Abstract
Schema theory proposes that memory for objects in scenes depends on the degree to which the objects appear to belong to a particular scene (consistent objects) or not (inconsistent objects). However, the degree to which an object will be consistent with a scene is likely to depend on whether the perception of the scene is itself consistent with our previous experience. Thus, for example, if a familiar scene is presented in an unfamiliar manner, the effect of object consistency on memory is likely to change. We wished to see whether this kind of effect can be used to investigate the degree to which a virtual environment (VE) is perceived as “normal”, i.e. consistent with our usual experience of similar environments, or “abnormal”, i.e. markedly different from our previous experience. Two initial studies measured the effect of rendering quality on memory for consistent and inconsistent objects in conditions of varying quality of radiosity (Experiment 1) and polygon count (Experiment 2). Participants interacted with the scenes wearing VGA resolution, head-tracked HMDs. They were then tested for memory for inconsistent and consistent objects. There was little effect of rendering quality except in one condition in which individual objects were hard to recognize. Experiment 3 therefore used a more extreme set of rendering types: wireframe with added color, and full radiosity. The proportion of inconsistent/consistent objects was varied, and object recognition tests ensured that all objects were easily recognized in all conditions. The results showed a significant interaction between rendering type, object type, and consistency ratio. This suggests that inconsistent objects are only preferentially remembered if the scene looks “normal” or if there are many such objects in an “abnormal” scene. We conclude that memory for objects can be used to assess the degree to which the context of a VE appears normal.
Funded by EPSRC grant number GR/S58386/01