The above internal representational scheme owes much to two prior lines of research. The first involves the form of the representation. As I alluded to earlier, my approach is a modification of the classic idea of a frame notation stored as a semantic network (e.g., [#!ai:quillian1!#,#!ai:minsky1!#]). These ideas were further elaborated by researchers such as Wilensky (knowledge:wilensky1) and Sowa (knowledge:sowa1), who theorized about the nature of the underlying representation. That is why my representational scheme is ``fully instantiated'' (all concepts are frame entities in the system; nothing is represented as a simple internal symbol). This, in turn, gives a great deal of flexibility to the knowledge system. As I will discuss in Chapter 6, a creative model needs this flexibility in order to be able to dynamically create concepts, which is a core aspect of my theory.
The other similar line of research with respect to my knowledge representation scheme is the large amount of work done on the topic of functional representations (e.g., [#!functional:sembugamoorthy-chandrasekaran-1986!#]). This line of inquiry concerns the content of the representations--what is needed to best describe an object such that it may be powerfully reasoned about? A popular approach is to utilize a structure-behavior-function (SBF) content for concepts (e.g., see [#!sbf:goel-1989!#] for an early discussion of the approach and [#!sbf:goel-bhatta-stroulia-1996!#] for a more recent version of the paradigm). This relates the structure of a device to certain core behaviors which it may perform; these combine into order for the device to achieve the function it is designed to accomplish. This is reminiscent of my need to describe how concepts achieve their functions--what I referred to earlier as the explanation which relies on the primary attributes. SBF conceptual models, however, have largely been restricted to artifacts. My theory extends the idea of explaining function to all types of concepts within the knowledge system. This is necessary as my function-driven morphological synthesis (see Section 6.2.3) depends on being able to access the function of any concept, but this also means that I cannot always rely on structure to provide part of the explanation as structure may not exist for non-physical objects. For example, it is unwieldy to attempt to describe an action in an SBF framework.