UBC Theses and Dissertations
English teachers’ perceptions of coherence Foster, Daniel George
Freshman essays written in an exam situation and holistically scored for coherence by teachers of English 100, untrained for this particular project, were analysed for a variety of features. The purpose of this analysis was to discover what features these readers were in fact responding to when asked to respond to coherence problems. The lack of training resulted in widely divergent holistic scores (the inter-rater reliability coefficient was .52), indicating that the readers were not responding to the same aspects of the essays. Because of this divergence, a wide variety of features were tested for, including some unlikely to be directly related to coherence. The following seventeen features were tested: stylistic features—words per T-unit, words per clause, clauses per T-unit, percentage of words in final free modifiers, and percentage of T-units containing final free modifiers; discourse theory and semantic features—ratio of inferred to explicit propositions, violations of the given/new contract per hundred words, deictic ties per hundred words, ratio of specific to general nouns, semantically anomalous sentences per essay, and organization score; traditional composition handbook concerns—occurrences of "to be" as main verb, passive constructions per T-unit, and percentage of total words in nominal constructions; physical features—-words per essay, paragraphs per essay, and words per paragraph. These features were correlated with the holistic scores by computer. The results of this analysis indicate that, for the most part, the readers did not respond to features which might be expected to relate directly to coherence but rather to nominalization rate and clause length. The lack of agreement found among readers further indicates the lack of a strongly shared, implicit notion of what constitutes coherence.
Item Citations and Data