Collaborative Writing in a Computer-Supported Classroom: Mediation, and Self-Assessed Beliefs and Attitudes About Writing

Collaborative Writing in a Computer-Supported Classroom: Mediation, and Self-Assessed Beliefs and Attitudes about Writing

Meng-ying Daphne Lin

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Curriculum, Teaching and Learning Ontario Institute for Studies in Education University of Toronto

Collaborative writing in a computer-supported classroom: Mediation and self-assessed beliefs and attitudes about writing

Doctor of Philosophy, 2015 Meng-ying Daphne Lin Department of Curriculum, Teaching and Learning University of Toronto

ABSTRACT

This classroom study investigated how computer-based writing tools and collaborative writing tasks mediated the development of individual writing competencies among 35 graduate students during a 4-week course in Taiwan focused on English tests like the TOEFL. I prepared a unique platform of writing tools for the course and then quantified the types of writing knowledge generated from students' collaborative writing (CW) and uses of these mediation tools, then associated these with the students’ CW performance and then individual writing performance on 3 drafts of 4 TOEFL writing tasks over the duration of the course. Data included collaborative writers’ dialogues and online browsing histories and all compositions produced during the course. Supplementary data were questionnaires and interviews on students’ perceived changes in communications, reflections on CW, and self-assessed beliefs about their writing abilities. A comparison class of 26 students, which I also taught a year later, served as a control condition with the same course content, and writing tasks written individually (rather than collaboratively).

Analyses identified symbolic and material mediation tools in Language-Related

Episodes (LREs) in CW writers’ dialogues and uses of Internet writing resources. Generalized

Estimating Equation (GEE) modeling then tested the relationship between tool-mediated ii

writing knowledge and CW performance, and then correlation analyses evaluated the relationship between CW performance and students’ individual writing development. GEE modeling and interpretive analyses investigated over time the students’ changes in communications, reflections on CW, and self-assessed beliefs about their writing abilities.

Students in the CW course improved their English writing significantly more than comparison students did. Their symbolic and material tool use resulted in two types of writing knowledge, linguistic forms and skills of expression, which together with practice writing collaboratively, were associated with the students’ writing development. The CW students expressed positive views towards the mediation tools, their abilities for verbalizing in English, and various cognitive and attitudinal indicators, whereas the control group did not change their beliefs about writing over time. The present results strengthen the claims from Sociocultural

Theory about the value of mediation for writing development in a foreign language and provide insights about CW for writing teachers.

iii

ACKNOWLEDGEMENTS

I gratefully acknowledge that the dissertation could not be done without the help of many selfless contributors.

I would like to show my deepest appreciation to my supervisor Professor Alister Cumming for his thoughtful mentorship in my six years of doctoral study at OISE in the University of

Toronto. Professor Cumming always provided insightful feedback and provoking ideas when I brought my questions to him. His providing immediate attendance for my academic needs made the earning of this degree possible.

My special thanks are due to my committee members, Professors Jim Cummins and

Eunice Jang, for their valuable advice to polish this dissertation even more before my oral examination. Also, I would like to thank Professors Nina Spada and Charlene Polio for their helpful discussions of several issues in this thesis.

I would like to show my appreciation to three raters: Kate Wu, Blossom Wang, and

Stephanie Lin for their willingness to offer time in marking compositions while they were doing their graduate studies. Thanks are also due to Masa Hsu and his crew of programmers for the developing and modifying the English Writing platform.

I am indebted to church members in Holy Word Church for their prayers to keep me strong in faith in seeking academic growth. With their help, I was able to adopt myself to the different environment from where I grew up. I wish to thank the School of Graduate Studies of the

University of Toronto and Ministry of Education in Taiwan for their generous financial assistance.

Lastly, but most importantly, I would like to show my deep gratitude to my parents for their encouragement and everlasting love to make this trip to Canada possible.

TABLE OF CONTENTS

ABSTRACT ...... ii ACKNOWLEDGEMENTS ...... iv LIST OF TABLES ...... viii LIST OF FIGURES ...... xi LIST OF APPENDICES ...... xii CHAPTER ONE ...... 1 INTRODUCTION ...... 1 Rationale and research questions ...... 1 Significance of the study ...... 4 CHAPTER TWO ...... 6 THEORETICAL FRAMEWORK AND EXPLANATORY THEORIES ...... 6 Activity Theory as a descriptive framework ...... 6 Theoretical framework applied to the present study ...... 8 Sociocultural Theory as an explanatory theory ...... 11 Mediation ...... 11 ZPD ...... 12 Imitation ...... 14 CHAPTER THREE ...... 15 METHOD ...... 15 Research context and program objectives ...... 15 The development of a writing platform ...... 21 Structure and function of English Writing platform ...... 24 Collaborative writers’ mediation ...... 27 Overview of the Research Design ...... 29 Data collection schedule ...... 29 Data collection and analysis for Question Set One ...... 31 Research Question 1.1 ...... 31 Research Question 1.2 ...... 35 Rating procedures and reliability ...... 36 Generalized Estimating Equations (GEE) modeling ...... 37 Choices of the correlation structure and model selection for GEE ...... 39 Modeling procedures ...... 40 Research Question 1.3 ...... 40 Data collection and analysis methods for Question Set Two ...... 43 Research Question 2.1 ...... 43 Research Question 2.2 ...... 44 Research Question 2.3 ...... 45 CHAPTER FOUR ...... 48 RESULTS ...... 48 Research Question 1.1 ...... 48 Dialogue on language forms ...... 48 Dialogue on skills of expression ...... 49 Online resource use on language forms ...... 50 Online resource use on skills of expression ...... 51 Research Question 1.2 ...... 53 v

Inter-rater reliability...... 54 Descriptive analysis of CW performance ...... 58 The association of mediation and weeks with students’ development ...... 59 On holistic writing performance ...... 59 On communicative quality ...... 60 On organization ...... 61 On argumentation ...... 62 On linguistic accuracy ...... 63 On linguistic appropriacy ...... 65 Observations ...... 65 Grouping effects ...... 67 Research Question 1.3 ...... 68 Descriptive analysis of collaborative and comparison writers’ performance in their before- and after-program tests ...... 69 Were the two writing groups comparable? ...... 71 Differences in collaborative writers and comparison writers’ over-time development ...... 72 Correlations between collaborative writing changes and individual writing development ...... 74 Research Question 2.1 ...... 75 What worked well? ...... 75 What did not work well?...... 77 Collaborative writers’ communications and relationships ...... 77 Research Question 2.2 ...... 81 Mediation and objects ...... 81 Objects to outcomes ...... 82 Idea expansion and verbalizing and wording in English ...... 82 Transfer of collaborative to individual writing ...... 83 Ideas broadened from alternative perspectives ...... 83 Other contributions of CW ...... 84 Division of labour ...... 84 Community ...... 85 Rules: Advantages ...... 86 Rules: Limitations ...... 86 Rules: Improvements ...... 87 Research Question 2.3 ...... 88 Descriptive analysis of students’ self-assessed beliefs about their writing abilities .. 88 Between-group contrasts in self-assessed beliefs about writing abilities ...... 90 Within-group contrasts in self-assessed beliefs about writing abilities ...... 91 CHAPTER FIVE ...... 94 DISCUSSION AND IMPLICATIONS ...... 94 The complexity of modeling process-product writing development ...... 98 L2 learners' development of self-assessed beliefs about their writing abilities ...... 100 Explanations of student writers’ development ...... 102 Mediation tool use towards writing goals ...... 103 Mediation-activated cognitive processes and cognitive distribution ...... 104 Collective scaffolding and group ZPD during CW ...... 105 vi

L1 mediated L2 writing processes ...... 107 L2 writing communities and imitation through repeated writing ...... 107 Implications ...... 109 Research implications ...... 109 Pedagogical implications ...... 112 Limitations of the study and suggestions for future studies ...... 113 REFERENCES ...... 117

vii

LIST OF TABLES

Table 2.1 ...... 10 Research Issues and How They Fit the Current Theoretical Framework ...... 10 Table 3.1 ...... 17 Participants in Two Classes in 2011 and 2012 ...... 17 Table 3. 2 ...... 19 Writing Task Schedule for Students in the Collaborative Writing Classroom ...... 19 Table 3.3 ...... 30 Data Collection Schedule ...... 30 Table 3.4 ...... 42 Data Collection and Data Analysis Methods for Research Questions 1.1 to 1.3 ...... 42 Table 3. 5 ...... 46 Subdivisions of Students' Self-Assessed beliefs about Their Own Writing Abilities (following Wey, 1998) ...... 46 Table 3. 6 ...... 47 Data Collection and Data Analysis Methods for Research Question Set 2 ...... 47 Table 4. 1 ...... 53 Number of Groups, and Their Frequencies of Making Reference to Tool-Mediated Writing Knowledge on Language Forms and Skills of Expression per Week ...... 53 Table 4. 2 ...... 55 Rater Katie and Rater Yu’s Inter-Rater Reliability Scoring 107 Collaboratively Written Compositions ...... 55 Table 4. 3 ...... 56 Rater Katie and Rater Brenda’s Inter-Rater Reliability Scoring 317 Collaboratively Written Compositions ...... 56 Table 4. 4 ...... 57 Rater Katie and Rater Brenda’s Inter-Rater Reliability Scoring 156 Comparison Written Compositions ...... 57 Table 4. 5 ...... 58 Means and Standard Deviations of Collaboratively Written Compositions in Weeks 1 and 2 .. 58 Table 4. 6 ...... 58 Means and Standard Deviations of Collaboratively Written Compositions in Weeks 3 and 4 .. 58 Table 4. 7 ...... 59 Parameter Estimates for the Model Effect of Mediation and Time on Holistic Scores of CW .. 59 Table 4. 8 ...... 61 Parameter Estimates for the Model Effect of Mediation and Time on Communicative Quality of CW ...... 61 Table 4. 9 ...... 62 Parameter Estimates for the Model Effect of Mediation and Time on Organization of CW ..... 62 Table 4. 10 ...... 63 Parameter Estimates for the Model Effect of Mediation and Time on Argumentation of CW .. 63 Table 4. 11 ...... 64 Parameter Estimates for the Model Effect of Mediation and Time on Linguistic Accuracy of CW ...... 64 Table 4. 12 ...... 65 viii

Parameter Estimates for the Model Effect of Mediation and Time on Linguistic Appropriacy of CW ...... 65 Table 4. 13 ...... 70 Collaborative Writers’ Means and Standard Deviations on Their Before and After Program Tests ...... 70 Table 4. 14 ...... 70 Comparison Writers’ Means and Standard Deviations on Their Before and After Program Tests ...... 70 Table 4. 15 ...... 72 GEE Analysis of Collaborative and Comparison Writers’ Before and After-Program Performance on Each Aspect of Writing ...... 72 Table 4. 16 ...... 73 GEE Analysis of Collaborative Writers’ Within-Group Development ...... 73 Table 4. 17 ...... 74 GEE Analysis of Comparison Writers’ Within-Group Development ...... 74 Table 4. 18 ...... 75 Correlations between CW and Individual Writing Changes among Writing Dimensions ...... 75 Table 4. 19 ...... 76 Categorized Answers to Question One on the After-CW Questionnaire: What Worked Really Well in This Task? ...... 76 Table 4. 20 ...... 77 Categorized Answers to Question Two: What Did Not Work Well in This Task? ...... 77 Table 4. 21 ...... 78 Means and Standard Deviations of Self-ratings of Collaborative Writers’ Communications and Relationships in Weeks Two, Three, and Four...... 78 Table 4. 22 ...... 80 Model Effect of Time on Self-ratings of Collaborative Writers’ Communications and Relationships...... 80 Table 4. 23 ...... 89 Means and Standard Deviations of Collaborative Writers’ Self-Assessed Beliefs about Their Writing Abilities ...... 89 Table 4. 24 ...... 89 Means and Standard Deviations of Comparison Writers’ Self-Assessed Beliefs about Their Writing Abilities ...... 89 Table 4. 25 ...... 91 GEE Modeling of Collaborative and Comparison Writers’ Self-assessed beliefs about their Writing Abilities ...... 91 Table 4. 26 ...... 92 GEE Modeling of Collaborative Writers’ Over Time Performance on Self-Assessed Beliefs about Writing Abilities ...... 92 Table 4. 27 ...... 93 GEE Modeling of Comparison Writers’ Over Time Performance on Self-Assessed beliefs about Writing Abilities ...... 93 Table J 1 ...... 139 Correlation Coefficients for Holistic Scores over 4 Weeks...... 139 Table J 2 ...... 139 Correlation Coefficients for Communicative Quality Scores over 4 Weeks ...... 139 ix

Table J 3 ...... 139 Correlation Coefficients for Organization Scores over 4 Weeks ...... 139 Table J 4 ...... 139 Correlation Coefficients for Argumentation Scores over 4 Weeks ...... 139 Table J 5 ...... 139 Correlation Coefficients for Linguistic Accuracy Scores over 4 Weeks ...... 139 Table J 6 ...... 140 Correlation Coefficients for Linguistic Appropriacy Scores over 4 Weeks ...... 140 Table K1 ...... 146 Complete Categorized Answers to Question One in the After-CW Questionnaire: What Worked Really Well in This Task? ...... 146 Table K2 ...... 147 Complete Categorized Answers to Question Two: What Did Not Work Well in This Task? . 147 Table L 1 ...... 149 Writing Group Interview in Chinese (English translation): Group a and b ...... 149 Table L 2 ...... 151 Writing Group Interview: Group c and d ...... 151 Table L 3 ...... 152 Writing Group Interview: Group e and f ...... 152 Table L 4 ...... 154 Writing Group Interview: Group g and h ...... 154 Table L 5 ...... 155 Writing Group Interview: Group i and j ...... 155 Table L 6 ...... 156 Writing Group Interview: Group k and l ...... 156 Table L 7 ...... 157 Writing Group Interview: Group m and n ...... 157 Table L 8 ...... 159 Writing Group Interview: Group o and p ...... 159 Table M 1 ...... 160 Collaborative Writing Scores in Week 1 and Week 2 ...... 160 Table M 2 ...... 161 Collaborative Writing Scores in Week 3 and Week 4 ...... 161

LIST OF FIGURES

Figure 2. 1. The structure of human activity systems (Engeström, 1987)...... 6 Figure 2. 2. The current CW system...... 8 Figure 3. 1. English writing platform: students’ entry to write...... 25 Figure 3. 2. English writing platform: instructors’ entry to administer writing tasks and questionnaire survey...... 26 Figure 3. 3. English writing platform: instructors’ entry to retrieve writing processes and artifacts...... 27 Figure 4. 1. A student’s consulting language forms online...... 51 Figure 4. 2. Holistic scores of group l and group n that always used mediation tools...... 68 Figure 4. 3. Holistic scores of group c and group p that not always used mediation tools...... 68

LIST OF APPENDICES

Appendix A ...... 129 Independent Writing Tasks ...... 129 Appendix B ...... 130 Transcription Conventions...... 130 (Excerpts were translated from an original Mandarin version) ...... 130 Appendix C ...... 131 Questionnaire on Mediation Tool Use ...... 131 (translated from an original Mandarin version) ...... 131 Appendix D ...... 133 Coding Table of Writing Groups’ References to ...... 133 Tool-mediated Writing Knowledge ...... 133 Appendix E ...... 135 Communicative Writing Rubrics ...... 135 Appendix F ...... 137 Holistic Rating Scales ...... 137 Appendix G ...... 139 Correlation Coefficients for Six Outcome Variables over Four Weeks ...... 139 Appendix H ...... 141 Post-writing Questionnaire ...... 141 Appendix I ...... 143 Protocols for Group Interviews ...... 143 (translated from an original Mandarin version) ...... 143 Appendix J ...... 144 Questionnaire on Self-assessed Beliefs about Writing Abilities ...... 144 (translated to a Mandarin version for the present research) ...... 144 Appendix K ...... 146 Complete Categorized Answers to Open-ended Questions ...... 146 in the After-CW Questionnaire ...... 146 (translated from an original Mandarin version) ...... 146 Appendix L ...... 149 Group Interviews ...... 149 (translated from an original Mandarin version) ...... 149 Appendix M ...... 160 Collaborative Writing Scores from Week 1 to Week 4 ...... 160

xii

CHAPTER ONE

INTRODUCTION

Rationale and research questions

The factors that mediate second-language (L2) writing performance are various and complicated, including L1 writing expertise, L2 linguistic knowledge, genre, time constraints, writing instruction and assessment, and L2 writing contexts (Leki, Cumming, & Silva, 2008).

The complex nature of writing has meant that writing instruction is relatively challenging and laborious. As a consequence, collaborative writing (CW) has become a popular instructional activity in language classrooms for sharing the burden of responsibility among students in addition to instructors or as an instructional technique itself to help L2 learners improve their writing and language acquisition (Alexander, 2009; Dobao, 1994, 2012; Fry & Griffin, 2010;

Mak & Coniam, 2008; Ranker, 2009, Storch, 1998, 2005; Wigglesworth & Storch, 2009).

Regarding the adoption of CW, a form of mediated-writing activity in classrooms, considerable research has focused on such related issues as L1 or target language use in collaborative dialogue (Storch & Aldosari, 2010; Swain & Lapkin, 2000), comparing group writing with individual writing (Wigglesworth & Storch, 2009), or the social aspects of language learning in collaborative tasks (Donato & McCormick, 1994; Storch, 2002; Swain &

Lapkin, 1998). Dobao (2012) indicated that little of the literature on CW has investigated the nature of mediation processes and their relationship with writing performance or development.

The goal of the present thesis study was to investigate whether the development of individual student’s writing abilities and self-assessed beliefs about writing (i.e., learning goal orientations, performance goal orientations, metacognition, self-efficacy, and effort) are built up through students’ uses of mediation tools during multiple CW tasks and students’ interactions in a

2 face-to-face CW classroom that involved consulting writing resources online. Mediation tools, according to Vygotsky (1987a), refer to physical and psychological tools that facilitate language learners’ mental development. The definitions of mediation tool use in this study refer to the use of online resources (physical tools) and collaborative dialogues (psychological tools) while collaborative writers are solving writing problems. The special focusi of the study were to investigate how writers use mediation tools, to describe what tool-mediated writing knowledge is produced during CW to solve writing problems, and to evaluate the effect of applying such writing knowledge on writing performance over time. In general, this study investigated the relationships between mediation, interaction, L2 writers’ development, and their self-assessed beliefs about writing.

In order to document and analyze the CW activity in this study, a computer-based CW system was created based on Activity Theory (AT) to account for the interrelations of every contextual component of CW in a computer-supported classroom (i.e., participants, tools, objects, outcomes, division of labour, community, and rules). The creation of the CW system derived from my observation of language teaching and learning in countries where English is a

Foreign Language (EFL). EFL teaching engages in qualitatively different practices from pedagogies for English L1 or L2 students in English-dominant countries such as Canada, the

U.S., or Australia. Most writing courses in EFL countries are offered as one of several courses within a broader foreign-language curriculum. In these circumstances, rules, constraints, and goals from institutions and educational programs, learners’ perceptions of their learning tasks, and their beliefs about their language abilities all influence students’ learning outcomes. A researcher cannot afford to overlook these fundamental aspects of the learning of L2 writing. I have chosen to examine them from an AT perspective that enables researchers to analyze human behavior with an “activity-based, context-sensitive approach” (Cole, 1996, p. 340).

I further interpret the research results through Sociocultural Theory (SCT). It guides empirical studies on collaborative language learning with theoretical accounts of how mediation contributes to learning improvement in relation to several linked components, such as the Zone of Proximal Development (ZPD), scaffolding, and imitation. I undertook the study with multiple data sources, qualitative data serving as supplementary to help explain quantitative results.

For the present research, I posed questions about and investigated collaborative writers’ uses of mediation tools in their composing processes, writing performance, communication relationships attitudes towards integrating mediation tools in a writing classroom, and self-assessed beliefs about writing abilities. My specific research questions were:

1. Issues about collaborative writers’ use of mediation tools and its relation to their writing performance.

1.1 What writing knowledge was generated through the mediation of collaborative writers’ dialogue and their consulting online writing resources?

1.2 How did the tool-mediated writing knowledge influence CW performance over time?

1.3 Did writers’ CW performance influence their individual writing performance?

2. Issues about collaborative writers’ communication relationships, their attitudes to the mediated writing, and self-assessed beliefs about their writing.

2.1 For collaborative writers, what aspects of CW worked well and what didn’t, and how did they change their communication relationships over time?

2.2 What did collaborative writers perceive about the mediated writing in the writing classroom?

2.3 Did the mediated writing make a difference in students’ self-assessed beliefs about their writing abilities?

The study was undertaken in a class that I taught where CW was practiced weekly over a

4 month with the hope of strengthening L2 students’ individual writing abilities. A similar research design on CW had been proposed by Lauer and Hendrix (2009), who assessed group discussions and repeated writing assignments with the intent to measure writing improvement, wherein students formulated a written answer, discussed it with their peers, and revised their drafts again. However, Lauer and Hendrix concluded that accurate and powerful statistical techniques would have to be applied to such repeated-measurement research. To respond to that call in the present study I adopted an advanced statistical technique, Generalized Estimating

Equation (GEE) analyses, to examine students’ writing development over time.

Significance of the study

The study is significant for several reasons. First, despite the persuasive results of CW research under the influences of Sociocultural Theory and its pedagogical implications, it has been unclear what types of writing knowledge students have employed to solve writing problems collaboratively when using mediation tools (e.g., verbal exchanges and online writing resources). The writing knowledge that collaborative writers consult reflect what tool-mediated action is really concerned about in view of learning and putting knowledge into practice to benefit students’ writing development (Vygotsky, 1987b). In addition, most prior research on

CW has focused on one type of mediation tool, verbal exchanges only. Second, the present study differs from prior studies in attempting not only to quantify and examine the direct contribution of tool-mediated writing knowledge to CW performance but also to associate students’ CW performance with their individual writing performance and development. Most prior CW research examined either the former or the latter only and so little research has explored the relationship between CW performance and students’ individual writing development. Third, the present study employed multiple data collection methods, such as

5 recording students’ dialogues and online browsing logs and administering questionnaires, permitting the triangulation of different, complementary data to explain changes in students’ writing, interactions, and beliefs. Last, the basis of the research in Activity Theory provides a theoretical framework to explain students’ writing development associated with the mediation of tool use.

CHAPTER TWO

THEORETICAL FRAMEWORK AND EXPLANATORY THEORIES

This chapter presents the theoretical background to the study. I review ideas and relevant publications about Activity Theory and Sociocultural Theory. These theories serve as a descriptive framework for this thesis’ research context (Activity Theory) and to explain the research results (Sociocultural Theory). A CW system that I developed for the research is also presented following the introduction of Activity Theory.

Activity Theory as a descriptive framework

As a number of recent publications have showed, Activity Theory (AT) is a useful tool for describing and analyzing activity settings, especially in its expanded version (Engeström,

Miettinen, & Punamaki, 1999; Wells, 1999). Figure 2.1 shows a basic structure of activity systems.

Figure 2. 1. The structure of human activity systems (Engeström, 1987).

Before Engeström expanded a micro-level activity system into macro-level analyses, concepts of activity systems concentrated on an individual actor operating with other actors and tools (Leont’ev, 1978). After the expansion, the new ideas about activity systems considered a

7 macro-level of activity that included the community where the activity was carried out. This expansion aimed to represent the social or collective elements in an activity system through the addition of the elements of community, rules, and division of effort while at the same time emphasizing the importance of analyzing their interactions with each other. Engeström also emphasized the importance of contradictions within activity systems as the driving force of change and thus development.

This model was useful for bringing together a wide range of information about the factors that impact on an activity. In addition, the model enabled scholars to portray collaborative process explicitly and the changes that occurred in an activity system. Therefore, within the framework, all interactions among subjects, objects, mediating artifacts, outcomes, rules, community, and division of effort are considered together. Some of the significant ideas about AT are that:

a. In order to achieve outcomes it is necessary to produce certain 'objects' which may

include knowledge, experiences, and actual physical products. Some products may

not be physical, such as thinking or verbal processes.

b. An activity is typically mediated by the tools used and artifacts that were considered

in relation to the activities, e.g., policy documents, samples, recipes, and facilities.

c. An activity is also mediated by the community in which the activity is being carried

out. The community may oppose or support the activity, and it may facilitate or

impede access to resources.

d. In addition, the community may support or impose rules on the subjects: those persons,

groups, or organizations that are undertaking the activity. There may also be 'rules'

about the kind of products, knowledge, and experiences that are approved or

acceptable and access to tools and artifacts and that are permitted to do which aspects

of the activity.

e. To the extent that a person is engaged with the community the subject may share

responsibility with the community for the achievement of the object. This is likely to

be realized through some form of division of labour.

Theoretical framework applied to the present study

Drawing on the research questions and AT outlined above, I modeled a CW system and specified conceptualizations under each AT component. Figure 2.2 illustrates the components and their specifications for the CW system. Within this theoretical framework, hypothetical explanations of my analysis focus mainly on issues derived from CW processes (e.g., mediation), which I have assumed contribute to participants’ writing development and foster the L2 students’ self-assessed beliefs about writing abilities, including perspectives like learning goal orientations, performance goal orientations, metacognition, self-efficacy, and effort.

Figure 2. 2. The current CW system.

Figure 2.2 can be elaborated as follows in respect to the present research: L2 learners composed under the rules of writing collaboratively so that they could ask questions and seek assistance when comprehension or composing was difficult. Also, they were only allowed to compose with computers in a classroom. The rules were set up by the instructor and also functioned through the guidelines of the curriculum for the collaborative student writers to follow along with any implicit, pragmatic rules or social conventions of interaction that the students themselves enacted with each other. All students were in a computer-supported writing classroom that included L2 learners at similar levels of English writing proficiency. As for the division of labor, performing writing tasks could be shared by means of peer-tutoring, problem solving, and decision making. Peer tutoring or near-peer role model effects could be regarded as a form of reciprocal teaching while students were practicing CW. As time went by, strengthened interpersonal relationships, in relation to problem solving and peer revision activities, were predicted to enhance the near-peer role model effect for all students. For the relatively advanced students in the writing group, their construction of language expert identities was also enhanced through practices akin to reciprocal teaching (Palinscar & Brown, 1984). Looking to the upper part of the triangle in Figure 2.2, subjects are all the participants in this study; mediating tools are symbolic cognitive artifacts like L1 or L2 dialogue records and physical artifacts like computer/Internet resources and written production from revisions. The object of this CW system was to foster students’ solidarity and inter-class collaboration by means of their performing CW tasks so that L2 learners could achieve the outcome of improving their individual writing performance and self-assessed beliefs about writing abilities. For the purpose of contrasting the outcomes of this CW system with those of a non-CW system, comparison writers only had lecture-based instruction, not CW, were introduced to the study but in a separate, parallel context in a class that I taught a year later.

Table 2.1 shows how the study’s research issues are framed within Activity Theory.

Table 2.1 Research Issues and How They Fit the Current Theoretical Framework Research objectives How they fit Activity Theory Collaborative writers’ tool-mediated Interaction among tools and artifacts, objects, and actions and their relation to their writing outcomes performance Collaborative writers’ ccommunication Relationships among division of efforts, relationships, their attitudes to community, rules, and outcomes tool-mediated writing, and self-assessed beliefs about writing abilities

I proposed three research sub-questions for each research issue. The first question was to investigate the relationship among tools and artifacts, objects, and outcomes. I attempted to examine the types of writing knowledge that were produced from collaborative writers’ conversations and their searches of online writing resources (tools). The second question was to investigate the association of the adoption of tool-mediated writing knowledge on CW performance (relating tools and artifacts to objects). The third was to test the association of CW performance and individual writing performance (objects to outcomes).

The three research questions for the second issue investigated were to observe the relationships among rules, community, division of efforts, and outcomes. I intended to examine collaborative writers’ attitudes to tool-mediated writing (rules and community), their communicative relationships (division of efforts), and self-assessed beliefs about writing abilities (outcomes). Following these conceptualizations two explanatory theories are provided below.

Sociocultural Theory as an explanatory theory

SCT gives communicative activities a central role in cognitive development. The main claim of SCT is that cognitive functions (e.g., memory, reasoning, analysis, and application) are mediated mental activities, that is, their development is mediated by cultural interactions.

