You are on page 1of 188

E. M.

MURILLO-OTHON

SEARCHING FOR “VIRTUAL”

COMMUNITIES OF PRACTICE IN

THE USENET DISCUSSION NETWORK

Volume I of II
PhD Volume I of II

E. M. MURILLO-OTHON

PhD

2006
2006
SEARCHING FOR “VIRTUAL” COMMUNITIES OF PRACTICE

IN THE USENET DISCUSSION NETWORK

Combining quantitative and qualitative methods

to identify the constructs of Wenger's theory

Volume I of II

Enrique M. MURILLO-OTHON

submitted for the degree

of Doctor of Philosophy

School of Management

University of Bradford

2006
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

SEARCHING FOR “VIRTUAL” COMMUNITIES OF PRACTICE

IN THE USENET DISCUSSION NETWORK

Combining quantitative and qualitative methods

to identify the constructs of Wenger's theory

Abstract: Communities of Practice (CoPs) are informal organisational groups


that spontaneously form around shared problems or disciplines. As natural
vehicles of knowledge-sharing and innovation, they are increasingly contemplated
in Knowledge Management projects. Yet CoPs have been mostly studied in stable
co-located settings, a rare circumstance in today’s turbulent workplace. This
research aims to extend CoP theory to the Internet by undertaking a systematic
search and a rigorous assessment of wholly Internet-based CoPs.

The thesis adopts Wenger’s (1998) CoP theory, specifically his constructs of
Mutual Engagement, Shared Repertoire, Joint Enterprise, Community, and
Learning/Identity-acquisition, explicitly included in the proposed model of
Internet-based CoP. The search concentrated on the Usenet discussion network;
2,842 newsgroups were examined through Netscan, an online-interaction analysis
tool, to locate eleven virtual communities exhibiting empirical traits associated
with focused energetic CoPs: professional topic, high interaction-volume, non-
conflictual focused discussions, and core-periphery structure.

A survey of community-members and a content analysis of discussions were


used to seek triangulating evidence of Wenger’s constructs, deemed present
when both methods concurred. Four communities, evenly split between
computer and non-computer topics, were assessed as Usenet-based CoPs
because they exhibited all Wenger constructs. Substantial construct evidence in
remaining communities suggests potential avenues for further research.

Establishing that extra-organisational CoPs spontaneously emerge in social


areas of the Internet provides new support to Wenger’s position that they are
naturally-occurring social structures. For individuals, they constitute a
constantly-available resource for enhancing professional practice, whereas
organisations can encourage employee engagement with extra-organisational
Internet-based CoPs to keep abreast of fast changing fields.

Keywords: communities of practice, virtual communities of practice, virtual


communities, Usenet, newsgroups, computer-mediated communication

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
I
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

BRIEF CONTENTS

PART ONE – THEORETICAL BACKGROUND 1


V
O 1. Introduction 2

L 2. Previous Research on Communities of Practice 19

U 3. Virtual Communities of Practice 54


M
E PART TWO – RESEARCH STRATEGY: DESIGN 95

4. Research Design 96
I
5. Methods 132

PART THREE – RESEARCH STRATEGY: EXECUTION 164

6. Usenet Search and Community Selection 165

7. Survey of Community Members 191


V
O 8. Content Analysis of Community Discussions 250

L
PART FOUR – STUDY FINDINGS 289
U
M 9. Overall Assessment of Selected Communities 290
E 10. Examining Normal Activity in Detected VCoPs 330

11. Conclusions 388


II
Afterword 423

References 424

Glossary of Internet Terms 442

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
II
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CONTENTS

VOLUME I

PART ONE – THEORETICAL BACKGROUND 1

1 - Introduction 2

1.1 - Communities of practice 5

1.2 - Research aim, objectives and strategy 9

1.3 - Outline of Thesis 14

1.4 - Summary 18

2 - Previous Research on Communities of Practice 19

2.1 - Initial formulation of the Community of Practice concept 20


and early studies

2.2 - Wenger’s social theory of learning 24

2.3 - Wenger’s theory of Communities of Practice 28

2.3.1 - MUTUAL ENGAGEMENT 31

2.3.2 - JOINT ENTERPRISE 34

2.3.3 - SHARED REPERTOIRE 35

2.3.4 - COMMUNITY 36

2.3.5 - LEARNING 37

2.3.6 - IDENTITY ACQUISITION 37

2.4 - Challenges to Wenger’s theoretical framework 38

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
III
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.4.1 - Alternative terminology for alternative but related phenomena 40

2.4.2 - Alternative terminology for the same phenomenon 42

2.4.3 - Critiques of Wenger’s (1998) theory 47

2.5 - Summary 51

3 - Virtual Communities of Practice 54

3.1 - Reviewing the varieties of “virtual” in the literature 55

3.1.1 - Network access 57

3.1.2 - Media richness 58

3.1.3 - Participatory character 60

3.2 - Traces of Wenger’s constructs in previous studies of “virtual” CoPs 61

3.3 - Challenges to the concept of “virtual” Communities of Practice 83

3.4 – Summary 93

PART TWO – RESEARCH STRATEGY: DESIGN 95

4 - Research Design 96

4.1 - The model of “virtual” Community of Practice used in this thesis 97

4.1.1 - Model specification 97

4.1.2 - Exemplary Traits 99

4.1.2.1 - Mid-sized stable group 99

4.1.2.2 - High-volume of participant interaction 99

4.1.2.3 - Core-periphery structure 99

4.1.2.4 - Highly-focused discussions 100

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
IV
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.1.2.5 - Topic is an identifiable profession 100

4.1.2.6 - High-Quality institutional documents 101

4.1.2.7 - Non-conflictive 101

4.1.3 - Essential Traits 102

4.1.3.1 - MUTUAL ENGAGEMENT 102

4.1.3.2 - SHARED REPERTOIRE 102

4.1.3.3 - JOINT ENTERPRISE 103

4.1.3.4 - COMMUNITY 103

4.1.3.5 - LEARNING/IDENTITY ACQUISITION 103

4.1.4 - Model discussion 104

4.2 - Research aim and objectives 105

4.3 - Research Questions 109

4.3.1 - Principal Research Question 109

4.3.2 - Preliminary Research Questions 110

4.3.3 - Exemplary Trait Research Questions 112

4.3.4 - Essential Trait Research Questions 113

4.4 - Focusing the study on the Usenet discussion network 114

4.5 - Research philosophy 121

4.6 - Research strategy 126

4.6.1 - Stage I: Model Specification 126

4.6.2 - Stage II: Strategy Formulation 126

4.6.3 - Stage III: Usenet Search 127

4.6.4 - Stage IV: Participant Survey 128

4.6.5 - Stage V: Content Analysis of Discussions 128

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
V
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.6.6 - Stage VI: CoP Assessment 129

4.6.7- The Funnel Research Strategy 129

4.7 - Summary 133

5 - Methods 134

5.1 - The Netscan analyser 134

5.2 - Social Network Analysis and the core-periphery model 139

5.3 - The Survey questionnaire: objectives and design 145

5.3.1 - What the Survey will measure 145

5.3.2 - Generation of the item pool 145

5.3.3 - Design of the web-based instrument 150

5.3.4 - Qualified review of the item pool 152

5.3.5 - Possible inclusion of validation items 152

5.3.6 - Pilot runs of the instrument 152

5.3.7 - Item evaluation 153

5.3.8 - Possible optimisation of scale length 154

5.4 - The Content Analysis: objectives and design 155

5.4.1 - Rationale and criteria for purposive sample selection 155

5.4.2 - The coding scheme 158

5.5 - Summary 162

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
VI
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

VOLUME II

PART THREE – RESEARCH STRATEGY: EXECUTION 164

6 - Usenet Search and Community Selection 165

6.1 - First Sub-stage: Netscan Search 167

6.2 - Second Sub-stage: Institutional document search and assessment 175

6.3 - Third Sub-stage: Core-Periphery Analysis 180

6.4 - Fourth Sub-stage: Final Selection of Survey Sample 187

6.5 - Summary 189

7 - Survey of Community Members 191

7.1 - Survey deployment 193

7.2 - Analysis of hypothesised structure questionnaire 200

7.2.1 - Item results 200

7.2.2 - Item reliability analysis 207

7.3 - Using Exploratory Factor Analysis to evaluate and improve scales 209

7.3.1 - Addressing critical EFA issues 210

7.3.1.1 - Factor extraction model 211

7.3.1.2 - Number of factors to retain 211

7.3.1.3 - Sample size 212

7.3.1.4 - Rotation technique 212

7.3.1.5 - Criterion of salient loading 213

7.3.2 - Applying EFA to Item-blocks 213

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
VII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

7.3.2.1 - LEARNING block 214

7.3.2.2 - MUTUAL ENGAGEMENT block 217

7.3.2.3 - SHARED REPERTOIRE block 220

7.3.2.4 - COMMUNITY block 221

7.4 - Questionnaire validation and refined scale results 224

7.5 - Content analysis of the community question 238

7.6 - Discussion of results 242

7.7 - Survey problems and limitations 246

7.8 - Summary 248

8 - Content Analysis of Community Discussions 250

8.1 - Theoretical selection of thread sample 251

8.2 - Nudist data import and pilot coding exercises 253

8.3 - The two-stage coding procedure: message-wise and code-wise 258

8.4 - General code counts and illustrative coded threads 264

8.5 - Discussion of results 275

8.5.1 - Detected manifestations of MUTUAL ENGAGEMENT 275

8.5.2 - Detected manifestations of SHARED REPERTOIRE 279

8.5.3 - Detected manifestations of COMMUNITY 281

8.5.4 - Detected manifestations of LEARNING 281

8.5.5 - Detected manifestations of JOINT ENTERPRISE 282

8.5.6 - Additional comments 282

8.6 - Limitations of the Content Analysis 285

8.7 – Summary 287

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
VIII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

PART FOUR – STUDY FINDINGS 289

9 - Overall Assessment of Selected Communities 290

9.1 - Summary of detected Exemplary and Essential Traits 291

9.2 - The CPLUS virtual community 295

9.3 - The TAXES virtual community 298

9.4 - The PHYSRES virtual community 302

9.5 - The XTRPRG virtual community 305

9.6 - The MEDTRAN virtual community 307

9.7 - The COBOL virtual community 310

9.8 - The VISOBJ virtual community 313

9.9 - The UKAGRI virtual community 316

9.10 - The CRYPT virtual community 319

9.11 - The CIVWAR virtual community 320

9.12 - The FINPLAN virtual community 323

9.13 - Commonalities between successful VCoPs 325

9.14 - Summary 329

10 - Examining Normal Activity in Detected VCoPs 330

10.1 - Rationale and method used in this chapter 331

10.2 - The TAXES VCoP 333

10.3 - The CPLUS VCoP 345

10.4 - The XTRPRG VCoP 361

10.5 - The PHYSRES VCoP 373

10.6 - Summary 386

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
IX
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

11 - Conclusions 388

11.1 - Assessing achievement of research aims and objectives 388

11.1.1 - Research Objective One 389

11.1.2 - Research Objective Two 390

11.1.3 - Research Objective Three 390

11.1.4 - Research Objective Four 391

11.1.5 - Research Objective Five 391

11.1.6 - Research Objective Six 392

11.1.7 - General Aim of the Thesis 393

11.2 - The Research Questions addressed 394

11.2.1 - Preliminary Research Questions 394

11.2.1.1 - Preliminary1 394

11.2.1.2 - Preliminary2 395

11.2.1.3 - Preliminary3 395

11.2.1.4 - Preliminary4 396

11.2.2 - Essential Trait Research Questions 396

11.2.2.1 - Essential1 397

11.2.2.2 - Essential2 398

11.2.2.3 - Essential3 399

11.2.2.4 - Essential4 399

11.2.2.5 - Essential5 400

11.2.2.6 - Essential6 400

11.2.3 - Principal Research Question 401

11.2.4 - Exemplary Trait Research Questions 401

11.2.4.1 - Exemplary1 402

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
X
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

11.2.4.2 - Exemplary2 402

11.2.4.3 - Exemplary3 402

11.2.4.4 - Exemplary4 403

11.2.4.5 - Exemplary5 403

11.2.4.6 - Exemplary6 403

11.2.4.7 - Exemplary7 403

11.3 - General conclusion and thesis contributions 406

11.4 - Study implications 409

11.4.1 - Implications for theory 409

11.4.2 - Implications for individuals 412

11.4.3 - Implications for organisations 414

11.4.4 - Implications for Management practice 415

11.5 - Limitations of this research 417

11.5.1 - Limitations with the sample selection procedure 417

11.5.2 - Limitations of the Survey 417

11.5.3 - Limitations of the Content Analysis 418

11.5.4 - Overall methodological evaluation 419

11.6 - Directions for further research 420

Afterword 423

References 424

Glossary of Internet Terms 442

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XI
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

LIST OF FIGURES

1.1 - Research Objectives and Strategy Stages 13

1.2 - Structure of Thesis 15

2.1 - Components of Wenger’s social theory of learning 27

2.2 - Observable Traits of a CoP 30

2.3 - Timeline of significant references in CoP theory development 39

2.4 - Contrasting “Organisational” and “KM” CoP references 53

3.1 - Articles about “virtual” CoPs published in Proquest + IEEE databases 55

4.1 - The VCoP Model 98

4.2 - Triangulation of Survey and Content Analysis results 124

4.3 - The Funnel Research Strategy 130

4.4 - Research Objectives and Strategy Stages 132

5.1 - Definitions of Netscan quantitative results 137

5.2 - Posting activity of top 60 posters in soc.history.medieval 143

5.3 - Graph of ideal-pattern for core-periphery model 144

6.1 - Overview of Stage III of the Research Strategy 166

6.2A - Graph of nDiff values for MEDTRAN 185

6.2B - Posting activity at the core of MEDTRAN for 52-week sample 185

6.3A - Graph of nDiff values for GENMETH 186

6.3B - Posting activity at the core of GENMETH for 52-week sample 186

7.1 - Stages IV and V of the Funnel Research Strategy 192

7.2 - A thank-you note from a Survey respondent 198

7.3 - Parallel Analysis for the LEARNING block 215

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

7.4 - Parallel Analysis for the MUTUAL ENGAGEMENT block 217

7.5 - Parallel Analysis for the SHARED REPERTOIRE block 220

7.6 - Parallel Analysis for the COMMUNITY block 222

7.7 - Link of validated scales to Wenger’s constructs 226

7.8 - Effect of Tenure on Shared sense of professional community 232

7.9 - Effect of Tenure on Members’ knowledge of each other 232

7.10 - Effect of Gender on Shared sense of professional community 233

7.11 - Effect of Gender on Members’ knowledge of each other 233

7.12 - Effect of Coreness on Members’ knowledge of each other 234

7.13 - Effect of Expertise on Acquiring new knowledge 235

7.14 - Effect of Expertise-by-Tenure on Collective problem-solving 236

7.15 - Effect of Coreness-by-Age on Shared criteria 236

7.16 - Effect of Coreness-by-Age on Sense of professional community 237

8.1 - Coded fragment from thread TAXES1 268

8.2 - Coded fragment from thread XTRPRG4 272

9.1 - The VCoP Model 291

9.2 - Excerpt from the Home Page of comp.lang.c++.moderated 295

9.3 - A response to the community question from the CPLUS newsgroup 296

9.4 - Charter of misc.taxes.moderated 298

9.5 - A thank-you note from a member of the TAXES VCoP 299

9.6 - Excerpt from the Home Page of sci.physics.research 302

9.7 - A response to the community question from the PHYSRES newsgroup 303

9.8 - Extract from the FAQ of the XTRPRG VCoP 305

9.9 - A response to the community question from the XTRPRG newsgroup 306

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XIII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

