Two experiments dissociated the roles of intrinsic orientation of a shape and participants' study viewpoint in shape recognition. In Experiment 1, participants learned shapes with a rectangular background that was oriented differently from their viewpoint, and then recognized target shapes, which were created by splitting study shapes along different intrinsic axes, at different views. Results showed that recognition was quicker when the study shapes were split along the axis parallel to the orientation of the rectangular background than when they were split along the axis parallel to participants' viewpoint. In Experiment 2, participants learned shapes without the rectangular background. The results showed that recognition was quicker when the study shape was split along the axis parallel to participants' viewpoint. In both experiments, recognition was quicker at the study view than at a novel view. An intrinsic model of object representation and recognition was proposed to explain these findings.
People experience an object visually from a single viewpoint at any given time. In most situations, the viewpoint from which an object is first experienced is different from the viewpoint from which it is later recognized. Recognition must therefore rely on a representation and a process that can accommodate the change of the viewpoints from study to recognition. The models proposed to conceptualize the nature of such mental representation and process can be divided into two categories: structural description models and view-based models (Hayward, 2003).
Structural description models claim that an object is represented as a set of parts with a specific structural description of the spatial relations among the parts, and recognition relies on matching the structure descriptions between the test object and the objects in memory (e.g., Biederman, 1987; Marr & Nishihara, 1978). For example, a hand can be represented as five fingers and one palm, with specific spatial relations among the parts. The view-based models claim that snapshots of an object are represented at all learning viewpoints and that recognition relies on the normalization between the recognition view and the closest snapshot in memory (e.g., Tarr & Pinker, 1989, 1990; Ullman, 1989).
The view-based models have found support in findings that object recognition performance is better at a familiar view than at a novel view, and that performance at a novel view decreases as the angular distance between the novel view and the closest study view increases (e.g., Cooper, 1975; Jolicoeur, 1988; Tarr & Pinker, 1989, 1990). In contrast, the structural description models have found support in findings that object recognition can be viewpoint independent when it relies on distinctive features that are also viewpoint independent (e.g., the number of parts; see Biederman & Gerhardstein, 1993). Recent studies have shown that object recognition can rely on both structure and view information (e.g., Foster & Gilson, 2002).
In this project, we proposed and tested an intrinsic model of object representation and recognition derived from the intrinsic model of spatial memory (Mou, Fan, McNamara, & Owen, 2008; see also Mou & McNamara, 2002; Mou, McNamara, Valiquette, & Rump, 2004), which claims that interobject spatial relations in a layout of objects are represented with respect to the intrinsic orientation1 (intrinsic reference direction) of the layout. The study location of the observer is also represented with respect to the intrinsic orientation of the layout. Recognizing a previously learned configuration requires the observer to (1) identify the intrinsic orientation of the test scene, (2) align the intrinsic orientation of the test scene with the represented intrinsic orientation of the layout, and (3) compare the spatial relations in the test scene with the represented spatial relations of the layout.
Mou et al. …