Furthermore, the social activities in which people participate can be transformed into their mental activities. As Stetsenko and Arievitch (1997) explained, “Psychological processes emerge first in collective behavior, in co-operation with other people, and only subsequently become internalized as the individual’s own ‘possessions’” (p.161). That is, cognitive functions are mediated through sociocultural activities that result in processes of internalization.

The principles of SCT recognize the importance of language learning as a socially situated activity. For example, working in collaborative situations allows learners to work and to pool their linguistic knowledge and to use scaffolding collectively (Donato, 1994). Furthermore, such activities may allow learners to work at a higher level of activity than would be the case where they are working alone (Donato, 1994; Storch, 2002a; Swain, 2000; Wigglesworth &

Storch, 2009). SCT also recognizes classroom instruction as the kind of context that Vygotsky suggested can lead to people’s development: "Development has its internal logic that does not lay down the path for instruction to follow, but in fact follows from instruction itself." (Lantolf

& Thorne, 2006, p. 239).

Mediation

Mediation refers to the processes by which human beings’ higher forms of mental activities develop. Mediation involves a relationship established through the use of physical and psychological tools. In relation to language learning, Donato and McCormick (1994) claimed

12 there are two types of mediation: artifact mediation (such as computers, text books, or workbooks) and social mediation, manifested through opportunities for interaction, discourse patterns among learners, and/or assistance provided by more competent others, such as teachers or peers. This empirical study quantitatively measures how artifact mediation (i.e., online writing resources) and social mediation (i.e., dialogues among collaborative writers) in CW affected students’ writing development in English.

ZPD

Vygotsky’s Zone of Proximal Development (ZPD) is an important theoretical construct for promoting the use of collaborative activities for language learning. Vygotsky defined the

ZPD as the gap between what a learner could accomplish alone and what that person could achieve with the support of more capable peers, adults and/or cultural artifacts (Vygotsky,

1987a; Warschauer, 1997). Specifically, the ZPD is conceived of as “the collaborative construction of opportunities for individuals to develop their mental ability” (Lantolf, 2000, p.

17). In addition, from Vygotsky’s viewpoint, human learning and development result from social interaction in which an individual learns to expand his or her current competence through the guidance of an expert or a more experienced individual. In other words, to encourage the

ZPD for learning, negotiation between an expert and a novice is required so that learners may engage in cognitive restructuring or elaboration for cognitive growth. Guidance within the ZPD begins with the idea that the path of learning is from social interaction to internalized independent functioning. Internalization along with mediation is one of the most important concepts in SCT. The process of internalization is a key element in the formation of mental functions.

Kinginger (2002) identified several interpretations of the ZPD. Among them, metalinguistic interpretation and scaffolding interpretation are highly relevant to the present study. Metalinguistic interpretations follow from collaborative interactions mediated through second language learning. Swain (2000) argued that learners externalize their thoughts through collaborative dialogue, a type of knowledge-building activity. Collaborative dialogue provides an insight into learners’ cognitive activities, which are the outcome of the inner speech taking place in learners’ minds.

Scaffolding is a widely discussed means of facilitating the ZPD. Scaffolding refers to the assistance which a more competent other can provide to learners to support their performing a learning activity. Each individual has learning potential that can be realized through assistance of others that are more capable. To do so, assistance is ‘scaffolded’ to a learner’s needs, allowing learners opportunities to practice and develop their behaviors and intellects by assisting them to access the ZPD. Wood, Bruner and Ross (1976) suggested that scaffolding involves several processes, such as encouraging learners to take an interest in the task, interpreting the task and maintaining learners’ interest in its objectives, lessening frustration during task completion, and offering demonstrations of how to achieve the goal of the task.

In relation to language learning, scaffolding is generally thought of as assistance provided through interactions between a competent other and a learner. However, scaffolding may not only be found in expert-novice interactions but also in novice-novice interactions

(Donato, 1994; Donato & McCormick, 1994; Storch, 2002a). Many writing researchers have investigated the influence of interaction during group work in writing classrooms (de Guerrero

& Villamil, 1994; Donato, 1994; Donato & McCormick, 1994; Villamil & de Guerrero,

1996)..The present thesis study further examines how novice-novice student writers scaffold or peer tutor each other on CW tasks to create a mutual ZPD.

Imitation

Vygotsky proposed that imitation is the key to internalization (Vygotsky, 1987a).

Imitation here does not refer to mindless mimicking but, “it involves goal directed cognitive activity that can result in transformations of the original model” (Lantolf & Thorne, 2007, p.

207). As Vygotsky (1987a, p. 210) stated, “development based on collaboration and imitation is the source of all the specifically human characteristics of consciousness that develop in the child.” Similarly, when adults learn a language, they often report practicing L2 patterns they have heard in class and engage in during their everyday activities such as walking on streets and jogging (Lantolf & Genung, 2002).

CHAPTER THREE

METHOD

The present study adopted complementary qualitative and quantitative research methods

(i.e., mixed methods) to investigate the relationships among writing processes and the outcomes of collaborative and individual writing performance (Polio, 2012). This chapter describes the data collection procedures and methods of data analysis that were employed to address my research questions following the explanatory theories presented in the previous chapter.

Research context and program objectives

The study took place in two four-week English intensive classes of 36 and 26 Taiwanese graduate students in the summer of 2011 and 2012. Most of the participants came from different fields of studies and were admitted to arts-related MA programs in Tainan National University of the Arts in Taiwan. The university had a policy of language requirements that graduate students had to meet. If they did not meet the language requirement, students could not graduate with a Master’s degree. The policy was practiced in all departments of the university. For example, the Department of Conservation of Cultural Relics and Museology stated in its graduate studies bulletin 2011-2012 that “First year graduate students have to join an intensive

English program providing that students have passed Test of English as a Foreign

Language, Test of English for International Communication, or General English Proficiency

Test” (Department of Conservation of Cultural Relics and Museology, 2011). This language prerequisite program acted as a substitute for students’ taking international standardized English exams. Among those standardized exams, the Test of English as a Foreign Language (TOEFL) was selected as a target exam by the school because of its worldwide popularity, so the program’s curriculum was designed to build up students’ skills and knowledge for writing the

TOEFL particularly. In this program, there was a before-program test to screen participants and an after-program test to evaluate if they could pass their courses. I was the instructor of the courses.

Participants in this study had received English education in Taiwan for at least seven years, starting from the age of 13 or younger up to the age of 20. English education in Taiwan is compulsory and English courses have appeared officially in high school curricula for six years.

After high school, most students go to university to take different academic majors; whether they still receive English language education depends on each of the universities’ policies on language learning or proficiency. Increasingly, most universities in Taiwan have come to regard

English as a compulsory course for freshmen to take even though courses’ titles and contents vary a lot. These courses normally include teaching and learning language skills, developing language exam-taking skills, and professional English writing. Table 3.1 illustrates the characteristics of participating students in the two classes in terms of their genders, majors, and language learning backgrounds.

Table 3.1 Participants in Two Classes in 2011 and 2012 Class in 2011 Class in 2012

(n=35) (n=26) females 78% 69% gender males 22% 31% art 47% 44% undergraduate music 25% 28% majors language and literature 11% 12% social science 17% 16% music 31% 31% visual arts 33% 16% intended graduate studies sound and image arts 20% 47% letters and cultural 16% 6% heritage 7 years from elementary 46% 20% school up to high school years of learning 10 years up to English 43% 44% college/university more than 10 years 11% 36% ESL program 6% 8% attending ESL living in English speaking programs and 0% 16% countries living in English speaking country none of them 94% 76% learning other yes 39% 52% languages no 61% 48% taking TOEFL yes 6% 0% exams no 94% 100%

In the 2011 and 2012 summer classes, I spent 120 hours with the students in each class.

The difference between the two classes was that the earlier (2011) class was implemented with

CW and writing instruction and the later (2012) class was implemented with individual writing and writing instruction, serving as a comparison group between the two classes. Students in the

CW classroom formed either dyadic or triadic groups on the basis of their choices or preferences to find their own writing partners. The CW groups remained the same throughout the whole

18 course over a month. For both classes of students, developing writing skills was a primary learning goal, and writing argumentative essays (i.e, independent writing) was one of the language learning components in the course syllabus. Students also learned how to write integrated writing tasks but those tasks were not analyzed as a focus of research in this thesis.

Students had two learning sections per day. One started from 9 to 12 in the morning and the other from 2 to 5 in the afternoon. Both lasted 5 days a week and 4 weeks in August, 2011 and

July, 2012. Notably, English grammar and writing skills from the same textbook, North

Star-Building Skills for the TOEFL iBT (Solórzano, 2006) was the basis for the curriculum and lectures (of 10 to 20 minutes) once or twice a week in both classes. The textbook is a three-level series, which links learning and assessment with a skill-building curriculum that incorporates authentic test material from Educational Testing Service (ETS). Deductive writing skills were also taught in both classes. The topics of lectures on writing skills specified in the textbook were:

a. Paraphrasing

b. Identifying and using rhetorical structure

c. Comparing and contrasting

d. Summarizing

e. Identifying and using cohesive devices

Among all the writing topics, these five topics about writing skills were presented in

lectures in different sequences on the basis of my perception of students’ writing needs in the

summer courses in 2011 and 2012 separately.

A total of four argumentative writing tasks were assigned to students as in-class writing

activities (See Table 3. 2). In addition, another writing task was administered in class as a

before-program test and after-program test, assuming the pretest should be similar to the

posttest (Polio, 2012). All writing prompts, including the before-program and after-program

tests were selected from various TOEFL preparation materials including FREE TOEFL iBT

Test Sample Questions, The Official Guide to the TOEFL® Test, and TOEFL® Practice Online

(see Appendix A for all 5 prompts). Each of the selected tasks was considered as equivalent to

the others as they appeared in the previous TOEFL tests according to the specifications shown

in the preparation materials. Participants were allowed one hour maximum to complete each

writing task.

Table 3. 2 Writing Task Schedule for Students in the Collaborative Writing Classroom

Writing task Week 1 Week 2 Week 3 Week 4 T 1 before-program test after-program test able 3. 2 Pre-CW1, CW1, post-CW1 2 3 Pre-CW2, CW2, post-CW2 displays 4 Pre-CW3, CW3, post-CW3 the 5 Pre-CW4, CW4, post-CW4 writing task schedule that the CW class followed. The top row is the number of the weeks and the columns show the writing tasks corresponding to the weeks. Except for a before-program test and an after-program test, each student in the CW classroom had to write one writing task three times in one week. Students in the CW classroom wrote three drafts of each composition, first individually, second collaboratively, and third individually. Pre-CW 1 was the writing task 1 that each individual participant had to draft before they practiced CW1. CW 1 was participants’ rewriting task 1 collaboratively in class. They were permitted to collaboratively combine and so edit their drafts. Post-CW 1 was individuals’ rewriting task one again after CW1. A before-program test and an after-program test were given to participants in the comparison class

20 as well. For the comparison class, participants wrote the same before and after-program tests and all four writing tasks in class individually. Because of their revisions, CW students spent slightly more time writing, about 2 hours more per week and 8 more hours in total, than the students in the control class did. Students in both classes signed consent forms agreeing to participate in the research, thereby avoiding potential Hawthorne effects from the research. The basis of comparison in my research design is therefore mediated writing in the CW class versus non-mediated writing in the comparison class. Table 3.3 outlines the entire data collection schedule for both writing classes, including the writing task schedule in detail.

This design of students’ writing and rewriting repeatedly reflects practices of writing instruction in some EFL writing classes. In addition, to investigate the proposal of SCT that collaboration can promote the transfer of knowledge from interpsychological interaction to a person’s self-regulated performance (DiCamilla & Anton, 1997), the design makes testing knowledge transfer possible by associating the performance differences between pre-CW and

CW with the performance differences between pre-CW and post-CW.

The goal of the study was to investigate whether the development of individual student’s writing abilities was built up through multiple writing and rewriting and students’ interactions and their uses of mediation tools. The comparison class did the same before- and after-program tests and four writing tasks in the same order, except that they only did each of the four writing tasks individually once a week. I provided grades and comments on CW compositions and only grades on pre-CW and post-CW compositions. For the comparison class, I graded and commented on individual compositions.

Before the summer program began most learners had difficulty in completing a writing task within a given time. Mediation was designed to help build students’ writing knowledge and confidence in completing a writing task. I expected that the writing knowledge students gained

21 from mediation in the class could be internalized and displayed in their individual writing performance.

The development of a writing platform

To integrate and facilitate my classroom teaching and the thesis study a writing platform was developed and tested, and afterwards used as an interface for writing instruction, learning, assessment, and research in both prerequisite TOEFL writing classrooms in 2011 and 2012.

Students in both classes composed on the platform whether they did collaborative writing or individual writing.

From instructional and assessment points of view, the writing platform facilitated the multi-task management for delivering writing tasks and for my receiving students’ compositions.

From a learning perspective, the platform allowed writers to consult online writing resources easily and to learn how to collaborate with other peer students. From a research perspective, the writing platform was a data collection instrument that kept records on each L2 writer over time, documenting changes both in their uses of mediation tools and in their writing performance in the computer-supported classroom.

A main goal of this thesis study was to investigate the relationships among mediation behaviors, the outcomes of CW, and individual writing development. Data sets collected during the course by means of the writing platform were used later for data analysis. This writing platform recorded two forms of mediation behaviors: students’ CW dialogues and their online browsing histories. Compositions from pretests, posttests, CW activities, and individual writing were also stored in the platform. Questionnaire surveys about the students’ dyadic or triadic relationships, on their attitudes towards computer-supported CW, and on their self-assessed beliefs about writing abilities were also conducted in this platform.

For the purpose of creating an effective writing platform, I adopted the guidelines for an instructional design model: Analysis, Design, Development, Implementation, and Evaluation

(ADDIE). This basic, generic model was derived from the work of numerous instructional designers (Kemp, Morrison, & Ross, 1998). However, as Smith, Newman, and Park (1997) observed, no one set of design principles satisfies all conditions because applications of any instructional model are context-sensitive. Consequently, in order to fit my needs, certain modifications were made to the four-dimension model used in the study—Analysis,

Development, Implementation, and Evaluation (Lin, 2001)—as described in detail below.

The analysis phase included three activities: (a) task, (b) resource, and (c) needs analysis.

This was a process of defining how many tasks, how much information, and what features and functions needed to be embedded into the platform. In terms of task analysis, the goal of the writing platform was to facilitate students’ composing four CW tasks in the computer-supported classroom. Therefore, to meet both instructional and assessment purposes, all the writing tasks, including pretest/posttest writing tasks and questionnaire surveys were carried out on the platform. With respect to resource analysis, the incorporation of a writing platform into a classroom made my role as an instructor no longer a knowledge dispenser but rather a facilitator to provide resources and aids to students to solve their problems. Because all students differ in writing styles and respond differently to writing activities, it was vital to offer writing resources to fit a variety of individual needs. To address such possible needs, I gathered online writing resources onto the platform, mostly links to an online dictionary, collocation tools, lists of common errors in English writing, and writing templates. Writing templates were model formats which included, for example, instruction on how to start a paragraph with opening patterns and what transitional words to use to make paragraphs coherent. For the needs analysis, the following features and functions were implemented in the platform:

a. To be able to allow a teacher to create writing tasks and questionnaire surveys online,

b. To be able to allow at least 30 students to compose at the same time without any

Internet work load problems,

c. To be able to keep a record of individuals’ online browsing history,

d. To be able to record students’ talk when they started to perform writing,

e. To have teacher-interface and writer-interface web pages in one platform,

f. To be able to generate a database that gathers information on all the students’ writing

processes, and

g. To be able to create CW accounts and individual writing accounts separately.

The development phase involved issues like (a) what combination of media and methods were the most cost-effective and (b) how materials were authored, reviewed, produced, and validated. Basically, information from the analysis phase formed the bases for the design phase.

First, I conveyed the functions and features that the platform had to be equipped with to a head programmer so that he could decide: (a) which programming language to use, (b) how many other programmers would be involved, (c) how to structure the entire platform, and then (d) how much work to dispense to other programmers to code. The last step in this phase was for me to inform programmers of the number and the order of writing tasks and questionnaires which were going to be embedded, and the nature of those items (e.g., open-ended questions or multiple-choice questions), in the questionnaire. The writing platform was created by three programmers in two weeks in July 2011, which involved one week in development, and the other week in testing its compatibility with other equipment (i.e., microphones and headsets) in a writing computer room. The writing platform was built on a server and through client software that had to be installed onto students’ computers so that CW processes (e.g., talk and browsing history) were recorded and delivered to the server’s databases. There were two web-page entries

24 for an instructor to administer writing tasks and questionnaires and to access students’ compositions and writing processes, and one page entry was for students to write. Programmers gave the platform a name: English Writing.

The implementation phase meant integrating the platform into a real teaching environment. The success of a computer-supported course depends on whether it was implemented as intended. To ensure this, I had written a course syllabus and created a manual for using English Writing before the course started.

The evaluation phase was intended to help student users to solve problems. I conducted an initial pilot study in the first day of my class. The main purpose of the pilot study was to evaluate the feasibility of the procedures and instruments proposed, such as whether an instruction manual was needed for using English Writing to perform CW, or if students could perform the writing tasks appropriately in dyads or triads, and thirdly if the Mandarin–translated questionnaires were able to elicit authentic answers to address the research questions. After the pilot study, a few programming deficits were fixed and it was integrated into CW activities in the class.

Structure and function of English Writing platform

The English Writing platform has one entry webpage that allowed students to write independently and collaboratively in the platform, as shown in Figure 3. 1. Meanwhile, there were two other instructors’ web pages, one of which managed the administration of writing tasks and questionnaire surveys (as shown in Figure 3. 2), and the other was a database that stored written compositions for me to retrieve (Figure 3. 3).

Figure 3. 1. English writing platform: students’ entry to write.

By logging into the instructors’ entry, I was able to manage students’ accounts and upload writing prompts and questionnaires. The webpage also allowed me to monitor if student writers submitted their writing pieces on time. Questionnaire results were also available through this webpage. Other functions like searching certain students’ compositions or questionnaires filled in on certain dates were easy to use in the platform (as shown in Figure 3. 2).

Figure 3. 2. English writing platform: instructors’ entry to administer writing tasks and questionnaire survey.

Once students logged into the students’ webpage with their individual accounts they immediately got access to the writing task they were assigned to write in that week, as Figure 3.

1 shows. For CW, each writing group was given a team account to log into. After they were recognized in the system with the team account, a CW task popped out on the screen.

Underneath a writing task was a writing area for students to jot down their ideas for writing; at the same time, if they wanted to browse websites to get some help with respect to their writing, they clicked one of the icons on the blue bar of the platform to open a browsing window. For example, a list of writing resources (e.g., writing templates, collocation tools, and common writing errors) was mailed to the students’ mailbox and they might use it to facilitate or correct their writing. The browsing window kept a record of all their browsing history. Once they finished writing and submitted their piece of writing to a database server, their writing was analyzed and incorporated into their browsing history in one webpage, as Figure 3. 3 indicates.

Only a course instructor could retrieve information from this webpage.

Figure 3. 3. English writing platform: instructors’ entry to retrieve writing processes and artifacts.

Collaborative writers’ mediation

Collaborative writers were offered a variety of mediating materials in this summer program. Material mediation included writing artifacts, print materials, and electronic resources.

Writing artifacts contained students' note-taking, writing samples, and the written compositions they produced in writing collaborative tasks (e.g., revisions made from writing partners’ drafts or their complete new co-constructed compositions). Print materials were topics about writing skills contained in a textbook that the program assigned to the class. These topics were: (a) paragraphing, (b) using detailed examples, (c) identifying and using rhetorical structures, (d) identifying and using cohesive devices, and (e) comparing and contrasting. Electronic resources comprised the largest proportion of the mediating materials, including various TOEFL writing

28 web site links and grammar slides. These were made into a list and were sent to students’ email boxes. The types of TOEFL web site links were:

a. Introduction to TOEFL,

b. Official TOEFL website,

c. Google dictionary,

d. Conjunctions for TOEFL writing,

e. Common sentence patterns for TOEFL writing,

f. Writing templates for TOEFL,

g. 185 TOEFL writing samples on 14 topics,

h. Integrated writing templates,

i. Independent writing templates,

j. TOEFL general writing templates,

k. General writing skills,

l. Experiences in taking TOEFL writing tests,

m. Essential vocabulary for understanding TOEFL tests, and

n. Websites for TOEFL test preparation exercises.

The topics on grammar slides were:

a. tenses,

b. adverb clauses of time,

c. nouns, and

d. subject-verb agreement.

While the collaborative writers were offered the mediating materials listed above and symbolic mediation (e.g., their dialogues), comparison writers were also provided with

29 mediation but not the web site links or peer-to-peer dialogues. These differences were consistent with the research questions guiding this thesis study.

Overview of the Research Design

To answer research questions, I adopted “tool-mediated action” as a unit of analysis related to Activity Theory. One tool-mediated action referred to a writing group’s using a certain tool (e.g., asking the other writer partner questions or consulting web sites) to solve a writing problem (regardless of whether the resolution of the problem was correct or effective or not). In order to facilitate data analysis, that a writing group took any tool-mediated action was coded as 1 and 0 if it did not.

Data collection schedule

Table 3.3 displays my data collection schedule in the order of the weeks that the five writing tasks were distributed to participants and the data sets that were collected corresponding to writing groups in August 2011 and July 2012.

Table 3.3

Data Collection Schedule

Week 1 Week 2 Week 3 Week 4 Befor Pre- Post Pre- Post Pre- Post Pre- After- CW class e-pro CW CW CW CW Post- CW -C CW -C CW -C CW progra gram 1 2 3 4 CW 4 1 W 1 2 W 2 3 W 3 4 m test test Discussion x x x x logs Online researchin x x x x g history Writing x x x x x x x x x x x x x x artifacts Post-writi ng x x x x questionna ire survey Group x interview Self-assess ed beliefs about x x writing abilities Comparis on writing class Writing x x x x x x artifacts Self-assess x x ed beliefs Note. Pre-CW 1=drafting before CW1; CW 1=CW on writing task 1; Post-CW 1=writings after CW1

Data collection and analysis for Question Set One

The writing platform was the major research instrument employed to address the first set of research issues. Before writing tasks were assigned as activities in class, students acknowledged in a consent form that their writing processes and performance would be documented in the platform for future analysis. Thirty-five students of collaborative writers and

26 comparison students agreed to participate in this thesis project. Full populations of students in the classes participated. Students in the collaborative class chose their own writing partners to form a triad or a dyad CW group. There were 16 groups in total, including 12 dyad groups and four triad groups. After each week’s writing, data sets were collected and retrieved from the writing platform to address the first set of research questions: students’ dialogue logs, web browsing histories, and writing artifacts (i.e., pre-CW drafts and post-CW compositions, CW compositions and before-program and post-program compositions).

Research Question 1.1

Writers’ self-reports on mediating tool use, dialogue logs, and web browsing histories were the data sets used to answer the research question regarding the themes emerging from collaborative writers’ dialogues and their types of searching from online resource while writing collaboratively. The purpose of categorizing them to answer Research Question 1.1 was to create a coding scheme and logical framework about mediating processes, grounded in the local context and activities, to form a basis to answer Research Question 1.2 (about the influences of mediating tools on writing processes).

For this thesis study, 16 groups of students, including 12 dyadic groups and four triadic groups, wrote four argumentative essays in sessions of one hour each over four weeks. On average, a writing group spent 10 minutes discussing a writing task and planning what to write

32 to respond to the writing tasks. The remaining minutes were spent drafting, checking online information, revising writing, and off-line talk. Their dialogues were recorded during each hour of writing. Their online browsing histories from the beginning to the end of the students’ composing were also kept in the management system of the writing platform. The total amount of dialogue logs for all the 16 groups of students over four sessions of CW was around 32 hours.

All of the dialogue logs were spoken in Mandarin, the participants’ first and dominant language

(and mine as well). I selected parts of Mandarin language-related episodes (LREs) from the dialogues and translated them into English and have presented the English excerpts in the next chapter. The English excerpts are presented in the transcription conventions created by Sacks,

Schegloff, and Jefferson (1974) (see Appendix B).

I created a mediation coding scheme to analyze the dialogue themes and web browsing histories with reference to (a) the mediation categories appearing in the students’ self-report questionnaire, (b) my verification of the data with recorded dialogues and web browsing histories (Appendix C), (c) segmentation of the mediation categories into LREs defined from existing literature, and (d) adherence to the number of variables for a specific statistical technique to be used to address Research Question 1.2. The triangulation of these different sources of data aimed to increase the validity of the coding scheme (Mathison, 1988).

The procedure for creating the new coding scheme was complicated and time-consuming.

First, I identified and narrowed down the self-reported categories on mediation tool use. The information in the self-reported mediation tool use survey (including themes of dialogues and types of online website visits) students wrote after each CW task was varied and disparate.

Self-reported dialogue themes from this survey included vocabulary, translation, collocations, accuracy, organization, and argumentation. The types of online resource use included translation and writing templates.

Secondly, in order to analyze those collective categories, LREs were screened out from students’ dialogue recordings using Swain and Lapkin’s (1998) definitions of LREs: “ any part of a dialogue where the students talk about the language they are producing, question their language use, or correct themselves or others” (p.326). In this verification stage, with the intent to simplify and facilitate the match of self-reported data with real data, I merged vocabulary, translation, collocation, accuracy, and writing templates into a category I called “language forms” (either on syntactical or lexical forms) and merged organization and argumentation into a different category I called “skills of expression” so that there were only two broad and clear mediation types: language forms and skills of expression. Interestingly, in the students’ dialogues there were often cases in which one writer asked the other to translate Mandarin words into English or to confirm grammar on the basis of the person’s memory of high school

English lessons. Online dictionaries were also often used to translate Mandarin words into

English. In examining the browsing history, I found 90% of the histories involved

Mandarin-English translation and 10% was on writing templates.

Thirdly, after verifying the self-reported data with the real discourse or searching data, I consulted published literature to identify how LREs were categorized in students’ goal-appropriate discourse to complete classroom language tasks. In Swain and Lapkin (1998), for all the student pairs in jigsaw activities, LREs were classified as either lexis-based or form-based. The lexis-based LREs involved students seeking vocabulary or choosing among competing vocabulary items. The form-based LREs involved students focusing on spelling or an aspect of morphology, syntax, or discourse. Following Swain and Lapkin (1998), Dobao (2012) also examined the relationships of LREs and writing correctness between pairs and groups in a jigsaw writing activity and then compared the writing performance of the pairs with that of the groups. The categorizations of LREs he used were lexis-focused LREs, form-focused LREs, mechanics-focused LREs, unresolved form-focused LREs, and incorrectly resolved

34 form-focused LREs. Other LRE categorizations like Kowal and Swain (1994) had meaning-based LREs and grammatical LREs in dictogloss tasks, and Storch (1998) used categories of grammar, meaning, discourse, intuition, none, and others to analyze students’ reconstruction tasks.

The reason for my including skills of expression (e.g., rhetorical organization and argumentation), in addition to conventional definitions of LREs, in analyzing the students’ dialogues was because these episodes appeared frequently in the data, probably because the students acknowledged they were central for writing argumentative compositions and also for how their writing would be evaluated in the real TOEFL test and textbook. The global evaluation of these writing dimensions were probably excluded from previous research literature on LREs because the task type that researchers analyzed were usually closed tasks focused on language learning rather than written compositions (e.g., jigsaw, dictogloss, and reconstruction tasks). As Ellis (2003) observed, “a closed task requires learners to reach a single, correct solution or one of a small finite set of solution” (p.89). On the contrary, an open task such as composing a written argument has no predetermined solution, so involves making free choices, debate, general discussion, or argumentation as in the writing tasks in this thesis study.

Finally, after the two-dimensional coding scheme was established, I consulted statistical references to check how many predictors can be accommodated with 61 participants as a sample in advanced regression analysis techniques. After comparing different rules of thumb from

Green (1991), Harris (1975), and Cohen and Cohen (1975), I adopted Harris's (1975) formula for yielding the absolute minimum number of variables per participants. Harris suggested that the number of participants should exceed the number of predictors by at least 50 (i.e., the total number of participants equals the number of predictor variables plus 50). On the basis of this rule, I judged that the independent variables from the potential coding scheme (e.g., language forms and skills of expression) together with the number of times students wrote (i.e., four) were

35 within the safe range of predictors and so could form three potential independent variables: (a) language form mediation (lexical or syntactic forms); (b) skills of expression mediation (how rhetorical structures were organized and how arguments and supporting ideas were presented); and (c) times of writing (four). Data are presented in Chapter 4 on the uses of mediation tools and the frequency of references to tool-mediated knowledge.