9.10 - Charter of sci.med.transcription 308

9.11 - A response to the community question from the MEDTRAN VCoP 308

9.12 - A message from a core member of comp.lang.cobol 312

9.13 - A message from comp.lang.clipper.visual-objects 314

9.14 - An excerpt from the FAQ of uk.business.agriculture 317

9.15 - A response to the community question from the UKAGRI newsgroup 318

9.16 - A response to the community question from the CRYPT newsgroup 319

9.17 - Two e-mail comments to the Survey from the CIVWAR newsgroup 321

9.18 - A warning from the moderators of soc.history.war.us-civil-war 322

9.19 - A message from the Moderators of misc.invest.financial-planning 324

9.20 - A moderator warning in misc.taxes.moderated 328

10.1A - Extract from thread TAXES1, Message 1 335

10.1B - Extract from thread TAXES1, Message 2 338

10.1C - Extract from thread TAXES1, Message 3 340

10.1D - Extract from thread TAXES1, Message 4 341

10.1E - Extract from thread TAXES1, Message 5 342

10.1F - Extract from thread TAXES1, Message 6 343

10.2A - Extract from thread CPLUS2, Message 1 347

10.2B - Extract from thread CPLUS2, Message 2 350

10.2C - Extract from thread CPLUS2, Message 3 352

10.2D - Extract from thread CPLUS2, Message 4 354

10.2E - Extract from thread CPLUS2, Message 5 355

10.2F - Extract from thread CPLUS2, Message 6 357

10.3A - Extract from thread XTRPRG5, Message 1 363

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XIV
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

10.3B - Extract from thread XTRPRG5, Message 2 365

10.3C - Extract from thread XTRPRG5, Message 3 366

10.3D - Extract from thread XTRPRG5, Message 4 367

10.3E - Extract from thread XTRPRG5, Message 5 368

10.3F - Extract from thread XTRPRG5, Message 6 369

10.3G - Extract from thread XTRPRG5, Message 7 370

10.3H - Extract from thread XTRPRG5, Message 8 371

10.3I - Extract from thread XTRPRG5, Message 9 372

10.4A - Extract from thread PHYSRES3, Message 1 375

10.4B - Extract from thread PHYSRES3, Message 2 376

10.4C - Extract from thread PHYSRES3, Message 3 378

10.4D - Extract from thread PHYSRES3, Message 4 380

10.4E - Extract from thread PHYSRES3, Message 5 382

10.4F - Extract from thread PHYSRES3, Message 6 384

10.4G - Extract from thread PHYSRES3, Message 7 385

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XV
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

LIST OF TABLES

1.1 - Recently proposed taxonomies of the Knowledge Management field 3

3.1 - Selected studies about “virtual” Communities of Practice 65

3.2 - Evidence of Wenger’s constructs in selected studies about “virtual” CoPs 66

5.1 - Netscan search for “invest” sorted by Posts (September 1-30, 2002) 136

5.2 - Newsgroups pre-selected from Table 5.1 sorted by PPRatio 139

5.3 - Smith’s interaction measures for soc.history.medieval 143

5.4 - A record of the item-generation process of the Survey instrument 149

5.5 - Response rates for the Survey pilots 153

5.6 - Initial theory-derived codes for the Content Analysis 161

6.1 - Mainstream Usenet hierarchies 169

6.2 - Original newsgroup population and progressive application of selection 171


criteria

6.3 - First Sub-stage newsgroup sample 174

6.4 - Results of Second Sub-stage 178

6.5 - Core-periphery analysis of finalist newsgroups 182

6.6 - Newsgroups selected to participate in the study 189

7.1 - Response rates for the main Survey 194

7.2 - Characteristics of the Survey sample (n =239) 195

7.3 - Item summaries for LEARNING block 201

7.4 - Item summaries for MUTUAL ENGAGEMENT block 203

7.5 - Item summaries for SHARED REPERTOIRE block 204

7.6 - Item summaries for COMMUNITY block 205

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XVI
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

7.7 - Item summaries for Expertise scale 206

7.8 - Descriptive statistics of hypothesised scales 208

7.9 - Reliability analysis for ‘Sharing useful information’ 208

7.10 - Reliability analysis for ‘Shared artifacts’ 209

7.11 - LEARNING block: pattern matrix 216

7.12 - MUTUAL ENGAGEMENT block: rotation sums of squared loadings 218

7.13 - MUTUAL ENGAGEMENT block: rotated factor matrix 219

7.14 - SHARED REPERTOIRE block: pattern matrix 221

7.15 - COMMUNITY block: rotation sums of squared loadings 222

7.16 - COMMUNITY block: rotated factor matrix 223

7.17 - COMMUNITY block: reliability analysis for factor 3 item-set 224

7.18 - Validated scales derived from Exploratory Factor Analysis 227

7.19 - Newsgroup-level statistics from validated scales 229

7.20 - Correlation between refined scales and Coreness, Expertise and Tenure 230

7.21 - 5-Way ANOVA of Survey data (significant effects only) 231

7.22 - Responses to the question Do you consider this newsgroup to be a 239


community?

7.23 - Detected instances in textual responses to open community question 241

7.24 - Evidence of Essential Traits detected by the Survey in each newsgroup 245

8.1 - Original thread sample and progressive application of selection criteria 252

8.2 - Descriptive statistics of selected threads 256

8.3 - Refined working definitions of theory-predicted codes 262

8.4 - Instances of Wenger’s constructs detected in each newsgroup 265

8.5 - Typical forms of the Shared artifacts sub-codes 279

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XVII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

8.6 - Essential Traits detected in each newsgroup by the Content Analysis 284

9.1 - Stage III operationalisation of Exemplary Traits 291

9.2 - Empirical manifestations of the Exemplary Traits detected during Stage III 293

9.3 - Constructs and Sub-constructs detected by the Survey and the Content 294
analysis in each community

11.1 - The Research Questions and specific study results used to address them 405

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XVIII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CONTENTS OF THE APPENDICES CD-ROM

Appendix A Research timeline

Appendix B Survey materials used at the pilot stage

Appendix C Review of Internet-based survey providers

Appendix D Survey materials used during main Survey deployment

Appendix E Complete results of Netscan Usenet searches

Appendix F BASIC routine used to import Usenet messages into MS Access

Appendix G Complete UCINET core-periphery results for 19 analysed newsgroups

Appendix H Reliability analysis of hypothesised scales

Appendix I Auxiliary Factor Analysis results

Appendix J Complete results of 5-way ANOVA

Appendix K Complete coded responses to the open community question

Appendix L Final version of EFA-validated questionnaire

Appendix M Thread sampling universe for Content Analysis sample selection

Appendix N Complete Nudist code counts for 44 analysed threads

Appendix O Nudist data project for 44 content analysed threads and Agent
newsreader with 44 original threads

Appendix P Nudist-generated reports of 44 coded threads

Appendix Q Original institutional documents of eleven studied newsgroups

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XIX
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

ACKNOWLEDGEMENTS

This research was partially supported through a research grant awarded by the School of

Management of Universidad Panamericana in Mexico City, which the author gratefully

acknowledges.

Surveyworld.net of The Netherlands graciously provided the researcher with free use of

their web-based survey solution, which was used in the pilot and main Survey runs.

I am grateful to the Faculty and Staff of the University of Bradford, who gave me

invaluable support during these six years as an extramural student.

I wish to thank Etienne Wenger, Molly Wasko and Robin Teigland for reading through

my complicated letters, and providing encouraging and thoughtful replies.

My greatest debt lies with my supervisor, David Spicer. He made a large contribution

to this research project, first by believing it it, then with his constant and cheerful

encouragement, and lately with many patient re-readings, and with his good-natured

insistence on high-standards. Thanks, Dave.

Mexico City, May 2006

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XX
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

I dedicate this work to my parents, Enrique and Rosa María.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XXI
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

PREFACE

A brief Glossary of Internet terms is included at the end of the thesis for the convenience

of the reader. Terms defined in the glossary are highlighted in the text with a dashed

underscore.

The thesis is structured around Wenger’s (1998) constructs whose presence is judged as

indicative of the existence of a CoP (see Section 2.3). Given the key role played by the

constructs, and their recurrence (mostly as nouns, but sometimes as verbs too) throughout

the thesis, they have been highlighted by writing them in SMALL CAPS FONT thus:

MUTUAL ENGAGEMENT, SHARED REPERTOIRE, JOINT ENTERPRISE, COMMUNITY AND

LEARNING/IDENTITY ACQUISITION.

Because of their excessive length, the Appendices are included in CD-ROM format in a

back pocket.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XXII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

If you have an apple and I have an apple and we exchange apples then you and I

will still each have one apple. But if you have an idea and I have an idea and we

exchange these ideas, then each of us will have two ideas.

George Bernard Shaw

We should not only use the brains we have, but all that we can borrow.

Thomas Woodrow Wilson

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
XXIII
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Part One

Theoretical Background

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
1
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CHAPTER ONE

Introduction

The study of Knowledge and the Firm has been a topic of increasing importance in

Management research since the early 90’s (Nonaka, 1991; Kogut and Zander, 1992;

Quinn, 1992; Starbuck, 1992; Stewart, 1994). Closely following the lead of the

Information Revolution, this focus on knowledge soon developed into a full-fledged

academic and consulting specialty, Knowledge Management (KM). As a rough

indicator of its rapid growth, Despres and Chauvel (2000) note that scientific articles

containing the keyword “knowledge management” in the ABI/INFORM database rose

from 3 in 1988 to 698 in 1998, an average yearly increase of over 100 percent.

In this short history, the KM field has spawned numerous sub-fields and conceptual

approaches, causing some authors to describe it as a fragmented field (Grover and

Davenport, 2001; Grey and Meister, 2003). The fragmentation has resulted in various

proposed taxonomies; five recent ones are shown in Table 1.1, and they in turn reflect a

diversity of views and methodologies. Nevertheless, they all identify the study of

Communities of Practice (CoPs) as a distinct KM sub-field. These have been defined as

“groups of people who share a concern, a set of problems, or a passion about a topic,

and who deepen their knowledge and expertise in this area by interacting on an

ongoing basis” (Wenger, McDermott and Snyder, 2002: 4). Examples might include a

group of nurses who discuss patient cases over their daily lunch meeting (Wenger,

1996), or photocopier repair technicians who regularly swap stories about the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
2
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

machines they fix (Orr, 1996), or master craftsmen who produce some of the finest

flutes in the world at three workshops in the Boston area (Cook and Yanow, 1993).

Table 1.1 – Recently proposed taxonomies of the Knowledge Management field


Article

Kakabadse, Subramani, Nerur


Despres and
Binney (2001) Earl (2001) Kakabadse and and Mahapatra
Chauvel (2000)
Kouzmin (2003) (2003)

Case studies in six


Methods

Database search Database search


companies and
and ethnographic Literature review Literature review and statistical co-
interviews with 20
thematic analysis citation analysis
CKO’s
Taxonomy

7 “major clusters” “6-element KM 5 “dominant KM 8 “domains of KM


7 “schools of KM”
of KM activity spectrum” models” research”

Business Transactional KM Systems school Cognitive model Knowledge as firm


intelligence capability

Benchmarking Analytical KM Engineering school Community model Practice of KM

Datawarehousing Asset Commercial school Network model Innovation and


Management KM change
Sub-fields included in the Taxonomy

Communities of Process-based Spatial school Quantum model Philosophy of


Practice KM Knowledge

Groupware and Developmental Strategic school Philosophical Situated Learning


virtual teaming KM model and Communities
of Practice

Innovation, Innovation and Cartographic Organizational


synergies and creation KM (with school Information
creativity CoPs included Processing and IT
explicitly) Support for KM

Learning, Organizational Knowledge


competencies, school (with CoPs Communication,
and employee included explicitly) Transfer and
development Replication

Organizational
Learning and
Learning
Organizations

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
3
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The concept of CoP has led to ground-breaking insights about workplace learning, one of

KM’s concerns. Researchers found that employees are constantly learning as they go about

their daily work, and much of this learning bears little relation –and is often at odds– with

formal training and canonical work procedures (Brown and Duguid, 1991; Wenger, 1998).

Rather, learning is a matter of becoming competent members of informal communities,

which through their shared practices provide a living repository for knowledge (Orr, 1990).

Thus CoPs are not confined to formal apprenticeships, but are natural social structures that

extend to all social settings where people work and accomplish things together (Wenger,

1998). Hence their relevance to business organisations concerned with knowledge.

The concept has attracted considerable interest in management research; many authors

treat it as a key element in the knowledge-based view of the firm (Kogut and Zander,

1996; Brown and Duguid, 1998; Tsoukas and Vladimirou, 2001; Grover and Davenport,

2001). As regards management practice, the explicit recognition and support of these

elusive communities is becoming the cornerstone of an increasing number of KM

programmes (Wenger and Snyder, 2000).

Research about CoPs stresses the importance of members interacting regularly with each

other because it is these interactions that build both the community and the shared

practice (Wenger, 1998). In pre-Internet days, interactions were typically face-to-face,

with all the advantages of rich communication this proximity entails (Cummings, Butler

and Kraut, 2002). Currently, with the rapid spread of Information and Communication

Technology (ICT), coupled with unrelenting globalisation, business organisations have

become “wired” and more far-flung, increasingly making stable worker co-location the

exception rather than the rule (Holtshouse, 1998). This raises the question of whether

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
4
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

geographically-distributed employees, whose only link with remote colleagues is an

electronic network, can function as a CoP. Some authors argue co-location is a necessary

condition for the kind of interaction a CoP needs to work (Brown and Duguid, 2000b;

Wasko and Teigland, 2004). Others point out the leanness of the Internet medium is not

an insurmountable obstacle to rich communication (Carlson and Zmud, 1999; Burke,

Aytes and Chidambaram, 2001). Yet others argue that if participants have a shared

practice, then they share the common ground and context that allows rich interaction to

take place (Wenger et al, 2002; Brown and Duguid, 2001).

