Several studies measure the effect of task on eye-movements. Examples include every-day activities like tea making (Land, Mennie, & Rusted,
1999), judgment of social status in a photograph (Yarbus,
1967), rating preference for a picture (Buswell,
1935), or rating facial attractiveness (Shimojo, Simion, Shimojo, & Scheier,
2003). Notwithstanding the realism of such tasks and the potential role of eye-movements as indicator of cognitive disorders such as autism (Pelphrey et al.,
2002), quantification of task related effects typically rely on variants of visual search. The usage of visual search in eye-movement research dates back at least to Buswell (
1935), who instructed an observer to “find a person looking out of one of the windows of the tower,†causing remarkable context-driven effect on eye-movement patterns. Most research on visual search, however, is based on reaction time measurements in well-controlled search displays. Following the pioneering work of Treisman and Gelade (
1980), the dependence of reaction times versus set size serves as measure of whether a search is serial (reaction time increases linearly with set-size) or parallel (reaction time is independent of set size). The slope of this relation determines the difficulty of the search. Using such a setting, more and more sophisticated models have been put forward to explain the relation of target features to those of attended items. Probably the most influential is “Guided Search†(Wolfe,
1994; Wolfe et al.,
1989): Akin to the saliency map, the stimulus is filtered in various feature channels, top-down attention enhances the channels present in the target, resulting in an “activation map†guiding search. If the target is different in just an individual elementary feature, the activation map has one peak, the target pops out, and search is parallel; if the target is defined by the conjunction of multiple features, the activation map has multiple peaks that have to be searched sequentially. Hence, Guided Search provides a mechanism accounting for Treisman and Gelade's (
1980) data and a framework for integrating bottom-up saliency and top-down feature biases.