Research Question 1.2

The main concern of this research question was to explore the relationships between the use of mediation and CW performance over time. Data sets were the number of groups that consulted or did not consult tool-mediated writing knowledge and the scores of 16 groups’ CW compositions on four writing tasks. To generate scores for the written compositions, I had them rated by independent raters using an established instrument for writing assessment. Generalized

Estimating Equations (GEE) modeling was then used to test the relationships between the predictors of mediation (i.e., language forms, skills of expression, and times of writing) and the outcome variables of writing performance (i.e., as scored in five dimensions from the rating scale and holistic ratings).

To analyze whether adoption of tool mediated knowledge influenced these students’ writing performance, I converted “language forms” and “skills of expression” into two binary variables. This binary coding was conducted on the entire 16 groups of writers over four weeks.

Together with times of writing (which was an ordinal variable, weeks 1 to 4), “language forms” and “skills of expression” were adopted to test how mediation tool use over time influenced writing performance. See the coding results in Appendix D.

Rating procedures and reliability

I adopted Hamp-Lyons’ (1991) and Hamp-Lyons and Henning’s (1991) multiple-trait rating scale and holistic rating scale to assess students’ writing (see Appendix E and Appendix

F). The former rating scale ranged from 0 to 9 on five writing dimensions: communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy. The latter also had 0 to 9 scales. My reasons for adopting these scales were, first, because the scoring rubrics were validated and have been widely used in previous research on argumentative academic writing in English (Hamp-Lyons & Henning, 1991) and, second, the multiple-trait rating scale could indicate the specific traits of L2 writing that the students might develop and which the holistic rating scale might underestimate. I acknowledged that another measure of writing—the triad model of complexity, accuracy, and fluency (CAF)—is more fine-grained for text analyses than Hamp-Lyon’s assessment scales are, and CAF provides clear definitions on each writing scale (Polio, 1997). Hamp-Lyon’s multiple-trait scale was appropriate in this research in that it was simple and clear for raters to be quickly familiar with applying rubrics to rating processes. Therefore, it led to efficient composition marking. Moreover, the holistic rating scales were similar in kind to those used for the TOEFL, which was the test that students in the present courses were preparing themselves to take.

For the sake of maintaining fairness and objectiveness (because I was the course instructor and knew the students), independent raters were invited to do the scoring. Among graduate students I knew, raters were carefully selected on the basis of their specialties, experiences in scoring writing, language background, and willingness to help. Two raters, Katie and Yu, and later on Brenda joined in the scoring. Their names like all names that appear in the thesis are pseudonyms. These raters’ backgrounds were similar: they were Chinese/English bilinguals and graduate students in the University of Toronto, and their fields of studies were

37 linguistics and language teaching. In addition, they all had English essay scoring experience in their teaching assistantship years. Following the raters’ agreement to participate, a training section was given a couple days before their start of marking compositions in which the raters reviewed task instructions, scoring rubrics, and anchor essays from ETS TOEFL independent writing samples illustrating the degree of diversity within each score level. Five essays were then randomly selected from the entire essay set collected from the actual participants for scoring practice. After all essays were rated, if two scores assigned by two raters were within one point difference, I averaged them into one global performance score. When two ratings differed by more than one point, a third expert rater was employed to resolve the discrepancies by averaging the score assigned by the third rater with the closest original score (following procedures recommended in Weigle, 1999). Using Pearson’s Product Moment Correlations, the inter-rater reliability was estimated for the multiple-trait rating scale (i.e., communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy) and holistic rating scale, as reported in Chapter 4.

To reduce the possibility of inadvertent rater biases about the sequence or conditions of writing, I kept the three raters blind about which data they were coding by giving them randomly selected compositions to mark at three times. The rater Katie completed all the 404 essays from the CW group and 156 essays from the comparison writing group. Yu completed one third of the essays from the CW group, and Brenda scored the other two-thirds of the CW and all the comparison writing.

Generalized Estimating Equations (GEE) modeling

After the scores were tallied and averaged across the raters, I used Generalized Estimating

Equations (GEE) as a longitudinal data analysis technique to explore the relationships between mediation categories, time, and CW performance. GEE, an extension of the quasi-likelihood

38 approach, is often used to analyze data in longitudinal, nested or repeated measures designs.

Liang and Zeger (1986) developed the technique to increase confidence in estimating correlated data responses in regression model parameters. Although GEE analysis is new to studies of SLA and writing, the method has received wide use in medical and life sciences such as epidemiology, gerontology, and biology (Ballinger, 2004).

Among other statistical techniques that feature in analyzing longitudinal data, GEE was selected in this thesis for its generating marginal models, which tell the user how much the average response would change for every one-unit increase in a covariate across the population.

The other reason was that the GEE algorithm has been incorporated into many major statistical software packages, including SPSS, SAS, and HLM.

According to Twisk (2003), traditional approaches like ANOVA and MANOVA have several limitations, such as missing data imputation and restrictive assumptions about correlation structure, for which they cannot precisely generate accurate estimates for longitudinal data. Unlike ANCOVA, ANOVA, and MANOVA, even if a dataset is incomplete,

GEE can still apply its formula to calculate complete values. The principle behind this operation of counting complete data only is that GEE uses the “all available pairs” method, in which all non-missing pairs of data are employed in estimating the working correlation parameters.

Another merit of GEE is that the GEE type of regression differs from the usual linear regression modeling (e.g., GZLM) in that it takes into account clustered or repeated data whereas traditional techniques (e.g., GZLM) do not. GEE also does not assume correlation coefficients are the same among performances across time. For example, in a longitudinal study, an L2 learner’s lexical performance in one task may be correlated with that in a second task; however, it would be inappropriate to assume that the correlation stayed the same between each performance. On the contrary, assuming that changes occur over time, a stronger correlation may be supposed to be found between subsequent task performances. In order to take into

39 account the dependency of observations GEE required me to specify an appropriate working correlation matrix before calculating differences over time.

Choices of the correlation structure and model selection for GEE

Since correlation structure was so important to GEE analysis in determining accurate estimates to select an appropriate correlation structure for a data set was a prerequisite to the main analysis. On the basis of published literature (Garson, 2012; Twisk, 2003) I selected an appropriate correlation structure for the data set using QIC values. Quasi-likelihood under independence criterion (QIC) is one of the goodness of fit values offered for GEE. A smaller value has a better model fit for QIC. QIC establishes the best working correlation structure assumption by running the model for different working correlation structure assumptions, choosing the assumption with the lowest QIC value.

After I first examined each dependent variable with “the smallest QIC” criterion, I found that an exchangeable matrix generated the smallest QIC for all the outcome variables. However, the correlation structure of dependent variables from repeated measures had the least possibility to be exchangeable, which assumed all the r ratios were the same. In addition, Twisk (2003) argued that results obtained from GEE with an exchangeable matrix were the same as those generated from repeated measure ANOVA. So this first step of applying QIC criterion took me back to a traditional technique that yielded potentially inaccurate parameter estimates. The above attempt to use QIC values as the references to select best models was not successful; as

Ballinger (2004) informed researchers, there is no universally accepted test for goodness of fit for GEE models.

Recognizing that QIC criteria did not fit to describe my data well, I took the other step that Twisk (2003) suggested of examining the dependency of observations in all the dependent variables. As long as dependent variables’ correlation matrix was examined and correctly

40 selected and employed to the procedures of GEE analysis, GEE could be assumed to generate correct estimates for each assigned parameter and therefore the best model for a data set.

After examining all of the dependent variables (e.g., holistic scores, communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy) with

Pearson correlation analysis, I selected “unstructured” as the appropriate correlation structure to be assigned to the GEE analyses. Appendix G lists all the tables of the six dependent variables’ correlation coefficients over 4 weeks (i.e., holistic ratings, communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy).

Modeling procedures

To model the GEE analyses to answer Research Question 1.2, I selected normal distribution with the identity link function, which assumed that the dependent variables were predicted directly and estimated parameters with maximum likelihood estimates. An unstructured working correlation matrix was employed after the correlation structure was examined, as described in the previous section. Predictors were “language forms” and “skills of expression” mediation and “weeks”. Dependent variables were writing performance on the multiple-trait writing dimensions and holistic scores. For a clear and orderly presentation of the analysis and results, I started with describing the differences that all the factors made on one general outcome variable: holistic scores. After that, I looked to see where differences were found on the other outcome variables: communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy.

Research Question 1.3

Following the identification of the relationships between the uses of mediation and the outcomes of CW performance over time, this third research question asked if CW changes (from

41 pre-CW to CW) were associated with their individual writing performance (from pre-CW to post-CW). Pearson correlations were conducted to test the association. Collaborative writers’ changes referred to the over-time contrast between student writers’ before-mediated performance and mediated performance. Students’ individual writing development was the over-time development after they experienced CW activities during four weeks.

Scores ranging from 0 to 9 on five writing dimensions from both classrooms were utilized. The scores of the comparison writing, representing students’ maturational growth, without collaborative, mediated writing, were used as the comparison with those of the students who did CW.

Table 3.4 outlines the data collection and data analysis methods in accordance with the sequence of research questions I attempted to answer overall in the first question set.

Table 3.4 Data Collection and Data Analysis Methods for Research Questions 1.1 to 1.3 Research Coding schemes Instruments Data sets Data analysis techniques issues or writing rubrics Students’ self-reported survey Coding scheme generated and Data screening and on mediation tool use, data verification; transcription frequency counts 1.1 discussion logs, and web conventions (Appendix B) search histories (Appendix C) The above data sets and Communicative writing rubrics CW written compositions (Hamp-Lyons & Henning, GEE for associating Collaborative Five writing (Appendix D and Appendix 1991) and holistic rating scales mediation categories with writers’ 1.2 tasks in a G) (Hamp-Lyons, 1991) CW performance tool-mediated writing (Appendix E and Appendix F) actions and platform CW compositions and Communicative writing (a) GEE comparing before relations to (Appendix individual compositions rubrics: scales from 0 to 9 on program performance with their writing A) from collaborative writers five traits and one holistic after-program performance performance rating (Appendix E and across two classes of Appendix F) students 1.3 (b) Pearson correlation coefficients between CW performance and individual writing performance

Data collection and analysis methods for Question Set Two

This set of research questions aimed to explore the relationships among writers’ division of labor, community, rules, and outcomes within the theoretical frameworks of Activity Theory and SCT. The data collection instruments to explore inquires of the second research question set were: questionnaires on collaborative writers’ dyadic or triadic communications (shown in

Appendix H), interview protocols on CW (see Appendix I), and questionnaires on self-assessed beliefs about writing abilities (displayed in Appendix J). Thirty-five collaborative writers responded to all the questionnaires and interviews. Twenty-six comparison writers were asked to respond only to the questionnaires on their self-assessed beliefs about writing abilities.

Research Question 2.1

The post-writing questionnaire was composed of 37 questions in Likert scales (with 19 reverse-scale questions) and two open-ended questions asking students what worked well and what did not work well. This questionnaire was administered in the writing platform right after students participated in each of four CW tasks (Appendix H). The Likert scale items in the questionnaire were developed by Lowry et al. (2005) who adopted a variety of general social survey (GSS) questions to a post-writing questionnaire to investigate the impact of process structures on CW teams. This questionnaire synthesizes several communication taxonomies related to:

a. collaborative writers’ communications, including discussion quality,

communication appropriateness, communication richness, and task discussion

effectiveness;

b. quality of relationships among people, which included agreement and teamwork;

c. process satisfaction, which asked participants to rate their perceived satisfaction with

their group’s writing processes

Numerical data were derived from the questionnaire and GEE was used in the analysis.

Two open-ended questions of Research Questions 2.1 were transcribed and categorized in

Appendix K.

Research Question 2.2

Group interviews were conducted in Mandarin according to the following protocols (see

Appendix I) in week four after all writing activities had ended. Questions in the protocols asked for the students’ reflections in relation to constructs of Activity Theory, specifically concerning the activity system for CW in which they had participated (see Figure 2. 2 in Chapter 2). The following questions were posed:

Mediation to object: How did mediation tools (e.g., discussion and online writing

resource) help with your CW?

Object to outcomes: Did CW help with your individual writing and how did it help you?

Division of labour: How did you collaborate with each other?

Community: What did you think of your partnership?

Rules: What advantages were there to practice CW in the classroom?

What limitations were there to practice CW in the classroom?

What improvements were there to practice CW in the classroom?

What were other issues related to the class or the program?

The full set of participants, sixteen groups, was interviewed by me for around 10 minutes each. I also kept notes on what they discussed and how they said they felt when they were talking to me or the other writer in the group. I transcribed the interview data, a sample of which appears in Appendix L (translated from the original Mandarin). The functions of these data were

45 to document students’ perceptions of CW and to support the analyses to answer Research

Question Set One. On the basis of students’ answers to questions about division of labour and community, novice/novice and novice/expert groups were also identified.

Research Question 2.3

To assess if CW in a computer-supported classroom made a difference in students’ self-assessed beliefs about their writing abilities, both collaborative writers and comparison individual writers completed self-assessed beliefs of their writing using a questionnaire adopted from Wey (1998). After participants finished the 4-scale questionnaire items, I analyzed their results in terms of learning goal orientation, performance goal orientation, metacognition, self-efficacy and effort. See Appendix J for the questionnaire and Table 3. 5 for the scoring subdivisions. Descriptive analyses and GEE were used to explore the data.

Table 3. 5 Subdivisions of Students' Self-Assessed beliefs about Their Own Writing Abilities (following Wey, 1998) scales sample items Item numbers -One of my primary goals in writing is to Learning Goal understand the concepts. 1,9, 11, 17,21,26 Orientation -One of my primary goals in writing is to acquire new skills. -One of my primary goals in writing is to do Performance Goal better than others. 2,8, 12, 18,23 Orientation - One of my primary goals in writing is to get a high grade. -I determine how to solve a writing task before I begin. Metacognition 3,4, 7, 13, 16,24 -I check how well I am doing when I solve a writing task. -I believe I will receive an excellent grade in a writing assignment. Self-Efficacy 6,15,20,22,27,28 -I'm confident that I can understand the basic concepts taught in writing courses. -I work hard to do well even if I don’t like a Effort writing task. 5,10,14,19,25 -I put forth my best effort on tasks.

Table 3. 6 outlines the data collection and data analysis methods according to the sequence of research questions I attempted to answer in the second question set.

Table 3. 6

Data Collection and Data Analysis Methods for Research Question Set 2

Research issues Instruments Data sets Coding schemes Data analysis techniques

Post writing Students’ responses to 2 Seven-dimensional GEE questionnaire on open-ended questions dyadic communication

Students’ dyadic RQ dyadic or triadic (Appendix K) and numerical taxonomy: scales from or triadic 2.1 communication data sets to multiple choice 1 to 7 communication, (Appendix H) questions attitudes to Group interview Transcripts of collaborative Students’ reflections on Interpretive analysis computer-suppor RQ protocols (Appendix writers’ interviews constructs of the current ted CW and 2.2 I) (Appendix L) CW activity system self-assessed Self-assessment Numerical data sets from Five-dimensional GEE beliefs about questionnaires in the comparison writers and self-assessed beliefs their writing RQ writing platform collaborative writers about writing 2.3 (Appendix J) taxonomy: scales from

1 to 5 (Table 3.5)

Note. RQ means research question.

CHAPTER FOUR

RESULTS

In this chapter, analyses, results, and classroom observations are reported. This chapter is divided into six sections according to the numerical order of the six research questions.

Research Question 1.1

The first research question investigated the tool-mediated writing knowledge generated from students’ uses of particular mediation tools, specifically collaborative writers’ dialogues and their online browsing histories. Two types of tool-mediated knowledge were identified: language forms and skills of expression. Language forms referred to either syntactical or lexical forms and they included vocabulary, translation, collocation, accuracy, and writing templates.

Skills of expression were how rhetorical structures were organized and how arguments and supporting ideas were presented, that is, as organization and argumentation. Illustrations of those two types of knowledge from students’ dialogues and online resource uses, retrieved from the study’s writing platform, are given below.

Dialogue on language forms

Eva and Cindy discussed a writing task they did together, focusing on issues about language forms and skills of expression in the four following extracts. The prompt for the task was: “Do you agree or disagree with the following statement? A teacher’s ability to relate well with students is more important than excellent knowledge of subject being taught. Use specific reasons and examples to support your answer.”

The first extracted dialogue occurred when they were reviewing what they had co-constructed, attending especially to syntactic and morphological forms:

E: add the word, that::: after think

C: add the word, if:::

to see if it is correct after teacher's marking the composition and:::

add the word, a:::

before teacher:::

even if (1.5)

teaches:::

very well, and to listen to:::

him to say.

They not only corrected grammatical errors but also discussed which word to use: arguments or aspects. Cindy insisted on using arguments. After making a decision on wording, they had a question relating to the syntactic category of the word “work”.

C: arguments

E: = or aspects

C: arguments

C: [...] Can the word work be a noun? If it’s not a noun, we can’t put the word here.

E: (1.5) Wait! Let me google it!

Dialogue on skills of expression

In addition to lexical and syntactic forms in English, Cindy and Eva discussed how to integrate their expressions into one common statement. The dialogue below shows them comparing each other’s arguments in their own pre-collaborative written compositions, which they later combined into one statement while writing collaboratively.

C: [...] Teachers were supposed to have certain knowledge to be teachers so if their

relationships with students were not good, students didn’t listen to them.

C: What main ideas did you write?

E: =I wrote that teachers should let students love them first

and teach them.

C: =Our argument was similar.

E: (…) Or we should write something like teachers should not only teach knowledge but

also teach social manners.

Later in the same writing task, Cindy and Eva completed writing their introductory paragraph and were ready to start to write their second paragraph. They were trying to organize their ideas:

C: ((murmuring and jotting something down)) There are three main ideas in the

following.

E: =After we constructed main ideas, we wrote supporting details.

According to my observation, students’ using L1 to accomplish L2 writing tasks highlighted that Mandarin and English played a complementary role for students to clarify grammatical issues and to manage or organize co-constructed arguments.

Online resource use on language forms

Electronic resources, mostly web site links, were provided to collaborative writers to facilitate their writing in the writing platform (as described in Chapter 3). According to the students’ browsing histories on consulting linguistic forms, Google translation was commonly used either to translate or to confirm acquired vocabulary. Figure 4. 1 displays one such instance wherein a student writer consulted Google translation for a sentence that he didn’t know how to put it into English.

Figure 4. 1. A student’s consulting language forms online.

Online resource use on skills of expression

TOEFL writing templates were often visited by the students, and a majority of collaborative writers used them to frame their arguments and ideas. Writing templates functioned as model formats that offered structures to student writers to follow. For example, two CW groups used similar transitional phrases from templates (e.g., in the first place, secondly, lastly, and in conclusion in group g; first of all, additionally, lastly, and to summary in group p) to support their writing flow of presenting their viewpoints towards an argument writing task. For example, to the writing task, “Young people or old people enjoy their life better”, writing group g wrote:

I disagree that young people enjoy life more than older people do…In the first

place, most of older people have the stable source of income and much time to

enjoy their life.…Secondly, their children have already grown up, they do not

take care them anymore….Lastly, I do not think that young people's activities do

not enjoy their life. Likely, singing songs until next morning or drinking. The

kind of activities destroy their health…In conclusion, according to my opinions, I

think the older people have much time and money to enjoy their life.

(Collaborative writing group g, 2011)

Writing group p wrote:

As far as I am concerned I disagree that young people enjoy life more than older

people do.….First of all …government policy for the welfare of the elderly is

getting better and better….Additionally, older people have more money than

young people…. Lastly, older people have more knowledge and experience of

life. …To summary, according to the above opinions, I don’t think that young

people enjoy life more than older people do. (Collaborative writing group p,

2011)

There were two interesting trends in the uses of mediation tools. One was that there were ten groups that did not employ either kind of tool-mediated writing knowledge during their CW over four weeks. Two groups did not employ any language forms, and eight groups did not consult any skills of expression (Table 4. 1). The other trend was that the student writers used those tools two or three times more frequently for generating writing knowledge about

“language forms” than they did for “skills of expression”. On average, each CW group consulted on “language forms” knowledge 22 times per week and “skills of expressions” at least seven times per week. Table 4. 1 tallies these frequencies. The most common situation for peer writers to consult one of these kinds of writing knowledge was when they were uncertain about correct answers to shared writing problems they encountered when they were writing collaboratively.

Table 4. 1 Number of Groups, and Their Frequencies of Making Reference to Tool-Mediated Writing Knowledge on Language Forms and Skills of Expression per Week

Week 1 Week 2 Week 3 Week 4 M f The number of 16 15 16 15 Language groups 22 times Forms Total per group 479 341 235 309 Frequencies The number of 10 15 15 16 Skills of groups 7.5 times Expression per group Total f 90 71 130 132

Through the mediation process students created situated writing knowledge that was specifically useful and meaningful to each writer in the group. Writing knowledge generated in each group was situated and unique; as Brown et al. (1989, p. 32) indicated, “knowledge is situated, being in part a product of the activity, context, and culture in which it is developed and used.” In addition to the situated nature of tool-mediated knowledge, each group generated different amounts of writing knowledge. Some writing groups did not consult writing knowledge on either language forms or skills of expressions; meanwhile other groups consulted language forms much more frequently than skills of expressions. Each individual writer in this study obtained different amounts of unique tool-mediated writing knowledge that was meaningful and useful for solving the person’s writing problems. That was something that classroom lectures cannot offer to individual writers.

Research Question 1.2

This research question asked whether the students’ uses of tool-mediated knowledge influenced their CW performance. Preliminary to carrying out the GEE modeling to investigate this question, I had first to establish scores for the main traits in students’ written compositions and to establish the reliability of the ratings for these scores.

Inter-rater reliability

As described in Chapter 3, three independent raters, Katie, Yu, and Brenda marked the qualities of the compositions written by students for this research. Rater Katie was responsible for marking both collaborative and comparison written compositions. Rater Yu evaluated a quarter of the collaborative written compositions; Rater Brenda completed scoring the rest of the collaborative written compositions and the entire sample of comparison written compositions.

Their inter-rater reliability is reported in three tables below in the order of CW and then comparison writing.

Utilizing Pearson’s r correlations, the resulting inter-rater reliability estimates for the first pair of raters (raters Katie and Yu), who both marked one portion of the CW, were relatively strong (r=.77 to .88, for 642 decisions derived from five writing scales and one holistic rating for 107 written compositions) as Table 4. 2 shows.

Table 4. 2 Rater Katie and Rater Yu’s Inter-Rater Reliability Scoring 107 Collaboratively Written Compositions CW (CW): Rater Katie Measures (r) M SD 1 2 3 4 5 6

1. CQ .85** 5.30 1.27

2. Org .88** 5.19 1.28 CW 3. Arg .87** 5.34 1.24 Rater 4. Lacc .80** 5.22 1.23 Yu 5. Lapp .77** 5.61 1.09

6. Holistic .86** 5.33 1.17

M 5.49 5.20 5.35 5.25 5.72 5.32

SD 1.19 1.32 1.17 1.09 1.16 1.13

Note. CQ=communicative quality; Org = organization; Arg = argumentation; Lacc = linguistic accuracy; Lapp = linguistic appropriacy. All correlations were significant at the 0.01 level (**2-tailed).

The inter-rater reliability estimates for the second pair of raters (raters Katie and Brenda), who marked the other major portion of CW were also strong (r=.85 to .93, for 1902 decisions derived from five writing scales and one holistic rating for 317 written compositions) as Table 4.

3 displays.

Table 4. 3 Rater Katie and Rater Brenda’s Inter-Rater Reliability Scoring 317 Collaboratively Written Compositions CW (CW): Rater Katie Measures (r) M SD 1 2 3 4 5 6

1. CQ .90** 5.46 1.22

2. Org .93** 5.26 1.33 CW: 3. Arg .91** 5.41 1.32 Rater 4. Lacc .85** 5.23 1.09 Brenda 5. Lapp .85** 5.71 1.11

6. Holistic .91** 5.44 1.18

M 5.39 5.15 5.32 5.13 5.76 5.25

SD 1.22 1.33 1.28 1.06 1.13 1.20

Note. CQ=communicative quality; Org = organization; Arg = argumentation; Lacc = linguistic accuracy; Lapp = linguistic appropriacy. All correlations were statistically significant at the 0.01 level (**2-tailed).

The strength of association between the third pair of raters who evaluated the comparison writing was strong too (r=.87 to .91, for 936 decisions derived from five writing scales and one holistic rating for 156 written compositions in Table 4. 4). Averaged scores from the three pairs of raters were used to carry out the main statistical calculations for answering

Research Questions 1.2 and 1.3.

Table 4. 4 Rater Katie and Rater Brenda’s Inter-Rater Reliability Scoring 156 Comparison Written Compositions Comparison Writing (BW): Rater Katie Measures (r) M SD 1 2 3 4 5 6

1. CQ .90** 5.72 1.27

2. Org .89** 5.38 1.33 BW 3. Arg .88** 5.49 1.37 Rater 4. Lacc .88** 5.33 1.27 Brenda 5. Lapp .87** 5.33 1.28

6. Holistic .91** 5.64 1.26

M 5.79 5.47 5.47 5.36 5.52 5.55

SD 1.26 1.34 1.28 1.22 1.19 1.23

Descriptive analysis of CW performance

Table 4. 5 Means and Standard Deviations of Collaboratively Written Compositions in Weeks 1 and 2 week 1 week 2 n=36 M (SD) Min Max M (SD) Min Max Communicative quality 5.16 (1.16) 2.50 7.50 5.60 (0.74) 3.50 6.50 Organization 4.96 (0.98) 2.50 7.00 5.33 (1.16) 2.50 7.00 Argumentation 5.12 (1.19) 2.00 7.00 5.58 (1.01) 3.00 7.00 Linguistic accuracy 5.24 (1.08) 2.00 6.75 5.57 (0.77) 4.50 7.00 Linguistic appropriacy 5.51 (1.32) 2.00 8.00 5.71 (0.55) 4.50 6.50 Holistic scores 5.00 (1.00) 2.50 6.75 5.53 (0.61) 4.00 6.25

Table 4. 6 Means and Standard Deviations of Collaboratively Written Compositions in Weeks 3 and 4 week 3 week 4 n=36 M (SD) Min Max M (SD) Min Max Communicative quality 5.83 (0.78) 4.50 8.00 6.06 (0.84) 4.50 7.00 Organization 5.63 (0.84) 4.00 7.00 5.96 (0.92) 4.00 7.50 Argumentation 5.76 (0.97) 4.50 8.00 5.94 (1.06) 4.25 8.00 Linguistic accuracy 5.51 (0.55) 4.50 6.00 5.88 (1.05) 3.50 7.75 Linguistic appropriacy 6.06 (0.92) 4.50 8.00 6.25 (0.69) 4.75 7.25 Holistic scores 5.70 (0.83) 4.50 8.00 6.03 (0.87) 4.75 7.25

The collaborative writers appear to have increased their mean scores slightly and progressively on all the writing scales over four weeks. That is, the more writing and CW activities the students experienced, the better they wrote in their co-constructed writing performance on different writing tasks. Among the scales, the students’ writing had the highest scores for linguistic appropriacy over all four weeks, and had the lowest scores on organization

59 in weeks one and two and in linguistic accuracy in weeks three and four. The GEE analyses, reported next, examined these apparent changes to determine whether tool-mediated knowledge and time were factors that contributed to the changes on separate writing dimensions.

The association of mediation and weeks with students’ development

On holistic writing performance

Results from the GEE analyses indicated that language forms (F), skills of expression (S) and weeks (W) were all statistically significant in the model in interpreting the students’ holistic writing scores (HS).

