This research is concerned with how people remember spatial and nonspatial information. The project will examine (a) memory for photographs, (b) the metric structure of spatial memories, and (c) constraints on theories of memory and retrieval. A series of eleven experiments will investigate memory for photographs and other two-dimensional projections of three-dimensional spatial layouts. The principal goal of these studies is to determine which three-dimensional properties of a scene are encoded in the mental representation of a photograph of that scene. In addition, these experiments will determine when two- and three- dimensional spatial relations are extracted during picture perception. The results from these studies will bear on current theories of visual perception. Many of these theories assume that spatial relations in depth are inferred from information given in two dimensions (e.g., retinal images). In other words, some three-dimensional spatial relations should not be available until certain two-dimensional spatial relations have been perceived. This prediction will be tested in this research. A second set of six experiments will investigate the metric structure of spatial representations. These experiments are expected to show that spatial memories often violate two fundamental axioms of metric (e.g., Euclidean) representation, segmental additivity and symmetry. The final set of four experiments will examine the range of priming effects. Priming refers to the facilitative effects that responding to an item (e.g., lion) may have on responding to a subsequent item (e.g., tiger). Several theories of memory predict that priming will occur only if items are directly associated (e.g., lion-tiger). Other theories predict that priming will occur even if items are separated by two or three associative steps (e.g., lion-stripes and mane-stripes, respectively). Two experiments will establish limits on the range of priming effects in lexical (word vs. nonword) decisions and in recognition, and two computer simulations will assess the extent to which neural-network models of memory are constrained by the findings.
