Schema approaches to scene structure

Historically, a central theoretical construct in the field of picture and scene memory has been the scene schema (Biederman, Mezzanotte, & Rabinowitz, 1982; Brewer &Treyens, 1981; Friedman, 1979; Intraub, 1997; Mandler & Ritchey, 1977; Pedzek, Whetstone, Reynolds, Askari, & Dougherty, 1989). The basic claim of schema theories is that episodic representations of scenes are structured according to prior experience with scenes of that type. For example, one’s memory for a particular kitchen scene will be strongly influenced by one’s memory for kitchens in general, a kitchen schema, which will govern the types of information retained in memory from that scene (see chapter 7 ofThe Visual World in Memory for the influence of schemas on memory for visual events). The standard description of a scene schema is an abstract representation of a particular scene category specifying the objects that are typically found in that type of scene and the typical locations of those objects (Mandler & Parker, 1976).

Two components are consistently present in schema accounts of scene memory: abstraction and distortion (for a critical review, see Alba & Hasher, 1983). First, scene memory is proposed to be highly abstract and conceptual in nature— that is, limited to the gist of the scene (Mandler & Ritchey, 1977; Potter et al., 2004). Scene details are initially activated during perceptual processing of the scene, but the details are quickly forgotten. In this claim, schema theories are quite similar to claims of gist-based representations in the change blindness literature (O’Regan, 1992; Rensink, 2000; Simons & Levin, 1997).The evidence that scene representations preserve significant visual detail, and are not limited to gist, has been reviewed exhaustively above. Thus, the schema theory claim of gist abstraction is not well supported by experimental evidence.

Second, the schema approach holds that memory for scene properties will be distorted by prior knowledge. Objects frequently found within a scene of that type (such as a dresser in a bedroom) will be remembered most frequently, because they have pre-existing “slots” in the schema. Incongruous or unexpected objects (such as a pig in a bedroom) will be remembered less accurately and will be normalized to default values in the schema. Although common sense would dictate that anomalous objects should be remembered most frequently from a scene (as they would be most salient), normalization is a central feature of schema theory (Bartlett, 1932). Brewer and Treyens (1981) tested the normalization claim by having participants remember the objects in a graduate student office, some of which were semantic- ally consistent (desk) and some inconsistent (skull). On a free-recall test, participants more frequently reported semantically consistent objects than inconsistent objects, supporting the claim of normalization. However, Brewer and Treyens provided no control over guessing, and the advantage for consistent objects could easily have been generated by a bias to guess that consistent objects had been present. For example, if asked to report which objects had been in a kitchen scene, one could guess that there was likely to have been a stove, even if one did not specifically remember a stove.

In contrast to the Brewer and Treyens (1981) result, subsequent studies controlling guessing have found the reverse effect: better memory for semantically inconsistent objects in scenes (Friedman, 1979; Hollingworth & Henderson, 2000, 2003; Pedzek et al., 1989). Although some researchers have proposed schema explanations to account for superior inconsistent-object memory, these have been somewhat ad hoc. For example, Friedman (1979) proposed that inconsistent objects are stored robustly as part of a “weird list” that is appended to the schema representation. This type of modification would render the schema approach all but unfalsifiable. In general, the absence of inconsistent-object normalization argues against the standard schema account of scene memory.

 
Source
< Prev   CONTENTS   Source   Next >