Visual saliency does not account for eye movements during visual search in real-world scenes

Semantic memory for contextual regularities within and across scene categories: Evidence from eye movements

“…There is little dispute that visual behaviors including gaze control are strongly influenced by a variety of top-down factors (see, e.g., Buswell, 1935;Henderson, Brockmole, Castelhano, & Mack, 2007;Torralba, Oliva, Castelhano, & Henderson, 2006;Yarbus, 1967). Preexisting representations of scene structure , likely constituent objects (De Graef, Christiaens, & d'Ydewalle, 1990;Friedman & Liebelt, 1981;Henderson et al, 1999), and likely locations of objects (Eckstein, Drescher, & Shimozaki, 2006;Henderson et al, 1999;Neider & Zelinsky, 2006;Torralba et al, 2006;Võ & Henderson, 2009) all constrain search and gaze behaviors.…”

Saliency on a natural scene background: Effects of color and luminance contrast add linearly

“…We had used this instruction in earlier studies and expected it to bias fixation allocation in a bottom-up driven mode and to operate in sharp contrast to explicit top-down tasks, such as search (Einhäuser, Rutishauser, & Koch, 2008;Henderson et al, 2007). A recent experiment (Steinwender & König, 2007) indeed showed that "study carefully" yielded the same result with respect to lowlevel features as the explicit instruction of "free-viewing," whereas, for example, "subjective assessment" yielded distinct fixation behavior.…”

“…All stimulus presentation and eye position recording was programmed in MATLAB (MathWorks, Natick, MA), using its psychophysics and eyelink toolbox extensions (Brainard, 1997; Cornelis- Castelhano, & Mack, 2007). More and more evidence has accumulated that the saliency map's fixation prediction is mostly indirect, which undermines the causal and mechanistic implications of the model.…”

The interaction of visual and linguistic saliency during syntactic ambiguity resolution

“…In goaloriented tasks, in which a target object is specified (e.g., visual search), saliency effects are weaker, if not neutralized. Attention is mostly directed to regions that are contextually related to the search target (e.g., looks to a TABLE when searching for a MUG, Henderson, Brockmole, Castelhano, & Mack, 2007). More recent studies, however, indicate that visual saliency can interact with goaloriented, top-down processes.…”