Abstract

Much of cognitive science is based on the Computational Theory of Mind hypothesis. The claim is that the mind is in part a computer and as such requires a representational medium--a language of thought--in which to represent information and to carry out computations. ;But the Computational Theory of Mind is much more than a bland commitment to internal representations. It requires that the system of representation have some very stringent properties. In this dissertation it is demonstrated that, depending on which version of the computational theory one prefers one is necessarily committed to a mechanism with most or all of the following seven properties: causally efficacious syntax, syntactic disjointness, syntactic differentiation, right causal connections, unambiguity, semantic differentiation, and the maintenance criterion. These properties are referred to as the CTM-properties because they are entailed by the Computational Theory of Mind. A subset of these properties apply to symbol systems in general and are referred to as the CTM$\sp\prime$-properties. ;The main thesis advanced is that, while it may be true that in certain cases the system of internal representation can be characterized as a language exhibiting the CTM-properties, there are also genuine cognitive situations where it needs to exhibit very un-CTM-like properties, in particular, properties associated with symbol systems of sketching. This leads to a dilemma which can only be solved by either circumscribing the domain of cognitive processes to be explained by a CTM-type symbol system or by reworking our notions of computation and representation. ;Two problem solving studies involving expert designers are carried out and described. The first study identifies some of the cognitive processes involved in open-ended, ill-structured problem solving situations like design and notes a correlation between certain of these cognitive processes and certain external symbol systems used by designers. Some of these symbol systems possess the CTM$\sp\prime$-properties while others do not. The second study manipulates the external symbol systems subjects are allowed to use along these dimensions. It is predicted that this manipulation will hamper some cognitive functions and not others. The results indicate statistically significant differences along the lines predicted