This thesis makes a contribution to the debate by performing a systematic theory-

informed search for working CoPs over a specific area of Internet known as Usenet.

The search successfully detected four Usenet-based communities exhibiting all

theoretical attributes of CoPs as defined and described by Wenger (1998).

Beyond this overview, Chapter One comprises four sections. The First provides a short

introduction to the concept of CoP. Section Two states thesis aim and objectives, and

briefly outlines the research strategy. Section Three describes the chapter structure of

the thesis. Section Four is a chapter summary.

1.1 – Communities of practice

Lave and Wenger (1991) are credited with the introduction of the term Community of

Practice. Their research involved ethnographic studies of traditional apprenticeship as

practiced in various countries and professions: butchers, Goa tailors, Yucatan midwives,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
5
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

quartermasters and non-drinking alcoholics. In each case, novices gradually became full

members of a community by learning its language, practices, and sensibilities. They

learned not through formal classes, but by being allowed to participate in the real work of

the community, albeit in a peripheral role. The authors called this mode of belonging

legitimate peripheral participation (LPP), and concluded it is the defining characteristic

of apprenticeship as a form of learning.

Because Lave and Wenger’s (1991) book pre-dates the emergence of KM, some consider

CoPs a sub-field of the Organisational Learning field (e.g. Easterby-Smith, 1997; Bell,

Whitwell and Lukas, 2002; Easterby-Smith and Lyles, 2003). However, these literature

reviews fail to mention Wenger’s (1998) book and its considerably more developed CoP

framework. Moreover, Wenger himself aligns the concept within the KM field (Wenger,

1999; Wenger, 2000a; Wenger, 2004b), as the previous section did.

Even though Lave and Wenger made LPP the focus of their research, and referred to

CoPs only peripherally, it was the latter concept that quickly caught the attention of

management researchers. First, Brown and Duguid (1991) argued organisational CoPs

were significant sites of innovation through their adaptive non-canonical practices.

Shortly afterwards came increased interest in knowledge-based theories of the firm

(Grant, 1996; Kogut and Zander, 1996; Spender, 1996a; Tsoukas, 1996) and widespread

agreement that, in the knowledge economy, knowledge is a key source of competitive

advantage and superior performance (Spender and Grant, 1996; Davenport and Prusak,

1998; Hansen, Nohria and Tierney, 1999; Von Krogh, Nonaka and Aben, 2001). This

launched KM, which went through a first wave of projects heavily biased toward

ICT-based solutions, with mostly disappointing results (McDermott, 1999a; Scarbrough,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
6
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2003; Thompson and Walsham, 2004). In the meanwhile, some companies reported

excellent knowledge outcomes from supporting and leveraging existing CoPs (Brown

and Grey, 1995; Hanley, 1998; Wright, 1999; Barrow, 2001). Hence, second-generation

KM projects display increased understanding of the limits of ICT, the social character of

knowledge, and the essential role of intra-organisational CoPs (Blackler, 2000; Tsoukas

and Vladimirou, 2001; Newell, Robertson, Scarbrough and Swan, 2002).

Managers have thus gained increasing awareness of the advantages of CoPs, specifically:

• they foster knowledge-sharing (Stewart, 1996)

• they foster innovation at all levels (Brown and Duguid, 1991)

• they are the seat of core competences (Brown and Grey, 1995; Manville and

Foote, 1996)

• they can assume ownership and stewardship of knowledge (Wenger, 2000a)

• they transfer best practices (Wenger and Snyder, 2000)

Yet, managers attempting to harness the power of CoPs have found they pose distinct

management challenges (Wenger and Snyder, 2000). CoPs are different from formal

organisational structures, such as teams or business units. These are created by

management decree, and have specific deliverables defined for them. By contrast, CoPs are

informal, emergent structures, formed by self-selected and committed members, who pick

their own leadership, negotiate a learning enterprise, and answer only to themselves. Active

efforts to manage or direct CoPs can be counter-productive, since their success depends on

the kind of personal passion and committment that cannot be mandated (Stewart, 1997;

Thompson, 2005). Yet CoPs can be supported, “nurtured” and indirectly guided, so that

they become reliable knowledge assets for the organisation (Wenger and Snyder, 2000). In

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
7
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

fact, recent studies describe some profit and non-profit organisations that by modelling

themselves after CoPs achieve pervasive innovation, high commitment from members, and

exceptional levels of collaboration (Lee and Cole, 2003; Evans and Wolf, 2005).

However, in the current business climate of rapid globalisation, flattened hierarchies and

mobile workers (Hindle, 2006), one condition for conventional CoPs is becoming

increasingly rare: stable employee co-location (Holtshouse, 1998). The most important

enabler of a CoP is regular participant interaction on the issues the community is

interested in, what Wenger (1998) calls MUTUAL ENGAGEMENT. In the past, interactions

were face-to-face, thus facilitating knowledge-sharing. With the workplace turmoil of the

21st century, this is becoming increasingly difficult, prompting one expert to ask

rethorically: “If the water cooler was a font of useful knowledge in the traditional firm,

what constitutes a virtual one?” (Prusak, 1997: xiii).

Hence the importance of critically examining the possibility of “virtual” CoPs, which this

study views as true CoPs whose interactions are regularly performed through computer

networks, particularly the Internet. This is a debated issue. Some authors, while readily

agreeing people use the Internet for sharing practices, do not believe true CoPs can be

fully Internet-based (notably Brown and Duguid, 2000b; Wasko and Teigland, 2004). By

constrast, other authors provide detailed descriptions of successful Internet-based

communities they characterise as CoPs (e.g. Saint-Onge and Wallace, 2002; Schlager,

Fusco and Schank, 2002; DeSanctis, Fayard, Roach and Jiang, 2003).

The fact that competent people like to talk shop, and the rapid spread of the Internet as

a universal communication medium, led the researcher to hypothesise that passionate

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
8
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

practitioners will take advantage of the Internet to seek and ENGAGE other practitioners,

eventually cohering into true Internet-based CoPs. Based on this assumption, the thesis

devises an original Research Strategy to first detect Internet-based communities

exhibiting observable attributes frequently associated with CoPs, and then intensively

examine the best candidates among them to assess whether or not they exhibit all

theory-predicted CoP attributes.

Studies about “virtual” CoPs are hardly new, although this thesis argues previous studies

have not rigorously applied Wenger’s (1998) theory, thus rendering claims about “virtual”

CoPs problematic. Specifically, no previous study applies Wenger’s constructs of MUTUAL

ENGAGEMENT, COMMUNITY, SHARED REPERTOIRE, JOINT ENTERPRISE, and LEARNING/

IDENTITY ACQUISITION to rigorously assess whether or not a virtual community is a true

CoP. This study aims to locate Internet communities with high “CoP-potential”, and to use

the presence of Wenger’s constructs as the theoretically-grounded criteria for assessing

them as CoPs. Moreover, the constructs will be used as the guidelines for developing the

virtual CoP model, the research questions and the research design.

1.2 – Research aim, objectives and strategy

As noted before, the research starts with the assumption that passionate practitioners will

use the Internet to engage with each other thus forming Internet-based CoPs. The thesis

will pursue this working hypothesis by constructing a theory-informed model of “virtual”

CoP to provide a concrete target, and then conducting a systematic Internet search for

virtual communities that display every trait specified by the model. Therefore:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
9
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The aim of this thesis is to extend Wenger’s (1998) theory of CoPs to the social

areas of the Internet by undertaking a systematic search for, and a rigorous

assessment of, working exemplars of “virtual” or Internet-based CoPs.

This overall aim can be organised into six specific and cumulative Research Objectives:

Research Objective One: To build a model of Internet-based CoP that is

consistent with Wenger’s theory of CoPs, and includes theory-informed criteria

that enable selection of virtual communities with high “CoP-potential”.

Research Objective Two: To design, under the guidance of the model, a

Research Strategy that will, first, enable detection of Internet-based communities

exhibiting high CoP-potential, and second, enable theory-grounded analysis of

selected communities to rigorously assess whether or not they truly are CoPs.

Research Objective Three: To conduct a highly comprehensive search of the

Internet, and to locate the virtual communities that display the strongest affinity

to the model of Internet-based CoP.

Research Objective Four: To question regular participants of selected

communities about the presence or absence of visible manifestations of

Wenger’s constructs.

Research Objective Five: To directly examine online interactions of regular

participants in selected communities to determine the presence or absence of

visible manifestations of Wenger’s constructs.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
10
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Research Objective Six: To assess the presence or absence of Wenger’s constructs

in each community, using concurring evidence from two independent methods, and

thereby establish whether any among selected virtual communities display all

theoretical properties of CoPs, and can therefore be characterised as such.

The singular research problem that launched this study –involving an Internet search

for particular communities, and a rigorous assessment of their CoP attributes–, is

addressed through a multi-stage Research Strategy, tightly linked to the Research

Objectives, and comprising the coordinated deployment of several research methods

and Internet tools. The Strategy and its links to the Objectives is briefly outlined here,

and is fully described in Chapter Four.

Addressing Objective One first requires a review of CoP theories and previous studies

about “virtual” CoPs. Stage I of the Research Strategy can then draw upon extant theory to

build a coherent and theoretically-grounded model of Internet-based CoP. This model, in

turn, provides a blueprint for designing the remaining Stages of the Strategy.

Stage II assumes the hypothesised model is an existing Internet entity, and develops

additional Strategy Stages to enable efficient detection of suitable virtual communities and

assessment of their CoP status. A key decision is to restrict the Internet search to the Usenet

discussion network; the rationale for this is provided in Chapter Four.

Stage III of the Strategy addresses the Third Research Objective, concerning the search for

suitable virtual communities in the Usenet network. The main instrument deployed in this

Stage is Smith’s (1999) newsgroup analysis tool, Netscan, which can thoroughly search

Usenet’s principal hierarchies for newsgroups with the best profile for this study.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
11
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Quantitative and qualitative criteria derived from the model are combined to make a

stringent newsgroup selection, and thus arrive at a narrow set of Usenet-based communities

exhibiting the highest affinity to the virtual CoP model. The methods used to perform the

search are described in Chapter Five, and their actual deployment, in Chapter Six.

Stage IV of the Research Strategy addresses Objective Four, and uses a web-based

Survey to question participants of selected communities about visible manifestations of

Wenger’s constructs. The design and development of the Survey instrument is

described in Chapter Five; the deployment and statistical analysis of results are

reported in Chapter Seven.

Stage V, which addresses Objective Five, performs a qualitative Content Analysis of a

sample of online discussions between core members of each community. The method

targets text-based manifestations of Wenger’s constructs. The design of the coding

scheme and the rationale for sample selection are described in Chapter Five; the actual

coding of discussions and the results of the analysis are reported in Chapter Eight.

Stage VI of the Strategy, which focuses on Objective Six, involves reviewing each of the

virtual communities, and making a synthesis of the quantitative and qualitative data

obtained to this point. In each community, independent evidence for Wenger’s constructs

is provided by the Survey and the Content Analysis, thus affording robust triangulation.

Evidence for the presence of each construct is deemed sufficient if both instruments

concur. Communities that display all Wenger constructs can be assessed as Usenet-based

CoPs. The review of each community and the results of the CoP-assessment are reported

in Chapter Nine.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
12
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The preceding outline emphasises the one-to-one relationship between Research

Objectives and Strategy Stages, the latter extending over core chapters of the thesis (see

Figure 1.1). This is further elaborated in the next section which provides a complete

outline of the thesis and its chapter organisation.

Research Objectives Research Strategy Thesis Chapter

Objective One Stage I

Build a theoretically-grounded Model Specification Chapter Four


model of Internet-based CoP

Objective Two Stage II

Develop an effective strategy for Strategy Formulation Chapter Four


locating and assessing Internet CoPs

Objective Three Stage III

Comprehensively search the Internet Usenet Search Chapter Five: design


and identify virtual communities with Chapter Six: results
the strongest affinity to the model

Objective Four Stage IV

Ask community participants about the Participant Survey Chapter Five: design
presence of Wenger constructs Chapter Seven: results

Objective Five Stage V

Directly examine participant interaction Content Analysis Chapter Five: design


for the presence of Wenger constructs Chapter Eight: results

Objective Six Stage VI

Assess the presence of Wenger CoP Assessment Chapter Nine


constructs, and identify communities
displaying all the properties of CoPs

Figure 1.1 – Research Objectives and Strategy Stages


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
13
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

1.3 – Outline of Thesis

The thesis is structured into four major Parts, as shown in Figure 1.2. Part One covers

the Theoretical Background of the study and includes Chapters One through Three.

Beyond this Introduction, Chapter Two reviews previous research about CoPs, starting

with a historical overview of the development of the concept. This is followed by an

explanation of Wenger’s (1998) social theory of learning, in which CoPs are but one

element. The next section concentrates on CoPs alone and presents Wenger’s

theoretical framework, with particular emphasis on the constructs he identifies as

essential elements of CoPs. The chapter ends with a comparison and contrast of

Wenger’s theory with other theories of CoPs.

Chapter Three examines the issue of “virtual” CoPs and is based on the selection and

review of 20 articles that explicitly or implicitly address the topic. The First Section

examines different meanings of the term “virtual” in the articles, and proposes a

classification of virtual communities based on three contingent variables of network

design and/or utilisation. The next section then looks at the articles through the

theoretical lens of Wenger’s constructs, to build a case about their feasibility in online

environments, and to establish the research need. The Third Section addresses

challenges to the notion of “virtual” CoP, particularly Brown and Duguid’s (2000a)

concept of Networks of Practice (NoPs), and Wasko and Teigland’s (2002; 2004)

empirical work on Electronic Networks of Practice (ENoPs).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
14
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Chapter One – Introduction


Theoretical Chapter Two – Previous Research on Communities of Practice
Background
Chapter Three – Virtual Communities of Practice

Chapter Four – Research Design


Design
Research Strategy

Chapter Five – Methods

Chapter Six – Usenet Search and Community Selection

Execution Chapter Seven – Survey of Community Members

Chapter Eight – Content Analysis of Community Discussions

Chapter Nine – Overall Assessment of Selected Communities

Study Findings Chapter Ten – Examining Normal Activity in Detected VCoPs

Chapter Eleven – Conclusions

Figure 1.2 – Structure of Thesis

Part Two, comprising Chapters Four and Five, addresses the design of the Research

Strategy used by the thesis.

