Rule-Governed Approaches to Language and Composition

Abstract

Composition theory generally has ignored grammar over the past 15 years, focusing instead on what has been described as “classifications of texts and relations among writers, readers, and subject matter.” Nevertheless, composition has been and continues to be strongly influenced by the model of language that is implicit in modern grammar. This model proposes that language is rule governed and, as a result, is deterministic. Transformational-generative grammar is the most well-known articulation of the model among composition specialists. This article describes the general features of the model and discusses some of the ways it has influenced composition. After assessing the various weaknesses of the rule-governed model, the article outlines a new model of language that is being developed in cognitive science by David Rumelhart, James McClelland, and others working in parallel distributed processing. This alternative model is associational and probabilistic and is grounded in connectionist theory and research. An association model of language provides composition specialists new perspectives on writers, research, and theory. The article concludes by suggesting possible ways to reconsider the act of composing and related theories.