The paper by S. Ullman explores the computational interpretation of structure from motion, focusing on how 3-D structure and motion can be inferred from 2-D transformations of projected images when no 3-D information is conveyed by individual projections. Ullman proposes a scheme that divides the image into groups of four elements, tests each group for rigid interpretation, and combines the results. This scheme is shown to correctly decompose scenes containing rigid objects in motion, recovering their 3-D structure and motion. The analysis is based on the "structure from motion" theorem, which states that the structure of four non-coplanar points can be recovered from three orthographic projections. The interpretation scheme is extended to cover perspective projections, and its psychological relevance is discussed. The paper also reviews experimental findings that support the use of principles similar to those underlying the structure-from-motion interpretation scheme in human visual perception.The paper by S. Ullman explores the computational interpretation of structure from motion, focusing on how 3-D structure and motion can be inferred from 2-D transformations of projected images when no 3-D information is conveyed by individual projections. Ullman proposes a scheme that divides the image into groups of four elements, tests each group for rigid interpretation, and combines the results. This scheme is shown to correctly decompose scenes containing rigid objects in motion, recovering their 3-D structure and motion. The analysis is based on the "structure from motion" theorem, which states that the structure of four non-coplanar points can be recovered from three orthographic projections. The interpretation scheme is extended to cover perspective projections, and its psychological relevance is discussed. The paper also reviews experimental findings that support the use of principles similar to those underlying the structure-from-motion interpretation scheme in human visual perception.