Chapter Four begins with the construction and discussion of the model of virtual CoP

that provides guidance and focus to this study. This is followed by a discussion of the

specific aims and objectives of the thesis, and by the detailed formulation of the

Research Questions. Next, the decision of focusing the search on the Usenet network is

justified, and previous studies of Usenet-based communities are reviewed. Ontological

and epistemological assumptions of the thesis are made explicit, and the logic of

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
15
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

triangulation is discussed. Finally, following the guidelines afforded by the model, a

particular Research Strategy is devised in order to achieve all Research Objectives.

Chapter Five provides a detailed discussion of the various tools and methods the

Research Strategy deploys in each Stage. Specifically, four methods are explained:

Smith’s (1999) Netscan tool and measures of online interaction; the application of

Social Network Analysis to newsgroups and the core-periphery model; the web-based

Survey of newsgroup members; and the Content Analysis of newsgroup discussions.

Part Three addresses the Execution of the Research Strategy, comprising Chapters Six

through Eight.

Chapter Six describes the search for Usenet communities with strong affinity to the

virtual CoP model. It begins with a broad Usenet search using Netscan, followed by

an assessment of the affinity of analysed newsgroups to the virtual CoP model, based

on Smith’s (1999) measures of interaction, a core-periphery analysis, and a

qualitative assessment of institutional documents (such as a Charter or a FAQ)

maintained by the different newsgroups. The output of Chapter Six is the reduced

subset of stable and persistent virtual communities to be included in the study, as well

as the specific members from each community to be included in Survey and Content

Analysis samples.

Chapter Seven describes the deployment of the web-based Survey to high-coreness

members of selected communities. It then describes the use of Exploratory Factor

Analysis to validate and refine Survey scales, and the logical link between the refined

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
16
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

scales and hypothesised indicators of the Wenger constructs. Statistical analyses of scales

are reported, and a list provided of Wenger constructs detected within each community.

Chapter Eight reports on the Content Analysis of 44 discussion threads (four from each

newsgroup) selected on theoretical grounds. These were imported into the textual

analysis software NUDIST, where a two-pass coding procedure was used to code every

Usenet message for the presence or absence of pre-defined manifestations of Wenger’s

constructs. Code-counts by thread and community are reported, and coding decisions

are illustrated with brief exercises. Typical manifestations of Wenger’s constructs are

described for different communities, and the chapter concludes with a list of constructs

detected in each community.

Finally, the Study Findings are collected under Part Four, which includes Chapters

Nine, Ten and Eleven.

Chapter Nine draws on the quantitative and qualitative data reported in previous

chapters to build a complete profile of each community, and assess the presence of each

of Wenger’s constructs, using concurring results of the Survey and the Content

Analysis. Four communities were found to exhibit all Wenger constructs, and were

therefore assessed as Usenet-based CoPs.

Chapter Ten then takes a different look at communities assessed as true CoPs by

performing ethnographic analysis of a short interaction episode, using the theoretical lens

of Wenger’s CoP theory. This results in a richer depiction of the day-to-day activities of

the communities and further confirms their characterisation as Usenet CoPs.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
17
Introduction
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The thesis is drawn to a close in Chapter Eleven which begins by reviewing the aim

and Research Objectives, and assessing to what extent they have been achieved. A

similar assessment is performed for the Research Questions. Thesis contributions are

discussed, as are implications for theory, for individuals and organisations. Study

limitations are discussed, and areas for future research suggested.

1.4 – Summary

This chapter introduced the concept of CoP as a sub-field of KM theory and practice.

These informal organisational groups are natural vehicles of knowledge-sharing and

innovation, and as such are increasingly valued by knowledge-based organisations. The

issue of Internet-based or “virtual” CoPs was also introduced, and placed against the

backdrop of contemporary business and technological trends.

The specific aims and Objectives of the study were mentioned, and a brief outline

provided of the multi-stage Strategy it will follow, with particular emphasis placed on

the correspondence between Strategy Stages and Research Objectives.

Having introduced the topic and the research issues, the chapter closed with an

overview of the chapter organisation of the thesis.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
18
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CHAPTER TWO

Previous Research on Communities of Practice

Communities of Practice (CoPs) have been studied by various disciplines, including

Sociology, Education and Management, but it goes beyond the scope of the thesis to

review all literatures. Hence, the review will focus on Management literature.

This chapter will achieve four purposes:

• It will present and critically discuss Wenger’s (1998) theoretical framework

of CoPs, and justify its adoption by this research on the grounds of it being

the most comprehensive and consistent extant theory.

• It will present Wenger’s constructs as visible traits that allow empirical

identification of a CoP, real or “virtual”.

• It will propose an operational definition of CoP based on Wenger’s

constructs.

• It will propose an operational definition of each construct by specifying

concrete observable manifestations.

The chapter is organised into five sections. The First reviews early CoP studies and the

evolution of the concept through the work of Wenger, Brown and Duguid, and others.

Section Two introduces Wenger’s theory of CoPs by first describing his broader social

theory of learning, of which CoPs are but one element. The Third Section then

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
19
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

describes CoPs in greater detail, particularly their constitutive elements, which play a

key role in this research. The Fourth Section reviews critiques of Wenger’s theory and

distinguishes between CoPs and similar phenomena described in the literature. The last

section is a chapter summary.

2.1 – Initial formulation of the Community of Practice concept and early studies

The concept of CoP was originally proposed by anthropologist Jean Lave and sociologist

Etienne Wenger in the context of their ethnographic analysis of various institutions of

apprenticeship. The focus of their research was social learning; they studied traditional

apprenticeships to understand how and why they were ancient and effective social

mechanisms for learning (Lave and Wenger, 1991). They proposed a theory of learning

whereby people learn by becoming acknowledged but peripheral members of social

communities where knowledge resides, not as abstract ideas, but as embodied and shared

practices. Hence learning would be the process of joining a community, and actually

taking part in its practices, beginning with the most basic and gradually mastering the

most complex, but always working alongside established members. The progression from

peripheral membership to full insider status they named legitimate peripheral

participation, and it is their main intended contribution and the title of their book (ibid).

Moreover, they coined the term communities of practice, to designate the communities

apprentices joined, which are the living repositories of practices. However, explaining

these communities was not their aim; they acknowledge that the concept of CoP “is left

largely as an intuitive notion” in need of further development (Lave and Wenger, 1991:

42), which is why their definition reads more like a description:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
20
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

A community of practice is a set of relations among persons, activity and world,

over time and in relation with other tangential and overlapping communities of

practice. A community of practice is an intrinsic condition for the existence of

knowledge, not least because it provides the intrinsic support necessary for

making sense of its heritage. Thus participation in the cultural practice in which

any knowledge exists is an epistemological principle of learning. The social

structure of this practice, its power relations, and its conditions for legitimacy

define possibilities for learning (i.e. for legitimate peripheral participation).

(ibid: 98)

Lave and Wenger’s focus on the process of joining CoPs rather that on the CoPs

themselves left important gaps in their ground-breaking research. For instance, they

have scant a word for insiders or full members of the community, and then only in

reference to the learning opportunities they provide for apprentices. They equate

learning with peripheral participation, thereby disregarding the learning that results

from interactions among insiders, even in the absence of apprentices. More recently,

this learning of insiders has been described as the main source of energy for the

community’s continued existence (Wenger, 2000b).

It was just this learning between insiders that caught the earlier attention of Julian Orr,

another anthropologist, who conducted an ethnographic analysis of Xerox photocopier

repair technicians as they went about their work of fixing machines (Orr, 1990). It was

supposed to be an individual job, and technicians were supposed to merely follow the

repair procedures specified in the manual. In practice, though, Orr discovered the “tech

reps” had an informal but strong community that met daily for breakfast to exchange

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
21
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

problem-solving tips. They did this by creating and telling stories about the machines

they fixed. This narration was an effective way of communicating complex and

difficult-to-articulate knowledge about the machines. Orr learned that the official

manual reps were supposed to follow was mostly useless for the task of actually

repairing a malfunctioning machine. The copiers were complex systems, using different

kinds of paper, working in different temperature and humidity conditions, and having

different maintenance histories. All these differences rendered them idiosyncratic and

unpredictable systems which no manual could comprehensively explain. Hence, it was

the technicians who put in the necessary adaptation, working as an informal but

effective distributed team, helping and educating each other through narration, and

collectively maintaining a detailed and updated knowledge-base about the machines

they serviced.

Lacking Lave and Wenger’s (1991) CoP concept, Orr (1990) referred to his community

of technicians as an occupational community, following van Maanen and Barley

(1984). Other CoP researchers would later cite Orr’s ground-breaking study as the

earliest ethnography of a CoP (Raelin, 1997; Brown and Duguid, 2001; Teigland,

2003). (The concept of occupational community is reviewed in the Fourth Section of

this chapter.)

Brown and Duguid (1991) were the first to write about the role of intra-organisational

CoPs, noting how, through their continuously-adapting non-canonical practices, they

were significant sites of innovation. Their often-cited article was the first to argue that,

despite their near-invisibility, CoPs constituted a valid and important concern for

management, especially in knowledge-based organisations.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
22
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The mid-nineties can be characterised as a period of “emerging excitement” about the

concept of CoP (Hirsch and Levin, 1999), with enthusiastic accounts appearing in

business magazines (e.g. Brown and Gray, 1995; Manville and Foote, 1996; Stewart,

1996). A typical definition from that period, suggested by two Knowledge Management

(KM) consultants, reads thus:

A group of professionals, informally bound to one another through exposure to

a common class of problems, common pursuit of solutions, and thereby

themselves embodying a store of knowledge (Manville and Foote, 1996: 80).

Interest in the concept of CoP is partly due to the fact that it puts a name to the familiar

human experience of belonging to, and participating in, a group of like-minded peers.

Indeed, Wenger (1999) claims CoPs are natural social structures, as old as humankind,

which would make the concept more than just the latest management fashion

(Abrahamson, 1996). In addition, the mid-nineties were a time marked by the contrast

between the early promise of the knowledge-based organisation (Drucker, 1988), and

the disappointing results of first-generation KM systems (McDermott, 1999a; Stewart,

2001; Scarbrough, 2003; Thompson and Walsham, 2004). Emerging CoP theory

compellingly attributed these failures to an overly-optimitic focus on technology that

was largely unaware of the essential role of intra-organisational CoPs (Wenger, 2000a;

Swan, Newell and Robertson, 2000a; Tsoukas and Vladimirou, 2001).

After his research with Lave, Wenger shifted his attention from the process of

induction of new members to the CoP itself. He based his new theorising on

ethnographic fieldwork conducted in 1989-90 in a medical claims processing centre

operated by a large US insurance company. The results of his research were published

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
23
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

several years later as an ethnography of claims processors (Wenger, 1998). The book

remains, to this day, the most detailed and comprehensive treatise on CoPs (Schwen

and Hara, 2003; Plaskoff, 2003), arguably making Wenger’s theory the de facto

standard. This is also suggested by its becoming the focus of an increasing number of

critiques (e.g. Fox, 2000; Contu and Willmot, 2000; Marshall and Rollinson, 2004;

Roberts, 2006).

Wenger points out, though, that he uses the concept of CoP mainly as an “entry point”

into a broader social theory of learning, or as a way to bring together social theory and

learning theory (Wenger, 1998; 2004a). Some elements of this theory identify and

illuminate specific traits of CoPs, which in turn can be used to build a model of

Internet-based CoP. Therefore, before describing Wenger’s concept of CoP, an

overview of his social theory of learning is provided.

2.2 – Wenger’s social theory of learning

Wenger begins by noting alternative learning theories have largely focused on the

mechanics of learning –brain research, genetics, evolutionary psychology, cognitive

science, information-processing models (Wenger, 2004a). Yet, by making

meaningfulness the ultimate aim of learning, Wenger’s theory addresses a different

level. A focus on meaningfulness, coupled with the premise that meanings are

negotiated between human beings, implies that the social nature of human beings is

not a circumstantial but an essential aspect of learning. This does not deny the

possibility or value of individual learning, but asserts that meaningfulness cannot be

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
24
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

determined in isolation. Brown and Duguid (2000b: 139) quote Marx in pointing out

that even Robinson Crusoe, the oft-quoted model of pure economic isolation, set

about organising his castaway life “like a true-born Briton” [by keeping] “a set of

books”.

Briefly, Wenger’s (2004a) reasoning is:

• The negotiation of new meanings –not just the acquisition of new skills or

information– is fundamental to human learning.

• The ability of human beings to negotiate meaning is rooted in their social

nature.

• The negotiation of meaning is embedded in the practice of specific human

communities. These communities and their practices provide material for

learning: language, artifacts, interpretation of the world. Moreover, they

socially-define what it means to be competent with respect to a practice.

• Learning so understood transforms both a person’s ENGAGEMENT with the

world and being in the world. LEARNING is therefore a social becoming, the

ongoing negotiation of an IDENTITY that is developed in the context of

participation (and non-participation) in various CoPs.

Hence, Wenger’s main interest is Learning, not CoPs, which play an instrumental role

in his theory, albeit an essential one. CoPs make possible a focus on meaningfulness

because they locate Learning in a social structure where the meaning of Learning is

negotiated (Wenger, 2004a). Moreover, CoPs are just one of four constitutive and

interconnected elements of the theory, specifically (1998: 5):

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
25
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• Meaning: a way of talking about our (changing) ability –individually and

collectively– to experience our life and the world as meaningful.

• Practice: a way of talking about the shared historical and social

resources, frameworks, and perspectives that can sustain mutual

engagement in action.

• Community [of practice]: a way of talking about the social configurations

in which our enterprises are defined as worth pursuing and our

participation is recognizable as competence.

• Identity: a way of talking about how learning changes who we are and

creates personal histories of becoming in the context of our communities.

For Wenger, then, knowing is a matter of displaying competences defined in social

communities, and therefore an act of identifying with, and belonging to, those

communities. However, at any given time the personal experience of meaning of

community members is not necessarily congruent with these socially-defined

competences. When competence and personal experience diverge, one will “pull” on

the other: either members will feel an urge to align their experience of meaning with

the competence defined by the community, or they will try to communicate their

experience to the community to change the way it defines competence. It is in this tug-

of-war where LEARNING, understood as a transformation of knowing takes place.

