Two Visual Systems and Their Eye Movements: Evidence from Static and Dynamic Scene Perception

Two Visual Systems and their Eye Movements: Evidence from Static and Dynamic Scene Perception Boris M. Velichkovsky, Markus Joos, Jens R. Helmert, and Sebastian Pannasch (velich {joosy helmert pannasch}@psychomail.tu-dresden.de) Dresden University of Technology Applied Cognitive Research/ Psychology III Mommsenstrasse 13 Dresden, D-01062 Germany Abstract The existence of two distinct visual pathways in the primate brain is a persistent theme for evolutionary, neurophysio- logical, motor control and neuropsychological research. As one of the most widely cited results in cognitive neuroscience, this distinction has survived decades of critical analysis under different guises (e.g. ambient vs. focal or visuomotor vs. cognitive). However, the interplay between these two processing streams in the solution of everyday tasks remains to be an unresolved issue. In particular, how do they guide eye movements, their most immediate output? Results from our recent study on hazard perception in a simulated driving environment demonstrated that specific combinations of eye movement parameters are indicative to an involvement of either of the two systems. In a further experiment, we tried to validate these parameters by testing assumptions about memory representations related to these two modes. After a short presentation of various real world scenes, subjects had to recognize cut-outs from them, which were selected according to their fixation parameters. Random cut-outs from not seen pictures (catch trials) were also presented. The results confirmed our hypothesis: cut-outs corresponding to presumably focal mode of processing were better recognized than cut-outs similarly fixated in the course of ambient exploration. Keywords: Active Vision; Dorsal and Ventral Streams; Ambient and Focal Attention; Scene Perception; Recognition; Eye Movements. Introduction In contemporary studies of visual cognition, one can discover several clusters of research that are only loosely connected to each other (for similar arguments, see Simons & Rensink, 2005). An intensive albeit still controversial discussion is, for instance, how and even whether visual information is retained across saccades while viewing a scene (Bridgeman, Van der Heijden, and Velichkovsky, Irwin's object file theory of transsaccadic memory emphasises the crucial role of visual attention in what local visual information from a scene is or is not represented (Irwin, 1992). Attending an object in a visual scene allows binding its features into a unified object description (Treisman, 1988). This object description is linked to a spatial position in a master map of locations, forming a temporary representation in visual short-term memory ( VSTM ). According to Irwin (1992) three to four discrete objects can be hold at a time in VSTM . Finally, object files are the primary content of transsaccadic memory providing local continuity from one fixation to the next. More recently, coherence theory (Rensink, 2000a, 2000b) proposed a similar explanation of the fact that despite the 'snapshot-like' character of visual information acquisition the world around us is experienced as being stable, coherent and richly detailed. Just as in object file theory visual attention is the premise to bind sensory features into a coherent object representation, which can be hold in VSTM preventing them from disruptions like saccades. Prior to focused attention, proto-objects with only limited temporary and spatial coherence are formed in parallel across the visual field but being volatile and replaced on appearance of a new stimulus at their position. On withdrawal of attention a coherent object resolves into its constituent proto-objects again leaving no or only little after-effect of attention (Rensink, 2000a). Both these visual transience hypotheses (Hollingworth & Henderson, 2002) rely on the idea, that a complete metric representation of a visual scene is neither possible nor necessary. O'Regan and Noe (2001) go even further saying (actually paraphrasing Brooks, 1991, p. 139) “the world serves as its own memory”. One drawback of these approaches to scene perception is, however, that they consider eye movements as mechanical events of no further interest. The classification is simply based on measuring whether the subject is holding their eyes stable at a designated position (fixation) or is doing a jerky eye movement (saccade). Presentation or extinction of visual stimuli (e.g. a scene or an object) is in the majority of cases executed in relation to the start or end of a saccade. The underlying processing, which may be reflected in the variation of fixation durations, is neglected. There is another line of research analyzing the duration of visual fixations in terms of task complexity, levels of processing or skills, especially for reading tasks (Velichkovsky, 1999). In a recent study by Unema, Pannasch, Joos, and Velichkovsky (2005) subjects viewed computer generated images of rooms containing different interior, in order to be able to answer questions about the distribution of objects within the room or about the presence/absence of particular objects. The authors found a clear shift of the ratio of fixation durations and saccadic amplitudes across the tasks and also over the viewing time. At the beginning of image inspection fixations with shorter durations and saccades with longer amplitudes were