A Cognitive Technology to Teach Composition Skills: Four Studies with the R-Wise Writing Tutor

Abstract

Four studies examined the effectiveness of user-adaptive computer-aided instruction that explicitly models the cognitive processes of composing for developmental writers, and is integrated with classroom composition instruction. The four school-year studies were designed to build on each other, each informing design improvements to a writing tutor named Reading and Writing in a Supportive Environment (R-WISE) as well as informing subsequent research designs. The first year study ( N = 852) compared traditional classroom controls with R-WISE treatment classrooms. The second year study ( N = 1,151) compared students using R-WISE with students using a word processor. The third year study ( N = 1,277) measured the effect of the teacher's instructional style, and the level of student control over the R-WISE software, on student performance. The fourth year study ( N = 617) replicated previous studies. The treatment groups for the four studies each posted significant gains over controls on most holistic and analytical measures of writing quality. The results of testing multiple design factors of R-WISE demonstrated the efficacy of long-term evaluation and enhancement of user-adaptive writing software in a field-based context.

Word processing programs and weaker writers/readers: a meta-analysis of research findings

“…Robo-writer (Zhang, 1993) was the simplest of the three interventions, and ''knew'' only that students were writing selfselected stories and employed an ''unintelligent'' feedback process for writing stories (i.e., text-to-speech voice synthesis, and vocabulary lists-1,000 word core list; four teacher-generated lists; a first names list; and a user-generated list stored as HyperCard TM stacks). improving Differently, R-WISE (i.e., Reading and Writing in a Supportive Environment; Carlson & Crevoisier, 1994;Rowley et al, 1998) provided ''semi intelligent'' support for generating (planning and outlining), translating and editing essay writing and compelled students to attend to the cohesion of their work.…”

“…Nine of the studies in this review employed whole class (MacArthur et al, 1995) or whole school assignment to treatments, and then examined student-level effects (e.g., Carlson & Miller, 1996;Rowley, Carlson, & Miller, 1998). Adjusting standard errors (SE) for these nine studies was necessary, as a portion of the total variance in such quasi-experiments was likely a consequence of grouping or clustering within treatments (schools or classrooms), with the total variance representing a sum of group and student variances.…”

Tutoring Adolescents in Literacy: A meta-analysis

“…• computer tutoring, if well-structured and informed by theory, can effectively help students that have limited access to face-to-face tutoring (Britt & Aglinskas, 2002;Franzke et al, 2005;Hough et al, 2007;Rowley & Meyer, 2003;Rowley, Carlson & Miller, 1998);…”

“…Although we recognized the differences in some outcome measures within a study, conceptually we treated them as a general construct of literacy (e.g., reading, writing, vocabulary) and used all outcome measures to arrive at one average effect size for each study. Therefore, each study contributed one average effect size (ES) to the subsequent analyses with the following exceptions: (1) Britt and Aglinskas (2002) consisted of three independent sub-studies, so we calculated an effect size for each sub-study, resulting in three effect sizes; (2) Chi et al (2001) included two separate sub-studies, resulting in two effect sizes; (3) Hough et al (2007) compared in-person and webcam tutoring -two different types of tutoring -so we calculated separate effect sizes for each type, resulting in two effect sizes based on pretest-posttest gain scores; Rowley and Meyer (2003) compared three experimental groups with the control group, each involving different amounts of tutoring, so each of the three experimental-control group comparisons produced an effect size; Rowley, Carlson, and Miller (1998) included four independent sub-studies, only two of which had sufficient information to calculate effect sizes.…”

“…The reasons for this trend are not clear. Most of the students at grades 8 and 9 came from two studies (Rowley, Carlson & Miller, 1998;Rowley & Meyer, 2003), which examined computer-based tutoring and produced small effect sizes with very large weights. Otherwise, tutoring seems to be quite effective for adolescents.…”