Hence, LEARNING is a negotiation between social competence and personal experience

(Wenger, 1998). Yet there is more at stake in this negotiation than defining a position

on the issues. Belonging to a community involves members’ IDENTITIES because they

identify with their community: the decision to belong to a CoP involves one’s sense of

self. Therefore LEARNING, whether it changes the experience of a member or the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
26
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

competence of the community, involves ACQUIRING AN IDENTITY: LEARNING is an act

of becoming (ibid).

To illustrate the connections between LEARNING and the four elements, and to highlight

the distinct processes that result in LEARNING, Wenger (1998: 5) proposed a diagram,

reproduced in Figure 2.1. The diagram depicts CoPs as just one element of the full

learning theory, and it is on this element that the next section will focus.

learning as
belonging
community
[of practice]
learning as
doing

practice Learning identity

learning as
becoming

meaning
learning as
[meaningful]
experience

FIGURE 2.1 – Components of Wenger’s social theory of learning


(after Wenger, 1998: 5)

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
27
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.3 – Wenger’s theory of Communities of Practice

Within his social theory of learning, Wenger joins the elements of practice and

community by describing three dimensions of practice as the source of coherence of a

community (of practice), i.e. as what makes that particular kind of community cohere.

These dimensions thus become constitutive or defining elements of CoPs (Wenger,

1998):

• MUTUAL ENGAGEMENT: members build the community and the practice

by conducting practice-related interactions with each other on a regular

basis.

• a JOINT ENTERPRISE: members collectively negotiate what their CoP is all

about (i.e. they negotiate meanings) and hold each other accountable to

this understanding.

• a SHARED REPERTOIRE: over time, members develop a set of shared

resources that allow them to more effectively ENGAGE.

Wenger’s use of the term practice is very specific to his theoretical characterisation of

a CoP. He views practice as an emergent social structure that members of a community

collectively develop through their negotiation of meaning in order to address the

challenges of their ENTERPRISE. The claims processors, for instance, developed an

indigenous practice to accomplish the job in a manner satisfying for themselves

(Wenger, 1998). This practice imported elements of broader professional practices,

such as law and medicine. Yet it also developed new elements that were an original

response to local challenges. Ultimately the practice was a complex, adaptable, ad-hoc

solution, wholly-owned by the community of claims processors.


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
28
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

These qualifications are important because “practice” has at least two other common

usages in the Management literature. First, practice is used to refer to a particular

procedure that has yielded good results elsewhere, which is the sense intended when

speaking of sharing or transferring “best practices” (e.g. O’Dell and Grayson, 1998;

Szulanski, 2003). Second, practice is also used to refer to the way things are done in a

particular discipline, that by definition transcends any local CoPs (Brown and Duguid,

2001). An attentive look at these three usages reveals that practice is always created

and validated by a community, or as Wenger notes, practice is always social practice.

Yet, it is important to specify the scope, because the communities can be big and loose

(such as the medical community) or small and tight (as are most CoPs), just as the

practice can be very general (as, say, the field of cardiology), or very particular (as that

of the claims processors).

As noted earlier, Wenger defines MUTUAL ENGAGEMENT, JOINT ENTERPRISE and

SHARED REPERTOIRE as dimensions of practice. Therefore, if practice is the indigenous

ad-hoc response of a CoP to the challenges of its ENTERPRISE, then MUTUAL

ENGAGEMENT is the “effective” dimension of practice, JOINT ENTERPRISE, the

“purposive” dimension, and SHARED REPERTOIRE the “instrumental” dimension. The

presence of these three dimensions in a group is a necessary and sufficient condition for

the existence of a CoP, which is why their empirical detection will be a key concern of

the Research Strategy. Moreover, assuming a CoP exists, two other previously

mentioned elements are necessarily implied: a COMMUNITY and, as the end result,

LEARNING/IDENTITY ACQUISITION. Together these five elements form a set of

observable properties or traits that identify a group as a functioning CoP, as illustrated

in Figure 2.2. Because this research addresses the problem of detecting and positively

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
29
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

identifying Internet-based CoPs, it will place considerable emphasis on these

observable traits, which will be referred to as Wenger’s constructs.

MUTUAL ENGAGEMENT LEARNING/IDENTITY


ACQUISITION

SHARED REPERTOIRE Community


of Practice

JOINT ENTERPRISE COMMUNITY

FIGURE 2.2 – Observable Traits of a CoP

Some authors point out that Wenger (1998) does not propose a formal definition of

CoP (Barab, MaKinster and Scheckler, 2003; Schwen and Hara, 2003). The closest he

comes to it, is the following:

Being alive as human beings means that we are constantly engaged in the

pursuit of enterprises of all kinds, from ensuring our physical survival to

seeking the most lofty pleasures. As we define these enterprises and engage in

their pursuit together, we interact with each other and with the world, and we

tune our relations with each other and with the world accordingly. In other

words, we learn. Over time, this collective learning results in practices that

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
30
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

reflect both the pursuit of our enterprises and the attendant social relations.

These practices are thus the property of a kind of community created over time

by the pursuit of a shared enterprise. It makes sense, therefore, to call these

kinds of communities communities of practice (Wenger, 1998: 45).

Because this rather lengthy description does not easily lend itself to the building of a

model, this thesis will operationalise Wenger’s theory by focusing on the constructs;

hence, the following operational definition of a CoP, real or virtual:

Operational Definition of a Community of Practice: A Community of

Practice is a stable group exhibiting all of Wenger’s constructs, i.e. MUTUAL

ENGAGEMENT, SHARED REPERTOIRE, JOINT ENTERPRISE, COMMUNITY and

LEARNING/IDENTITY ACQUISITION.

The observable traits included in the definition will be referred to as the Essential

Traits, and play a key role in the construction of the model of Internet-based CoP

described in Chapter Four. Of course, the definition begs a discussion and operational

definition of each of Wenger’s constructs. This is addressed in the following sub-

sections.

2.3.1 – MUTUAL ENGAGEMENT

Wenger considers MUTUAL ENGAGEMENT nothing less than the root cause of a CoP:

Although the concept of engagement need not be defined in terms of

specifiable communities of practice, the two concepts are closely linked

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
31
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

since mutual engagement will give rise to communities of practice over time

(Wenger, 1998: 174).

Membership in a community of practice is therefore a matter of mutual

engagement. That is what defines the community (ibid: 73).

Thus MUTUAL ENGAGEMENT is arguably the most important among Wenger’s

constructs, to the extent that the aim of this research could be re-phrased as a search for

actual instances of sustained MUTUAL ENGAGEMENT taking place over the Internet,

followed by the analysis of the COMMUNITIES and SHARED REPERTOIRES produced by

this ENGAGEMENT.

Still, Wenger does not give a precise definition of MUTUAL ENGAGEMENT, only a broad

description. The ENGAGEMENT of members of a CoP involves doing things together,

talking, producing artifacts, and solving problems (Wenger, 2000b). In general, the

construct refers to the interactions members of a CoP have with each other. Potential

manifestations would be:

• collective problem-solving

• sharing practice-related information

• debating practice-related issues

Since a wide variety of interactions may qualify as valid instances of MUTUAL

ENGAGEMENT, the list above does not aim to be exhaustive, only representative. These

are specific and observable activities which indicate the presence of the ENGAGEMENT

construct. Some are fairly fundamental, such as collective problem-solving or


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
32
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

debating issues, because they presupose a specific discipline or domain of knowledge,

plus the ability (and the need) of members to effectively address the problems and

challenges of this domain in a collaborative effort. In particular, the task-oriented

nature of collective problem-solving clearly differentiates CoPs from other types of

communities, such as fan clubs, or friendship groups. Communities that merely talk

about a shared interest, or whose primary purpose is socialisation, are not CoPs. CoPs

get things done.

Wenger (1998) argues the scope of MUTUAL ENGAGEMENT is limited, as is the number of

members that can have direct ENGAGEMENT with each other, which places a limit on the

size of a CoP. While technological developments, such as the Internet, have the potential

to expand the limits of the scope of ENGAGEMENT, these developments involve trade-offs

in the complexity of ENGAGEMENT that can be supported (ibid).

In addition, MUTUAL ENGAGEMENT admits different levels of complexity and/or

intensity which results in multiple levels of community involvement and membership.

From this perspective, a community of practice is a node of mutual engagement

that becomes progressively looser at the periphery, with layers going from core

membership to extreme peripherality. The interactions of all these levels affords

multiple and diverse opportunities for learning. Different participants

contribute and benefit differently, depending on their relations to the enterprise

and the community (Wenger, 1998: 118).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
33
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.3.2 – JOINT ENTERPRISE

Again Wenger (1998) does not offer an explicit definition; still, he often uses as

equivalent the concept of domain, particularly in practitioner-oriented articles. This

domain is defined as:

the area of knowledge that brings the community together, gives it its identity,

and defines the key issues that members need to address. A community of practice

is not just a personal network: it is about something. Its identity is defined not just

by a task, as it would be for a team, but by an "area" of knowledge that needs to

be explored and developed (Wenger, 2004b: 3).

Even though the two terms are equivalent, the earlier name has advantages. The term

“JOINT ENTERPRISE” highlights the fact that the topic of a CoP is something collectively

negotiated by its members and to which they hold each other accountable. In a very real

sense their ENTERPRISE is “created” by them and “belongs” to them, and it does not

necessarily coincide with a pre-existing topic or discipline in which the CoP itself

might be embedded. For instance, in his ethnography of claims processors, Wenger

(1998) notes that the JOINT ENTERPRISE of the unit he studied was not just to effectively

process claims, but also whatever it took to create an acceptable atmosphere to work in.

In fact, members held each other more stringently accountable to the latter goal than to

the former.

CoPs can be defined by disciplines, problems or by situations (Wenger, 2004b), but

their ENTERPRISE is as indigenous as their practice, i.e. it is always a collectively

negotiated response of a particular group of people to the conditions and challenges

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
34
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

they face. Furthermore, it is a dynamic response, as their understanding of what their

ENTERPRISE is about is continually negotiated. Defining a JOINT ENTERPRISE is thus an

ongoing process, not a static agreement. It is neither a charter nor a statement of

purpose, although these artifacts might exist and give clues as to what the JOINT

ENTERPRISE might be.

JOINT ENTERPRISE is visibly manifested by the community caring about an identifiable

domain of knowledge.

2.3.3 – SHARED REPERTOIRE

Over time, the ENGAGEMENT of members results in the development of a set of

communal resources that allow members to more effectively pursue their ENTERPRISE.

These resources include routines, tools, ways of doing things, stories, symbols, and

artifacts that the community has produced or adopted in the course of its existence, and

which have become part of its practice. They can be very heterogeneous: they gain their

coherence from the fact they belong to the practice of a COMMUNITY pursuing a

negotiated ENTERPRISE (Wenger, 1998).

Sometimes Wenger uses the terms SHARED REPERTOIRE and practice interchangeably,

particularly in practitioner-oriented writings (e.g. Wenger et al, 2002). More precisely,

though, SHARED REPERTOIRE is the instrumental dimension of the community’s

practice, whereas the practice itself is the holistic, emergent and indigenous response of

the community to the challenge of its ENTERPRISE (Wenger, 1998). The distinction is

important, because specific practices (as in “best practices”) can be part of a SHARED

REPERTOIRE. Representative manifestations of SHARED REPERTOIRE include:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
35
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• shared artifacts

• shared criteria

• shared practices

2.3.4 – COMMUNITY

Even though the COMMUNITY construct is the most immediately visible trait, it is not

conceptually the first. Wenger sees MUTUAL ENGAGEMENT as actually preceding the

formation of a CoP, which is why COMMUNITY is not a constitutive element on the

same level as the previous three. Prior to MUTUAL ENGAGEMENT there is just a group of

people, maybe even having similiar jobs or interests (e.g., an occupational community),

but who have not yet sustained sufficient interactions to develop an indigenous

practice.

Wenger (1998) emphasises the term COMMUNITY should not be romanticised. MUTUAL

ENGAGEMENT creates strong personal interrelationships, but these need not be

harmonious for a CoP to exist, particularly when people are compelled to interact

because of an outside mandate, as in the case of the claims processors. In voluntary

CoPs, participant “chemistry” plays a bigger role, but they are not immune to conflict

either, and may even find it beneficial (Wenger et al, 2002).

Visible manifestations of COMMUNITY include:

• members’ knowledge of each other

• shared sense of community

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
36
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.3.5 – LEARNING

Wenger’s complex notion of LEARNING was described in the previous section (see

Figure 2.1, p. 27). This construct is somewhat different from the others because it

happens in people’s heads and is not as visible as say MUTUAL ENGAGEMENT. Still,

suitable research instruments can be devised. People form CoPs to learn together about

topics that interest them. Thus members of a CoP should, if asked, report LEARNING as

a primary result of participating. Alternatively, LEARNING should also be in evidence if

members’ interactions are recorded and qualitatively analysed. Observable

manifestations of LEARNING include:

• acquiring new knowledge

• acquiring new skills

• acquiring/enacting a professional IDENTITY

This last manifestation connects with the remaining construct.

2.3.6 – IDENTITY ACQUISITION

The kind of personal investment required to belong to a CoP results in people’s

IDENTITIES being engaged, reaffirmed, and ultimately shaped by such belonging. It is

no exageration to say that a person chooses to join a CoP literally to be someone.

We identify with some communities strongly and not at all with others. We

define who we are by what is familiar and what is foreign, by what we need to

know and what we can safely ignore, […] by the communities we do not belong

to as well as by the ones we do (Wenger, 2000b: 239).


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
37
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

As noted earlier, Wenger considers IDENTITY ACQUISITION as another way of referring

to LEARNING, specifically, “learning as becoming”. Thus IDENTITY ACQUISITION is not

really an independent construct, but only a different manifestation of LEARNING. Yet,

because participants in the study may perceive identifying with a community as a

different matter from learning, the two will be addressed separately by the research

instruments. A visible manifestation of IDENTITY ACQUISITION is discussion of career

issues.

This concludes the discussion of Wenger’s (1998) CoP theory. The concept has been

used widely, yet most authors that cite Wenger use a condensed definition, with little or

no mention of the constructs (e.g. O’Dell and Grayson, 1998; Davenport and Prusak,

1998; Lesser and Everest, 2001; Cohen and Prusak, 2001; Brown and Duguid, 1998;

2001). There are also authors who have challenged Wenger’s view, or have introduced

different terminology. These are reviewed next.