Table 4. 7 Parameter Estimates for the Model Effect of Mediation and Time on Holistic Scores of CW

Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 2.95 0.44 (2.09, 3.82) p=.00* Language forms Forms_M1* 1.50 (.81, 2.20) p=.00* Forms_M0* 0 Skills of expressions Skills_M1* 0.76 0.33 (.12, 1.41) p=.02* Skills_M0* 0 Weeks week 4 0.86 0.25 (.36, 1.35) p=.00* week 3 0.48 0.37 (-.24, 1.20) p=.19 week 2 0.50 0.36 (-.22, 1.21) p=.17 week 1 0 . . *M1was with mediation; M0 was without mediation; 0=reference point Outcome variable: holistic scores of CW performance

Table 4. 7 shows that tool-mediated knowledge (both language forms and skills of expression) and practice over time contributed significantly to the holistic scores on compositions written through CW. The type of mediated knowledge also made a difference.

Student writers who used “language forms” mediation performed 1.5 points more than the students who did not use “language forms” mediation. Similarly, student writers who used

“skills of expression” mediation performed .76 points more than students who did not use this type of mediation. As for time, compared to week 1, collaborative writers performed significantly better in week 4. Even though students made some positive changes to their writing in week 2 and 3, those changes were not statistically significantly different from week 1. In comparison with week 1, the distinct improvement occurred to writing in week 4 when students increased .86 scores in their CW performance. These results show that student writers’ adoption of tool-mediated knowledge had positive influences on the holistic scores of their collaboratively written compositions, but discernible changes in the quality of their compositions needed time and practice to be realized.

On communicative quality

GEE analyses indicated that language forms (F) and skills of expression (S) were statistically significant in a model interpreting the communicative quality (CQ) of compositions written collaboratively.

Table 4. 8 Parameter Estimates for the Model Effect of Mediation and Time on Communicative Quality of CW

Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 2.36 0.58 (1.22, 3.51) p=.00* Language forms Forms_M1* 2.02 0.30 (1.44, 2.60) p=.00* 0 Forms_M0* Skills Skills_M1* 1.14 0.40 (0.35, 1.92) p=.01* 0 Skills_M0* Weeks week 4 0.61 0.31 (0.01, 1.21) p=.05 week 3 0.32 0.41 (-0.49, 1.12) p=.44 week 2 0.33 0.40 (-0.46, 1.13) p=.41 0 week 1 *M1 was with mediation; M0 was without mediation; 0=reference point Outcome variable: communicative quality of CW performance

Table 4. 8 demonstrates that the use of tool-mediated knowledge, both skills of expression and language forms, was associated significantly with the communicative quality of collaboratively written compositions. Student writers who used “language forms” mediation performed 2.02 points above the student writers who did not use “language forms” mediation.

Student writers who used “skills of expression” mediation performed 1.14 points beyond the student writers who did not use it. Even though the compositions written collaboratively showed minor differences in communicative quality in weeks 2, 3 and 4, the degree of improvement was not significantly different from week 1.

On organization

Organization was evaluated in terms of if the message of the writing was logically organized and the logic enabled the message to be followed effortlessly. Time in weeks (W) was the only statistically significant predictor in this GEE model for interpreting changes in the organization scores (ORG) on compositions that students had written collaboratively. The

62 results in Table 4.9 indicate that the student writers improved the rhetorical organization of their written compositions with practice and over several times writing collaboratively, but it took four weeks to show this difference. Their uses of tool-mediated knowledge did not appear to be associated with these changes in the rhetorical organization of the written compositions.

Table 4. 9 Parameter Estimates for the Model Effect of Mediation and Time on Organization of CW

Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 4.78 0.84 (3.13, 6.43) p=.00* Language forms Forms_M1* -0.04 0.97 (-1.93, 1.86) p =.97 0 Forms_M0* Skills Skills_M1* 0.29 0.44 (-0.57, 1.15) p =.51 0 Skills_M0* Weeks week 4 0.87 0.28 (0.33, 1.42) p =.00* week 3 0.57 0.30 (-0.03, 1.16) p =.06 week 2 0.39 0.40 (-0.40, 1.18) p =.33 0 week 1 *M1 was with mediation; M0 was without mediation; 0=reference point Outcome variable: organization of CW performance

On argumentation

Argumentation was assessed in terms of if relevant arguments were presented with main ideas and effective supporting material. Skills of expression (S) was the only statistically significant predictor in this GEE model interpreting changes in the argumentation scores (ARG) on compositions that students had written collaboratively. Table 4. 10 shows that across four weeks of CW, the groups of student writers who used mediation involving skills of expression were judged to have 1.01 points better qualities of argumentation in their written compositions than did the students who did not use skills of expression mediation.

Table 4. 10 Parameter Estimates for the Model Effect of Mediation and Time on Argumentation of CW

Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 3.17 0.84 (1.51, 4.82) p =.00* Language forms Forms_M1* 1.23 0.71 (-0.16, 2.63) p =.08 0 Forms_M0* Skills Skills_M1* 1.01 0.38 (0.27, 1.76) p =.01* 0 Skills_M0* Weeks week 4 0.54 0.29 (-0.02, 1.10 ) p =.06 week 3 0.34 0.36 (-0.37, 1.05) p =.35 week 2 0.39 0.43 (-0.46, 1.23) p =.37 0 week 1 *M1 was with mediation; M0 was without mediation; 0=reference point Outcome variable: argumentation of CW performance

On linguistic accuracy

Linguistic accuracy was evaluated in terms of errors of vocabulary, spelling, punctuation, or grammar. Language forms (F), skills of expression (S), and weeks (W) were all significant predictors in this GEE model interpreting student writers’ linguistic accuracy (Lacc) scores on their written compositions.

Table 4. 11 Parameter Estimates for the Model Effect of Mediation and Time on Linguistic Accuracy of CW

Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 1.28 0.79 (-0.27, 2.83) p =.11 Language forms Forms_M1* 3.10 0.38 (2.35, 3.86) p =.00* 0 Forms_M0* Skills Skills_M1* 1.22 0.36 (0.51, 1.92) p =.00* 0 Skills_M0* Weeks week 4 0.38 0.17 (0.04, 0.72) p =.03* week 3 -0.21 0.33 (-0.66, 0.62) p =.95 week 2 0.27 0.40 (-0.52, 1.05) p =.51 0 week 1 *M1 was with mediation; M0 was without mediation; 0=reference point Outcome variable: linguistic accuracy of CW performance

Table 4. 11 shows that uses of tool-mediated knowledge (both language forms and skills of expression) and time (in weeks) had positive associations with changes in the linguistic accuracy of the student writers’ compositions written collaboratively. Student writers who used

“language forms” mediation performed 3.10 points better than student writers who did not use

“language forms” mediation. In turn, student writers who used “skills of expression” mediation performed 1.22 points better than the student writers who did not use “skills of expression” mediation. Comparing the scores on linguistic accuracy in week 1 to week 4, the collaborative writers on the whole produced compositions that had significantly better (.38 higher scores) linguistic accuracy in week 4, indicating again that this kind of improvement took time to materialize.

On linguistic appropriacy

Mediation involving language forms (F) was the only statistically significant predictor in the GEE model interpreting changes in the linguistic appropriacy (Lapp) scores of compositions by the students who wrote collaboratively. As Table 4.12 shows, over four weeks of writing, students who used “language form” mediation performed 2.08 points better in the linguistic appropriacy of their compositions than the students who wrote collaboratively without using

“language form” mediation. These changes did not differ significantly over time nor were they associated with mediation that involved “skills of expression”.

Table 4. 12 Parameter Estimates for the Model Effect of Mediation and Time on Linguistic Appropriacy of CW Parameter Estimates SE 95% CI (lower, upper) Sig. Intercept 2.61 0.60 (1.43, 3.79) p =.00* Language forms Forms_M1* 2.08 0.35 (1.39, 2.78) p =.00* 0 Forms_M0* Skills Skills_M1* 1.12 0.62 (-0.08, 2.33) p =.07 0 Skills_M0* Weeks week 4 0.54 0.31 (-0.07, 1.15) p =.09 week 3 0.23 0.30 (-0.37, 0.82) p =.45 week 2 0.05 0.27 (-0.48, 0.57) p =.87 0 week 1 *M1 was with mediation; M0 was without mediation; 0=reference point Outcome variable: linguistic appropriacy scores of CW performance

Observations

Writers in this study mostly adopted a deductive writing approach to accomplish their writing. Results from the model effect described above suggest that these L2 writers’ development on deductive writing organization was fostered by weekly practice. Two rationales

66 for these writers to adopt this particular type of organization are: instruction on this organization approach and the nature of deductive writing.

Writers widely using a deductive writing approach over four weeks may have resulted from the writing instruction they received in the classroom. Deductive writing displays main ideas at the beginning of the written composition and then examples or evidence follow the main ideas. Deductive logic has a relatively easy writing convention for writers to follow. Given that their writing had to be done in a timely manner, the students may naturally have taken this approach because it required less time to master the approach to compose than an inductive writing approach would have. Deductive organizational patterns are also flexible for collaborative writers to co-develop an argument in their writing compared to an inductive organizational pattern. For collaborative writers, adopting deductive writing allowed all the writing members to integrate group members’ ideas readily. These L2 writers reported that organization worked well in CW week two and week three (Table 4. 19). In their interviews, the students reported that what they benefited most from collaborative writing was members’ sharing different perspectives, and that facilitated their writing with more supporting ideas (see interview data in Appendix L). As times of practicing writing increased, this easy and flexible approach was readily undertaken in every writing task, and the writers’ development of a deductive writing approach was fostered with times of CW.

The development of argumentation was associated with one factor, skills of expression.

As collaborative writers’ consulted knowledge of skills of expression, either from verbal interactions or online consultations, they gradually added appropriate supporting ideas or examples to their writing. Writers also expressed in their after-collaborative writing questionnaire that content discussion and integrating members’ ideas worked well during collaborative writing (Table 4. 19). Furthermore, some interviewees pointed out that the

67 argumentative writing skills had been reinforced by collaborative writers’ sharing supporting ideas and examples (see interview data in Appendix L).

For these L2 writers, grammatical errors were the most common and most difficult problems in writing in English. Accordingly, the development of linguistic accuracy involved all factors’ contributions. Without mediation and weekly practices, it was hard for the L2 learners to write lexically and syntactically correct sentences. It was evident that comparison writers who only received classroom instruction did not develop their English proficiency distinctly in respect to linguistic accuracy.

Given the difficulty of achieving linguistic accuracy, collaborative writers were more concerned with vocabulary than with grammar in revising their writing; for some writers, editing was done in the sense that they tried for the most precise words they could get. Also, writers’ reporting on their not working well in vocabulary and grammar actually reflected that even though their writing improved, they were trying hard to get correct uses of linguistic forms.

Grouping effects

Even though the above GEE modeling on every aspect of writing performance indicated collaborative writers’ improvement, possible grouping effects could be observed and inferred from whether collaborative writers consulted mediation tools or not during CW. For example,

CW scores from groups l and n (that always used mediation tools) were actually better over time than average CW scores (as shown in Figure 4. 2). However, scores from groups c and p (that sometimes consulted mediation tools) did not always maintain their scores above the average

CW scores (see Figure 4. 3). Collaborative writing scores over time are listed in Appendix M.

On the whole, grouping effects were so variable that they defied systematic analysis.

Figure 4. 2. Holistic scores of group l and group n that always used mediation tools.

Figure 4. 3. Holistic scores of group c and group p that not always used mediation tools.

Research Question 1.3

Three analyses were conducted to investigate whether tool-mediated knowledge encouraged groups’ ZPDs to facilitate their individual writing development. Test one contrasted collaborative writers’ holistic performance in the before- and after-program tests with comparison individual writers’ performance. Test two evaluated both groups of students’ over-time differences in relation to particular writing dimensions. Test three correlated CW changes with collaborative writers’ individual writing changes.

The purpose of contrasting collaborative writers’ holistic scores with comparison writers’ in their before-program tests was to make sure that the writing performance from both classes was statistically comparable from the outset. As Polio (2012) suggested, a pretest was needed to determine whether the intact classes were equivalent. If both classes started with similar writing performance in the before-program test, and then collaborative writers performed better than comparison writers did in the after-program test, writing knowledge transferred from tool-mediated knowledge to students’ individual writing development can possibly be inferred.

By comparing collaborative writers’ improvement with comparison writers’ improvement, I aimed to distinguish students’ development from gaining tool-mediated writing knowledge or classroom instruction. Finally, correlations between collaborative writers’ individual writing changes over time were conducted to determine if their association was positive, and so to see if it can be inferred that tool-mediated writing knowledge fostered collaborative writers’ individual writing development (as indicated by the collaborative writers’ after-program writing development). GEE and Pearson Correlation Coefficients were used to address these inquiries.

Descriptive analysis of collaborative and comparison writers’ performance in their before- and after-program tests

Performance differences between the two classes are shown in the Table 4. 13 and Table

4.14. In the before-program test, collaborative writers scored on average 0.28 less than the comparison writers did in communicative quality, 0.13 less in organization, 0.06 less in argumentation, 0.27 more in linguistic accuracy, 0.77 more in linguistic appropriacy, and 0.17 less in holistic scores. In the after-program test, collaborative writers on average scored higher than the comparison writers did, for example, 0.50 more in communicative quality, 0.58 more in organization, 0.83 more in argumentation, 0.67 more in linguistic accuracy, 1.12 more in

70 linguistic appropriacy, and 0.57 more in holistic scores. To evaluate if those observed differences were statistically significant, GEE was adopted.

Table 4. 13 Collaborative Writers’ Means and Standard Deviations on Their Before and After Program Tests CW group n=36 Before-program test After-program test M (SD) Min Max M (SD) Min Max Communicative 4.47 (1.04) 2.50 6.50 5.91 (0.70) 4.50 7.75 quality Organization 4.13 (0.95) 2.00 6.00 5.78 (0.91) 4.00 8.00 Argumentation 4.32 (1.16) 2.00 6.50 5.97 (0.74) 4.75 7.75 Linguistic 4.65 (1.15) 2.50 7.50 5.49 (0.64) 4.50 7.00 accuracy Linguistic 5.22 (0.86) 3.50 8.00 6.03 (0.79) 4.75 8.00 appropriacy Holistic scores 4.31 (0.92) 2.50 6.25 5.82 (0.73) 4.00 7.75

Table 4. 14 Comparison Writers’ Means and Standard Deviations on Their Before and After Program Tests Comparison writing group n=26 Before-program test After-program test M (SD) Min Max M (SD) Min Max Communicative 4.75 (1.20) 2.00 6.75 5.41 (0.93) 3.00 7.00 quality Organization 4.26 (1.22) 2.00 6.75 5.20 (1.22) 2.00 7.00 Argumentation 4.38 (1.16) 2.00 6.50 5.14 (1.07) 2.00 6.50 Linguistic accuracy 4.38 (1.09) 2.00 6.50 4.82 (1.08) 2.50 7.00 Linguistic 4.45 (1.10) 2.00 6.75 4.91 (0.99) 2.00 7.00 appropriacy Holistic scores 4.48 (1.13) 2.00 6.75 5.25 (0.97) 2.75 7.00

Were the two writing groups comparable?

GEE results shown in Table 4.15 indicate that the student writers in the two classes did not start differently (p=.51) in holistic writing performance, but the students had statistically significant different performances in the after-program tests (p=.01). Students in the CW class performed better than the comparison writing class did on holistic scores in the after-program test (p = .01). This trend appeared as well in other dimensions of writing for the two classes in the before- and after- program tests, specifically communicative quality (before-program: p=.32; after-program: p=.02), organization (before-program: p=.65; after-program: p=.04), argumentation (before-program: p=.85; after-program: p=.00), and linguistic accuracy

(before-program: p=.33; after-program: p=.00). As for linguistic appropriacy, the two classes of writers started differently and also ended differently (before-program: p=.00; after-program: p=.00).

Table 4. 15 GEE Analysis of Collaborative and Comparison Writers’ Before and After-Program Performance on Each Aspect of Writing Mean Outcome Parameter SE 95% CI (lower, upper) Sig. variables Difference CW -0.18 0.27 (-0.70, 0.34) p=.51 before BW 0* Holistic rating CW 0.57 0.22 (0.13, 1.01) p=.01* after BW 0* CW -0.28 0.29 (-0.85, 0.28) p=.32 before Communicative BW 0* quality CW 0.50 0.21 (0.08, 0.91) p=.02* after BW 0* CW -0.13 0.28 (-0.68, 0.42) p=.65 before BW 0* Organization CW 0.58 0.28 (0.04, 1.13) p=.04* after BW 0* CW -0.06 0.29 (-0.63, 0.52) p=.85 before BW 0* Argumentation CW 0.83 0.24 (0.36, 1.30) p=.00* after BW 0* CW 0.28 0.28 (-0.28, 0.83) p=.33 before Linguistic BW 0* accuracy CW 0.67 0.23 (0.21, 1.12) p=.00* after BW 0* CW 0.76 0.25 (0.27, 1.26) p=.00* before Linguistic BW 0* appropriacy CW 1.11 0.23 (0.66, 1.57) p=.00* after BW 0* *0=reference performance *SE=standard error *95%CI=95% Wald confidence interval for difference

Differences in collaborative writers and comparison writers’ over-time development

In comparing results from the after-program test with the before-program test (see Table

4.16), collaborative writers improved their writing performance on every dimension of writing by scores of 1.51 more holistically (p=.00), 1.44 (p=.00) more in communicative quality, 1.65

73 more in organization (p=.00), 1.65 more in argumentation (p=.00), 0.84 more in linguistic accuracy (p=.00), and 0.81 more in linguistic appropriacy (p=.00).

Table 4. 16 GEE Analysis of Collaborative Writers’ Within-Group Development

Outcome 95% CI Parameter Estimate SE Sig. variables (lower, upper)

Intercept 4.31 0.15 (4.01, 4.60) p=.00* Holistic rating after 1.51 0.15 (1.21, 1.81) p=.00* before 0* Intercept 4.47 0.17 (4.13, 4.80) p=.00* Communicative after program 1.44 0.19 (1.08, 1.80) p=.00* quality before program 0* Intercept 4.13 0.16 (3.83, 4.44 ) p=.00*

1.65 0.19 (1.28, 2.03 ) p=.00* Organization after before 0* Intercept 4.32 0.19 (3.95, 4.69 ) p=.00* Argumentation after 1.65 0.17 (1.31, 1.99 ) p=.00* before 0* Intercept 4.65 0.19 (4.28, 5.02) p=.00* Linguistic accuracy after 0.83 0.18 (0.48, 1.19) p=.00* before 0* Intercept 5.22 0.14 (4.93, 5.49) p=.00* Linguistic 0.81 0.15 (0.53, 1.10) p=.00* appropriacy after before 0* *0=reference performance *SE=standard error *95% CI=95% Wald confidence interval

Comparison writers also improved their scores on every writing scale (p=.00) except for linguistic accuracy (see Table 4.17). They scored 0.77 more in holistic scores (p=.00), 0.66 more in their communicative quality (p=.00) in the after-program test, 0.94 more in organization (p=.00), 0.76 more in argumentation (p=.00), and 0.46 more in linguistic appropriacy (p=.02).

Table 4. 17 GEE Analysis of Comparison Writers’ Within-Group Development

Outcome 95% CI Parameter Estimate SE Sig. variables (lower, upper)

Intercept 4.48 0.22 (4.05, 4.90 ) p=.00* HS after 0.77 0.18 (0.41, 1.18 ) p=.00* before 0* Intercept 4.75 0.23 (4.30, 5.20) p=.00* CQ after program 0.66 0.2 (0.27, 1.06 ) p=.00* before program 0* Intercept 4.26 0.23 (3.80, 4.72 ) p=.00* ORG after 0.94 0.22 (0.51, 1.38 ) p=.00* before 0* Intercept 4.38 0.22 (3.94, 4.81 ) p=.00* ARG after 0.77 0.21 (0.35, 1.19 ) p=.00* before 0* Intercept 4.38 0.21 (3.96, 4.79) p=.00* Lacc after 0.44 0.22 (0.01, 0.87 ) p=.05 before 0* Intercept 4.45 0.21 (4.04, 4.87) p=.00* Lapp after 0.46 0.2 (0.07, 0.85) p=.02* before 0* *0=reference performance *SE=standard error *95% CI=95% Wald confidence interval

Correlations between collaborative writing changes and individual writing development

Pearson correlations showed positive associations between changes in collaborative writing over four weeks of the course and individual writing (as indicated in the post-program test) in all dimensions of writing (see Table 4. 18): holistic scores (r=.63, n=144 , p=.00), communicative quality (r=.66, n=144 , p=.00), organization (r=.65, n=144 , p=.00), argumentation (r=.59, n=144 , p=.00), linguistic accuracy (r=.68, n=144 , p=.00), and linguistic

75 appropriacy (r=.57, n=144 , p=.00). The bigger the CW changes were, the better the individual writing development was.

Table 4. 18 Correlations between CW and Individual Writing Changes among Writing Dimensions CW (N=144) Measures (r) 1 2 3 4 5 6 M SD 1. HS .63* 1.02 1.39 2. CQ .66* 1.01 1.43 Individual 3. ORG .65* 1.14 1.45 Writing

(N=144) 4. ARG .59* 0.97 1.54 5. Lacc .68* 0.48 1.39 6. Lapp .57* 0.57 1.47 M 0.73 0.75 0.84 0.72 0.64 0.46 SD 1.25 1.39 1.43 1.44 1.37 1.32 Note. CQ=communicative quality; Org = organization; Arg = argumentation; Lacc = linguistic accuracy; Lapp = linguistic appropriacy. All correlations were statistically significant at the 0.01 level (**2-tailed).

Research Question 2.1

This research question investigated how the CW student writers changed their communication relationships over time. The CW questionnaire was administered after students'

CW practices in weeks two, three, and four. Thirty-nine questions were included in the questionnaire: two open-ended questions and 37 Likert-scaled questions. Open-ended questions asked students to indicate what aspects of writing they thought worked and did not work well during each specific week of CW. The rest of the 37 Likert-scaled questions covered seven dimensions of students’ communication relationships. GEE analyses examined students’ over-time changes in their CW relationships.

What worked well?

Table 4. 19 and Table 4. 20 summarize these results and show the percentage ratio of each category (of what the students said worked well) to the entire set of categories that I

76 established. The collaborative writers felt the most satisfied with their discussions of content

(25%), organization (16%), and perspective-taking (16%) in the second week of CW. Content discussions involved the students communicating about what ideas to write. Discussions about organization concerned the logic or relevance of main ideas to support an argument.

Perspective-taking concerned which side of arguments was adopted, what the opposing position might be, and whether members in a writing group took the same perspective towards an argument or not. The top three categories of CW discussed in the third week remained the same as in week two but in a different order of priority: organization (29%), content discussion (14%), and perspective-taking (10%). In week four, except for frequent content discussions (19%), the student writers shifted their focuses to working well on "verbalizing ideas from Chinese to

English" (17%) and "integration of writing members' ideas" (16%).

Table 4. 19 Categorized Answers to Question One on the After-CW Questionnaire: What Worked Really Well in This Task? week 2 Times week 3 times week 4 Times discussion/ 17 17 13 organization content discussion negotiation (25%) (29%) (19%)

11 8 verbalizing/word 12 organization content discussion (16%) (14%) choice (17%) perspectives/ 11 perspectives/ 6 11 integration of ideas consensus (16%) consensus (10%) (16%)

28 27 34 others others others (43%) (47%) (48%) 67 58 70 total total total (100%) (100%) (100%) Note. Only the categories that were reported in the top three frequencies are listed in the Table and the rest of the categories were “others”. See Appendix K for full categorizations.

What did not work well?

In regards to the aspects of CW that the students did not perceive to work well, grammar and vocabulary were reported most frequently in all weeks of writing. In addition, verbalizing in

English was reported in week two and timing was reported in weeks three and four. Interestingly, these student writers first identified ideas, flow, and coherence as dissatisfying issues during

CW in week three (see Table K2 in Appendix K). Overall, differences reported between weeks two and three were that students shifted their attention more to global issues of writing. In week four, students first identified discussion engagement and online resource use (e.g., too much reliance on mediation tool use) as issues not working well during collaboration. In general, these students' reports on what did not work well indicated their awareness of what they lacked in CW, extending from language forms, texts, and mediation tool use to discussion processes.

Table 4. 20 Categorized Answers to Question Two: What Did Not Work Well in This Task? week 2 times week 3 times week 4 times grammar 17 (24%) grammar 17 (22%) grammar 11 (17%)

vocabulary 17 (24%) vocabulary 15 (20%) vocabulary 9 (14%)

verbalization 8 (11%) timing 11 (15%) timing 8 (12%)

others 29 (41%) others 32 (43%) others 37 (57%)

75 65 total 71 (100%) total total (100%) (100%) Note. Only the categories that were reported in top three frequencies are listed in the Table and the rest of the categories were “others”. See Appendix K for full categorizations.

Collaborative writers’ communications and relationships

Descriptive analyses of responses to the Likert-scale items on the Post-writing

Questionnaire showed only minor differences in the means and standard deviations across seven

78 aspects of the student writers’ perceptions of their communications and relationships in groups over time. The seven dimensions of within-group communications and relationships were: quality of discussion, process satisfaction, agreement, communication richness, task discussion effectiveness, communication appropriateness, and teamwork. Ranges in the averaged means of all seven dimensions over three weeks, shown in Table 4.21, were: 4.66 (SD=0.96) to 5.81

(SD=1.00) in week two, 4.94 (SD=1.10) to 5.66 (SD=1.15) in week three (Table 4. 21), and 5.01

(SD=1.05) to 5.52 (SD=0.98) in week four.

Table 4. 21 Means and Standard Deviations of Self-ratings of Collaborative Writers’ Communications and Relationships in Weeks Two, Three, and Four week 2 week 3 week 4 M Min Max M Min Max M Min Max n=36 (SD) (SD) (SD) Quality of 4.66 4.94 5.05 2.25 7 3 7 2.75 7 discussion (0.96) (1.10) (1.13) Process 5.13 5.26 5.39 3.2 7 2.2 7 3.8 7 satisfaction (0.96) (1.09) (0.93) 5.53 5.62 5.47 Agreement 4 7 3.4 7 3.6 7 (0.96) (1.00) (1.03) Communication 4.69 5.00 5.01 3 7 2.5 7 3.25 7 richness (1.00) (1.19) (1.05) Task discussion 4.91 5.03 5.10 3.38 7 3.38 7 3.62 7 effectiveness (0.96) (0.96) (0.92) Communication 5.44 5.33 5.32 3 7 3.67 7 4 7 appropriateness (0.93) (1.04) (0.96) 5.81 5.66 5.52 Teamwork 3.62 7 3 7 4 7 (1.00) (1.15) (0.98)

GEE analyses examined if any dimension of the student writers’ perceptions of their communications and relationships in groups changed over time. Only two of seven dimensions

79 showed statistically significant differences over time (please refer to Table 4.22). Collaborative writers rated their communication richness 0.31 points better in week three compared to week two. In contrast, teamwork was rated 0.45 points less in week four compared to week two.

Possible explanations for the teamwork score reduction with time were that collaborative writers adopted fixed-mode cooperation, which made some group members do the same task during

CW, like typing or correcting grammar, and such fixed-mode cooperation did not help improve team members’ perceptions of their interactions.

Table 4. 22 Model Effect of Time on Self-ratings of Collaborative Writers’ Communications and Relationships

95% CI (lower, Outcome variables Parameter Estimate SE Sig. upper)

Intercept 4.66 0.16 (4.35, 4.97) p=.00* week 4 3 2.54 (-1.98, 7.98) p=.28 Quality of discussion week 3 0.29 0.15 (-0.01, 0.58) p=.06 week 2 0* Intercept 5.13 0.16 (4.82, 5.44) p=.00* week 4 0.12 0.2 (-0.28, 0.51) p=.57 Process satisfaction week 3 0.13 0.13 (-0.13, 0.39) p=.34 week 2 0* Intercept 5.53 0.16 (5.22, 5.84) p=.00* week 4 -0.21 0.19 (-0.58, 0.17) p=.29 Agreement week 3 0.09 0.12 (-0.14, 0.33) p=.44 week 2 0* Intercept 4.69 0.16 (4.37, 5.02) p=.00* week 4 0.17 0.17 (-0.16, 0.50) p=.30 Communication richness week 3 0.31 0.13 (0.06, 0.56) p=.02* week 2 0* Intercept 4.91 0.16 (4.60, 5.22) p=.00* week 4 0.05 0.21 (-0.35, 0.46) p=.80 Task discussion effectiveness week 3 0.12 0.13 (-0.13, 0.37) p=.33 week 2 0* Intercept 5.44 0.15 (5.13, 5.73) p=.00* week 4 -0.26 0.21 (-0.67, 0.15) p=.21 Communication appropriateness week 3 -0.1 0.12 (-0.34, 0.14) p=.40 week 2 0* Intercept 5.81 0.16 (5.49, 6.13) p=.00* week 4 -0.45 0.2 (-0.37, 0.06) p=.03* Teamwork week 3 -0.16 0.11 (-0.37, 0.06) p=.16 week 2 0* *0=reference performance *SE=standard error *95% CI=95% Wald confidence interval

Research Question 2.2

Group interviews elicited the student writers’ reflections on CW. Interview protocols included questions linked to the main components of Activity Theory: mediation, object, outcome, division of labour, rules, and community. Appendix L shows the full set of interview data.

