Abstract
Most objects that we encounter can be decomposed into parts, arranged according to a representation of the spatial relationships among those parts (Biederman, 1987). Our subjective impression is that we immediately and accurately grasp spatial relations when we identify an object. We tested whether these relations are constructed after the individual parts are recognized, by asking participants to identify the parts within an object, or their relations, in a rapidly presented display. We briefly flashed (50 ms) a display containing four squares (one per quadrant), each split either horizontally or vertically into two colors. Participants were given a 2-alternative forced choice that tested their encoding of one of three properties: the colors in the cued square, the direction of its split, or the spatial relationship among those colors. The two alternatives were always correct for the two non-tested properties, and participants always knew the relevant property in a given block. In the color block, the foil shared only one of the two colors with the target, encouraging participants to encode the colors of the target. In an orientation block, the foil was split in a different orientation from the target, encouraging participants to encode the orientation of the split within the square. In the spatial relation block, the foil was different from the target only in terms of which color was presented in which half of the target, encouraging participant to encode the spatial relation between the colors. The results showed a relatively high accuracy in the color feature (M = 84%) and the spatial structure (M = 85%) blocks, but a sharp drop in accuracy for the spatial relation block (M = 68%), suggesting that constructing spatial relationship representations requires that object parts enter a time-consuming binding process.