Visual search, selection, and attention

To investigate how the brain selects the target for an eye movement, multiple stimuli that can be distinguished in some way must be presented. This experimental design is referred to as visual search. The visual search paradigm has been used extensively to investigate visual selection and attention (reviewed by Wolfe, 1998). In a visual search task, multiple stimuli are presented among which a target is discriminated. Search is efficient (with fewer errors and faster response times) if stimuli differ along basic visual feature dimensions, such as color, form, or direction of motion. In contrast, if the distractors resemble the target or if no single feature clearly distinguishes the stimuli, then search becomes less efficient (more errors, longer response times) (e.g., Duncan and Humphreys, 1989).

Saccade target selection cannot be discussed without consideration of the allocation of visual attention. In fact, it can be argued that visual target selection and the allocation of visual attention amount to the same thing. Visual attention is the topic of Chapters 101, 102, and 103. The focus of visual attention can be directed away from the focus of gaze. So the link between shifting gaze and directing attention is not obligatory (Crawford and Muller, 1992; Eriksen and Hoffman, 1972; Jonides, 1980; Klein et al., 1992; Posner, 1980; Remington, 1980; Reuter-Lorenz and Fendrich, 1992; Shepherd et al., 1986). Nevertheless, several lines of evidence indicate that covert orienting of visual attention and overt orienting of saccades depend on a common selection mechanism. First, perceptual sensitivity is reduced and saccade latency is elevated if attention is directed away from the target for a saccade (Deubel and Schneider, 1996; Hoffman and Subramaniam, 1995; Kowler et al., 1995). Second, directing attention in space can influence the trajectory of saccades (Kustov and Robinson, 1996; Sheliga et al., 1995). Third, bottom-up visual factors influence visual selection for attention and saccades in the same way. The visual conspicuousness of an oddball stimulus can drive covert (e.g., Theeuwes, 1991) and overt (Theeuwes et al., 1998) selection, and nontarget elements that resemble a designated target can get inadvertently selected covertly (e.g., Kim and Cave, 1995) and overtly (Bichot and Schall, 1999b; Findlay, 1997; Motter and Belky, 1998; Zelinsky and Sheinberg, 1997). Fourth, cognitive strategies can override both covert (e.g., Bacon and Egeth, 1994) and overt (e.g., Bichot et al., 1996) oddball selection. Target selection is also influenced by implicit memory representations arising through short-term priming of location or stimulus features for covert (e.g., Maljkovic and Nakayama, 1994, 1996) and overt (Bichot and Schall, 1999b; McPeek et al., 1999, 2000; MePeek and Keller, 2001) orienting. In addition, experts are more likely than novices to fixate conspicuous but irrelevant parts of a visual image from their field of expertise (e.g., Chapman and Underwood, 1998; Nodine and Krupinski, 1998; Nodine et al., 1996). Finally, the pattern of visual fixation can be influenced by verbal instructions (Yarbus, 1967).

To explain observations like these, most models of visual search postulate the existence of a map of salience derived from converging bottom-up and top-down influences (e.g., Cave and Wolfe, 1990; Itti and Koch, 2000; Koch and Ullman, 1985; Treisman, 1988; Wolfe, 1994). Other investigators have noted that the salience map is a useful construct to organize our understanding of how the brain selects the target for an eye movement (Findlay and Walker, 1999; Kusunoki et al., 2000; Thompson et al., 2001). Salience refers to how distinct one element of the image is from surrounding elements. This distinctness can occur because the element has visual features that are very different from those of its surrounding (a ripe red berry in green leaves). The distinctness can also occur because the element is more important than others (the face of a friend among strangers). The distinctness derived from visual features and importance confers upon that part of the image greater likelihood of receiving attention and a gaze shift. In the models of visual search referred to above, one major input to the salience map is the maps of the features (color, shape, motion, depth) of elements of the image. Another major input is top-down modulation based on goals and expectations. The representation of likely targets that is implicit in and dependent on the feature maps becomes explicit in the salience map. Peaks of activation in the salience map that develop as a result of competitive interactions represent locations that have been selected for further processing, and thus covert orienting of attention and possibly but not necessarily a saccadic eye movement.