Mediation and objects

Generally, the interviewees expressed positive views towards the help of mediation tools in their writing. Most of the students said they used two to four websites to help themselves write. One writer reported strategically using more than four tools. Two students in one group said that they had never employed any material mediation tools to write.

The types of websites mostly involved translation, dictionaries, sentence patterns, writing templates, and grammatical conjunctions. One group of students reported they tried to use collocation websites, and found them useful but hard to use effectively in a timely manner because they were only allowed one hour to write. The role of mediation tools was often to check or brush up students’ memories of acquired vocabulary and to learn new writing skills and linguistic knowledge.

One student writer stated that checking an online dictionary reminded her of acquired vocabulary but she could not acquire new vocabulary from it. Sometimes she checked the same word up to four times, and its usage still needed clarification through the Internet again. She said she used new words in the CW but she was not be able to use them later in her individual writing because they were passive rather than active vocabulary. On the contrary, another writer in another writing group indicated that, “we checked one word up in one writing task again and again. However, after several times of writing, we didn’t need to check it again” (Group O in

Table L8 of Appendix L). Another group of students said that they made notes of sentence patterns retrieved from websites to facilitate their writing the next time. They tried to use different sentence patterns in different writing tasks. In terms of acquiring new patterns, they responded that they “can remember something from the use of mediation tools” (Group G in

Table L4 of Appendix L). Another writer said he used new words in his individual writing once the words had been used twice before. He also said he reduced his times consulting mediation websites, such as sentence patterns, Google translation, and conjunctions as he did more writing, except for the online dictionary because “different writing tasks required different fields of vocabulary” (Group B in Table L2 of Appendix L).

Some collaborative writers reported that they rarely used online resources. They said that they built up their writing knowledge and skills from their own verbal language only. Even though most of the students used mediation tools to facilitate their CW, some students expressed concerns that too much reliance on the tools caused them to worry about being able to do

“writing without tools”.

Objects to outcomes

Idea expansion and verbalizing and wording in English

The principal contribution of CW to individual writing reported by the students was in idea expansion, which prompted the students to make more precise or accurate their verbalization of wording in English. Most interviewees indicated that idea expansion directly produced positive influences on their individual writing skills.

In pursuing the expansion of their ideas, the students said they learned how to group their ideas into paragraphs and how to articulate and complete their sentences. Also, with increased practice writing, the students said they gradually grasped certain writing skills that

83 allowed them to express themselves and match their wording with ideas that they wanted to express. As one writer stated, “It was good to have CW to complete ideas; otherwise there was no difference between the first time of writing [pre-CW] and the third time of writing

[post-CW]” (Group N in Table L7 of Appendix L).

Transfer of collaborative to individual writing

Some of the students said they followed the train of thought employed during their CW to apply it to their subsequent individual writing (regardless of what they had written the first time performing the task individually). One of the students stated her reason for adopting and modifying her CW to her individual writing: “I was not good at grammar so I wrote my individual writing on the basis of CW” (Group G in Table L4 of Appendix L).

Another student reported that, “the line of thinking got clearer and clearer after CW”

(Group E in Table L3 of Appendix L). A different student said, “I didn't switch my standpoint in any of my individual writing. …I was glad to have CW to enrich my examples” (Group E in

Table L3 of Appendix L). A different student observed that she sometimes applied the words obtained from material mediation during CW to her individual writing: “I can remember some words and phrases in templates and I used them again in my own writing” (Group P in Table L8 of Appendix L).

Ideas broadened from alternative perspectives

A majority of the student writers indicated that group discussion made their thoughts clearer than they were while thinking individually. When they were writing collaboratively, ideas about alternative lines of thought broadened their conceptions. Consequently, the students’ logic, rhetorical structure, and organization of their writing became clear too. As one student

84 said, “I became better in writing supporting examples. It [CW] also helped me develop different perspectives towards a topic” (Group E in Table L3 of Appendix L).

Other contributions of CW

Other contributions of CW mentioned during the interviews were the student writers’ building up their abilities for structuring paragraphs and increasing grammatical accuracy. One writer mentioned, “I could not write anything in the first time [of writing] but after CW, I felt so good to be able to write an entire paragraph. It relieved my nervousness” (Group G in Table

L4 of Appendix L). A different student said, “Some basic writing skills like typing in English were also trained through subsequent times of writing” (Group I in Table L5 of Appendix L)

Division of labour

The tasks reported being assigned to group members while writing collaboratively included: checking grammar or choice of words, typing and integrating ideas into paragraphs, providing examples, and consulting an online dictionary.

A majority of the student writers reported they performed writing tasks in a

“fixed-mode” of cooperation, meaning that the students retained the same kind of cooperation to co-construct their writing no matter what a person was responsible for doing. Normally, this responsibility was for one of the aspects of writing that a student was good at, such as correcting spelling or grammar or providing examples for an argument. Once a student’s role in CW was identified and established, the person did the same kind of task again and again while writing collaboratively. That is, these students gradually figured out and cooperated to establish routine role(s) in their CW.

Fixed-mode cooperation functioned to streamline how the students wrote together. The students collectively added words, phrases, and ideas to a writing template; or alternatively,

85 students selected a writing sample from one of the students’ pre-collaborative-writing drafts and modified it. As two students said, “We compared our first draft to decide one writing sample…we read each other's draft again to see if we can enrich the writing sample with more examples” (Group B in Table L2 of Appendix L). Another type of fixed-mode cooperation described was, “We compared our writing first to see if we shared the same points of view and if yes, then we tried to integrate everyone's ideas into one writing sample that we agreed with”

(Group E in Table L3 of Appendix L).

Sometimes, group members took turns to complete aspects of the writing tasks; however, this mode of collaborative involved little mutual cooperation within the group. For example, one student said she typed the first and the third tasks of CW, whereas her two writing partners did the typing on the second task. When they discussed what viewpoints to choose for the CW, they opted for the train of thought that had the most main ideas and supporting examples if there were two alternative schools of thought among the three students. When this student typed, she said she sometimes wanted the other two students to check if her wording was correct, but the other two colleagues provided "no comments" most of the time. Seemingly, they did not want to contradict each others' suggestions. I observed during their writing that this group tended to be the quietest group in the class. Three of these students said that they would go back to their original ideas when they wrote the same writing task again as post-CW, perhaps because two of the students normally shared the same perspective on an argument whereas the third student held the opposite perspective most of the time. That latter student admitted that she did CW for the sake of CW.

Community

Despite the students having been placed into the same class of high-intermediate English proficiency, their writing abilities varied slightly. Some of the students said they wished they

86 had partners with more knowledge of English to help them. Other students observed how their partners benefited from their guidance in English, for example, “She improved from only knowing a word to knowing how to put words together to make sentences and then to paragraphs. She barely knew a word and sometimes she asked me for help at the very beginning” (Group P in Table L8 of Appendix L). Four novice/expert groups (i.e., group b, d, k, and p) and 12 novice/novice groups were identified in the interview.

Rules: Advantages

Quite a few groups emphasized their acquiring writing knowledge from mediation tools and using it in their individual writing. In addition, several student writers expressed that discussing and checking online information made them feel secure. As one person said, “It was helpful to memorize the spelling of a word or its pronunciation if I checked it up once again”

(Group M in Table L7 of Appendix L).

Two students said that thought they could not write without computers anymore. One of them specifically pointed out that paper-and -pencil writing was good too but it cannot beat computers in offering varieties of resources. Another continued, “Using computers to check spelling and English usage was beneficial to student writers in that it can easily tell you where the error was and the writer can memorize it by correcting it. This was what paper-and pencil writing cannot do” (Group B in Table L1 of Appendix L).

Rules: Limitations

Students said they needed an English writing expert in their group to solve writing problems, especially for grammatical usage. Some group members observed that the way they tended to solve problems while writing was to select an English usage that they all knew.

Triadic groups in CW activities could not benefit every member of the groups because of uneven divisions of labour. One comment on groups of three observed that at the beginning this configuration seemed to be helpful but later on seemed to have limited benefits because two writers took charge of the most of the writing tasks, leaving the other student just responsible for typing what they told her to write. The student admitted she benefited from sharing different points of views only, so she concluded that, “Three persons as a group was no good. Two persons were better.” (Group E in Table L3 of Appendix L).

Internalization of English language forms seemed hard because of too much reliance on tool use. “Computers and material mediation was not always good,” one student said, because

“you can easily remember the word you just wrote down [with pens] but with computers, you can't do that and you will always want to check up information. Therefore, internalization was hard to happen” (Group E in Table L3 of Appendix L).

Other issues brought up by interviewees were distraction caused from talking, time limitations for writing, and instability of the computer system. Some students also mentioned that they felt they needed time to get used to using material sources of mediation such as the collocation tools.

Rules: Improvements

In addition to recommending that longer time be allowed for writing in the classroom and improving the computers’ stability, one student suggested allowing students to change their writing partners. Also, another writer suggested that the instructor should have mini-meetings with writing groups.

Research Question 2.3

A questionnaire was administered to determine if CW was associated with any changes in the students’ self-assessed beliefs about their writing abilities. The questionnaire consisted of

28 items addressing learning goal orientation, performance goal orientation, metacognition, self-efficacy, and effort. These items were answered on a 4-point Likert scale with 1 representing "almost never" and 4 representing "almost always".

Descriptive analysis of students’ self-assessed beliefs about their writing abilities

Tables 4.24 and 4.25 indicate that both groups of writers increased their self-assessed beliefs about their writing abilities between the beginning and end of the program. However, the comparison writers had higher average self-assessed beliefs before the program started than the collaborative writers did. The collaborative writers made significant changes in their mean self-assessed beliefs about writing abilities after four weeks of tool-mediated writing activities.

For instance, they increased from 14.36 (SD=3.74) in learning goal orientation before CW to

17.75 (SD=3.02) after they finished the course. This change was the greatest among all the scales assessed. The comparison writers, after four weeks of in-class lectures and individual

(rather than collaborative) writing, increased their self-assessed beliefs about each aspect of their writing abilities too. However the following GEE analyses show that the increases in their self-assessed beliefs that the collaborative writers made were bigger than those that comparison writers made. The GEE analyses considered between-group contrasts first and then within-group differences.

Table 4. 23 Means and Standard Deviations of Collaborative Writers’ Self-Assessed Beliefs about Their Writing Abilities before program after program n=36 M (SD) Min Max M (SD) Min Max Learning goal 14.36 (3.74) 6 24 17.75 (3.02) 12 24 orientation Performance goal 9.33 (2.86) 5 16 11.17 (3.54) 6 20 orientation Metacognition 13.64 (2.86) 6 21 16.83 (3.53) 8 24 Self-efficacy 12.58 (3.52) 6 21 15.42 (3.03) 10 24 Efforts 14.03 (2.88) 9 20 16.36 (2.54) 10 20 Note. Min = minimum, Max = Maximum

Table 4. 24 Means and Standard Deviations of Comparison Writers’ Self-Assessed Beliefs about Their Writing Abilities before program after program n=26 M (SD) Min Max M (SD) Min Max Learning goal 16.6 (4.04) 8 23 18.32 (3.15) 8 22 orientation Performance goal 10.04 (3.91) 5 18 10.92 (4.12) 5 19 orientation Metacognition 15.4 (2.60) 12 21 17.00 (2.78) 13 24 Self-efficacy 14.36 (2.94) 11 22 15.64 (3.44) 11 24 Efforts 15.96 (2.56) 11 20 16.6 (2.00) 12 19 Note. Min = minimum, Max = Maximum

Between-group contrasts in self-assessed beliefs about writing abilities

The GEE results in Table 4. 25 show that, except for performance goal orientation, the two groups of student writers differed significantly in their before-program self-assessed beliefs about their learning goal orientation, metacognition, self-efficacy and effort but not in their after-program self-assessed beliefs. That is, the comparison student writers self-assessed these four abilities higher than the collaborative student writers assessed themselves on the same abilities. The possible explanation for the differences might be that more (i.e., 36%) comparison students had over 10 years of learning English experience than the (i.e., 11%) CW students had

(Table 3.1). More years of learning means more learning opportunities, and individuals with learning goal orientations are likely to seek learning opportunities (Kozlowski, et al., 2010).

Wey’s (1998) validation of those scales of self-assessed beliefs revealed that learning goal orientation actually had stronger effects on metacognition, self-efficacy and effort than performance goal orientation did. Therefore it could be that the difference in years of English learning caused those differences. The collaborative student writers made greater increases in these self-assessments over the four weeks of the program so that no statistically significant difference was found between the two groups in after-program self-assessed beliefs about writing.

In sum, the collaborative and comparison student writers started differently in their self-estimations of their writing abilities. The comparison group self-assessed themselves before starting their course 2.14 points higher on learning goal orientation, 1.59 points higher on metacognition, 1.65 points higher on self-efficacy, and 1.78 points higher on effort. For performance goal orientation, there was no difference found (p=.53) between groups in the before-program and after-program tests.

Table 4. 25 GEE Modeling of Collaborative and Comparison Writers’ Self-assessed beliefs about their Writing Abilities

Mean 95% CI (lower, Outcome variables Parameter SE Sig. Difference upper) before CW -2.14 0.98 (-4.07, -0.21) p=.03* Learning goal program BW 0* orientation after CW -0.25 0.82 (-1.86, 1.35) p=.76 program BW 0* before CW -0.55 0.89 (-2.29, 1.19) p=.53 Performance goal program BW 0* orientation after CW 0.75 0.99 (-1.19, 2.69) p=.45 program BW 0* before CW -1.59 0.7 (-2.96, -0.22) p=.02* program BW 0* Metacognition after CW 0.16 0.82 (-1.45, 1.77) p=.84 program BW 0* before CW -1.65 0.81 (-3.24, -0.06 ) p=.04* program BW 0* Self-efficacy after CW 0.23 0.85 (-1.44, 1.90) p=.79 program BW 0* before CW -1.78 0.69 (-3.14, -0.42) p=.01* program BW 0* Effort after CW -0.11 0.6 (-1.28, 1.07) p=.86 program BW 0* *0=reference performance *SE=standard error *95%CI=95% Wald confidence interval for difference CW = collaborative writers BW = comparison writers

Within-group contrasts in self-assessed beliefs about writing abilities

As Table 4.26 presents, the collaborative student writers increased each aspect of their self-assessed writing abilities over time. These student writers self-assessed themselves 3.38 more in learning goal orientation, 1.83 points more in performance goal orientation, 3.19 points more in metacognition, 2.83 in self-efficacy, and 2.33 in effort after the course was completed compared to their self-assessments prior to the course. In contrast, Table 4.27 shows that the

92 comparison student writers' changes in their self-assessed beliefs about their writing abilities were minor and not statistically significantly different before and after the program.

Table 4. 26 GEE Modeling of Collaborative Writers’ Over Time Performance on Self-Assessed Beliefs about Writing Abilities 95% CI (lower, Outcome variables Parameter Estimate SE Sig. upper) Intercept 14.36 0.61 (13.16, 15.56 ) p=.00* Learning goal orientation after program 3.38 0.5 (2.41, 4.37) p=.00* before program 0* Intercept 9.33 0.47 (8.41, 10.25) p=.00* Performance goal 1.83 0.57 (0.71, 2.95) p=.00* orientation after program before program 0* Intercept 13.64 0.47 (12.72, 14.56) p=.00* Metacognition after program 3.19 0.46 (2.28, 4.11) p=.00* before program 0* Intercept 12.58 0.58 (11.45, 13.72) p=.00* Self-efficacy after program 2.83 0.42 (2.02, 3.65) p=.00* before program 0* Intercept 14.03 0.47 (13.10, 14.96) p=.00* Effort after program 2.33 0.41 (1.53, 3.13) p=.00* before program 0* *0=reference performance *SE=standard error *95% CI=95% Wald confidence interval CW = collaborative writers BW = comparison writers

Table 4. 27 GEE Modeling of Comparison Writers’ Over Time Performance on Self-Assessed beliefs about Writing Abilities Estimate 95% CI (lower, Parameter SE Sig. Outcome variables upper) Intercept 16.50 0.77 (15.00, 18.00) p=.00* Learning goal 1.12 0.79 (-0.42, 2.65) p=.16 orientation after program before program 0* Intercept 9.89 0.75 (8.41, 11.36) p=.00* Performance goal 0.62 0.65 (-0.65, 1.88) p=.34 orientation after program before program 0* Intercept 15.23 0.52 (-0.19, 2.42) p=.00* Metacognition after program 1.12 0.66 (-0.19, 2.42) p=.09 before program 0* Intercept 14.23 0.57 (13.12, 15.35) p=.00* Self-efficacy after program 0.81 0.7 (-0.56, 2.17) p=.25 before program 0* Intercept 15.81 0.5 (14.82, 16.80) p=.00* Efforts after program 0.15 0.68 (-1.18, 1.49) p=.82 before program 0* *0=reference performance *SE=standard error *95% CI=95% Wald confidence interval

CHAPTER FIVE

DISCUSSION AND IMPLICATIONS

This chapter starts with a summary of the research findings. Results from the second set of research questions functioned as partial explanations for findings addressed in the first set of research questions. Then I consider the theoretical interpretations of the results before making recommendations for future research and for pedagogical policies and practices.

The study was conducted in two intensive English classes of 36 and 26 Taiwanese graduate students who were admitted to arts-related MA programs in 2011 and 2012. The intensive English courses were to equip students with skills for writing TOEFL. Lectures on a

TOEFL textbook and a writing platform were involved in the course instruction and the research design. As the instructor of the two courses, I had students practice CW in the 2011 course, and the 2012 course functioned as a comparison course in which students did individual writing. All student writers wrote one writing task as pretests before their course started, and they did the same writing task as posttests after their course ended. For the CW class, sixteen groups of collaborative writers practiced collaborative rewriting activities four times in a writing platform in class; student writers in the comparison class practiced individual writing four times under the same conditions. The only difference between the two classrooms was the students’ writing conditions. The collaborative rewriting activity was designed so that students individually wrote drafts before they collaboratively rewrote the same task, and then they did individual composition again. Comparison writers performed a writing task once, and they did four times of individual writing to contrast with the CW class. I made this contrast to see if the students in the CW condition improved their individual writing more than comparison writers did. If they did, this result could provide evidence that collaborative writers’ individual performance did

95 benefit from tool-mediated writing knowledge that had been generated from their performing collaborative rewriting tasks over four weeks. Then, accordingly, the correlation of collaborative writing performance with individual writing development was tested to see if there was a positive correlation between them.

The aim of the first set of research questions was to investigate if students’ individual writing development can be fostered through the mediation of their tool use when performing in-class CW tasks over four weeks. To realize this purpose, three linked research questions were addressed by: (a) creating a coding scheme of tool-mediated writing knowledge from the student writers’ reports on mediation tool use and data on their actual uses of mediation, (b) examining the association between the use of tool-mediated writing knowledge and CW performance, and then (c) examining the correlation between CW performance and individual writing performance.

Results for the first research question suggested that students used mediation tools to generate writing knowledge as “language forms” and “skills of expression”. The students’ collaborative dialogues and their online browsing histories while writing were then analyzed in terms of the frequencies of their consulting these two types of writing knowledge. The groups that consulted these mediated sources of writing knowledge utilized tools to consult “language forms” more often than “skill of expression” in each writing task, although there were ten groups that did not adopt any tool-mediated writing knowledge over time.

To evaluate whether the student writers’ uses of tool-mediated knowledge of “language forms” and “skills of expression” and their weekly writing practice contributed to changes in their CW, I used GEE to analyze the extent to which these three factors (language forms, skills of expression, and weekly writing practice) contributed to changes in the students’ writing performance. The quality of the students’ compositions written before, during, and after four weeks of instruction was evaluated by independent raters in terms of six dimension of writing:

96 holistic scores, communicative quality, organization, argumentation, linguistic accuracy, and linguistic appropriacy.

Results from the GEE analyses confirmed that “language forms”, “skills of expression”, and “weeks of practices” explained changes in the holistic quality of collaboratively written compositions over the four weeks of the course. Furthermore, the students’ adoption of

“language form” mediation resulted in significantly greater writing improvement than did students’ adoption of mediation involving skills of expression or their simply practicing writing over the period of the course. Specific dimensions of improvement in the students’ writing performance related to these contributory factors were reported as follows. Students’ uses of both language forms and skills of expression contributed to the communicative quality of their written compositions. The weeks that students spent practicing writing contributed to the organization of their written compositions. Their uses of mediation involving skills of expression contributed to the argumentation in their writing. Improvements in the linguistic accuracy of the students’ CW were positively associated with all three factors. Improvements in ratings of the linguistic appropriacy of the students’ collaboratively written compositions were positively related to the students using mediation related to language forms.

Comparing the initial English writing abilities of the students who wrote collaboratively with those of the students in the comparison writing class, it was found that both groups of students started their courses with similar proficiency in every dimension of English writing except for linguistic appropriacy. Comparing theses student writers’ compositions in their after-program test to their before-program test, both groups of students developed over time their writing in different dimensions. Collaborative writers improved significantly in every dimension of English writing. Comparison writers also showed development in most dimensions of English writing except for linguistic accuracy. The students who wrote

97 collaboratively improved more in every writing dimension in the after-program test than did the comparison writers (who wrote individually rather than collaboratively).

The correlation between the development of CW performance and individual writing development was positive. For the collaborative student writers, their uses of tool-mediated knowledge and four weeks of writing practice, shown to be associated with changes in their CW, might potentially have facilitated their individual writing development.

The second set of research questions concerned issues arising from Activity Theory about the student writers’ communications and relationships while collaborating to write (i.e., rules and communities), their approaches to organizing their CW (i.e., division of labour), and their self-assessed beliefs about their English writing abilities after writing collaboratively (i.e., outcomes).

In regard to the student writers’ questionnaire responses about their communications and relationships, differences appeared in the richness of their communications and teamwork over time. In their evaluating what went well and what did not in performing CW tasks, the students indicated that their discussions of content, organization, and perspective-taking for their CW worked well in the second and third weeks of the course. In the fourth week, the students indicated that they felt they worked well in discussing content, verbalizing ideas from Chinese into English, and the integration of group members' ideas into the CW. At the same time, the student writers also observed that the CW did not work well in establishing the grammar, vocabulary, and verbalizations for their writing in week two and in their grammar, vocabulary, and timing of tasks in weeks three and four.

During group interviews the student writers reflected on several aspects of their CW.

They opined positively about the assistance of mediation tools in writing. Most of the student writers stated that they felt their written argumentation was strengthened by group discussions, which facilitated their abilities to verbalize ideas in English and provided additional supporting

98 examples beneficial to writing. The students observed that they usually adopted a fixed-mode of collaboration in which particular group members adopted and sustained the same roles for CW.

The students’ English proficiency in the writing groups varied a bit, though most groups combined novice writers in terms of their writing experience together, and some students expressed the view that more expert writers in their groups would have helped them to solve problems while writing more effectively or efficiently. The students reemphasized the advantages and improvements made through their adoption of mediation tools in this writing classroom. Suggestions were made to permit longer time periods for writing, improving the stability of the computers system, and changing writing partners during the course.

Regarding the students’ self-assessed beliefs about English writing abilities, the collaborative writers increased theirs over time significantly in every dimension of their writing: learning goal orientation, performance goal orientation, metacognition, self-efficacy, and effort.

In contrast, the changes expressed in their self-assessed beliefs about their English writing abilities by the comparison students were so minor that no significant difference appeared between their self-assessments before and after the program.

The complexity of modeling process-product writing development

My analyses have demonstrated empirically the relations between writing practice and learning opportunities, particularly involving collaborative work and uses of mediational tools, and the development of English writing abilities in the context of a natural course of foreign language teaching. Many L2 writing studies have attempted to relate writing development to various factors, and most of them have reached mixed and complicated results (Cumming &

Riazi, 2000; Hyland, 2011; Norris & Manchon, 2012). Such results include varied and dynamic

L2 development patterns that are associated with different proficiencies of L2 student writers, various orientations to L2 learning, or types of classroom practices experienced

(Cumming & Riazi, 2000; Hyland, 2011). Furthermore, Verspoor, Schmid, and Xu’s (2012) investigation of more than 60 L2 developmental variables found non-liner development of those linguistic features at the sentence, phrase, and word levels. Their results suggest that many factors cause individual differences and variability in writing development. L2 writing research also involves various dynamic systems. For example, studies exploring the process-product relations of L2 writing have focused on student-oriented variables (e.g., motivations, L1 writing skills, personal characteristics, or prior writing experience), on process variables (e.g., writing instruction, or writing activities), and on types of writing done (e.g., business writing or academic writing) (Cumming, 2013a; Leki, Cumming, & Silva, 2008). The dynamic systems that are generated from the numerous interactions between variables reflect the actual complexity of issues involved in learning and teaching L2 writing. Interactions are to be expected in achievement for different aspects of L2 writing among learners at different levels of

L2 proficiency and with different conditions and purposes for L2 writing instruction.

Given the complicated nature of L2 writing in relation to writing instruction and writing performance, the framework of Activity Theory (AT) provides a theoretical basis which relates classroom activities to L2 writing development. A number of case studies on individual or small groups of students have begun to investigate writing development through the lens of AT

(Cumming, 2006; Nassaji & Cumming, 2000; Yang, Baba, & Cumming, 2004). However, the present study is unique in modeling and evaluating two whole classes’ writing development fostered by various forms of mediation over a full course of instruction.

The difficulties and challenges of capturing writing development for a whole class center on what factors are essential, especially to investigate development resulting from CW, a specific type of writing activity that involves various sociolinguistic, personal, and interpersonal factors. A theoretically informed framework can, nonetheless, help illustrate how and why development happens. The adoption of Activity Theory enabled the present study to model key

100 factors during CW and L2 writing development. By doing so, interaction effects caused by multiple minor factors that defy simple causal relationships can be avoided too (Cumming &

Riazi, 2000). After my initial identification of influential L2 writing factors, the statistical technique GEE provided a unique, innovative, and relevant research approach to trace learners' writing changes with using mediation tools over time and to model these complex relations.

L2 learners' development of self-assessed beliefs about their writing abilities

Given that both groups of writers received the same writing instruction and improved their scores on an after-program writing test, it is interesting that they expressed themselves differently on their self-assessed beliefs about writing. Collaborative writers’ beliefs about writing were strengthened but the comparison writers were not. Such differences might result from factors associated with the assistance of mediation tools (i.e., collaborative dialogue and use of writing resources) that have been commonly reported in recent studies of self-efficacy, metacognition, and goal setting.

Self-efficacy beliefs are defined as students' judgments of their ability to organize and execute action that is required to achieve performance (Pajares & Valiante, 2006). According to

Bandura (1997), self-efficacy is affected by one’s mastery performance, experience of watching others perform tasks (also called peer modeling), verbal messages, and social persuasions received from others. Recent studies on the sources of efficacy within the context of second/foreign language learning have included factors like classroom interaction between learners (Greta, 2009), social and cultural setting (Wang & Pape, 2007), short tutoring sessions involving explicit explanation of structural rules (Mathews, 2010) and strategies (Magogwe &

Oliver, 2007; Su & Duo, 2012; Wong, 2005). These reported self-efficacy sources suggested collaborative writers’ self-efficacy improvement may be a result of CW granting students more interaction from collaborative dialogues, peer-tutoring, and online explanations of language

101 rules to deal with writing difficulties than individual writing did. In addition, Boscolo (2009) has suggested that writing development itself is one of the most influential sources through which students form their self-efficacy beliefs. Consequently, the improvement of CW performance as a result of mediation tool use seems also to have fostered learners increasing their self-efficacy

(see Table 4.6).