2.4 – Challenges to Wenger’s theory of CoPs

Before addressing authors who challenge Wenger’s theory directly, it is interesting to

check whether the social phenomenon he describes has been previously studied under a

different guise. This would not be surprising, since Wenger (1999) notes CoPs are a

natural human phenomenon, pointing to prehistoric tribes and medieval guilds as

historic examples. Figure 2.3 provides a timeline of significant references about CoPs

and similar social structures.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
38
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Constant (1980)
1980
(communities of technological practitioners)
1981

1982

1983
van Maanen and Barley (1984) 1984
(occupational communities)
1985

1986

1987

1988

1989

Orr (1990) 1990

Lave and Wenger (1991); Brown and Duguid (1991) 1991

1992
Cook and Yanow (1993) Trice (1993)
1993
(organisational culture) (occupational subcultures)
1994
Boland and Tenkasi (1995) Brown and Gray (1995)
1995
(communities of knowledge)
Manville and Foote (1996); Orr (1996); Stewart (1996) 1996

1997
Wenger (1998); Brown (1998); Brown and Duguid (1998)
1998
Gherardi, Nicolini and Odella (1998)
Wenger (1999); McDermott (1999a) 1999
Wenger (2000a; 2000b) Brown and Duguid (2000a; 2000b)
2000
Wenger and Snyder (2000) (networks of practice)
Tsoukas and Vladimiriou (2001) Brown and Duguid (2001) 2001

Wenger, McDermott and Snyder (2002) 2002

2003

Wenger (2004a; 2004b) 2004

Figure 2.3 – Timeline of significant references in CoP theory development

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
39
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.4.1 – Alternative terminology for alternative but related phenomena

Brown and Duguid (2001) credit Edward Constant with independently arriving at the

CoP concept before Lave and Wenger’s (1991) research. However, Constant’s (1980)

concept of communities of technological practitioners is actually closer to Brown and

Duguid’s own concept of Network of Practice (which is wholly defined by a shared

practice), than to Wenger’s concept of CoP (which is defined by direct ENGAGEMENT).

Constant’s aim is to build a Kuhnian model to explain a technological revolution,

specifically the advent of turbojets. Yet his communities are defined as people sharing

a narrow technical specialty, rather than people sustaining direct ENGAGEMENT.

Utilization of a community of practitioners as a primary unit of historical

analysis nevertheless does promise to generate basic insights for the history of

technology. A community of technological practitioners, moreover, may be

analyzed in turn as an aggregation of individuals or of firms, just as a

scientific specialty may be analyzed as an aggregation of individuals or of labs

and departments (Constant, 1980: 9, emphasis added).

Van Maanen and Barley’s (1984) notion of occupational community, designates

another similar phenomenon, defined as:

a group of people who consider themselves to be engaged in the same sort of

work; who identify (more or less positively) with their work; who share a set of

values, norms, and perspectives that apply to, but extend beyond work related

matters; and whose social relationships meld the realms of work and leisure

(1984: 295).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
40
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Crucially absent from this definition is any mention of MUTUAL ENGAGEMENT, which

for Wenger (1998) is the only necessary condition for the existence of a CoP,

irrespective of members’ occupations. Van Maanen and Barley do not require members

of occupational communities to actually interact, only to have the same occupation.

The two concepts are clearly different: a CoP, whose members do not necessarily have

the same occupation, is not a special or “local” case of an occupational community.

The concept of occupational subculture is another that superficially bears some

resemblance to the concept of CoP. According to Trice (1993), occupational

subcultures comprise unique clusters of ideologies, beliefs, cultural forms, and

practices that arise from shared educational, personal and work experiences of

individuals who pursue the same profession within the overarching organisational

culture of a single workplace. However, Trice’s concept, like Constant’s and Van

Maanen and Barley’s, does not require or guarantee that members of these groups

actually ENGAGE with each other. Therefore, though they may share a profession, they

cannot cohere into a CoP without regular ENGAGEMENT, although they would qualify

for Brown and Duguid’s (2001) concept of Network of Practice, which is examined in

the next chapter.

Knorr Cetina’s (1999) concept of epistemic culture is described by Brown and Duguid

(2001) as equivalent to a Network of Practice. Like Wenger, Knorr Cetina does not

provide a definition, only a rich description. Brown and Duguid (2001: 205) point out

Knorr Cetina’s concept makes no distinction between groups of scientists working

closely together on a regular basis, and same-discipline scientists who rarely meet or

know each other. Hence, a local portion of an epistemic culture may qualify as a CoP,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
41
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

but not the complete culture. As before, the crucial distinction is that a CoP is defined

by direct ENGAGEMENT, while both NoP and epistemic culture require only a common

practice, however specialised.

Finally, Brown and Duguid (2001) have proposed the similar concept of NoP to explain

the “leaky” and “sticky” properties of knowledge. However, because of its close

relationship to “virtual” CoPs, discussion of the concept of NoP will be deferred until

the next chapter (Section 3.3).

2.4.2 – Alternative terminology for the same phenomenon

Various researchers consider Orr’s (1990) ethnography of “tech reps” as the earliest

published study of a CoP (Raelin, 1997; Brown and Duguid, 2001; Teigland, 2003). Yet

Orr (1996) used the earlier concept of occupational community, which is inadequate

because it cannot distinguish the properties of the small, tight-knit community of tech

reps, who ENGAGE directly on a regular basis, from those of the larger technician force of

Xerox, who rarely if ever ENGAGE directly.

Cook and Yanow’s (1993) case study of three Boston workshops that produce some of the

world’s finest concert flutes has been cited by Stamps (1997) and Wenger et al. (2002) as

an example of CoP. The article presents a vivid description of activities at these workshops,

an account where all the constitutive elements of a CoP can be easily identified. Cook and

Yanow do not give a name to the group formed by flute craftsmen, and the intent of the

article was not to depict these workshops as CoPs. Rather, the authors were trying to

show how an organisation, as opposed to an individual, can be said to ‘learn’ by seeing

it as a system of shared meanings associated with and carried out through cultural

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
42
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

artifacts. They use the term organisational culture to refer to a quality that would allow

an organisation to ‘learn’ in a similar way a brain would allow an individual to learn. The

article has some limitations: it is based on a study of very small organisations with a

homogeneous culture of craftsmanship (even the owners of the workshops were

craftsmen, and held a position in the production line). The authors use the term

organisation indistinctly to refer both to these small workshops and to large corporations

such as IBM and Saab. Furthermore, they use the term organisational knowing to

indistinctly refer to the ability of a succesful basketball team to play the game, and the

ability of Saab to make a car, thus establishing a problematic equivalence between two

very different abilities. Cook and Yanow base their argument on the human ability to act

in groups and to share a culture. However, they do not address the complexities that scale

brings to such collective action. Thus their argument about group learning only seems

conclusive at the level of a homogeneous CoP, which in their research coincided, quite

exceptionally, with a full organisation. The generalisation to larger organisations

comprising a multiplicity of heterogeneous CoPs is not straightforward, nor is this issue

convincingly addressed in a later commentary by one of the authors (Yanow, 2000). In

fact, Brown and Duguid (1998: 98) argue that “organising knowledge across hybrid

communities [of practice] is the essential activity of organisational management”, an

activity that is far from trivial.

Boland and Tenkasi (1995) explicitly use the concept of CoP, adding nuances of their

own. They propose the term community of knowledge to describe communities of

specialised or expert knowledge workers in knowledge-intensive firms. They

acknowledge their concept is congruent with Lave and Wenger’s (1991) CoP even if

restricted to expert communities in knowledge-intensive firms. Also, they use the term

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
43
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

perspective where Wenger would use practice or SHARED REPERTOIRE. Their research is

about the requirements of communication systems that can adequately support

perspective-making and perspective-taking, what Wenger would designate MUTUAL

ENGAGEMENT, developing a SHARED REPERTOIRE and acquiring an IDENTITY. Thus

Boland and Tenkasi’s “community of knowledge” might be better designated a CoP of

experts, to avoid the unnecessary introduction of a new concept into the literature.

Some authors have differentiated themselves from Wenger by making absence of

management direction or support a necessary condition for a true CoP, and proposing

new names for groups that display all the properties of CoPs, yet develop with

management’s blessing.

For instance, Büchel and Raub (2002) introduced the concept of “knowledge networks”

which they argue extends beyond the traditional concept of CoP. The authors propose

four variants of this concept: “hobby” networks, “professional learning” network,

“business opportunity” network and “best practices” network. They argue only the first

two conform to the traditional concept of CoP, but it is the last two that can result in

organisational benefits. Their position is that once a CoP receives management support

it ceases to be “informal” and “voluntary”, and therefore ceases to be a CoP.

Similarly, Barret, Cappleman, Shoib and Walsham (2004: 1) establish a distinction

between communities “which are […] voluntary in terms of participation, and those

with a more managed membership”, with only the former being considered CoPs. They

use the umbrella term “knowledge communities” to cover voluntary and managed

communities, and they also include constellations of CoPs (Wenger, 1998).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
44
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Stork and Hill (2000) also present informality as a necessary condition of a CoP. They

describe a community of Information Technology (IT) managers at Xerox, which

began with an initial roundtable 2-day meeting, decided to meet again in two-months,

and thereafter met every six weeks. They took the name Transition Alliance, and

agreed their domain was to orchestate a major transition from Xerox’s proprietary IT to

more open industry standards. Senior management fostered the launch of the Alliance,

and was supportive of it, but did not try to control it or request deliverables. For

instance, attendance to meetings was not mandatory. Still, Stork and Hill argue that

because the group was deliberately established by senior management, it did not qualify

as a CoP, but as a new organizational group they labelled a “strategic community”.

However, the large amount of freedom this group enjoyed from the start, the fact that

members all shared the same practice and were all stakeholders in the IT transition, and

the crucial fact that they interacted regularly both in and between meetings indicates

that, from Wenger’s perspective, the group did evolve into a true CoP.

These various authors’ insistence on informality as a defining condition of a true CoP is

contradicted by Wenger’s (1998) study of claims processors, which showed

ENGAGEMENT to be the essential condition, not informality. Any workgroup ENGAGED

in a specific domain of knowledge will over time evolve into a CoP; that is, it will

develop an indigenous practice that allows it to get the job done, even if the workgroup

is formally established by management. Still, because a CoP defines itself through

ENGAGEMENT, its boundaries will not necessarily match institutional boundaries,

because membership is not defined by institutional categories. It is in this sense, that

Wenger describes CoPs as essentially informal, but he explicitly rejects the view that

CoPs can never have a formal status (Wenger, 1998).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
45
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Further evidence against informality as a defining condition of a CoP is provided by

several cases of strategically-important CoPs supported, and even started (in a socially

sensitive way) by management, such as the Eureka knowledge repository at Xerox

which evolved directly from Orr’s ethnography of service reps (Brown and Duguid,

2000b), and the PLL-design CoP at National Semiconductor described by Brown and

Grey (1995). Swan, Scarbrough and Robertson (2002) provide another example, a case

study of a successful CoP convened by administrators of a health organisation as a

vehicle for a radical innovation in the treatment of prostate cancer known as

brachytherapy. These managers envisioned their role as facilitating the construction of

a new multidisciplinary community engaged around brachytherapy practice. They

explicitly addressed constitutive dimensions of CoPs such as a well-defined domain of

knowledge, identity enhancement, networking, and knowledge brokering. Thus, even

though managers were relatively powerless before established medical professionals,

they were able through indirect means to effect important changes.

These studies support Wenger’s position that CoPs can assume “knowledge stewarding”

responsibilities in organisations if management is socially sensitive, and is careful not to

stiffle their self-organising drive (Wenger, 1999; 2000a). Intra-organisational CoPs are

ubiquitous, a consequence of ENGAGEMENT being their root cause (Wenger, 1998). The

problem is that not all CoPs are equally relevant to managers; most are only important to

members, helping them to cope with a particular class of problems at work. Much more

exceptional are CoPs with the potential to have a strategic impact on the business, whose

main interest is aligned with managers’ concerns, and who are actually recognised and

supported. The KM agenda of detecting or “launching” such CoPs has spawned much

writing, which has often been more popularising than rigorous.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
46
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.4.3 – Critiques of Wenger’s (1998) theory

Gherardi, Nicolini and Odella (1998) reject the view of CoPs as a community with

defined boundaries, established behavioural rules and canons. They argue CoPs are just

one of the forms of organising, specifically, organising for the execution and

perpetuation of a practice. Hence, they intentionally emphasise the term ‘practice’ over

‘community’.

In other words, referring to a community of practice is not a way to postulate

the existence of a new informal grouping or social system within the

organization, but is a way to emphasise that every practice is dependent on

social processes through which it is sustained and perpetuated, and that

learning takes place through the engagement in that practice (1998: 279).

It is possible that Gherardi et al (1998) are taking exception not with Wenger directly but

with popularising authors who emphasise the traditionally harmonious connotation of the

term community. In any case, Wenger’s (1998) position is that a stable group really does

exist, which is held together by sustained ENGAGEMENT, and does not require or produce

harmonious personal relationships. As to the name of this stable group, Lave and Wenger

(1991) picked the term community, but Brown and Duguid (2001) remark they could

have chosen cadre or commune just as well.

Contu and Willmott (2000; 2003) have critiqued Wenger for what they see as a shift

from an emancipatory discourse in his seminal 1991 book with Lave, to a discourse of

performance in his 1998 ethnography of claims processors and later publications.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
47
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Similarly, Fox (2000) and Marshall and Rollinson (2004) critique the cursory treatment

of power issues in Wenger (1998).

In point of fact, Wenger argues CoPs cannot be “managed” in the usual sense of the

word; they can be manipulated or coerced into submission, but “managing” the practice

of a CoP, in the narrow sense of exercising control over it, is not possible:

[T]he power –benevolent or malevolent– that institutions, prescriptions or

individuals have over the practice of a community is always mediated by the

community’s production of its practice. External forces have no direct power

over this production, because in the last analysis (i.e., in the doing through

mutual engagement in practice), it is the community that negotiates its

enterprise (Wenger, 1998: 80).

The issue has been empirically addressed by Thompson (2005), who used participant

observation and interviews to study a co-located CoP at a large IT hardware and services

firm. The 40-member group was formally established as a creative Web-design agency,

exempt from the commercial and procedural restrictions of the parent organisation. It

enjoyed heavy corporate sponsorship of IT infrastructure and culturally symbolic artifacts

(pool tables, video games, bean bags, etc.) conducive to a relaxed, informal and creative

work environment. The author reports strong group identification and epistemic

interaction, relying on Wenger’s (1998) framework to assess the emergent tight-knit

group as a CoP. However, the organisation tried to capitalise on the group’s success with

the addition of 140 new in-training participants, which required formal documentation of

procedures (hitherto unnecessary because of the group’s small size), and other

prescriptive measures such as controls on billable vs non-billable activities. This brought

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
48
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

about the demise of the original CoP, as members quickly withdrew identification and

commitment from the new, more formalised structure. These findings confirm Wenger’s

(1998) position that CoPs can be supported or “nurtured” but not controlled.

