Shapes - taken as well-defined collections of lines - are fundamental building blocks in architectural drawings. From doodles to shop drawings, shapes are used to denote ideas and represent elements of design, many of which ultimately translate into actual objects. But because designs evolve, the shapes representing a design are seldom static - instead, they are perpetually open to transformations. And since transformations involve relationships, conventional methods of describing shapes as sets of discrete endpoints may not provide an appropriate foundation for schematic design. This paper begins with a review of the perception of shapes and its significance in design. In particular, it argues that juxtapositions and inter-relationships of shapes are important seedbeds for creative development of designs. It is clear that conventional representation of shapes as sets of discrete lines does not cope with these -emergenti subshapes, the most basic of which arise out of intersecting and colinear lines. Attempts to redress this by using “reduction rules” based on traditional point-and-line data structures are encumbered by computational problems of precision and shape specification. Basically, this means that some “close” cases of sub-shapes may escape detection and their specifications are difficult to use in substitution operations. The paper presents the findings of a computer project - Emergence II - which explored a'relational'description of shapes based on the concept of construction lines. It builds on the notion that architectural shapes are constructed in a graphic context and that, at a basic compositional level, the context can be set by construction lines. Accordingly, the interface enables the delineation of line segments with reference to pre-established construction lines. This results in a simple data structure where the knowledge of shapes is centralized in a lookup table of all its construction lines rather than dispersed in the specifications of line segments. Taking this approach, the prototype software shows the ease and efficiency of applying “reduction rules” for intersection and colinear conditions, and for finding emergent sub-shapes by simply tracking the construction lines delimiting the ends of line segments.