptual analysis followed by central operations to produce a response. Since central processing of the two tasks cannot be done concurrently, the processing of one task presumably occurs in parallel with the perceptual analysis of the other task. Thus, in a dual task condition our brain seems to switch the central processing from one task to another, while buffering the essential information for both tasks concurrently (Pashler & Johnson, 1998). Different levels of perceptual processing have been examined using dual task methodology to determine the type of information that can be processed at a pre-attentive level. The present study focuses on two kinds of processing natural scenes, namely the categorization of objects in scenes and the perception of the scene’s gist. Our aim is to examine what types of information one can extract from a natural scene before focusing attention on certain aspects of it. An overview of previous research examining these two types of processes is presented next. Perception of Objects in a Scene According to the classical view, perception without attention is possible when detecting primitive, low-level features such as motion, orientation, texture and brightness. Objects, on the other hand, consist of superior stimuli that require focused attention and a higher level of processing (Braun, 2003 cited in Evans & Treisman, 2005). Experimental evidence from studies using the dual task paradigm, Inattentional Blindness, and Rapid Serial Visual Presentation (RSVP), challenge the aforementioned view,