This study was conducted to compare the writing performance (writing products and writing processes) of the TOEFL iBT integrated writing task (writing from source texts) with that of the TOEFL iBT independent writing task (writing from prompt only). The study aimed to find out whether writing performance varies with task type, essay scores, and academic experience of test takers, thus clarifying the link between the expected scores and the underlying writing abilities being assessed. The data for the quantitative textual analysis of written products was provided by Educational Testing Service (ETS). The data consisted of scored integrated and independent essays produced by 240 test takers. Coh-Metrix (an automated text analysis tool) was used to analyze the linguistic features of the 480 essays. Statistic analysis results revealed the linguistic features of the essays varied with task type and essay scores. However, the study did not find significant impact of the academic experience of the test takers on most of the linguistic features investigated. In analyzing the writing process, 20 English as a second language students participated in think-aloud writing sessions. The writing tasks were the same tasks used in the textual analysis section. The writing processes of the 20 participants was coded for individual writing behaviors and compared across the two writing tasks. The writing behaviors identified were also examined in relation to the essay scores and the academic experience of the participants. Results indicated that the writing behaviors varied with task type but not with the essay scores or the academic experience of the participants in general. Therefore, the results of the study provided empirical evidence showing that the two tasks elicited different writing performance, thus justifying the concurrent use of them on a test. Furthermore, the study also validated the scoring rubrics used in evaluating the writing performance and clarified the score meaning. Implications of the current study were also discussed.

The aim of this thesis is to investigate the effect of the writing prompt on test takers in terms of their test taking processes and the final written product in a second language writing assessment context. The study employs a mixed methods approach, with a quantitative and a qualitative strand. The quantitative study focuses on an analysis of the responses to six different writing prompts, with the responses being analyzed for significant differences in a range of key textual features, such as syntactic complexity, lexical sophistication, fluency and cohesion. The qualitative study incorporates stimulated recall interviews with test takers to learn about the aspects of the writing prompt that can have an effect on test taking processes, such as selecting a prompt, planning a response, and composing a response. The results of the quantitative study indicate that characteristics of the writing prompt (domain, response mode, focus, number of rhetorical cues) have an effect on numerous textual features of the response; for example, fluency, syntactic complexity, lexical sophistication, and cohesion. The qualitative results indicate that similar characteristics of the writing prompt can have an effect on how test takers select a prompt, and that the test time constraint interacts with the prompt characteristics to affect how test takers plan and compose their responses. The topic and the number of rhetorical cues are the prompt characteristics that have the greatest effect on test taking processes. The main conclusion drawn from the study findings are that several prompt characteristics should be controlled if prompts are to be considered equivalent. Without controlling certain prompt characteristics, both test taking processes and the written product will vary as a result of the prompt. The findings raise some serious questions regarding the inferences that may legitimately be drawn from writing scores. The findings provide clear guidance on prompt characteristics that should be controlled to help ensure that prompts present an equivalent challenge and opportunity to test takers to demonstrate their writing proficiency. This thesis makes an original contribution to the second language writing assessment literature in the detailed understanding of the relationships between specific prompt characteristics and textual features of the response.