Academic journal article Memory & Cognition

Selective Attention during Scene Perception: Evidence from Negative Priming

Academic journal article Memory & Cognition

Selective Attention during Scene Perception: Evidence from Negative Priming

Article excerpt

In two experiments, we examined the role of semantic scene content in guiding attention during scene viewing. In each experiment, performance on a lexical decision task was measured following the brief presentation of a scene. The lexical decision stimulus named an object that was either present or not present in the scene. The results of Experiment 1 revealed no priming from inconsistent objects (whose identities conflicted with the scene in which they appeared), but negative priming from consistent objects. The results of Experiment 2 indicated that negative priming from consistent objects occurs only when inconsistent objects are present in the scenes. Together, the results suggest that observers are likely to attend to inconsistent objects, and that representations of consistent objects are suppressed in the presence of an inconsistent object. Furthermore, the data suggest that inconsistent objects draw attention because they are relatively difficult to identify in an inappropriate context.

A typical visual scene contains many objects set against a meaningful background. A goal of an observer viewing such a scene is to identify the scene and the objects contained within it, a process that apparently occurs quite rapidly (see, e.g., Potter, 1976). Because visual attention plays such an important role in perception, understanding how people perceive and represent scenes depends critically on understanding the factors that influence attention during scene viewing. The present set of experiments examines the representation of objects in complex and meaningful scenes, and the role of attention in supporting the development of that representation.

Much recent work on scene representation has been stimulated by demonstrations that observers often fail to notice changes in scene content. Grimes (1996; see also McConkie & Zola, 1979), for example, demonstrated that observers routinely fail to detect even very large scene changes, as long as those changes are made during a saccade, when vision is suppressed (Volkmann, 1986). Rensink, O'Regan, and Clark (1997) have further shown that a saccade is unnecessary for generating such "change blindness." In their approach, subjects see two alternating versions of a single scene (one original and one changed version). The versions are alternated in rapid succession, separated by a brief blank field; subjects are typically quite poor at detecting the change. Change blindness is also observed when the change is accompanied by the abrupt appearance of visual stimuli elsewhere in the display (O'Regan, Rensink, & Clark, 1999) or by an eyeblink (O'Regan, Deubel, Clark, & Rensink, 2000).

Several accounts of change blindness have emerged that differ with respect to the amount of perceptual scene information represented (see Mitroff, Simons, & Levin, 2004, for a review). One prominent account attributes change blindness to a sparse and fragile representation of the prechange scene. Rensink and his colleagues (e.g., Rensink, 2000; Rensink et al., 1997), for example, have argued that mental scene representations consist of a small number of objects that are the current focus of attention. Although some representation of those objects may persist following the withdrawal of attention, Rensink argues that what survives is a representation of object type, rather than a perceptual representation. By this account, failure to detect a change reflects either a failure to encode the changed object (because it is not the focus of attention), or a failure to retain the representation across the change. Irwin and his colleagues (e.g., Irwin & Andrews, 1996; Irwin & Gordon, 1998; Irwin & Zelinsky, 2002) have similarly argued that memory across a saccade consists of a small set of object token representations in visual shortterm memory (VSTM), combined with some residual activation in a semantic network. The set of objects represented are those objects that have recently been attended. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.