Effects of collaboration and isomorphic models on transfer: An L2 English writing investigation

Dennis Mitchell Koyama, Purdue University

Abstract

How can feedback become a productive resource for students? Much of the research investigating the role of feedback in second language (L2) writing has set out to find an answer to this question. In the preparation for future learning (PFL) framework, collaborative interactions engage learners in innovation tasks that push learners to reach beyond their individual dispositions and to build knowledge and understanding through each other’s contributions. Specifically, innovation tasks necessitate the use of prior knowledge in learner attempts to construct, or at a minimum offer, solutions to problems that are unfamiliar or unknown to them (Sears, 2006). A primary objective for having students innovate their initial solution is to prepare them to perceive and appreciate an expert solution. This approach is opposed to a more traditional efficiency model of learning tasks that embroils canonical solutions that can be rote and practice for mastery without generating novel ideas or solutions (Sears). The PFL framework posits that the innovation of ideas and solutions through dyadic interaction, with each other and with materials, can be a valuable catalyst for learning transfer of the deep structures of knowledge that apply across multiple contexts. Evidence of transfer for deep structures within the PFL framework can be seen by assessing learner performance on complex tasks (often called, target transfer tasks) that were not contained in the learning materials of a task, and not a focus of the individual or collaborative tasks (Schwartz et al.). This change in perspective from SPS to PFL has important implications for how language classrooms might design, employ, and measure effective learning tasks. This dissertation explored the usefulness of an expert model and a structured task in an L2 writing classroom. Two interaction levels—individual and collaborative—were examined for their facility of descriptive language related to data integration of graphical information from model feedback in a controlled pre/posttest experiment with international university students enrolled in an L2 English composition course. Two approaches to coding the data were taken. The first approach employed a coding scheme that provided a percentage of content overlap with the expert model—an indicator of factual recall and transfer. This was done by a line-by-line coding scheme (Glaser, 1978). The second approach considered how well the essays “fit” the expected data integrations provided in the model—an indicator of transfer of deep writing structure based on the relative balance of global versus local integrations. This was calculated with a Chi-square test of fit. The transfer of deep structures was further measured through an analysis of if students could identify a data interaction that did not exist in the model description. The results showed that learners in the dyad condition significantly outperformed learners in the individual and control conditions on content overlap and expected data integrations. The dyad condition also surpassed a truth-wins comparison, which provides a comparison of actual dyads to the theoretical pooling of knowledge individuals (Lorge & Solomon, 1955), and dyads were the only condition to include the target transfer item in their posttest revisions, indicating dyads were able to understanding complex data integrations in ways not available to learners in the individual and control conditions. (Abstract shortened by ProQuest.)

Degree

Ph.D.

Advisors

Silva, Purdue University.

Subject Area

English as a Second Language|Education

COinS