Changes in goals (learning goal and performance goal), self-efficacy, and effort are strongly related (Bandura & Locke, 2003). Students’ changes in their goals may reflect their increasing or decreasing their self-efficacy because students with higher self-efficacy tend to set challenging goals or vice versa. To fully explain why the present collaborative writers changed their goals, no single or direct sources explain the reasons except that sources of self-efficacy and self-efficacy itself could be taken into account. Accordingly, classroom interaction, peer modeling, classroom instruction, and writing improvement could all explain their changes in goals. Other reported factors contributing to goal orientations in the present study are related to teachers’ instructional activities (e.g., encouraging multiple drafts and providing mediation tools or feedback in this study) and teachers’ expectations (Yang, et. al., 2004). Changing goal orientations can affect student writers’ performance. Goal changes may direct students’ efforts to activities that are pertinent to the goal and also empower learners to invest greater effort for higher goals to sustain their efforts. In addition, changing goal orientations can guide student writers to use skills and strategies that are relevant to a task (Locke & Latham, 2002).

Metacognition generally refers to the use of higher order thinking, reasoning, and learning skills during learning, one example of which includes problem-solving strategies.

Collaborative writers’ abilities in solving writing problems were fostered in the present study through their being provided with immediate assistance from peer-tutoring and online resources.

These elements may have facilitated the CW students’ metacognition. In addition, changes in metacognition might be related to these students using their L1 writing knowledge and strategies,

102 which affects L2 metacognition (Raoofi, et al., 2014). For instance, Yau (2009) in her study of the use of L1 and L2 metacognition found that there was a strong link between learners’ L1 and

L2 metacognitive strategies: greater use of L1 metacognition contributed to L2 metacognition.

Storch and Wigglesworth (2003) concluded that student writers’ L1 could be a useful tool, especially in more meaning-focused activities, such as joint composition tasks. They also noted that a shared L1 could enable students to clarify issues of vocabulary, to discuss the structure of the composition in more depth, and to gain control over writing tasks. In other words, that CW writers’ languaging in Mandarin to retrieve English syntactic and lexical knowledge in order to solve writing problems or even using online translation may have facilitated the students’ metacognition in English writing.

Explanations of student writers’ development

The present study indicates that CW, assisted with other mediation tool use on a weekly basis, contributed to more growth in student writers’ improvement than individual writing did over the same time period. This result coincides with one of Vygotsky’s most important claims: that learning collaboratively with others, particularly in instructional settings, precedes and shapes development (Lantolf & Thorne, 2007). To explain writing development in the present study’s context, a number of theoretical accounts derived from Sociocultural Theory are discussed below in relation to mediation and writing practices in L2 writing classrooms: mediation tool use towards writing goals, mediation activating cognitive processes and cognitive distribution of L2 writing, collective scaffolding and group ZPD during CW, L1 mediating L2 writing processes, and L2 writing communities and imitation through repeated practice.

103

Mediation tool use towards writing goals

According to Vygotsky (1987a), learners’ mental functions are carried out through goal-directed mediated activity. The types of writing knowledge that students used mediation tools to attend to was in relation to the students’ goal-oriented use of mediation tools. According to Cumming (2006, p. 90), "a goal identifies the object or outcome one aims for and is also the standard by which one measures one's performance." The present CW students’ actions of using mediation tools were taken to realize their goals for writing. Reciprocally, goals also shaped the students' actions towards mediation tool use. Cumming (2006) observed that goals for writing improvement have a force that is often a clear intention, linked to an educational program’s goal, an instructor’s goal, and students’ own goals. I assume these processes occurred in the present study.

The program’s curriculum aimed to introduce and foster students’ knowledge writing tasks for the TOEFL. To pass the program or not affected whether students could graduate on time or not from their degree programs. In class, the writing component of the TOEFL itself was a focus of activity, including how writing was evaluated, introduced in the textbook, and explained in the course lectures. Students were informed that the same dimensions of English writing performance were going to be assessed in the class as they would really be in the

TOEFL test. Writing dimensions included addressing a topic effectively, organization, explanations, coherence, use of language, and word choice, and so forth. Given that the intensive program lasted 4 weeks, as an instructor my goal for the collaborative writers was to maximize their writing performance through practice and use of mediation tools while also leading students to meet the educational program’s goal.

In their responses to questionnaire surveys the student writers expressed strong learning goals. Their goals had objects like learning how to write argumentative writing tasks in English, so, to realize these goals, students took specific actions like seeking assistance from peers or

104 teachers or using tools or resources (Cumming, 2006). Findings presented in Table 4. 1 in

Chapter Four showed that students discussed linguistic forms more frequently (22 times for each group a week) than skills of expression (7.5 times for each group a week), indicating that they were concerned mostly with the accuracy of their use of English linguistic forms. In addition, in order to solve writing problems in a timely manner, each group of students used mediation tools to consult answers to their writing problems, knowing these were going to be assessed in a goal-driven context.

Mediation-activated cognitive processes and cognitive distribution

In Sociocultural Theory, mediation is considered to play a critical role in the construct of activity and generation of higher mental functions (e.g., memory, reasoning, application, and creativity). In the present study, higher mental functions were carried out when students applied tool-mediated writing knowledge to solve writing problems. Mediation tools in this study were resources provided on the writing platform, writing compositions, weekly practices, group communications, and writers’ peer-tutoring. The use of mediation tools activated writers’ cognitive functions, for example, for students’ reporting their using online dictionary to recall their memory of learned but forgotten words, to check spellings and grammatical problems, or to acquire new phrases. Writing on a weekly basis helped writers apply what they gained from the previous writing task to the next writing task. Group communications allowed students to share their writing knowledge and to provide feedback to create more enriched contexts for writing than were available to students who wrote individually.

By means of mediation, the activation of cognitive process facilitated writers to memorize, internalize, and apply what they had learned to different writing tasks. With immediate assistance of mediation tools, CW writers’ problems could be solved and clarified, and writing knowledge was distributed across students through verbal interactions (i.e., peer

105 discussions and tutoring), between students and tools through students’ consulting online resources (to recall their memory of vocabulary or to learn new words), and across time through weekly practice (i.e., students’ repeatedly rewriting the same tasks in CW and transferring accumulated knowledge to different writing tasks). The CW writers also changed their beliefs about their writing.

Collective scaffolding and group ZPD during CW

The nature of assistance among members of writing groups is, according to Storch

(2002b), an essential factor influencing whether collaborative work succeeds. Storch (2002b) found that a greater amount of writing knowledge and development occurred in the dialogue of expert/novice groups than in that of novice/novice groups. I observed some similar trends; for example, a competent writer in this study, Helen, expressed her observing Emy’s developing lexical and grammatical ability from words, to phrases, and then to entire sentences.

Nonetheless, results from the present thesis study indicate that writing knowledge and development happened in a class with mostly novice-novice students who were provided with material mediation during CW. While student writers were composing collaboratively, they identified each other’s merits in English writing (e.g., spelling, grammar or providing supporting ideas) and peer tutored each other with and about their strengths in writing. When in doubt about language use in CW, the students often consulted online assistance. With such collectively scaffolded assistance through peer dialogue and online resources, students created their specific group ZPDs in their CW groups. As many researchers have observed, scaffolding is generally thought of as assistance provided through interaction between a competent other and a learner, but scaffolding can also happen in novice-novice L2 interactions (Donato, 1994; Donato &

McCormick, 1994; Storch, 2002a). The idea of fostering language learning through peer-peer interaction assumes that within peer-peer interaction, peers can be concurrently experts and

106 novices (Gutierrez & Adela, 2006; Wells, 1999). Furthermore, “peers working within the ZPD of each other can support learning through, for example, questioning, proposing possible solutions, disagreeing, repeating, and managing activities and behaviors (social and cognitive)”

(Swain, Brooks, & Tocalli-Beller, 2002, p. 173).

Dialogic collective scaffolding in CW activities, according to Storch (2013), has several attributes in assisting language learning. According to Storch, one important feature is timeliness. In the present study feedback from and with collaborative writers was immediate and available during CW. As the first LRE excerpt on language forms on page 50 in Chapter Three demonstrated, one peer student noticed and corrected grammatical errors during CW. Unlike teachers’ feedback on students’ writing, which normally happens long after students have completing a writing task, this mediation was enacted immediately during CW. The other attribute of peer dialogue claimed in Storch (2013) is that suggestions or feedback occurring in

CW are negotiated and acceptable to all students in writing groups. Feedback incorporated into a co-constructed text assumes that all students in the writing groups have ownership of the text they are producing. This sense of ownership encourages more input to the text. In contrast, teachers’ feedback on written compositions offers few opportunities for students to negotiate or make continuing commitments to writing because the feedback is usually made to finished written texts a considerable time after they have been composed. As Storch (2013, p. 39) stated, collective scaffolding is timely and acceptable in “responding to the learners’ expressed need for assistance.” Collaborative writers can guide and support each other by pooling their writing knowledge to reach a better outcome than they might be able to achieve if they compose individually.

107

L1 mediated L2 writing processes

In studying learner-learner's interactional talk in L1, a number of studies have investigated the role of L1 in performing L2 linguistic tasks (Anton & DiCamilla, 1998; Scott & de la Fuente, 2008; Storch, 1998; Swain & Lapkin, 1998). According to Anton and DiCamilla

(1998), students’ L1 plays a role of psychological tool, enabling learners to scaffold members of a group. Similarly, Swain and Lapkin (2000) documented how classroom L2 learners pushed their linguistic development forward by talking in either their L1 or L2 about new features of the

L2. Even though there is no direct evidence of student writers in this study improving their writing because of their using their L1 in CW, interviewees did point out that L1 verbal interactions eased their paraphrasing and verbalizing ideas in English. As an instructor of the course, I realized that students’ using their L1 served learners to search for vocabulary items, to find alternatives for their writing, and to explain about grammatical rules and conventions. Such functions of the L1 have also been documented by researchers investigating L2 collaborative writing (Swain & Lapkin, 2000; Storch & Aldosari, 2010; Shehadeh, 2011) as well as individual writing (Cumming, 2013b; Murphy & Roca de Larios, 2010). Opportunities to use their L1 as a resource were presumably available more frequently to students in the CW condition than in the comparison class. Ortega (2009) suggested that L1 use by high proficiency L2 writers serves as a mediation tool to enable quality planning, revising, and monitoring processes.

L2 writing communities and imitation through repeated writing

Writing compositions on a weekly basis provided opportunities for students to apply what they learned from previous writing to the next writing tasks. For collaborative writers, writing practice on a weekly basis created learning opportunities through writing in a defined community of peer learners, through imitating writing during CW (Lantolf & Thorne, 2007),

108 and through repeated learning from writing similar writing tasks over time (Ahmadian, 2012;

Kim & Tracy-Ventura, 2013; Perin, 2002).

Storch (2005) suggested that collaborative tasks can improve essays by helping L2 writers predict readers' problems with a text. A writer is not working alone through cognitive processes but as a member of a community. Collaborative writing puts texts, writers, and readers into communities. Current concepts of learning to write see writing as a social practice, embedded in social and institutional contexts. Writing is personal and individual but "it is also interactional and social, expressing a culturally recognized purpose, reflecting a particular kind of relationship and acknowledging an engagement in a given community" (Hyland, 2011, p.31).

Collaborative writing offers opportunities for both readers and writers to participate in constructing written compositions.

Another important mechanism regarding writing improvement in this study is imitation, especially learners’ imitative writing behaviors as a means of acquiring a new language (Lantolf

& Thorne, 2007). In reporting how they collaborated with each other and their relationships within their writing groups (i.e., divisions of labour and community) students observed that they served as each other’s peer models with their own specialties during CW. In addition to refining their own writing, imitation might have occurred when one student read the other student’s writing (providing that they took turns to write). Imitation also occurred when students recalled items from their memory during CW, including writing knowledge they retrieved from their dialogues or consulting online writing resources. One student writer indicated that he followed the train of thoughts in CW and then expanded and applied it to his re-writing the same task

(e.g., Rater Katie remarked that some writing pieces were similar in structures, wording, and argumentation, except for fluency). This application of previous writing to the next writing suggests another learning mechanism, repeated activities, to explain the present study’s findings about writing development.

109

Repeated learning may have occurred in this study not only in improving writing fluency but also in situations where students adopted the same sentences from CW to their individually re-writing the same task (e.g., for the purpose of avoiding grammatical errors). Regarding studies on the benefits of students’ learning from repeating tasks, Baba and Nitta (2011, 2014) have reported that repetition of a task in the classroom in Japan fostered learners’ writing quality, including fluency. Casanave (2004) also found that in addition to writing more fluently, some

L2 students wrote more accurately over time and others wrote with more detail, depth, and expressiveness. Larsen-Freeman (2009) has emphasized the significance of repetition by indicating that “using a task more than once is what drives learning…When it comes to language learning, revisiting the same, or similar, territory again and again is essential” (p. 584).

The present study illustrated that with weekly practice, re-writing the same task in a week granted opportunities for students to reproduce a “better” and “enriched” quality of composition.

The statistical models on page 61, 63, and 64 in Chapter Four evidenced that weekly practice played a role in student writers’ development of holistic performance, organization, and linguistic accuracy.

Implications

Research implications

Definitions of L2 writing development are diverse and they vary with theoretical accounts and methodological decisions that researchers take (Norris & Manchon, 2012). The present thesis research adopted one theoretical framework, Activity Theory, to define and analyze salient factors that may influence CW performance. In turn, an explanatory theory,

Sociocultural Theory, was used to analyze the relations among texts, learning, and writing development in an English as a foreign language context. According to Norris and Manchón

110

(2012), it is important that theory-derived frameworks sketch out boundaries in valid empirical studies for researchers to know the targets of and paths towards L2 writing development. After identifying relevant factors, to increase the validity of the study further, I adopt a mixed-methods approach to confirm the interpretations, using the results of qualitative analyses to serve as complementary explanations to support the results of quantitative analyses (Norris &

Ortega, 2012). One attempt that the present study made that is different from other mediation-focused research was to quantify the types of knowledge derived from mediation tools rather than to record the tools themselves being used to generate knowledge (e.g., the frequencies of using a dictionary or certain language websites). This knowledge-oriented approach allowed me to investigate a deeper and more extensive level of tool use than other L2 writing researchers have used to analyze L2 writing development. In addition, this approach reflects the situation that technological tools available nowadays are multifunctional, offering

L2 writers multifaceted and integrated assistance like definitions, collocations, synonyms, and example sentences in one online search. However, the availability of all that information at once suggests that tool-mediated writing and language learning is also becoming complicated.

The present study used mixed research methods, supplementing quantitative data with qualitative data to examine sociocultural aspects of mediation on CW performance. This approach differs from common practices of the process-product design in sociocultural inquiry, which has seldom quantitatively measured language acquisition (Norris & Ortega, 2003) Few sociocultural studies have employed quantitative research methods (especially modeling processes and products) to investigate L2 writing development. Additional empirical research methods need to be proposed and evaluated (Polio, 2001) so that the complex nature of L2 writing development under the lens of SCT can be depicted fully and from diverse perspectives and results can be compared across different groups of students, with different factors, and in different cultures, societies, and types of educational programs. Another issue in relation to

111 research methods, particularly the subscale of the rating rubrics adopted in this study, was that the construct of linguistic appropriacy was not well defined. Linguistic appropriacy proved to be problematic in that the raters had difficulties in reaching agreement on what appropriacy was according to the rubric’s specifications. I could have opted to use TOEFL’s own scoring rubrics for this research, but I opted instead to use an independent measure of writing ability, which had been validated in previous studies (Hamp-Lyons, 1991; Hamp-Lyons & Henning,

1991), because students in the classes were reviewing the TOEFL scoring rubrics as part of the writing platform and for their exam preparation.

To investigate what factors influenced L2 writing development and to what extent, the present study used GEE techniques to examine the relationships between processes and products of CW practices. GEE is a non-parametric statistical technique that deals with correlated or repeated data and accurately produces covariates for predictors. It is an ideal research technique to model learning processes and outcomes, as evident in its adoption widely in professional fields other than applied linguistics or language education. Hopefully, the application of GEE techniques to future educational studies will arise from the beginning explorations of the present study.

Anderson (1982) concluded from a meta-analysis of learning studies in diverse domains that after learners had engaged in learning and practice for more than 100 hours, they acquired cognitive skills to a certain level. DeKeyser (2007) has elaborated on Anderson’s (1982) three stages of cognitive skill development—declarative, procedural, and autonomous—pointing out that turning knowledge (e.g., tool-mediated writing knowledge) into behaviors (e.g., to apply that knowledge to solve writing problems), or declarative knowledge into procedural knowledge requires more than just a few trials and needs plenty of practice (e.g., multiple writing) to gradually automatization knowledge and skills. Anderson’s (1982) conclusion about the time required to acquire cognitive skills corresponds to the findings in this thesis: collaborative

112 writers showing their significant development in holistic performance, organization, and linguistic accuracy in week four after they received 105 hours of instruction and mediated writing practices, not in week three. More longitudinal studies are needed to explore and support how mediated learning leads to internalize relevant knowledge and what the effective hours are for L2 learning.

Lastly, the present writing platform can be used in future research and applied, for example, to distance education and to study how other writers’ discourses in different educational contexts mediate their writing performance.

Pedagogical implications

In Asian classrooms where English is taught and studied as a foreign language (EFL) (e.g.,

Thailand, Malaysia, Hong Kong, and Taiwan) pedagogy often emphasizes teacher-oriented lectures and accuracy in language use. This focus may arise because with a large class size, one instructor can hardly manage instruction on a one-to-one basis. Therefore, adopting CW for classroom activities may be one of the easiest and most effective ways to help student writers practice and gain writing knowledge and skills. In terms of cross-cultural differences in group interactions, collaborative writers in this study seemed to follow conventional rules of the politeness to interact with each other (Brown & Levinson, 1987). According to Carson and

Nelson (1994), within-group relationships in collectivist cultures (e.g., Japan and Taiwan) tend to work toward harmony during group writing tasks. Group work in more individualistic or different cultures may apply other patterns. Even so, fostering successful mediation during CW turns out to be important in facilitating students’ writing development. As a few collaborative writers stated in their interviews, they felt they needed an expert to guide their writing.

Instructors providing guidance during L2 learners’ writing process may be a first priority when teachers practice CW in classrooms.

113

The present study also shed some light on the factors that contribute to different aspects of L2 writing development. Among all the writing dimensions, the development of holistic performance and linguistic accuracy proved to involve the combination of many contributory factors (i.e., language forms, skills of expressions, and weekly practice). When L2 learners are learning how to write, it is important for instructors to provide mediation tools, especially lexical assistance and practice opportunities for students to cope with difficulties in their writing and learning processes.

Finally, EFL writing is different from ESL (English as a Second Language) writing in that EFL learners have less exposure to the target language than ESL learners do. In addition,

EFL writers tend to have similar prior learning experiences and lower motivation to write in the target language (Manchón, 2009; Ortega, 2009). These attitudes and factors are changing, however, as there are increasing needs for writing English in work places, government institutions, business networks, and other professional purposes (Cumming, 2013a). CW research which considers social factors in writing development can inform L2 educators how to prepare themselves to guide EFL writers who have such various purposes for using and developing English writing abilities.

Limitations of the study and suggestions for future studies

The present study has several limitations. The study was exploratory in assuming mediation takes a central role to account for L2 student writers’ development. Learning how to write in a second language involves more factors than mediation tools; for example, the effort and hours that students devote to writing could also make adifference. I acknowledge that the present collaborative writers spent 8 hours more in total writing than did comparison writers, which might have contributed to differences in their writing development. My analytic methods reduced L2 writing and development to several simple factors. In doing so, the study overlooked

114 other potential factors that might contribute to writing development (Cumming & Riazi, 2000;

Norris & Manchón, 2012). In terms of longitudinal research design, I recognize that four weeks of observation is a brief period to observe students’ writing development. In addition, the generalization of students’ writing being affected by their CW is derived from correlation analyses (rather than analyses that can establish causality). A further limitation is that the practice of CW was combined with repeated revisions of writing tasks, so these two conditions for writing cannot be distinguished from each other in interpreting my results. Finally, within the rules of the collaborative writing, writing groups showed considerable variation in using mediation tools: Some groups consulted different types of writing knowledge or relied on the teaching of the peer-tutor to resolve their disagreements about writing mostly at lexical or sentence levels (as commonly occurs in L1 and L2 revising, Fitzgerald, 1987), and other groups did little consulting at all. Interactions also varied with grouping variables like dyadic/triadic groups and novice/competent groups and with the roles that they adopted during CW (i.e., division of labour) (Storch , 2013). Those group variations should be evaluated in future research because they alter collaborative writing activity dynamically and affect explanations of what contributesto each group’s writing development. Indeed, in future analyses with the present data, I could, for instance, try to tease out grouping effects by classifying groups by their types of collaboration (e.g., high, medium, and low) and attempt to relate these categories to writing achievement.

My suggestion for future research on L2 writing is to investigate how other aspects and elements of Activity Theory influence student writers’ development. Assuming that students’ individual writing development is positively associated with their CW performance, researchers should examine what types of division of effort bring out maximal writing performance, in what kinds of educational programs, and in what kinds of communities of students in terms of their writing proficiency and writing experience. CW rules, such as how much time is allowed for

115 writing in classrooms, or different types of collaboration in or outside of classes, might influence writing performance as well. The observation period for such research should lengthy, such as one academic year or even two, so researchers can observe, document, and analyze that, what, and how tool-mediated writing knowledge is transferred to learners’ individual writing performance.

By analyzing mediation with a quantitative method the present study did not investigate the variability in individual writer’s development. According to Lantolf and Thorne (2007), a characteristic of L2 acquisition is variability in the development of any learner and across learners. Learners acquire different subsystems of a target language depending on the types of mediation they obtained and the specific goals for which they use the language. Similarly,

Poehner (2009) argued that students as a group operate in a collective ZPD, whereby they all benefit, although to different degrees. As the group-as-whole develops in classrooms, the individuals comprising the group also develop (Lantolf & Poehner, 2010, p. 24). Future research using case studies should be conducted to examine individual development within group-as-whole development in EFL writing courses. Variability can appear in the ways that each student writer contributes his or her knowledge to a group and how this knowledge contributes to others’ performance.

Another important issue to explore is how novice-novice writers provided with mediation tools can create their group ZPD and how that might lead to students’ individual writing development. I wonder about differences in the effects on writing development of novice-novice groups with assistance or of expert-novice groups. The differences might indicate the best way to arrange students in a CW classroom to maximize their learning writing.

Finally, for researchers who want to use GEE techniques, close attention needs to be paid to the development of goodness-of-fit tests for GEE models (Ballinger, 2004). Over the past years, even though new methods have been introduced, there is still no universally accepted

116 test for goodness of fit for GEE models in use that extends beyond binary dependent variables

(Ballinger, 2004). An alternate approach to yield accurate estimates is to choose an appropriate correlation structure for GEE as I did in the present research.

117

REFERENCES

Ahmadian, M. J. (2012). Task repetition in ELT. ELT Journal, 66(3), 380-382.

Alexander, J. (2009). Gaming, student literacies, and the composition classroom: Some

possibilities for transformation. College Composition and Communication, 61(1), 35-63.

Anderson, J. (1982). Acquisition of cognitive skill. Psychological Review, 89(4), 369-406.

Anton, M., & DiCamilla, F. (1998). Socio-cognitive functions of L1 collaborative interaction in

the L2 classroom. Canadian Modern Language Review, 54(3), 314-342.

Baba, K., & Nitta, R. (2011). Dynamic effects of repeating a timed writing task in two EFL

university courses: Multi-element text analysis with Coh-Metrix. In P. M. McCarthy & C.

Boonthum (Eds.), Applied natural language processing and content analysis:

Identification, investigation, and resolution (pp. 397–411). Hershey, PA: IGI Global.

Baba, K., & Nitta, A. (2014). Phase transitions in development of writing fluency from a

complex dynamic systems perspective. Language Learning, 64 (1), 1-35.

Ballinger, G. A. (2004). Using generalized estimating equations for longitudinal data analysis.

Organizational Research Methods, 7(2), 127-150.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W. H. Freeman and

Company

Bandura, A., & Locke, E. A. (2003). Negative self-efficacy and goal effects revisited. Journal of

Applied Psychology, 88(1), 87-99.

Boscolo, P. (2009). Engaging and motivating children to write. In R. Beard, D. Myhill, M.

Nystrand & J. Riley (Eds.), The SAGE handbook of writing development (pp. 300-312).

London: Sage.

118

Brooks, L., & Swain, M. (2009). Languaging in CW: Creation of and response to expertise. In A.

Mackey & C. Polio (Eds.), Multiple perspectives on interaction (pp. 58-89). New York:

Routledge.

Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning.

Educational Researcher, 18(1), 32-42.

Brown, P., & Levinson, S. (Eds.)(1987). Politeness: Some universals in language usage. New

York: Cambridge University Press.

Carson, J., & Nelson, G. (1994). Writing groups: Cross-cultural issues. Journal of Second

Language Writing, 3(1), 17-30.

Casanave, C. P. (2004). Controversies in second language writing: Dilemmas and decisions in

research and instruction. Ann Arbor: University of Michigan.

Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation analysis for the

behavioral sciences. New York: John Wiley.

Cole, M. (1996). Cultural psychology: A once and future discipline. Cambridge, MA: Harvard

University Press.

Cumming, A. H. (Ed.) (2006). Goals for academic writing: ESL students and their instructors.

Philadelphia: John Benjamins.

Cumming, A. (2013a). Writing. In P. Robinson (Ed.), Routledge encyclopedia of second

language acquisition (pp. 703-707). London: Routledge.

Cumming, A. (2013b). Multiple dimensions of academic language and literacy development. In

L. Ortega, Cumming & N. Ellis (Eds.) Agendas for language learning research.

Supplement 1 to Language Learning, 63, 130-152.

Cumming, A., & Riazi, A. (2000). Building models of adult second-language writing instruction.

Learning and Instruction, 10(1), 55-71. de Guerrero, M., & Villamil, O. (1994). Social-cognitive dimensions of interaction in L2 peer

119

revision. The Modern Language Journal, 78, 484-496.

DeKeyser, R. (2007). Skill acquisition theory. In B. VanPatten & J. Williams (Eds.), Theories in

Second Language Acquisition: An Introduction (pp. 97-114). London: Lawrence

Erlbaum Associates.

Department of Conservation of Cultural Relics and Museology. (2011) Graduate studies bulletin.

Retrieved on June 10, 2014 from http://giccrmo.tnnua.edu.tw/ftp/20121108044906.pdf

DiCamilla, F. J., & Anton, M. (1997). Repetition in the collaborative discourse of L2 learners: A

Vygotskian perspective. Canadian Modern Language Review, 53(4), 610-635.

Dobao, A. (2012). CW tasks in the L2 classroom: Comparing group, pair, and individual work.

Journal of Second Language Writing, 21(1), 40-58.

Donato, R. (1994). Collective scaffolding in second language learning. In J. P. Lantolf & G.

Appel (Eds.), Vygotskian Approaches to Second Language Learning (pp. 33-56).

Norwood, NJ: Ablex.

Donato, R., & McCormick, D. (1994). A sociocultural perspective on language learning

strategies: The role of mediation. Modern Language Journal, 78, 453-464.

Ellis, R. (2003). Task-based language learning and teaching. Oxford, UK: Oxford University

Press.

Engeström, Y., Miettinen, R., & Punamaki, R.-L. (Eds.). (1999). Perspectives on activity theory.

Cambridge, UK: Cambridge University Press.

Fernandez Dobao, A. (2012). CW tasks in the L2 classroom: Comparing group, pair, and

individual work. Journal of Second Language Writing, 21(1), 40-58.

Fitzgerald, J. (1987). Research on revision in writing. Review of Educational Research, 57(4),

481-506.

Flower, L. S., & Hayes, J. R. (1977). Problem-solving strategies and the writing process.

College English, 39(4), 449-461.

120

Fry, S. W., & Griffin, S. (2010). Fourth graders as models for teachers: Teaching and learning

6+1 trait writing as a collaborative experience. Literacy Research and Instruction, 49(4),

283-298.

