Architects, graphic designers, and others frequently develop designs by picking out and transforming subshapes of two-dimensional or three-dimensional shapes. Shape grammars formalize this aspect of design by specifying rules of the form /to b$: the left-hand side a describes a type of subshape that may be picked out, while the right-hand side b describes what that type of subshape may become. Designs in the language specified by a shape grammar are derived by recursively applying the shape transformation rules to a starting shape. To apply a shape-transformation rule automatically, a computer system, must have the capacity to recognize instances of the type of subshape specified on the left-hand side of the rule. Sometimes such instances are explicitly input by the designer, and explicitly represented in a data structure: in this case, recognition is a relatively straightforward task. But there may also be'emergent'instances that were not explicitly input, and are only indirectly represented in the data structure. These emergent instances are potentially numerous, and may be extremely difficult to discover. This thesis focuses on mechanisms for picking out and transforming subshapes. The first three chapters place the issue in its broadest context by arguing that different designers--bringing different knowledge and attitudes to the task--will pick out and pay attention to different subshapes in a drawing. This contention is supported by introducing some of the relevant literature on perception, problem-solving, and creativity. Chapter 4 introduces shape grammars to provide a more formal framework for investigating this topic. Chapter 5 describes the properties and limitations of Topdown--a computer program which supports design by applying the rules of a shape grammar, but does not provide for recognition of emergent subshapes. Chapter 6 introduces ECART, a computer program which supports efficient recognition and transformation of emergent subshapes, and demonstrates how its performance transcends that of Topdown. Examination of the results produced by ECART suggest that a designer's conceptual filter--the repertoire of subshape types that he or she can recognize in a drawing--plays a crucial role in the development of design ideas.