We can look directly at something but, if we are not paying attention to it, we sometimes remember little about it and perhaps not even notice what is right in front of us. The relationship between attention and perception is well-established, and there is an extensive literature of studies that were conducted to examine this issue using a wide variety of experimental paradigms. A central theoretical theme of my work is that in order to understand the nature of attentional processing and how it is controlled, we must also understand the operations carried out at the "interface" between sensory analysis and attentional analysis.

     My views about these operations are influenced by and consistent with the tri-levelframework of visual perception developed by David Marr, Shimon Ullman, Zenon Pylyshyn, and others. In brief, visual analysis has been described in terms of three levels of processing. Low-level vision is data-driven, parallel,and provides input to higher-level sensory processes. High-level vision is knowledge-dependent, often serial, and often initiated voluntarily in accordance with our perceptual goals. Most intriguing and difficult to study, in my opinion, is intermediate-level vision. It is at this level of analysis, Shimon Ullman has argued, that visual routines may occur. According to the hypothesis, combinations of operations such as visual marking are carried out in response to bottom-up signals from low-level vision and/or top-down signals from high-level vision. For this reason, intermediate-level visual operations are sometimes stimulus-driven and other times goal-driven; and are sometimes carried out in parallel and other times carried out serially. My view is that understanding the nature of operations at this intermediate level may hold the key to understanding attentional processing and its relationship with perception.

back

next