Garson, G. D. (2012). Generalized Linear Models & Generalized Estimating Equations.

Retrieved from http://www.amazon.ca/Generalized-Estimating-Equations-Statistical-

Associates-ebook/dp/B009434OUQ

Greta, G. (2009). Investigating second language learner self-efficacy and future expectancy of

second language use for high-stakes program evaluation. Foreign Language Annals,

42(3), 505-540.

Green, S. B. (1991). How many subjects does it take to do a regression analysis? Multivariate

Behavioral Research, 26(3), 499-510.

Gutierrez, G., & Adela, G. (2006). Sociocultural theory and its application to CALL: A study of

the computer and its relevance as a mediational tool in the process of collaborative

activity. ReCALL, 18(2), 230-251.

Hamp-Lyons. (1991). Reconstructing "academic writing proficiency" In L. Hamp-Lyons (Ed.),

Assessing second language writing in academic contexts (pp. 127–153). Norwood, NJ:

Ablex.

Hamp-Lyons, & Henning, G. (1991). Communicative writing profiles: An investigation of the

transferability of a multiple-trait scoring instrument across ESL writing assessment

contexts. Language Learning, 41, 337-373.

Harris, R. J. (1975). A primer of multivariate statistics. New York: Academic Press.

Hyland, K. (2011). Learning to write: Issues in theory, research, and pedagogy. In R. Manchon

(Ed.) Learning-to-write and writing-to-learn in an additional language (pp. 17-35).

Amsterdam: John Benjamins.

Kemp, J. E., Morrison, G. R., & Ross, S. M. (1998). Designing effective instruction. Upper

121

Saddle River, NJ: Merrill.

Kessler, G. (2009). Student-initiated attention to form in wiki-based CW. Language Learning &

Technology, 13(1), 79-95.

Kim, Y., & Tracy-Ventura, N. (2013). The role of task repetition in L2 performance development:

What needs to be repeated during task-based interaction? System, 41(3), 829-840.

Kinginger, C. (2002). Defining the zone of proximal development in US foreign language

education. Applied Linguistics, 23, 240-261.

Kowal, M., & Swain, M. (1994). Using collaborative language production tasks to promote

students' language. Language Awareness, 3(2), 73-93.

Kozlowski, Steve W. J., Salas, Eduardo, & Society for Industrial and Organizational,

Psychology. (2010). Learning, training, and development in organizations. New York:

Routledge.

Lantolf, J. P. (Ed.). (2000). Sociocultural theory and second language learning. Oxford: Oxford

University Press.

Lantolf, J. P., & Genung., P. B. (2002). I'd rather switch that fight: An activity-theoretical study

of power, success, and failure in a foreign language In C. J. Kramsch (Ed.), Language

acquisition and language socialization: Ecological perspectives (pp. 175-196). New

York: Continuum.

Lantolf, J. P., & Poehner, M. E. (2010). Dynamic assessment in the classroom: Vygotskian

praxis for second language development. Language Teaching Research, 15(1), 11-33.

Lantolf, J. P., & Thorne, S. L. (2006). Sociocultural theory and the genesis of second language

development. Oxford: Oxford University Press.

Lantolf, J. P., & Thorne, S. L. (2007). Sociocultural theory and second language learning. In B.

VanPatten & J. Williams (Eds.), Theories in second language acquisition: An

introduction (pp. 201-224). New York: Routledge.

122

Larsen-Freeman, D. (2009). Adjusting expectations: The study of complexity, accuracy and

fluency in second language acquisition. Applied Linguistics, 30, 579–589.

Lauer, T., & Hendrix, J. (2009). A model for quantifying student learning via repeated writing

assignments and discussions. International Journal on Teaching and Learning in Higher

Education, 20(3), 425-433.

Leki, I., Cumming, A. H., & Silva, T. (2008). A synthesis of research on second language

writing in English. New York: Routledge.

Leont’ev, A. N. (1978). Activity, consciousness, and personality. Englewood Cliffs, N.J.:

Prentice-Hall.

Liang, K.-Y., & Zeger, S. L. (1986). Longitudinal data analysis using generalized linear models.

Biometrika, 73(1), 13-22.

Lin, M.-Y. (2001). A study of designing a course website to facilitate Sociolinguistics learning.

(Unpublished masters thesis), Providence University, Taiwan.

Locke, E. A., & Latham, G. P. (2002). Building a practically useful theory of goal setting and

task motivation: A 35-year odyssey. The American Psychologist, 57(9), 705-717.

Long, M. (1996). The role of the linguistic environment in second language acquisition. In W. C.

Ritchie & T. K. Bhatia (Eds.), Handbook of second language acquisition (pp. 413-468).

San Diego: Academic Press.

Lowry, P. B., Nunamaker, J. F., Curtis, A., & Lowry, M. R. (2005). The impact of process

structure on novice, virtual collaborative writing teams. Jr IEEE Transactions on

Professional Communication, 48(4), 341-364.

Magogwe, J. M., & Oliver, R. (2007). The relationship between language learning strategies,

proficiency, age and self-efficacy beliefs: A study of language learners in Botswana.

System, 35, 338–352.

Mak, B., & Coniam, D. (2008). Using wikis to enhance and develop writing skills among

123

secondary school students in Hong Kong. System, 36(3), 437-455.

Manchón, R. (2009). Writing in foreign language contexts: learning, teaching, and research.

Buffalo, NY: Multilingual Matters.

Mathison, S. (1988). Why triangulate? Educational Researcher, 17(2), 13-17.

Matthews, P. H. (2010). Factors influencing self-efficacy judgments of university students in

foreign language tutoring. The Modern Language Journal, 94(4), 618–635.

Murphy, E., & Roca de Larios, J. (2010). Searching for words: One strategic use of the

mother tongue by advanced Spanish EFL learners. Journal of Second Language Writing,

19, 2, 61-81.

Nassaji, H., & Cumming, A. (2000). What’s in a ZPD? A case study of a young ESL

student and teacher interacting through dialogue journals. Language Teaching Research,

4, 95-121.

Norris, J., & Manchon, R. (2012). Investigating L2 writing development from multiple

perspectives: Issues in theory and research. In R. Manchon (Ed.), L2 writing

development: Multiple perspectives (pp. 221-244). Boston: De Gruyter Mouton.

Norris, J. M., & Ortega, L. (2003). Defining and measuring L2 acquisition. In C. Doughty & M.

H. Long (Eds.), Handbook of second language acquisition (pp. 717-761). New York:

Blackwell.

Norris, J & Ortega, L. (2012). Assessing learner knowledge. In Susan Gass & Alison Mackey

(Eds.), Handbook of second language acquisition, (pp. 573-589). NewYork: Routledge.

Oi, K. (1999). Comparison of argumentative styles: Japanese college students vs. American

college students: An analysis using the Toulmin model. JACET Bulletin 30, 85-102.

Ortega, L. (2009). Studying writing across EFL contexts: Looking back and moving forward. In

R. Manchon (Ed.), Writing in foreign language contexts: Learning, teaching and

research (pp. 232-255). Bristol, UK: Multilingual Matters.

124

Pajares, F. (2003). Self-efficacy beliefs, motivation, and achievement in writing: A review of the

literature. Reading & Writing Quarterly, 19(2), 139-158.

Pajares, F., & Johnson, M. J. (1994). Confidence and competence in writing: The role of

self-efficacy, outcome expectancy, and apprehension. Research in the Teaching of

English, 28(3), 313.

Pajares, F., & Valiante. (2006). Self-efficacy beliefs and motivation in writing development. In

C. A., MacArthur, S. Graham & J. Fitzgerald (Eds.), Handbook of writing research (pp.

158-170). New York: Guilford Press.

Palinscar, A. S., & Brown, A. L. (1984). Reciprocal teaching of comprehension-fostering and

comprehension-monitoring activities. Cognition and Instruction, 1(2), 117-175.

Perin, D. (2002). Repetition and the informational writing of developmental students. Journal of

Developmental Education, 26(1), 2.

Poehner, M. E. (2009). Group dynamic assessment: Mediation for the L2 classroom. TESOL

Quarterly, 43(3), 471-491.

Polio, C. (1997). Measures of linguistic accuracy in second language writing research.

Language Learning, 47(1), 101-143.

Polio, C. (2001). Research methodology in second language writing: The case of text-based

studies. In T. Silva & P. Matsuda (Eds.), On second language writing (p. 91-116).

Mahwah, NJ: Erlbaum.

Polio, C. (2012). How to research second language writing. In A. Mackey & S. M. Gass (Eds.),

Research methods in second language acquisition: A practical guide (pp. 139-157).

Malden, MA: Wiley.

Ranker, J. (2009). Student appropriation of writing lessons through hybrid composing practices:

Direct, diffuse, and indirect use of teacher-offered writing tools in an ESL classroom.

Journal of Literacy Research, 41(4), 393-431.

125

Raoofi, S., Chan, S. H., Mukundan, J., & Rashid, S. M. (2014). Metacognition and

second/foreign language learning. English Language Teaching, 7(1), 36.

Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization

of turn-taking for conversation. Language, 50(4), 696-735.

Scott, V. M., & de la Fuente, M. J. (2008). What's the problem? L2 learners' use of the L1 during

consciousness-raising, form-focused tasks. The Modern Language Journal, 92(1),

100-113.

Shehadeh, Ali. (2011). Effects and student perceptions of collaborative writing in L2. Journal of

Second Language Writing, 20(4), 286-305.

Shell, D. F., Murphy, C. C., & Bruning, R. H. (1989). Self-efficacy and outcome expectancy

mechanisms in reading and writing achievement. Journal of Educational Psychology,

81(1), 91-100.

Smith, P. A., Newman, I. A., & Parks, L. M. (1997). Virtual hierarchies and virtual networks:

Some lessons from hypermedia usability research applied to the World Wide Web.

International Journal of Human-Computer Studies, 47(1), 67-95.

Solórzano, H. (2006). Northstar: Building skills for the TOEFL® iBT high intermediate. New

York: Pearson Longman.

Stetsenko, A., & Arievitch, I. (1997). Constructing and deconstructing the self: Comparing

post-vygotskian and discourse-based versions of social constructivism. Mind, Culture,

and Activity, 4(3), 159-172.

Storch, N. (1998). A classroom-based study: insights from a collaborative text reconstruction

task. ELT Journal, 52(4), 291-300.

Storch, N. (2002). Patterns of interaction in ESL pair work. Language Learning, 52(1), 119-158.

Storch, N. (2005). Collaborative writing: Product, process, and students' reflections. Journal of

Second Language Writing, 14(3), 153-173.

126

Storch, N. (2013). Collaborative writing in L2 classrooms. Buffalo: Multilingual Matters.

Storch, N., & Aldosari, A. (2010). Learners' use of first language (Arabic) in pair work in an

EFL class. Language Teaching Research, 14(4), 355-375.

Storch, N., & Wigglesworth, G. (2003). Is there a role for the use of the L1 in an L2 setting?

TESOL Quarterly, 37(4), 760-760.

Su, M. H., & Duo, P. C. (2012). EFL learners’ language learning strategy use and perceived

self-efficacy. European Journal of Social Sciences, 27(3), 335-345.

Swain, M. (2000). The output hypothesis and beyond: Mediating acquisition through

collaborative dialogue. In J. P. Lantolf (Ed.), Sociocultural theory and second language

learning (pp. 97-114). Oxford: Oxford University Press.

Swain, M., Brooks, L., & Tocalli-Beller, A. (2002). Peer-peer dialogue as a means of second

language learning. Annual Review of Applied Linguistics, 22, 171-185.

Swain, M. & Lapkin, S. (1998). Interaction and second language learning: Two adolescent

French immersion students working together. The Modern Language Journal, 82(3),

320-337.

Swain, M., & Lapkin, S. (2000). Task-based second language learning: The uses of the first

language. Language Teaching Research, 4(3), 251-274.

Swain, M., & Lapkin, S. (2002). Talking it through: Two French immersion learners’ response to

reformulation. International Journal of Educational Research, 37(3), 285-304.

Twisk, J. W. R. (2003). Applied longitudinal data analysis for epidemiology: A practical guide.

New York: Cambridge University Press.

Verspoor, M., Schmid, M., & Xu, X. (2012). A dynamic usage based perspective on L2 writing.

Journal of Second Language Writing, 21 (3), 239-263.

Villamil, O., & de Guerrero, M. (1996). Peer revision in the L2 classroom: Social-cognitive

activities, mediating strategies and aspects of social behavior. The Journal of Second

127

Language Writing, 5(1), 51-75.

Vygotsky, L. S. (1987a). The collected works of L. S. Vygotsky, Volume 1: Problems of general

psychology. New York: Plenum Press.

Vygotsky, L. S. (1987b). Thinking and speech (N. Minick, Trans.). In R. W. Rieber & A. S.

Carton (Eds.), The collected works of L. S. Vygotsky: Vol. 1. Problems of general

psychology (pp. 39-285). New York: Plenum Press (Original work published 1934).

Wang, C., & Pape, S. J. (2007). A probe into three Chinese boys' self-efficacy beliefs learning

English as a second language. Journal of Research in Childhood Education, 21(4),

364-377.

Warschauer, M. (1997). Computer-mediated collaborative learning: Theory and practice.

Modern Language Journal, 81(3), 470-481.

Weigle, S. C. (1999). Investigating rater/prompt interactions in writing assessment: Quantitative

and qualitative approaches. Assessing Writing, 6(2), 145-178.

Wells, C. G. (1999). Dialogic inquiry: Towards a sociocultural practice and theory of education:

New York: Cambridge University Press.

Wertsch, J. V. (1985). Vygotsky and the social formation of mind. Cambridge, Mass.: Harvard

University Press.

Wey, S.-C. (1998). The effects of goal orientations, metacognition, self-efficacy and effort on

writing achievement. (Unpublished doctoral dissertation), University of Southern

California, Los Angeles.

Wigglesworth, G., & Storch, N. (2009). Pair versus individual writing: Effects on fluency,

complexity and accuracy. Language Testing, 26, 445-466.

Wong. M. S. L. (2005). Language learning strategies and language self-efficacy: Investigating

the relationship in Malaysia. Regional Language Centre Journal, 36(3), 245-269.

Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of

128

child psychology and psychiatry, and allied disciplines, 17(2), 89-100.

Yang, L., Baba, K., & Cumming, A. H. (2004). Activity systems for ESL writing improvement:

Case studies of three Chinese and three Japanese adult learners of English. In D.

Albrechtsen, K. Haastrup & B. Henriksen (Eds.), Writing and vocabulary in foreign

language acquisition. Special issue of Angles on the English-speaking World, 4, 13-33.

Copenhagen: Museum Tusculanem Press.

Yau, Jia-ling C. (2009). Reading characteristics of Chinese-English adolescents:

Knowledge and application of strategic reading. Metacognition and Learning, 4(3), 217.

129

Appendix A

Independent Writing Tasks

Writing task 0: Pretest and posttest Do you agree or disagree with the following statement? Always telling the truth is the most important consideration in any relationship. Use specific reasons and examples to support your answer. Writing Task 1 Some young adults want independence from their parents as soon as possible. Other young adults prefer to live with their families for a longer time. Which of these situations do you think is better? Use specific reasons and examples to support your answer. Writing Task 2 Do you agree or disagree with the following statement? People benefit more from traveling in their own country than from traveling to foreign countries. Use specific reasons and examples to support your answer. Writing Task 3 Do you agree or disagree with the following statement? A teacher’s ability to relate well with students is more important than excellent knowledge of subject being taught. Use specific reasons and examples to support your answer. Writing Task 4 Do you agree or disagree with the following statement? Young people enjoy life more than older people do. Use specific reasons and examples to support your answer.

130

Appendix B

Transcription Conventions

(Excerpts were translated from an original Mandarin version)

Walter; Speakers' names are separated from their utterances by semicolons, followed by a few blank spaces. ?; A question mark instead of a name or initial indicates that no good guess could be made as to the identity of the speaker. ??; Multiple question marks followed by semicolon indicate that the speaker's identity is not clear but there are reasons to believe that it is someone different from the last unidentified speaker. ?Walter; a question mark before the name of the speaker stands for a probable but not safe guess regarding the identity of the speaker. (1.5) Numbers between parentheses indicate length of pauses in seconds and tenths of seconds. ... Three dots indicate an untimed pause. [...] Three dots between square brackets indicate that some material of the original transcript or example has been omitted or that the transcript starts or ends in the middle of further talk. = Equal signs indicate 'latching,' that is, two utterances that follow one another without any perceptible pause. [ A square bracket between turns indicates the point at which overlap by another speaker starts. // Double obliques indicate the point at which overlap by the next speaker begins. (don't) Words between parentheses in the transcripts represents the best guess of a stretch of talk which was difficult to hear. ( ? ? ) Blank spaces inside parentheses with occasional question marks indicate uncertain or unclear talk of approximately the length of the blank spaces between parentheses. (( )) Material between double quotes provides extralinguistic information, e.g. about bodily movements. so::: colons indicate the lengthening of the last sound.

131

Appendix C

Questionnaire on Mediation Tool Use

(translated from an original Mandarin version)

Group members:______/______Your name: ______Participants have to print out their Web browsing history and make reference to it to answer the following questions. 1. How many times in total did you search for answers to your writing questions on the Internet? Please count times on the basis of your Web browsing history. (Times)______2. Following question one, what types of search did you do while you were writing (e.g., lexical meaning, translating Chinese into English, grammar, collocation, etc.)? Please write down your search types in reference to your Web browsing history. Please refer to your Web browsing history and match the types of searching to times of searching. Fill in your types of searching Times 1. (Examples) translating Chinese to English 8 2. 3. 4. 5. 6. 7. 3. Have you ever asked your partner(s) for help if you had writing problems? If you did, how many times and what types of questions did you ask for his/her help? Did you get answers from him/her? Fill in types of questions Did I get answers? Times 1. (Examples) the meaning of the prompt Yes 2 2. 3. 4. 5. 6.

132

7. 4. Did you have any writing questions that you didn’t know how to solve and didn’t solve them? How many times? What types of questions were they? Fill in your writing questions that you didn’t know how to solve Times them 1. (Examples) 2 2. 3. 4. 5. 6. 7. 5. How many times did you read the prompt? ______6. How many times did you listen to the prompt? ______

133

Appendix D

Coding Table of Writing Groups’ References to

Tool-mediated Writing Knowledge

groups' making reference to tool-mediated knowledge or not frequency (1=yes, 0=no) language language groups time skills of expressions skills of expressions forms forms cwa 1 1 0 2 0 cwb 1 1 0 2 0 cwc 1 1 0 40 0 cwd 1 1 1 55 20 cwe 1 1 1 21 12 cwf 1 1 0 45 0 cwg 1 1 1 50 2 cwh 1 1 1 58 23 cwi 1 1 1 24 2 cwj 1 1 1 15 2 cwk 1 1 1 35 9 cwl 1 1 1 28 13 cwm 1 1 0 37 0 cwn 1 1 1 10 4 cwo 1 1 1 27 3 cwp 1 1 0 30 0 cwa 2 1 1 55 3 cwb 2 0 0 0 0 cwc 2 1 1 35 1 cwd 2 1 1 15 4 cwe 2 1 1 12 7 cwf 2 1 1 20 7 cwg 2 1 1 40 12 cwh 2 1 1 17 4 cwi 2 1 1 13 1 cwj 2 1 1 12 5 cwk 2 1 1 35 3 cwl 2 1 1 15 6 cwm 2 1 1 2 3

134 cwn 2 1 1 17 6 cwo 2 1 1 33 5 cwp 2 1 1 20 4 cwa 3 1 1 18 3 cwb 3 1 1 23 11 cwc 3 1 1 20 8 cwd 3 1 1 25 5 cwe 3 1 1 8 4 cwf 3 1 1 15 5 cwg 3 1 0 14 0 cwh 3 1 1 15 20 cwi 3 1 1 17 3 cwj 3 1 1 10 3 cwk 3 1 1 9 12 cwl 3 1 1 10 7 cwm 3 1 1 2 11 cwn 3 1 1 3 5 cwo 3 1 1 26 11 cwp 3 1 1 20 22 cwa 4 0 1 0 8 cwb 4 1 1 8 3 cwc 4 1 1 30 12 cwd 4 1 1 30 15 cwe 4 1 1 12 5 cwf 4 1 1 13 7 cwg 4 1 1 20 5 cwh 4 1 1 30 6 cwi 4 1 1 13 4 cwj 4 1 1 16 9 cwk 4 1 1 30 4 cwl 4 1 1 5 10 cwm 4 1 1 3 5 cwn 4 1 1 6 7 cwo 4 1 1 43 20 cwp 4 1 1 50 12

135

Appendix E

Communicative Writing Rubrics

(From Hamp-Lyons, 1991, pp. 147-148) Scales Communicative Quality Organization Argumentation Linguistic Accuracy Linguistic Appropriacy 9 The writing The writing Relevant arguments are The reader sees no errors of There is an ability to displays an ability to displays presented in an interesting vocabulary, spelling, manipulate the linguistic communicate completely way, with main ideas punctuation, or grammar system with complete in a way that gives the logical prominently and clearly appropriacy. reader full organizational stated, with complete satisfaction. structure, effective supporting enabling the material; arguments are message to be effectively related to the followed writer’s experience or views effortlessly. 8 The writing display an The writing displays a Relevant arguments are The reader sees no There is an ability to ability to communicate logical organizational presented in an interesting significant errors of manipulate the linguistic without causing the structure that enables the way, with main ideas vocabulary, punctuation, or systems appropriately. reader any difficulties. message to be followed highlighted, effective grammar. easily. supporting material and they are well related to the writer’s own experience or views. 7 The writing displays an The writing displays good Arguments are well The reader is aware of but There is limited ability to ability to communicate organizational structure presented with relevant not troubled by occasional manipulate the linguistic with few difficulties for that enables the message to supporting material and an errors of vocabulary, systems appropriately, but the reader. be followed throughout. attempt to relate them to the spelling, punctuation, or this doesn’t intrude on the writer’s experience or views. grammar. reader 6 The writing displays an The writing is organized Arguments are presented, The reader is aware of errors There is limited ability to ability to communicate well enough for the but it may be difficult for the of vocabulary, spelling, or manipulate the linguistic although there is message to be followed reader to distinguish main grammar－but only systems appropriately, but occasional strain for the throughout. ideas from supporting occasionally. this intrudes only reader. material; main ideas may not occasionally. be supported; their relevance may be dubious; arguments may not be related to the writer’s experience or views.

136

Scales Communicative Quality Organization Argumentation Linguistic Accuracy Linguistic Appropriacy 5 The writing displays an The writing is organized Arguments are presented, The reader is aware of errors There is limited ability to ability to communicate well enough for the but may lack relevance of vocabulary, spelling, manipulate the linguistic although there is often message to be followed clarity, consistency, or punctuation, or grammar that systems appropriately, which strain for the reader. most of the time. support; they may not be intrude frequently. intrudes frequently. related to the writer’s experience or views. 4 The writing shows a The writing lacks a clear Arguments are inadequately The reader finds the control There is inability to limited ability to organizational structure presented and supported; of vocabulary, spelling, manipulate the linguistic communicate, which puts and the message is difficult they may be irrelevant; if the punctuation, and grammar systems appropriately, which a strain on the reader to follow. writer’s experience or views inadequate. causes severe strain for the throughout. are presented, their reader. relevance may be difficult to see. 3 The writing does not The writing has no Some elements of The reader is aware There is little or no sense of display an ability to discernable organizational information are presented, primarily of gross linguistic appropriacy, communicate, although structure and a message but the reader is not inadequacies of vocabulary, although there is evidence of meaning comes through cannot be followed. provided with an argument, spelling, punctuation, and sentence structure. spasmodically. or the argument is mainly grammar. irrelevant. 2 The writing displays no No organizational structure A meaning comes through The reader sees no evidence There is no sense of ability to communicate. or message is recognizable. occasionally, but it is not of control of vocabulary, linguistic appropriacy. relevant. spelling, punctuation, or grammar. 1 A true nonwriter who has not produced any assessable strings of English writing. An answer that is wholly or almost wholly copied from the input text or task is in this category. 0 This rating should be used only when a candidate did not attend or attempt this part of the test in any way.

137

Appendix F

Holistic Rating Scales

(From Hamp-Lyons and Henning, 1991, pp. 370-373) The writing displays an ability to communicate in a way which gives the reader full satisfaction. It displays a completely logical organizational structure which enables the message to be followed effortlessly. Relevant arguments are 9 presented in an interesting way, with main ideas prominently and clearly stated, with completely effective supporting material; arguments are effectively related to the writer's experience or views. There are no errors of vocabulary, spelling, punctuation, or grammar and the writing shows an ability to manipulate the linguistic systems with complete appropriacy. The writing displays an ability to communicate without causing the reader any difficulties. It displays a logical organizational structure which enables the message to be followed easily. Relevant arguments are presented in an interesting 8 way, with main ideas highlighted, effective supporting materials and they are well related to the writer's own experience or views. There are no significant errors of vocabulary, spelling, punctuation or grammar and the writing reveals an ability to manipulate the linguistic systems appropriately. The writing displays an ability to communicate with few difficulties for the reader. It displays good organizational structure which enables the message to be followed without much effort. Arguments are well presented with relevant 7 supporting material and attempt to relate them to the writer's experience or views. The reader is aware of but not troubled by occasional minor errors of vocabulary, spelling, punctuation or grammar, and/or some limitations to the writer's ability to manipulate the linguistic systems, appropriately. The writing displays an ability to communicate although there is occasionally strain for the reader. It is organized well enough for the message to be followed throughout. Arguments are presented but may be difficult for the reader to 6 distinguish main ideas from supporting material; main ideas may not be supported; their relevance may be dubious; arguments may not be related to the writer's experience or views. The reader is aware of errors of vocabulary, spelling, punctuation or grammar, and/or limited ability to manipulate the linguistic systems appropriately, but these intrude only occasionally. The writing displays an ability to communicate although there is often strain for the reader. It is organized well enough for the message to be followed most of 5 the time. Arguments are presented but may lack relevance, clarity, consistency or support; they may not be related to the writer's experience or views. The reader is aware of errors of vocabulary, spelling, punctuation or grammar which intrude frequently, and of limited ability to manipulate the linguistic systems appropriately. The writing displays a limited ability to communicate which puts strain on the reader. It lacks a clear organizational structure and the message is difficult to follow. Arguments are inadequately presented an supported; they may be irrelevant; 4 if the writer's experience or views are presented their relevance may be difficult to see. The control of vocabulary, spelling, punctuation and grammar is inadequate, and the writer displays inability to manipulate the linguistic systems appropriately, causing severe strain for the reader. 137

138

The writing does not display an ability to communicate although meaning comes through spasmodically. The reader cannot find any organizational structure and cannot follow a message. Some elements of information are presented but the 3 reader is not provided with an argument, or the argument is mainly irrelevant. The reader is primarily aware of gross inadequacies of vocabulary, spelling, punctuation and grammar, the writer seems to have no sense of linguistic appropriacy, although there is evidence of sentence structure. The writing displays no ability to communicate. No organizational structure 2 or message is recognizable. A meaning comes through occasionally but it is not relevant. There is no evidence of control of vocabulary, spelling, punctuation or grammar, and no sense of linguistic appropriacy. A true non-writer who has not produced any assessable strings of English 1 writing. An answer which is wholly or almost wholly copied from the input text or task is in this category. Should not be used where a candidate did not attend or attempt this part of the 0 test in any way (i.e., did not submit an answer paper with his/her name and candidate number written on).