Wenger’s treatment of power issues is, in fact, consistent with his broader theoretical

framework. He sees CoPs as wielding power because it is they that socially-define

competence and identities (Wenger, 2000b). Thus, a person decides whether he or she

wants to belong to a particular community (i.e. learn its practice), but has a limited

capacity as an outsider (or even as full member) to change the practice of the

community. On the other hand, a CoP is powerless before an individual who does not

recognise its authority:

In other words, being recognized as competent only matters to the extent that

one identifies with the communities that can confer legitimacy to learning

(Wenger, 2004a: 12).

It ought to be noted that this competence-defining role of CoPs is also the source of

their greatest weakness: the danger of becoming insular (Wenger, 2000b), and losing

touch with the broader organisation and market environment (Thompson, 2005).

Finally, a recent critique by Lindkvist (2005) misreads Wenger by attributing to him

the position that practices produce CoPs:

The recent, elaborate work of Wenger (1998), one of the two inventors of the

term, should constitute an authoritative source to focus on. A first observation

is then that [CoPs] refer to practices that give rise to mutual engagement,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
49
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

joint enterprise and a shared repertoire (Wenger, 1998, pp. 73ff.). If this

happens we may speak about the kind of practices that deserve the additional

‘community’ designation. This raises the question of what connotations

previous usage of the term community brings to the [CoP] notion and to what

extent these are still informative. Moreover, how do we distinguish between

empirical practices that are likely to produce such community and those that

are not? (Lindkvist, 2005: 1192).

For Wenger (1998), though, practice is the indigenous ad-hoc response of a CoP to the

challenges of its ENTERPRISE, and it is produced collectively by the members of the

community through MUTUAL ENGAGEMENT. Thus, practices do not produce CoPs, it is

CoPs that produce practices, which is why Wenger views all practice as social practice

(and, cannot, therefore, conceive an “empirical practice” that is not the product of one

or more pre-existing CoPs).

Wenger’s (1998) theory thus seems to give coherent replies to the various critiques it has

received. Moreover, there is no competing theory: no other author has developed a

complete and detailed theoretical framework to explain the readily acknowledged social

phenomenon of CoPs. This section reviewed distinctly-named similar phenomena, and

alternative designations for the same phenomenon. The literature affords no compelling

reasons for using a different concept, or adopting the perspective of a different author.

Therefore Wenger’s (1998) theory of CoPs is selected as the foundation of this research.

The next chapter will examine the issues involved in extending the theory to the social

areas of the Internet.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
50
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

2.5 – Summary

The aims of this chapter were to formally introduce the concept of CoP, critically review

existing theories, and propose an operational definition for use in this thesis. The chapter

opened with a historical overview, followed by a description of Wenger’s social theory of

learning, of which CoPs are but one element. The chapter then focused on Wenger’s

(1998) theory of CoPs, with a strong emphasis on their constitutive elements. The last

section reviewed extant challenges to Wenger’s theory, and showed they are coherently

addressed by his theoretical framework.

As a consequence of the preceding discussion, Wenger’s (1998) theory of CoPs has

been identified as the most detailed and consistent currently available. No other

author matches the rich description Wenger offers of the constitutive elements of

CoPs: MUTUAL ENGAGEMENT, SHARED REPERTOIRE, JOINT ENTERPRISE, COMMUNITY,

and LEARNING/ IDENTITY ACQUISITION. Hence, the decision to adopt Wenger’s model

in this research. Furthermore, an operational definition of CoP was proposed, using

Wenger’s constructs as the elements that determine whether a particular group, real or

virtual, can be assessed as a CoP. Each of Wenger’s constructs was discussed and

visible manifestations were specified in order to build operational definitions. These

results will contribute to the model of Internet-based CoP that will be proposed in

Chapter Four. First, though, Chapter Three will critically examine the popular notion

of a “virtual” CoP.

One last observation about CoP literature: it was found to be roughly aligned along two

distinct groups, which might be labelled the “Organisational approach” and the “KM

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
51
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

approach”. The former emphasises theory development by describing pre-existing,

informal organisational CoPs; the latter emphasises the business value of CoPs, and

aims to identify, support, and launch organisational CoPs for KM purposes. The two

approaches are displayed in Figure 2.4 which contrasts their different positions on key

CoP issues through several representative references.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
52
Previous Research on Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

“Organisational” Approach “KM” Approach

These references view CoPs as emergent, self- These references view CoPs as hidden
organising, informal groups who set their own resources that should be identified and
learning agenda, and operate mostly beyond supported by management, and charged with
management control. Hence, the positions they pursuing knowledge initiatives that have
emphasise are: strategic value for the organisation. Hence,
the positions they emphasise are:

CoPs as organisational emergent structures. CoPs as organisational knowledge assets.

Brown and Duguid (1991); Brown (1998); Brown and Gray (1995); Stewart (1997);
Wenger (1998); Gherardi and Nicolini (2000) Hanley (1998); McDermott (1999b);
Wenger (1999; 2000a); Wenger et al (2002)
All competencies of the organisation reside in Core competencies of the organisation
CoPs. reside in CoPs.

Brown and Duguid (1991); Wenger (1998; Brown and Gray (1995); Manville and
2000a); Tsoukas and Vladimirou (2001) Foote (1996); Wenger (1999);
Saint-Onge and Wallace (2002)
CoPs cannot be managed conventionally. CoPs, as assets, should be managed.

Brown and Duguid (1991); Wenger (1998); Stewart (1997); Stork and Hill (2000);
Wenger and Snyder (2000); Thompson (2005) Barrow (2001); Büchel and Raub (2002);
Swan, Scarbrough and Robertson (2002)
A focus on “mundane” or “ordinary” CoPs. A focus on “strategically important” CoPs.

Orr (1990); Wenger (1998); Gherardi and Brown and Gray (1995); Stewart (1997);
Nicolini (2000); Tsoukas and Vladimirou (2001) Wenger (1999; 2004b); Büchel and
Raub (2002); Stork and Hill (2000); Swan et
al (2002); Saint-Onge and Wallace (2002)
A focus on description. A focus on design.

Orr (1990); Wenger (1998); Gherardi and McDermott (1999b; 2000); Wenger et al
Nicolini (2000; 2002) (2002); Saint-Onge and Wallace (2002)
A critical emancipatory perspective. A managerial results-oriented perspective.

Gherardi, Nicolini and Odella (1998); Hanley (1998); McDermott (1999a);


Contu and Willmot (2000; 2003); Fox (2000) Wenger (1999; 2004b); Stork and Hill (2000);
Lesser and Storck (2001); Barrow (2001);
Wenger et al (2002)
Social Learning. Managing Knowledge.

Lave and Wenger (1991); Raelin (1997); Stewart (1997); McDermott (1999a);
Gherardi, Nicolini and Odella (1998); Wenger (1999; 2004b); Barrow (2001);
Wenger (1998; 2000b; 2004a); Brown and Saint-Onge and Wallace (2002);
Duguid (2001); Tsoukas and Vladimirou (2001) Wenger et al (2002)

Figure 2.4 – Contrasting “Organisational” and “KM” CoP references

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
53
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CHAPTER THREE

Virtual Communities of Practice

The previous chapter reviewed theories of conventional co-located CoPs. This chapter

will examine the theoretical possibility of “virtual” CoPs, and review previous studies

that directly or indirectly address the topic. The review aims to establish both the

research need and the originality of this study. In addition, extant research provides

relevant lessons that will be applied to the construction of the model of virtual CoP that

will provide guidance and focus to this thesis.

The chapter is organised into four sections. The First examines the notion of “virtual”

in connection to CoPs by using three variables of network design and utilisation to

classify the different varieties of “virtual” found in the literature. Section Two

reviews previous articles about “virtual” CoPs to determine whether Wenger’s

theoretical constructs are explicitly or implicitly reported, and thereby establishes the

lack of studies that fully apply Wenger’s theory. Section Three examines theoretical

and empirical critiques of the notion of virtual CoP, most particularly Brown and

Duguid’s concept of Network of Practice (NoP). The final section is a chapter

summary.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
54
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

3.1 – Reviewing the varieties of “virtual” in the literature

Given the exploding popularity of the CoP concept, it soon followed that theorists,

practitioners and business writers began to explore the notion of a “virtual” CoP, i.e. a

community whose interaction was carried out through electronic networks, particularly

the Internet. “Virtual” CoPs are currently a popular topic. This can be seen in the graph

displayed in Figure 3.1, which plots the number of articles published yearly in the (non-

overlapping) Proquest and IEEExplore databases, which include in the body of the

article the search strings “virtual communities of practice” OR “online communities of

practice” OR “electronic communities of practice”.

50
43
39
40

30 28

20

10
10
5 5
2 1
0
1998 1999 2000 2001 2002 2003 2004 2005

Figure 3.1 – Articles about “virtual” CoPs published in Proquest + IEEE databases

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
55
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The increase in publications partly reflects the interest among KM practitioners who

nowadays consider CoP support and/or development an important element of

organisational KM programmes. For instance, professional KM conferences usually

include a seminar or workshop on launching and supporting “virtual” CoPs.

There are two problems with this current enthusiasm. The first lies in the ambiguity

of the over-used term “virtual”. For instance, Kimble, Hildreth and Wright (2001),

Baym (2000), and Saint-Onge and Wallace (2002) provide three different accounts of

succesful “virtual” CoPs. Yet, the first applies the term “virtual” to the work unit

formed by two co-located workgroups of the same company, one in the US, the other

in the UK, who communicate using phone, e-mail and video link; the second applies

the term to a stable, persistent, extra-organisational Usenet community; the third calls

“virtual” a CoP based on a proprietary intranet, designed to enable field interaction of

insurance agents from a Canadian company. The only common denominator is that

day-to-day interactions take place without co-location using electronic media.

However, as the example shows, the media used to support these virtual groups are

highly diverse. This suggests developing at least an elementary typology –the task of

this section– to classify previous studies about “virtual” CoPs, and clearly identify the

type this research will examine.

The second problem is an intuitive use of the concept of community of practice,

which suggests few of the “virtual” CoPs described in the literature are true CoPs as

defined and described by Wenger. As the previous chapter made clear, Wenger

emphasises that regular MUTUAL ENGAGEMENT between members is the key

requirement for the coalescing and continued existence of a CoP, and his theoretical

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
56
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

framework is the result of detailed examination of ENGAGEMENT in co-located CoPs

(Lave and Wenger, 1991; Wenger, 1998). Yet, “virtual” implies interaction without

co-location, which raises a theoretical issue few studies address. Specifically, the

literature review found few studies that make MUTUAL ENGAGEMENT an important or

necessary condition for the continued existence of the community. Of 20 “virtual”

CoP studies included in this review, only three (Saint-Onge and Wallace, 2002;

DeSanctis et al, 2003; Teigland and Wasko, 2004) recognised the role of

ENGAGEMENT. This renders problematic the intuitive characterisation as CoPs of

some virtual groups in the literature, an issue which is more closely examined in the

next section.

Within the “virtual” CoP literature, there are substantial differences in the

characteristics of described electronic networks and in the circumstances of the groups

using them. These differences led to the identification of three contingent factors of

network design and/or utilisation: Network access, Media richness and Participatory

character; these are discussed next.

3.1.1 – Network access

With respect to access, electronic networks are basically of three kinds. The most

restricted networks are proprietary information systems, developed and managed by

corporations to support information-sharing among employees in a secure environment.

Second are subscription-based services, such as private discussion forums set up by

various professional societies, which can be accessed though the Internet but require a

login and password. Finally there are fully open networks such as the Internet and

Usenet. Proprietary networks are more reliable and secure, but connect fewer people.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
57
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Public networks can reach a much greater number of more diverse people, but pose

greater privacy and security risks and have lower signal-to-noise ratios (Kollock and

Smith, 1996; Hahn, 2000). The advantage of networks that reach a larger number of

more heterogeneous people is that they are more likely to contain natural “organisers”

who are capable of mobilising other network participants to achieve collective action

(Oliver and Marwell, 1988).

Network access is important because unrestricted access can make topical-focus more

difficult for network-based communities. Usenet-based communities, for instance,

oftentimes must invest a considerable amount of time and energy to preserve the focus

of the discussion in the face of such Usenet threats as trolls and spam (Kollock and

Smith, 1996; Kling and Courtright, 2003). Moderated newsgroups offer an interesting

variant in terms of network access; though discussions are still conducted in the open,

posting to the newsgroup must be done through a moderator who filters messages

whose topic or tone are inappropriate to the charter of the newsgroup.

3.1.2 – Media richness

The communication capabilities of various kinds of networks offer a range that can go

from plain text messages to videoconferencing. According to media richness theory

(Daft and Lengel, 1986), higher medium bandwith can support interactions requiring

more complex communication. This theory would predict that videoconferencing is a

better medium than plain text messages for supporting complex distributed

ENGAGEMENT. However, recent studies have not supported the central proposition of

media richness theory, which is that performance improves when media richness is

matched to task equivocality (Dennis and Kinney, 1998; Kock, 2001).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
58
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Carlson and Zmud (1999) have proposed the alternative channel expansion theory, which

suggests certain experiences positively influence how an individual develops richness

perceptions for a given channel. Specifically, experience with the channel, experience

with the messaging topic, experience with the organisational context and experience

with communication co-participants. In two studies including a cross-sectional and a

multi-wave survey they found empirical support for the communication channel and

communication partner experiences.

Burke et al (2001) performed two longitudinal studies to measure cohesiveness and

process satisfaction among groups supported by media of various richness. The studies

included a total of 238 participants organised in 62 groups, which interacted once a

week over a four-week period to perform a complex collaborative task. Results in both

studies confirmed Daft and Lengel’s (1986) prediction that richer media would initially

result in greater cohesion and process satisfaction. However, both cohesion and

satisfaction exhibited constant increases over time for all media, rich and lean. Thus,

even the leanest media, which in these studies was asynchronous computer-mediated

communication (similar to newsgroups), allowed steady increases in cohesion and

process satisfaction over time, confirming an earlier suggestion by Walther (1992) that

contextual richness will develop in lean environments, perhaps more slowly than in rich

environments, but will develop nonetheless.

Lee (1994) argues from an interpretive perspective that richness is not an invariant

property of a communication medium (his study focused on e-mail) but an emergent

property of the interaction between the medium and its organisational context. This

idea is applied by Robertson, Sørensen and Swan (2001), who present a case study of a

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
59
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

fast-paced consultancy where e-mail played a crucial role in project work. Specifically,

e-mail was used by project leaders preparing client proposals to seek particular in-house

knowledge, while consultants used it to sell their particular expertise, and to record the

transaction when “hired”. Thus, in the shared context of this organisation, simple yet

timely e-mail messages were innovatively used to convey unusually rich meaning.

