The Level of Attention: Mediating Between the Stimulus and Perception

Abstract

Current conceptions of visual processing make good use of the metaphor of levels of vision. At the very least, there are meaningful distinctions to be made between early vision, mid-level vision, and high-level vision. Without getting too committed to the details, early vision is a level of local processing of simple stimulus attributes like the orientation and motion of line segments. Perhaps these can be considered to be the atoms of vision. If so, then mid-level vision is concerned with the molecules-larger pieces put together out of the early vision atoms. Like the "wetness" of water, these mid-level molecules may have properties that are not easy to predict from their early vision precursors. Examples might include Gestalt observations about the whole being greater than the sum of its parts (Kellman, 1998; Rock & Palmer, 1990) or the work of Adelson (refXX-this volume), Gilchrist (refXX-this volume) and others on the apparent brightness of surfaces. At a still higher level, the molecules of mid-level vision give rise to recognizable objects. Once upon a time, perhaps in the first flush of excitement about single-unit recordings in the visual system (Barlow, 1995), we might have thought of these levels in a fairly