W. examines some issues in scene description generation and understanding as part of a larger effort to model via computer programs our understanding of the sensory-motor world. The project is based on an integrated vision/language system.