Abstract

Cognitive scientists generally subscribe to an information-processing model of mind and implement this model through computational methods. Information processing is understood to be generation and composition of informational primitives into more complex pieces of information in response to signals extracted from the environment. Computationalism offers a powerful methodology for carrying out information processing, with abstract tokens standing in for pieces of information, and new information structures being created through application of rules. This purely syntactic model of cognition is unable, however, to explain the nature of semantic information. Modifications of Shannon information theory and applications of principles of natural selection fail to provide a non-syntactic account of the nature and origin of semantic information. Purely syntactic explanations of semantic information fail to capture the representational capabilities of true cognitive agents. Relying on computational explanations of cognitive behavior leads not only to an explanatory gap, but also to practical failings. These failings constitute the specific and general frame problems. To avoid these problems, a model of cognition that focuses on the adaptive capabilities of its realizing hardware must be adopted. Evolutionary and ecological models only provide pieces of a final theory. A high-level model for cognition can be found in the field of complex and self-organizing systems, a branch of dynamic systems theory. One possible characterization of the lower-level mechanisms responsible for the high-level behavior is given by Edelman's Theory of Neuronal Group Selection. What emerges is a theory with strong similarities to behaviorism and teleological functionalism, although without the deficiencies of either theory. This theory offers the foundation for a new understanding of the nature of representation and information