LEVELS OF COHESION: DISTINGUISHING BETWEEN TWO GROUPS OF COLLEGE WRITERS (COMPOSITION, ANALYSIS, APPLIED LINGUISTICS)

Cohesion has been defined in terms of grammatical and lexical devices that link sentences together to create a sense of connected discourse. As explicit linguistic markers, cohesive devices have been studied largely as a discourse phenomenon. However, cohesive devices can and do make connections within sentences as well. To date, few studies have investigated the relationship between cohesive devices and judgments of writing quality, and its relationship, in turn, to levels of cohesion. This dissertation attempts to do so by addressing the following questions in two separate studies: (1) Do cohesive devices contribute to the overall judged writing quality of student essays? (2) Do better and poor writers differ in their use of cohesive devices within and across sentences?;The results showed no significant differences in mean ratings between essay versions in the pass group, but there were significant differences in the fail group. The fail essays corrected for cohesive errors alone received lower ratings than the other essay versions. Post hoc analyses on two smaller subsamples showed that correction of cohesive devices still had no effect on rating in the pass essays but that it had a differential effect on the fail essays.;The second study analyzed the cohesion in the original essays by identifying (1) types of cohesive devices and (2) the distance of cohesive tie (one of five levels). A modified version of Halliday and Hasan's cohesion classification (1976) was used to identify cohesive devices, which were coded within sentences and between sentences.;The pass essays exceeded the fail essays in the frequency of one type of cohesive device, determiners, and in the use of cohesion within sentences. The higher incidence of determiners and cohesion within sentences reflected the greater syntactic density of the pass essays. A trade-off hypothesis between syntactic and discourse-based forms of cohesion was proposed to explain the results.;In the first study, 30 college essays written by native speakers of English were systematically corrected for mechanical/grammatical errors unrelated to cohesion and for cohesive errors. These corrected versions, along with the original essays, were then evaluated holistically by trained readers. The holistic ratings were analyzed to determine whether there were differences between groups (pass and fail essays) and essay version.