Detecting high-level and low-level properties in visual images and visual percepts

Abstract

In this article we provide further evidence that visual mental imagery and visual perception share modality-specific mechanisms, and we find that representing visual information in a mental image (activating stored information to create a picture-like mental representation) preserves relatively low-level visual detail. Subjects either saw or visualized simple pictures, and evaluated them for the presence or absence of six types of non-accidental properties. These properties varied from very `low-level' ones, such as T junctions, to very `high-level' ones, such as global symmetry. The question was whether both sorts of information are equally accessible in percepts and mental images. If mental images are equivalent to descriptions of perceptual units and their organization, as some have argued, then subjects should have greater difficulty accessing low-level properties in a mental image compared to the difficulty they experience when the drawing is visible. The results of two experiments were clearcut: Subjects could evaluate high-level properties more easily than low-level ones, but this difference was the same in imagery and perception. These findings suggest that mental images preserve relatively low-level visual features, and are not simply descriptions of a pattern.