A category adjustment approach to memory for spatial location in natural scenes.

Abstract

Memories for spatial locations often show systematic errors toward the central value of the surrounding region. This bias has been explained using a Bayesian model in which fine-grained and categorical information are combined (Huttenlocher, Hedges, & Duncan, 1991). However, experiments testing this model have largely used locations contained in simple geometric shapes. Use of this paradigm raises 2 issues. First, do results generalize to the complex natural world? Second, what types of information might be used to segment complex spaces into constituent categories? Experiment 1 addressed the 1st question by showing a bias toward prototypical values in memory for spatial locations in complex natural scenes. Experiment 2 addressed the 2nd question by manipulating the availability of basic visual cues (using color negatives) or of semantic information about the scene (using inverted images). Error patterns suggest that both perceptual and conceptual information are involved in segmentation. The possible neurological foundations of location memory of this kind are discussed.

Location memory for dots in polygons versus cities in regions: Evaluating the category adjustment model.

“…Because of our observed data, we believe that Holden et al's (2010) conclusion that both conceptual and perceptual knowledge was being used to categorize their upright scenes is tme at a very different level than our data are. Theirs was a categorization task; this is a different "level" of conceptual information than what we are talking about when we discuss knowledge about locations and other aspects of cities extant in long-term memory in the present case.…”

“…The main goal was to assess whether individuals tend to recall locations as being closer to the category prototype than the locations actually were. Following Holden et al (2010), rather than examining all the responses with respect to a fixed point across all locations (e.g., the centroid, or center of mass, of the polygons), we examined the magnitude and direction of the error vectors fi-om the correct locations themselves. That is, we converted each participant's response for each dot/city to a vector originating at the associated correct location and ending at the estimated location.…”

“…Further, the resultant vector represents the mean difference angle for a single participant (see Holden et al, 2010, for funher details). Because each panicipant's data represent a mean enor angle across cities, it was appropriate to analyze them using Hotelling's one-sample second-order analysis of angles (Zar, 1996; second-order refers to the averaging across the items).…”

“…For example, there are occasions when the model seems to apply to principally perceptual (recent) memory (e.g., the original Huttenlocher et al, 1991, article and many others) and other occasions when, at least in discussion, the model should apply to location memory more generally, including that which is derived fi-om extant knowledge (e.g.. Holden, Curby, Newcombe, & Shipley, 2010).…”

Bayesian combination of two-dimensional location estimates

“…Extending Bayesian cue combination to spatial tasks is important because location estimation is used in many tasks with both humans (e.g., Friedman & Montello, 2006;Friedman, Montello, & Burte, 2012;Holden, Curby, Newcombe, & Shipley, 2010;Huttenlocher, Hedges, & Duncan, 1991) and other animals (cf. Cheng et al, 2007).…”