D. Marr's article discusses the theory of early visual information processing, focusing on the computation of a primal sketch. The primal sketch is a primitive but rich description of intensity changes in an image, expressed in a vocabulary of intensity change types such as EDGE, SHADING-EDGE, EXTENDED-EDGE, LINE, and BLOB. This description is derived from the intensity array using fixed techniques and is called the primal sketch. The primal sketch is then grouped in a way appropriate for later recognition. The article argues that symbolic processes in the perceptual system can define 'place-tokens' in an image and group them according to certain rules. Homomorphic techniques fail to account for many of these grouping phenomena, which require mechanisms of construction rather than detection. The necessary grouping of elements in the primal sketch may be achieved by mechanisms that use processes inferred from above and the ability to select items by first-order discriminations acting on the elements' parameters. These mechanisms rarely use downward-flowing information about the contents of the image being processed. The article concludes that 'non-attentive' vision is implemented by these grouping operations and first-order discriminations acting on the primal sketch. The class of computations so obtained differs slightly from the class of second-order operations on the intensity array. The extraction of a form from the primal sketch using these techniques amounts to the separation of figure from ground. The article argues that most of the separation can be carried out using techniques that do not depend on the particular image in question. Therefore, figure-ground separation can normally precede the description of the shape of the extracted form. Higher-level knowledge and purpose are brought to bear on only a few of the decisions taken during the processing. This relegates the widespread use of downward-flowing information to a later stage than is found in current machine-vision programs, implying that such knowledge should influence the control of, rather than interfere with, the actual data-processing that is taking place lower down. The article also discusses the extension of the method to two dimensions and the combining of orientation-dependent descriptions. It concludes that the primal sketch is a fundamental step in visual processing, and that the grouping of elements in the primal sketch is essential for later recognition. The article also discusses the importance of place-tokens in visual processing and how they can be defined and grouped in various ways. The article argues that the grouping processes responsible for place-tokens read and write into the same storage, and that the way a place-token is defined is separate from the way it is grouped, illustrating the principle of modular design. The article also discusses the importance of orientation, intensity, fuzziness, and various measures of the size of an item in the image, as well as spatial proximity and collinearity in grouping processes.D. Marr's article discusses the theory of early visual information processing, focusing on the computation of a primal sketch. The primal sketch is a primitive but rich description of intensity changes in an image, expressed in a vocabulary of intensity change types such as EDGE, SHADING-EDGE, EXTENDED-EDGE, LINE, and BLOB. This description is derived from the intensity array using fixed techniques and is called the primal sketch. The primal sketch is then grouped in a way appropriate for later recognition. The article argues that symbolic processes in the perceptual system can define 'place-tokens' in an image and group them according to certain rules. Homomorphic techniques fail to account for many of these grouping phenomena, which require mechanisms of construction rather than detection. The necessary grouping of elements in the primal sketch may be achieved by mechanisms that use processes inferred from above and the ability to select items by first-order discriminations acting on the elements' parameters. These mechanisms rarely use downward-flowing information about the contents of the image being processed. The article concludes that 'non-attentive' vision is implemented by these grouping operations and first-order discriminations acting on the primal sketch. The class of computations so obtained differs slightly from the class of second-order operations on the intensity array. The extraction of a form from the primal sketch using these techniques amounts to the separation of figure from ground. The article argues that most of the separation can be carried out using techniques that do not depend on the particular image in question. Therefore, figure-ground separation can normally precede the description of the shape of the extracted form. Higher-level knowledge and purpose are brought to bear on only a few of the decisions taken during the processing. This relegates the widespread use of downward-flowing information to a later stage than is found in current machine-vision programs, implying that such knowledge should influence the control of, rather than interfere with, the actual data-processing that is taking place lower down. The article also discusses the extension of the method to two dimensions and the combining of orientation-dependent descriptions. It concludes that the primal sketch is a fundamental step in visual processing, and that the grouping of elements in the primal sketch is essential for later recognition. The article also discusses the importance of place-tokens in visual processing and how they can be defined and grouped in various ways. The article argues that the grouping processes responsible for place-tokens read and write into the same storage, and that the way a place-token is defined is separate from the way it is grouped, illustrating the principle of modular design. The article also discusses the importance of orientation, intensity, fuzziness, and various measures of the size of an item in the image, as well as spatial proximity and collinearity in grouping processes.