Key points are not available for this paper at this time.
A question of continuing interest to researchers in writing is what internal characteristics distinguish essays ranked high and low in overall quality. Empirical research at the college level has for the most part taken two approaches to this question, examining errors1 and syntactic features2 while generally ignoring the features of texts that extend across sentence boundaries.3 Neither the error approach nor the syntactic approach has been entirely satisfactory. For example, Elaine Maimon and Barbara Nodine's sentence-combining experiment suggests that, as is true when other skills and processes are learned, certain kinds of errors accompany certain stages in learning to write.4 Because the sources of error in written discourse are often complex and difficult to trace, researchers can conclude little more than what is obvious: low-rated papers usually contain far more errors than high-rated papers. With regard to syntax, Ann Gebhard found that with few exceptions the syntactic features of highand low-rated essays written by college students are not clearly differentiated. Indeed, research in writing quality based on conventions of written English and on theories of syntax, particularly transformational grammar, has not provided specific directions for the teaching of writing. Such results come as no surprise in light of much current research in written discourse. This research-published in such fields as linguistics, cybernetics, anthropology, psychology, and artificial intelligence-addresses questions, concerned with extended discourse rather than with individual sentences, questions about how humans produce and understand discourse units often referred to as texts.5 One such effort that has attracted the attention of researchers in writing is M. A. K. Halliday and Ruqaiya Hasan's Cohesion in English.6 Although Halliday and Hasan do not propose a theory of text structure or examine how humans produce texts, they do attempt to define the concept of text. To them a text is a semantic unit, the parts of
Witte et al. (Fri,) studied this question.