Academic journal article Memory & Cognition

Contributions of Category and Fine-Grained Information to Location Memory: When Categories Don't Weigh In

Academic journal article Memory & Cognition

Contributions of Category and Fine-Grained Information to Location Memory: When Categories Don't Weigh In

Article excerpt

Several studies have shown that people's memory for location can be influenced by categorical information. According to a model proposed by Huttenlocher, Hedges, and Duncan (1991), people estimate location by combining fine-grained item-level information in memory with category-level information. When the fine-grained information is inexact, category-level information is given greater weight, which leads to biased responses. We manipulated the distribution of locations presented in order to alter the usefulness of category information, and we manipulated background texture in order to alter accuracy of fine-grained memory. The distributional information reduced bias without altering overall accuracy of responding, whereas the background texture manipulation affected accuracy without changing bias. Our results suggest that category information may weigh in only when it is actively processed.

Memory for places can be critical to the survival of humans and nonhumans alike. Not surprisingly, the underlying mechanisms and the factors that influence location memory have been the focus of extensive research in cognitive science, developmental psychology, animal behavior, and neuroscience. One common finding is that spatial memory rarely relies on a single source of information; rather, multiple sources of spatial information are often encoded (Cheng, Shettleworth, Huttenlocher, & Rieser, 2007; Huttenlocher, Hedges, & Duncan, 1991; Spetch & Kelly, 2006). Understanding how multiple cues are integrated to control spatial search or estimation is an important step in the study of spatial memory.

Cheng et al. (2007) suggested that the integration of spatial information may be considered in terms of a functional Bayesian framework. According to their framework, spatial information from multiple cues is likely to be combined, as long as the information is not too discrepant among the different cues along the relevant dimension(s). They suggested that information from multiple cues is sometimes combined in a near optimal fashion, using a weighted average. According to Bayesian principles, the optimal weight for each cue should be proportional to the variance of the spatial information encoded by that cue. Cheng et al. (2007) proposed that a weighted combination of information may occur when spatial estimation is based on two or more sources of current information, on current information together with prior information, or on current information together with categorical information. The latter situation is the focus of our investigation.

Huttenlocher and colleagues conducted a series of studies to investigate how location memory for specific items is influenced by spatial categories (e.g., Huttenlocher, Hedges, Corrigan, & Crawford, 2004; Huttenlocher et al., 1991; Huttenlocher, Newcombe, & Sandberg, 1994). In one of their tasks (Huttenlocher et al., 1991), people were shown a series of dots within a circular space and were required to reproduce the dots' locations from memory. Although the dots were randomly distributed throughout the circle, people showed a bias to remember them as having been located toward the midpoint of each of four quadrants formed by imagined horizontal and vertical lines through the center of the circle. Huttenlocher et al. (1991) suggested that people naturally categorize the space into these four quadrants and that a dot's location is represented not just in terms of fine-grained metric information, but also in terms of the spatial category (i.e., the quadrant) in which it was located. Both types of information are assumed to be nonbiased, and there is evidence suggesting that fine-grained information and categorical information are weighted independently (e.g., Hund & Plumert, 2002). However, bias is introduced during retrieval and reproduction, when the fine-grained item information is combined with the category information. The locational center of the quadrant is assumed to be the central value of the category and, hence, the locational prototype. …

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.