139

Appendix G

Correlation Coefficients for Six Outcome Variables over Four Weeks

Table J 1 Correlation Coefficients for Holistic Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 0.06 0.56 0.20 2. Week 2 0.06 1.00 0.39 -0.08 3. Week 3 0.56 0.39 1.00 -0.09 4. Week 4 0.20 -0.08 -0.09 1.00

Table J 2 Correlation Coefficients for Communicative Quality Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 -0.08 0.09 0.26 2. Week 2 -0.08 1.00 0.20 -0.14 3. Week 3 0.09 0.20 1.00 0.27 4. Week 4 0.26 -0.14 0.27 1.00

Table J 3 Correlation Coefficients for Organization Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 -0.17 0.41 0.08 2. Week 2 -0.17 1.00 0.35 0.24 3. Week 3 0.41 0.35 1.00 0.05 4. Week 4 0.08 0.24 0.05 1.00

Table J 4 Correlation Coefficients for Argumentation Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 -0.08 0.61 0.42 2. Week 2 -0.08 1.00 0.27 0.20 3. Week 3 0.61 0.27 1.00 0.36 4. Week 4 0.42 0.20 0.36 1.00

Table J 5 Correlation Coefficients for Linguistic Accuracy Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 0.42 0.62 0.59

140

2. Week 2 0.42 1.00 0.44 0.47 3. Week 3 0.62 0.44 1.00 0.52 4. Week 4 0.59 0.47 0.52 1.00

Table J 6 Correlation Coefficients for Linguistic Appropriacy Scores over 4 Weeks 1 2 3 4 1. Week 1 1.00 0.49 0.63 0.54 2. Week 2 0.49 1.00 0.50 0.35 3. Week 3 0.63 0.50 1.00 0.26 4. Week 4 0.54 0.35 0.26 1.00

141

Appendix H

Post-writing Questionnaire

Date:______Group:______Part I: Please write your reflections on the following questions.

1. What worked really well in this task? Why?

A. ______B. ______C. ______

2. What did not work well in this task? Why?

A. ______B. ______C. ______

Part II: If you agree with the degree of your communication and relationship in dyadic writing, please tick the box next to the number.

Quality of discussions: Overall quality of discussions was poor 1□ 2□ 3□ 4□ 5□ 6□ 7□ good Outcome of the discussions was satisfactory 1□ 2□ 3□ 4□ 5□ 6□ 7□ unsatisfactory Execution of the discussions was incompetent 1□ 2□ 3□ 4□ 5□ 6□ 7□ competent Development of discussion contents was careful 1□ 2□ 3□ 4□ 5□ 6□ 7□ careless

Process satisfaction: “Our group-decision making process was. . .” Inefficient 1□ 2□ 3□ 4□ 5□ 6□ 7□ efficient Uncoordinated 1□ 2□ 3□ 4□ 5□ 6□ 7□ coordinated Fair 1□ 2□ 3□ 4□ 5□ 6□ 7□ unfair

142

Confusing 1□ 2□ 3□ 4□ 5□ 6□ 7□ understandable Satisfying 1□ 2□ 3□ 4□ 5□ 6□ 7□ unsatisfying

Agreement: “We experienced. . .” Consensus 1□ 2□ 3□ 4□ 5□ 6□ 7□ discord Conflict 1□ 2□ 3□ 4□ 5□ 6□ 7□ harmony High levels of participation 1□ 2□ 3□ 4□ 5□ 6□ 7□ low levels of participation Good team work 1□ 2□ 3□ 4□ 5□ 6□ 7□ bad team work Disagreement on our duties and roles 1□ 2□ 3□ 4□ 5□ 6□ 7□ agreement on our duties and roles

Communication richness: “In terms of our group’s communication, it can be said that. . .” Responses lacked details 1□ 2□ 3□ 4□ 5□ 6□ 7□ were filled with details Messages were very vivid 1□ 2□ 3□ 4□ 5□ 6□ 7□ unclear Forms of expression had high variety 1□ 2□ 3□ 4□ 5□ 6□ 7□ high redundancy The amount of information was lean 1□ 2□ 3□ 4□ 5□ 6□ 7□ rich

Task discussion effectiveness: The quality of the discussions was poor 1□ 2□ 3□ 4□ 5□ 6□ 7□ good The discussions were ineffective 1□ 2□ 3□ 4□ 5□ 6□ 7□ effective Discussion outcome unsatisfactory 1□ 2□ 3□ 4□ 5□ 6□ 7□ satisfactory The context of the discussions were carelessly developed 1□ 2□ 3□ 4□ 5□ 6□ 7□ carefully developed Issues were examined effectively 1□ 2□ 3□ 4□ 5□ 6□ 7□ ineffectively Participation in the discussions was unevenly distributed 1□ 2□ 3□ 4□ 5□ 6□ 7□ evenly distributed Ideas in the discussions were uncritically examined 1□ 2□ 3□ 4□ 5□ 6□ 7□ critically examined The amount of information exchanged was sufficient 1□ 2□ 3□ 4□ 5□ 6□ 7□ insufficient

Communication appropriateness: The discussions were appropriate 1□ 2□ 3□ 4□ 5□ 6□ 7□ inappropriate The discussions were on topic 1□ 2□ 3□ 4□ 5□ 6□ 7□ off topic The discussions were very unsatisfying 1□ 2□ 3□ 4□ 5□ 6□ 7□ very satisfying

Teamwork: “In general I would say that my team members. . .” Were direct and honest with each other 1 □ 2□ 3□ 4□ 5□ 6□ 7□ indirect and dishonest Did not accept criticism 1□ 2□ 3□ 4□ 5□ 6□ 7□ accepted criticism Resolved disagreements cooperatively 1□ 2□ 3□ 4□ 5□ 6□ 7□ uncooperatively Functioned as a team 1□ 2□ 3□ 4□ 5□ 6□ 7□ did not function as a team Were cooperative and considerate 1□ 2□ 3□ 4□ 5□ 6□ 7□ uncooperative and inconsiderate Confronted problems constructively 1□ 2□ 3□ 4□ 5□ 6□ 7□ unconstructively Were poor listeners 1□ 2□ 3□ 4□ 5□ 6□ 7□ good listeners Were unconcerned about each other 1□ 2□ 3□ 4□ 5□ 6□ 7□ concerned about each other

143

Appendix I

Protocols for Group Interviews

(translated from an original Mandarin version)

Questions on computer-supported CW (CSCW)

1. Did mediation tools help your CW? 2. Did CW help your individual writing? And how? 3. How did you collaborate with each other? 4. What did you think of your partnership in terms of learning how to write? 5. What advantages of CSCW were there in the intensive program? 6. What limitations of CSCW were there in the intensive program? 7. What improvements to CSCW were there in the intensive program? 8. What were other things related to the class or the program?

144

Appendix J

Questionnaire on Self-assessed Beliefs about Writing Abilities

(translated to a Mandarin version for the present research)

Name : ______ID Number______Class: ______Date: ______Directions: A number of statements which students have used to describe themselves are given below. Read each statement and indicate how you generally think or feel about writing tasks by marking your answer sheet. There is no right or wrong answers. Do not spend too much time on any one statement. Remember, give the answer that seems to describe how you generally think or feel about a writing task. Almost Never =1, Sometimes =2, Often =3, Almost Always =4 1. One of my primary goals in writing is to understand the concepts. 2. One of my primary goals in writing is to do better than others. 3. I determine how to solve a writing task before I begin. 4. I check how well I am doing when I solve a writing task. 5. I work hard to do well even if I don’t like a writing task. 6. I believe I will receive an excellent grade in a writing assignment. 7. I carefully plan my course of action. 8. One of my primary goals in writing is to not look foolish/stupid. 9. One of my primary goals in writing is to acquire new skills. 10. I put forth my best effort on tasks. 11. One of my primary goals in writing is to improve my knowledge. 12. One of my primary goals in writing is to get a high grade. 13. I check my work while I am doing it. 14. I work as hard as possible on writing tasks. 15. I'm confident that I can understand the basic concepts taught in writing courses. 16. I almost always know how much of a writing task I have to complete. 17. One of my primary goals in writing is to learn the material. 18. One of my primary goals in writing is to look capable to my peers and friends. 19. I concentrate as hard as I can when doing a writing task. 20. I'm confident I can do an excellent job on writing assignments.

145

21. Even if I did poorly on the writing, I know I learned something from the experience. 22. I expect to do well in writing assignments. 23. One of my primary goals in writing is to get a higher score than other students. 24. I make sure I understand what has to be done and how to do it. 25. A writing task is useful to check my knowledge. 26. I think I learned something from the writing that will be useful to me. 27. I'm certain I can master the skills being taught in writing courses. 28. Considering the difficulty of writing, the teacher, and my skills, I think I will do well in writing assignments.

146

Appendix K

Complete Categorized Answers to Open-ended Questions

in the After-CW Questionnaire

(translated from an original Mandarin version)

Table K1 Complete Categorized Answers to Question One in the After-CW Questionnaire: What Worked Really Well in This Task? Week 2 times week 3 times week 4 times discussion/negoti 17 17 13 organization content discussion ation (25%) (29%) (19%)

11 8 verbalizing/word 12 organization content discussion (16%) (14%) choice (17%) perspectives/cons 11 perspectives/consen 6 11 integration of ideas ensus (16%) sus (10%) (16%) verbalizing 5 verbalizing 4 Tasking 1 tasking 4 tasking 5 Completeness 3 integration of 4 integration of ideas 4 Resources 3 ideas completeness 4 completeness 2 writing time 4

submitting resources 3 resources 3 writing/computer 1 system writing time 3 fluency 1 Organization 9

perspective/consens effectiveness 2 3 perspective/consensus 9 us

writing process fluency 1 writing process 1 3 (listing) submitting 1 clear topics 1 Atmosphere 1 writing/computer recording 1 editing 1

147

computer systems 2

Table K2 Complete Categorized Answers to Question Two: What Did Not Work Well in This Task? week 2 times week 3 times week 4 times grammar 17 (24%) grammar 17 (22%) Grammar 11 (17%) vocabulary 17 (24%) vocabulary 15 (20%) Vocabulary 9 (14%) verbalization 8 (11%) timing 11 (15%) Timing 8 (12%)

content content timing 4 richness/exam 8 7 richness/examples ples content sentence richness/exampl 7 varieties/patter 3 Spelling 3 es ns sentence 7 collocation 1 Perspectives 6 varieties/patterns collocation 2 spelling 1 Verbalization 5 recording 1 perspectives 2 idea integration 2 computer 2 verbalization 5 Flow 1

idea carefulness 1 3 Coherence 2 integration creativity 1 flow 3 Timing 1

online resources(including brain 1 coherence 2 4 with less consulting them)

fixed writing spelling 1 direction 1 1 process perspectives 2 organization 1 Proficiency 1

148

discussion(engagem tasking 1 2 ent) topic 1 Length 1 understanding

Rhetoric 1

149

Appendix L

Group Interviews

(translated from an original Mandarin version)

Table L 1 Writing Group Interview in Chinese (English translation): Group a and b Groups Group A: Emily and Rachel Group B: Jon, Kelly, Daniel Object: After 3 times of CW, writing process: thinking was enlarged by the other side of what do you learn from it? ideas Writing process: assimilation of ideas; after discussion, the Jon: He relied on his own ideas rather than ideas discussed Object to outcomes: Does CW second writing was better than the first time of writing. in CW when he wrote his post-CW individual writing. help with your individual Their discussion facilitated the second writing by the Daniel: he used CW ideas to apply to his individual writing. writing? alternative ideas which made verbalizing in English (or Kelly: he integrated his ideas with others in CW to write his argumentation) easy. Idea expansion! individual writing. No fixed students and they reached agreement and then They compared their first artifacts and then decided one Division of labour: How do you writing. They concentrated on contents of a writing task writing sample. They read each other's writing to see if they collaborate with each other? more than words, sentences and grammar. can enrich the CW with more examples. Community and Division of labour: Roles of students and takes turns Kelly consulted synonyms. readers? Daniel: I consulted various writing knowledge: google translation, sentence patterns and conjunctions; By the third time, he can remember the sentences and put the pattern to his writing. He reduced the use of sentence patterns, google translation and conjunctions with times of writing. For dictionary, he cannot reduce his times of using it because Mediation and outcomes: do mediation tool use: translation and dictionary. Chuen: too different writing tasks required different fields of search tools help you with much reliance on the tools and was afraid of writing vocabulary. Tips for Google translation: his ideas were from writing without tools non-authentic translated Chinese from translation literature. He keyed in translated (literature) Chinese to get more authentic English. Jon: He didn’t like tools like common patterns, which yielded fixed answers. He had a free mind to develop his own expressions so he used google translation more. Rules: advantages Discussion and checking online info made people feel Jon and Daniel thought they can't write without computers 149

150

Groups Group A: Emily and Rachel Group B: Jon, Kelly, Daniel secured. anymore. Jon specifically pointed out that paper-and -pencil writing was good too but it can't beat computers in offering varieties of resources. Kelly: Using computers to check spelling and English usages were beneficial to student writers in that it can easily tell you where the error was and the writer can memorize the correct by correcting it. This was what paper-and pencil writing cannot do. Jon: For TOEFL exam takers, it was good. But not everyone needed to take it. Rules: disadvantages Too rely on material mediation Daniel: too much reliance on mediation Rules: limitation Nope "on topic" lectures and then writing to make lecturing and Making a project was better than examinations (e.g., Rules: improvements writing correspondent. English-English vocabulary). others Time was relatively okay

151

Table L 2 Writing Group Interview: Group c and d Groups Group C: Hui and June Group D: Stephanie and Sarah Object: After 3 times of CW, what do you learn from it? Improvements: Stephanie: structure, grammar (Sarah was Object to outcomes: Does CW good at that) and vocabulary (Stephanie checked up the help with your individual Internet resources) Sarah: contents of the writing were writing? enriched. Sarah typed the first and the second writing. Stephanie They tried hard negotiating ideas and finally reached typed the third writing. Sarah sometimes corrected her Division of labour: How do you agreement but had difficulties in writing up sentences that grammar and off-topic sentences. Stephanie did think-aloud collaborate with each other? reflected their ideas. but Sarah thought her English was actually with Chinese grammar. Community and Division of labour: Roles of writers and They took turns readers? Mediation and outcomes: do search tools help you with yes. yes especially dictionary writing Google translation, yahoo dictionary (lower chances) and Rules: advantages use google translation writing templates Rules: disadvantages time pressure Rules: limitation Rules: improvements The difficulty task was the "teacher topic" the easy topic others was "living at home or outside"; they had questions in translating idioms.

152

Table L 3 Writing Group Interview: Group e and f Groups Group E: Edward, Charlotte and Kelly Group F: Lillian and Rita Edward: I turned better in writing supporting examples. It Object: After 3 times of CW, also helped me develop different perspectives towards a what do you learn from it? topic. Charlotte: the line of thinking got clearer and clearer after Object to outcomes: Does CW CW. Edward: I didn't switch his point, either the positive help with your individual or the negative point of view from pre-CW writing to writing? post-CW writing. I was glad to have CW to enrich his examples. Kelly: We compared our writing first to see if we shared the Division of labour: How do you same points of view and if yes, then to integrate every one's collaborate with each other? ideas into one writing sample that we agreed with. Community and Division of Kelly was the writer; Edward and Charlotte were labour: Roles of writers and discussants. They were also the decision makers when a sometimes taking turn readers? group of three persons were discussing ideas. Mediation and outcomes: do Google translation, they would recheck the same word search tools help you with several times. Edward: They opened windows for sentence writing patterns, conjunctions, and writing templates. Edward: sharing ideas and sometimes he would prepare complementary: one was weak in listening so the other one Rules: advantages more examples after school for the next day's individual was responsible for typing or checking meanings from writing. Kelly: didn't do so. Google translation Kelly, at the very beginning, she thought this method was fun but later on thought she didn’t benefit much from this CW because she didn’t join the discussion and she was responsible for typing what they told her to write. Kelly admitted she benefited from sharing different points of No real expert in the pair: unknown things still remained views. Kelly: Three persons as a group were no good. Two Rules: disadvantages unknown, especially grammar usage they would debate the persons were better. usage till they saw something they all both knew Edward: Computers/material mediation was not always good. You can easily remember the word you just wrote down but with computers, you can't do that and you always wanted to check up info. Therefore, internalization was hard to happen. writing task design (repetitive writing on the same task) Rules: limitation Time was okay. tiring tasks; pressure and time limitation

153

Groups Group E: Edward, Charlotte and Kelly Group F: Lillian and Rita Rules: improvements time can be longer Was mentioning both points of views okay in real writing others high intermediate: a little bit difficult test??

154

Table L 4 Writing Group Interview: Group g and h Groups Group G: Yi and Iris Group H: Belle, Frank, Ava Iris: she could not write anything in the first time but after CW, she felt so good to be able to write an entire paragraph. Object: After 3 times of CW, CW relieved her nervousness. For the second time and the what do you learn from it? third time, she could not know if her writing is correct. She doesn't know if what she adopted from online writing resources was correct or not. Object to outcomes: Does CW Yi: she was not good at grammar. help with your individual Both of them wrote their individual writing on the basis of easy to know writing skills with several times of writing writing? CW. Belle: searching google for writing information Frank: Division of labour: How do you They would discuss what they wrote in the first time of correcting grammar; Ava: typing and integrating ideas and collaborate with each other? writing. paragraphs; sometimes changing the role with Belle Community and Division of Writers (Yi) and readers (Iris): incorporation type; they labour: Roles of writers and took turns as well were old friends. readers? Google translation, sentence patterns, conjunctions, and writing templates. They made notes on sentence patters Mediation and outcomes: do from those mediation tools to facilitate their writing next search tools help you with time. They tried to use different patterns in writing tasks. writing They can remember something from the use of mediation tools. complementary: new ideas from others for independent Rules: advantages They liked mediation tools. writing Rules: disadvantages Time was a little bit short. time limitation Rules: limitation Rules: improvements individual mini meeting others Iris felt the textbook was too difficult.

155

Table L 5 Writing Group Interview: Group i and j Groups Group I: Mary and Anita Group J: Hilda and Jan Object: After 3 times of CW, Writing process: starting from different points of views but writing process: thinking was enlarged by the other side of what do you learn from it? ending up with similar statement, different ideas Object to outcomes: Does CW Yes (didn't know how to type in English at the very help with your individual beginning). Some basic writing skills like typing in English writing process: assimilation of ideas writing? were also trained through increasing times of writing. complementary: They stayed the same kind of cooperation, talked more in the first time, always different opinions in Division of labour: How do you independent writing, "especially in staying at home or collaborate with each other? leaving home" didn’t know how to integrate each other's ideas to one paragraph Community and Division of labour: Roles of writers and took turns to write readers? Mediation and outcomes: do material mediation (Writing resources): they visited quite search tools help you with often at very beginning: template; yahoo dictionary: to get writing collocation "prepositions"; google translation=dictionary Rules: advantages Idea completion (to learn from each other) material mediation (use of collocation): need to get use of it Rules: disadvantages talk distraction and time Rules: limitation time limitation computer system: stable system, Time pressure: more time writing partnership: change partners once in a while: one Rules: improvements to write can be an expert and the other can be a novice writing others

156

Table L 6 Writing Group Interview: Group k and l Groups Group K: Hazel and Karen Group L: Maria and Doris Object: After 3 times of CW, Karen: reading turned faster Discussion made thoughts clear what do you learn from it? Object to outcomes: Does CW They made reference to CW and made ideas more help with your individual Hazel: Sure! He can understand more about the reading. completed. writing? When they had difficulty in word selection, they discussed it according to their understanding of the words and Division of labour: How do you knowledge. If it was grammar, they checked up online Hazel was a grammar corrector collaborate with each other? resources or with others. They had agreement of what to write from the same perspective under the same topic. They thought in Chinese. Community and Division of Writers and readers: Hazel was the grammar corrector. Expert (Doris: active in CW) and novice (Maria)/Reader labour: Roles of writers and Karen listened to Hazel 's suggestions of grammar and and writer: They took turns typing. readers? vocabulary. Mediation and outcomes: do search tools help you with Yes, dictionaries. Google translation and other writing resources writing Computer writing was better because writers can check up Rules: advantages correct info. CW was good because others can correct you if you misunderstood the prompt. Rules: disadvantages System can be stable. Two persons was a better group. Rules: limitation Time Rules: improvements longer time others

157

Table L 7 Writing Group Interview: Group m and n Groups Group M: Nancy, Jenny and Angel Group N: Jenny and Joe Object: After 3 times of CW, Jenny: Thinking/logics/structure/organization got more and what do you learn from it? more clear. Grammatical accuracy and thinking completeness: It was Object to outcomes: Does CW good to have CW to complete ideas; otherwise there was no help with your individual difference between the first writing and the third writing. To writing? have complete sentences and let the wording matched with ideas. Taking turns: Nancy typed the first and the third time. The rest of the two writers did the second time. They would prefer to the camp of thinking that had the most main ideas and supporting examples, providing that there were two schools of thoughts from the three writers. They didn’t say They used one template from beginning to the end. no to others' suggestions. Nancy was the one who always However, they appreciated each other's perspectives and typed and she sometimes wanted her other writers to check Division of labour: How do you wording so that variations were added on to the writing if her wording was correct. But the other two colleagues collaborate with each other? when revisions were made. Different opinions: Joe remained "no comments" most of the time. Nancy and compromised. Jenny was responsible for typing and she had Jenny would go back to her original ideas when she wrote more different perspectives towards to an opinion. the same writing task again as post-CW writing because their main ideas of the writing came from the majority of thoughts in the group and Nancy held the opposite perspective to the same writing task normally. She admitted that she did CW for the sake of CW. Writers and readers: Nancy as the expert to grammar and vocabulary in this group. Nancy typed the article most of the time; the other two were more like readers to Community and Division of supplement/enrich contents and helpers to check up online labour: Roles of writers and expert and novice group info. Readers discussed but the writer seemed to have readers? different ideas about the same topic but she was the writer so she can only wrote their ideas and latter on when she was writing the same task, she put on her own ideas. Checking up online dictionary reminded Nancy of her Mediation and outcomes: do acquired vocabulary but she can't get acquire new search tools help you with vocabulary from checking up online info. Sometimes writing Nancy figured that they checked up a word four times and still needed to check up something to make sure its usage.

158

Groups Group M: Nancy, Jenny and Angel Group N: Jenny and Joe She still used the word in the paragraph but she wasn’t able to use it later on in her writing because that particular word was her passive word rather than active word. Jenny checked parts of the words and vocabulary. Collocation dictionary was helpful. Angel didn't know if there were no computer and mediation tools how she can write. Nancy: It was helpful to memorize the spelling of a word or its pronunciation if I checked it up once again. They also They rarely used online resources. They built up their Rules: advantages used the online collocation corpus and found it useful. They writing knowledge and skills from verbal language. knew how to use it. Sometimes writers typed their own ideas without discussing ideas with others. Overly reliance on the Internet; Nancy was confused with Rules: disadvantages grammar so she put on one usage first; without Internet, Nancy won’t be able to spell words. Rules: limitation Time time limitation Rules: improvements English writing experiences: Nancy had no English classes others after she was a high school student. Her English writing Jenny was good at vocabulary experiences in high school were taking writing exams.

159

Table L 8 Writing Group Interview: Group o and p Groups Group O: Liz and Rose Group P: Emy and Helen Object: After 3 times of CW, Liz: more writing ideas what do you learn from it? Object to outcomes: Does CW help with your individual more ideas from CW; writing? Complementary: the first and second time Helen was responsible for typing after their discussion. The last time Division of labour: How do you They discussed their points in the first writing and created Emy typed the article. Emy was able to think of details. collaborate with each other? other points to put them into CW. Helen's ideas were the main sources of CW. Helen felt that their CW was getting smooth-sailing especially in the third time. She felt that Emy improved a lot. Community and Division of Reader and writer (Rose): Rose had more ideas when she labour: Roles of writers and took turns was in front of computers so she was always the writer. readers? (Liz had better senses in conjunction use and she was the Mediation and outcomes: do one checked it up) Liz: We checking one word up in one search tools help you with writing task again and again. However, after several times writing of writing, we didn’t need to check it again. Easy to complete ideas (to learn from each other) Emma sometimes applied the words that were obtained from material mediation in CW to her individual writing; I can Rules: advantages remember some words and phrases in templates (e.g., first, second, and finally) and I used them again in my own writing. Rules: disadvantages Rules: limitation Time Liz: To change writing partners one expert and one novice. Teachers' explanations can be broken down into sectional: Rules: improvements Rose: She didn’t like to change partners and thought main idea development followed by examples collaboration was important. Simple-sentence writing was ok. Helen didn't have much trouble in delivering her ideas with correct sentences. Helen: she thought Emy improved from only knowing a others word to writing all words to sentences to paragraphs. She barely knew a word and sometimes she asked Helen for help.

160

Appendix M

Collaborative Writing Scores from Week 1 to Week 4

Table M 1 Collaborative Writing Scores in Week 1 and Week 2 CW CW week 1 week 2 CQ cw1 Org cw1 Arg cw1 Lacc cw1 Lapp cw1 HS cw1 CQ cw2 Org cw2 Arg cw2 Lacc cw2 Lapp cw2 HS cw2 cwa 5.5 6 5.75 5 5.5 5 5.75 5 5 5 5 5.25 cwb 5.5 5.5 6.5 5.5 8 5.5 6 7 7 6 6 6 cwc 2.5 2.5 2 2 2 2.5 6 7 6 6 5.5 6 cwd 4 4 4 5 6 4 6 5.25 6 5 6 6 cwe 7.5 6 7 6.75 7.5 6.75 5.75 4.5 5 4.5 6 5 cwf 5 4.5 4 5.5 5 4.5 6 5 6.5 6.5 5.5 5.75 cwg 4.25 4 4.25 4.5 4.75 4 5.5 5.25 5.75 5 5.25 5.5 cwh 4.5 4.5 4.75 5.5 5.5 4.75 5.5 5 5 6.5 6.5 5.75 cwi 5.5 5.25 5.25 5.75 5.5 5.25 6.25 6 6.25 7 6 6.25 cwj 5.75 5.5 6 6 5.5 5.75 5.5 6 6 5.5 5.25 6 cwk 4.5 4.75 5 4.5 5 5 5 5 5.5 5.5 4.5 5 cwl 5 4.75 5 5.5 6 5 6 6.5 6.5 5 5 6 cwm 5.5 5 5 6.5 5 5 3.5 2.5 3 4.5 6 4 cwn 7 7 6.5 5.5 5.25 6.75 6.5 6 6 6.5 6.5 6 cwo 5.25 5.25 4.75 4 5 4.75 5.25 5 6 5.5 5.5 5.25 cwp 4.25 4.25 4.75 4.75 4.75 4.5 6 5.5 5 5.5 6 5.5 Scores averaged 5.09 4.92 5.03 5.14 5.39 4.94 5.66 5.41 5.66 5.59 5.66 5.58

161

Table M 2 Collaborative Writing Scores in Week 3 and Week 4 CW CW week 3 week 4 CQ cw3 Org cw3 Arg cw3 Lacc cw3 Lapp cw3 HS cw3 CQ cw4 Org cw4 Arg cw4 Lacc cw4 Lapp cw4 HS cw4 cwa 5.25 5 5 5 5 5 5 5 5 3.5 4.75 4.75 cwb 8 7 8 6 8 8 7 7 8 6.25 7 7 cwc 5.5 6 5.75 5.25 5.5 5.5 4.5 6 4.5 4.5 6.25 4.75 cwd 5.5 5 5 5.75 5.75 5.5 5 4 5 5 7 5 cwe 5.5 5.5 6.5 4.5 6.5 5.5 6.25 6.25 5.25 6.5 6 6 cwf 5 4.75 5 5.75 5.75 5 7 7.5 7 6.75 6.75 7.25 cwg 6 5.25 5.5 5.75 6.25 5.75 6 5.5 5 5 5.5 5.5 cwh 6 6.5 5.25 6 7 5.75 6.25 5.75 6.5 6 5.75 6.25 cwi 5.75 6 7 4.75 4.75 6 6 5.75 5.75 6 6 5.75 cwj 4.5 4 4.5 4.5 4.5 4.5 6 6.5 6.5 5.5 6.5 6.5 cwk 5.5 5 4.75 5.25 5.5 5.5 4.75 4.75 4.25 5 5.5 4.75 cwl 6 6 6.5 6 6 6 7 6 5.5 6.5 7 6.75 cwm 6 5 5.5 6 5.5 5 6.5 6.5 6 7.75 7.25 6.5 cwn 6 7 6 5.75 6 6.25 7 6.5 7 6.5 6.5 7 cwo 6 6.25 6 5.75 6.5 6 5 4.5 5.5 5 5.5 5 cwp 5.75 5 4.75 6 7 5.25 6.75 7 7.25 6.75 6.25 7 Scores averaged 5.77 5.58 5.69 5.50 5.97 5.66 6.00 5.91 5.88 5.78 6.22 5.98