Finally, there is anecdotal evidence that the ubiquitous Internet, despite its relatively

narrow bandwith, can support MUTUAL ENGAGEMENT in highly complex practices.

For instance, teams of pathologists use a robotic microscope connected to an Internet

server to view and discuss histologic samples in real time for cancer diagnosis (Sacile

et al, 1999). Chin, Myers and Hoyt (2002) describe the Virtual Nuclear Magnetic

Resonance Facility, an Internet-based collaboratory with a suite of tools for

computer-supported collaborative work that allow geographically distributed

scientists to precisely set the controls of remote nuclear magnetic resonance

spectrometers, conduct experiments, analyse results and communicate with other

researchers, all without leaving their homes or offices.

3.1.3 – Participatory character

The usage of electronic networks for documenting valuable knowledge is mandatory at

some workplaces, encouraged at others, and entirely voluntary for individuals outside

of the workplace. This results in different degrees of interest and commitment on the

part of participants.

Whether participation is mandatory or voluntary can affect the quality of ENGAGEMENT.

A mandatory network can compel some MUTUAL ENGAGEMENT and therefore, over

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
60
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

time, a CoP. However, it may turn out to be a rather unmotivated CoP. McDermott

(1999a) gives the example of a large consumer products company where employees

were directed by management to document key work processes in an electronic

database. It took a year to populate the database, but then it was little used. Most people

found it too generic to be useful, and it became an expensive “information junkyard”.

Stewart (2001) provides an interesting counter-example. He notes that at Price

Waterhouse Coopers, consultants widely regard the “Kraken”, a simple Lotus Notes

mailing list, as a superior source of knowledge compared to KnowledgeCurve, the

official state-of-the-art intranet. Yet, the “Kraken” is entirely dependent on the

goodwill of volunteer participants to read and provide answers to questions. This

echoes the experience of many Internet-based communities where high-quality

knowledge-sharing has been observed, even though (or perhaps, because) they are

based entirely on volunteers.

By using these three contingent factors to classify relevant references on virtual

learning communities, a more precise characterisation is achieved and the particular

focus of this thesis becomes clearer. The classification is performed next.

3.2 – Traces of Wenger’s constructs in previous studies of “virtual” CoPs

Studies of “virtual” CoPs comprise a large and heterogeneous collection of articles.

Furthermore, the literature review found several studies that are not about “virtual”

CoPs, and yet shed light on the problem. Therefore, this section will address several

purposes:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
61
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• To assemble a selection of previous studies that have made contributions

to the problem of “virtual” CoPs.

• To use the network typology of the previous section to classify selected

studies.

• To critically review each study, placing special emphasis on the definition

of CoP used, on the evidence (anecdotal or rigorous) of Wenger’s

constructs, and on the type of interactions achieved by the community.

• To build a theoretical case, using previous references, regarding the

feasibility of “virtual” CoPs that exhibit all of Wenger’s constructs.

• To identify gaps and highlight unaddressed issues in the “virtual” CoP

literature.

The first step is to build a comprehensive list of previous empirical studies that have

made contributions to the topic of “virtual” CoPs. As Figure 3.1 showed, the number of

published articles about “virtual” CoPs is large and growing fast, but these articles are

not equally useful or relevant. Thus, the list of included references is the result of

several selection criteria applied to a larger universe of examined studies (by mid-2004,

nearly all papers with the keywords “virtual” and “communities of practice” had been

examined, at least once). Specifically, studies deemed relevant for this research met the

following criteria:

• The study describes a successful learning group whose day-to-day

interaction is carried out through electronic networks.

• Described groups have a professional orientation (i.e. they are not about

hobbies or socialising).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
62
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• Described groups have stable membership and exhibit persistence over

time, thus qualifying as “communities”.

The first criterion needs little comment. The second one does: Wenger et al (2002)

argue that a CoP is more than a community of interest –people who like French movies,

for instance– because members of a CoP get actual work done together, they are

practitioners who, as they engage with each other, create a SHARED REPERTOIRE of

community resources. Thus the numerous studies of virtual communities focused on

hobbies or socialising (e.g. Watson, 1997; Erickson, 1997) were discarded. An

exception was Baym’s (2000) online ethnography of soap opera fans, because it was a

quality piece of research and specifically characterised its community as a Usenet-

based CoP, the same objective as this thesis.

The third criterion led to the exclusion of numerous studies about virtual teams convened

on a temporary basis to accomplish specific projects (e.g. Jarvenpaa and Leidner, 1999;

Rogers, 2000; Chalk, 2001; Smeds and Alvesalo, 2003; Davenport, 2004). The compulsory

nature, professional or academic, of many such projects resulted in impressive online

collaboration, but for a limited time. The disbandment of the team at the end of the project

clearly distinguishes them from persistent communities.

An explicit focus on the topic of “virtual” CoPs was not part of the selection criteria,

because, as previously mentioned, several useful papers were found describing persistent

online communities with various CoP-like traits, and yet not characterised as CoPs. On

the other hand, a few papers that explicitly address the topic of “virtual” CoPs were

discarded because of their intuitive and uncritical use of the concept. For instance,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
63
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Herrmann (1998) states “on-line community” and “on-line community of practice” are

equivalent “as it is assumed that one cannot study a community without focusing on the

practices that constitute it” (1998: 16). Johnson (2001) offers a survey of current articles

about online CoPs, but he uncritically accepts the claims of reviewed studies, about this

or that online community being a CoP. Moreover, by omitting a definition of CoP, the

article mixes real-world with virtual communities, and persistent communities with

temporary classroom teams. By contrast, Dubé, Bourhis and Jacob (2005) offer a robust

survey of 18 intentionally launched virtual CoPs at 14 organisations. Arguing CoPs are

structurally very diverse, they propose a typology of 21 “basic” or “structuring”

characteristics which they use to typify the CoPs in their survey. However, though citing

Wenger et al’s (2002) definition, they do not apply Wenger’s constructs, and thereby

render questionable the CoP-characterisation of some of their groups, particularly the

larger ones (several hundred members) where direct ENGAGEMENT becomes impossible.

The results of the study-selection process are displayed in Table 3.1, which also shows

the classification of each community according to the three-factor typology of network

utilisation proposed in the previous section.

The next step is to review each article, using Wenger’s constructs as the unifying

theoretical lens for examining this heterogeneous collection. Explicit –though rare–

mentions of Wenger’s constructs will be recorded in Table 3.2. So will more indirect

evidence reported in the articles, such as community descriptions revealing aspects that

can be interpreted as the presence of a Wenger construct. In addition, Table 3.2 will

record whether each paper explicitly describes its community as a CoP, explicitly

discards it as a CoP or makes no mention either way.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
64
Table 3.1 – Selected studies about “virtual” Communities of Practice
Article authors Article title Network access Medium richness Participation
Thomsen (1996) At work in cyberspace: exploring practitioner use of the PRForum. subscription forum low (listserv) voluntary
Murray (1996) Nurses’ computer-mediated communications on NURSENET: a case public forum low (listserv) voluntary
study.
Bowers (1997) Constructing international professional identity: what psychiatric nurses public forum low (listserv) voluntary
talk about on the Internet.
Millen and Dray (2000) Information sharing in an online community of journalists. public forum low (listserv) voluntary
Robey et al (2000) Situated learning in cross-functional virtual teams. private various media low to high mandatory
Lueg (2000) Where is the action in virtual communities of practice? public forum low (newsgroup) voluntary
Baym (2000) Tune in, log on: soaps, fandom and online community. public forum low (newsgroup) voluntary
Wasko and Faraj (2000) It is what one does: why people participate and help others in electronic public forum low (newsgroup) voluntary
communities of practice.
Teigland (2000) Communities of practice in a high-growth Internet consultancy: Netovation public forum low (newsgroup) voluntary
vs.on-time performance.
Kimble et al (2001) Communities of practice: going virtual private various media low to high mandatory
Wasko and Teigland (2002) The provision of online public goods: examining social structure in a private low (listserv) voluntary
Network of Practice
Hara and Kling (2002) IT support for communities of practice: an empirically-based framework. private low (listserv) voluntary
Wenger et al (2002) Cultivating communities of practice: a guide to managing knowledge. private high (teleconference) mandatory
Saint-Onge and Wallace (2002) Leveraging communities of practice for strategic advantage. proprietary network mid (proprietary intranet) voluntary
Schlager et al (2002) Evolution of an on-line education community of practice. subscription forum mid (staffed MUD) voluntary
Pan and Leidner (2003) Bridging communities of practice with information technology in pursuit of subscription forum low (bulletin board) voluntary
global knowledge sharing.
Ardichvili et al (2003) Motivation and barriers to participation in virtual knowledge-sharing proprietary network mid (proprietary intranet) voluntary
communities of practice.
DeSanctis et al (2003) Learning in online forums. public forum low (Yahoo group) voluntary
Lee and Cole (2003) From a firm-based to a community-based model of knowledge creation: public forum low (listserv) voluntary
the case of the Linux kernel development
Teigland and Wasko (2004) Extending richness with reach: participation and knowledge exchange in public forum low (newsgroup) voluntary
electronic networks of practice.
Table 3.2 – Evidence of Wenger’s constructs in selected studies about “virtual” CoPs
Article MUTUAL SHARED JOINT COMMUNITY LEARNING / Characterised
ENGAGEMENT REPERTOIRE ENTERPRISE IDENTITY ACQ. as a CoP
Thomsen (1996) a a a a no mention
Murray (1996) a a a a no mention
Bowers (1997) a a a a no mention
Millen and Dray (2000) a a a a no mention
Robey et al (2000) a a a a a yes
Lueg (2000) a a a no
Baym (2000) a a a yes
Wasko and Faraj (2000) a a a a a yes
Teigland (2000) a a a no
Kimble et al (2001) a a a a a yes
Wasko and Teigland (2002) a a a no
Hara and Kling (2002) a a a a yes
Wenger et al (2002) a a a yes
Saint-Onge and Wallace (2002) a a a a a yes
Schlager et al (2002) a a a a yes
Pan and Leidner (2003) a a yes
Ardichvili et al (2003) a a yes
DeSanctis et al (2003) a a a a a yes
Lee and Cole (2003) a a a a a yes
Teigland and Wasko (2004) a a a no
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

A solid article, though not directly addressing the “virtual” CoP issue, is Thomsen’s

(1996) description of the PRForum, a listserv addressed to public relations

practitioners. A content analysis of two-weeks of postings found most messages were

information exchanges (56%), with debate the second most frequent category (31%).

This was followed up with a short survey of participants, which revealed three primary

purposes of the forum: to facilitate the exchange of information and advice, to debate

issues affecting the profession, and to cultivate and foster a sense of self-validation and

enhanced efficacy, both at a personal and professional level. The community is not

characterised as a CoP, yet the reported purposes of the forum provide evidence of

Wenger’s constructs of MUTUAL ENGAGEMENT, JOINT ENTERPRISE and

LEARNING/IDENTITY. In addition, the cited responses from the survey mentioned as

typical members’ appreciation for their professional virtual community, providing

evidence of the COMMUNITY construct as well.

Murray (1996) performed a discourse analysis of a sample of messages from

NURSENET, an unmoderated mailing list devoted to nursing. In addition, semistructured

e-mail interviews were conducted with a small number of subscribers. The focus of the

study is examining communicative practices of this on-line group, such as politeness and

turn-taking. Still, the primary reason for participating, is “to keep myself in touch with the

current state of the nursing profession, and to keep myself up to date with new and

possibly more appropriate techniques than the ones I currently employ in my nursing

practice” (1996: 230). This was interpreted as evidence of LEARNING and JOINT

ENTERPRISE. Furthermore, the discourse analysis revealed a frequent discussion topic was

to challenge prevailing ideas within nursing and to discuss ways in which practice or

education might be improved, or new theory generated. This suggests debating issues,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
67
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

hence MUTUAL ENGAGEMENT. Finally, evidence of COMMUNITY comes from the report

that a “sense of community” is evident in the list, because participants feel safe and

confident in being able to present themselves to others.

Bowers (1997) describes his experience as manager of an umoderated mailing list

devoted to psychiatric nursing. He performs a content analysis of the topics, as measured

by thread heads, to determine topics that are universally appealing to psychiatric nurses.

The author argues the list can contribute to the formation of an international professional

IDENTITY for psychiatric nurses, thus citing one of Wenger’s constructs. Furthermore, the

content analysis revealed the list provided a forum for sharing knowledge, information,

ideas and expertise (hence MUTUAL ENGAGEMENT and LEARNING), and for giving and

receiving help and support (hence COMMUNITY). Finally, the list’s narrow focus on a

specialised professional topic was interpreted as evidence of JOINT ENTERPRISE.

Millen and Dray (2000) study the creation of shared collective goods in a community

of journalists based on a mailing list. A content analysis of community archives found

most postings are technical answers to problems (51%). The group exhibits a distinct

professional culture, although the authors do not follow up on this, nor use the CoP

label. They do identify knowledge-sharing as the group’s raison d’etre, thus providing

evidence of MUTUAL ENGAGEMENT and LEARNING. The narrow professional focus of

the discussion suggests a JOINT ENTERPRISE. The authors characterise the group as an

online professional community, and were struck by strong member commitment to help

each other, which suggests COMMUNITY. Finally, the article emphasises the value of the

discussion archives as a searchable knowledge resource, and the most valuable

collective good of the community, thus yielding evidence of SHARED REPERTOIRE.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
68
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Robey, Khoo and Powers (2000) interviewed 22 workers and managers in three cross-

functional regionally-distributed teams in a large U.S. company. They assessed the teams

as virtual CoPs because they conducted much of their day-to-day interaction through

phone, e-mail, voice-mail or videoconferencing, and their LEARNING was situated in the

indigenous practices team-members developed to function effectively in this context.

Wenger’s (1998) constructs are not named explicitly but the numerous quotations

reproduced from the interviews reveal the presence of ENGAGEMENT (because team

members worked collaboratively with remote colleagues using electronic media),

LEARNING (evinced by the induction of new team members to the practices of this work

mode), SHARED REPERTOIRE (e.g. a common practice of producing spreadsheets from a

shared mainframe database, and using them as e-mail attachments), JOINT ENTERPRISE

(the shared disposition to make the virtual team arrangement work) and COMMUNITY

(evinced by the effort team members make to build relationships when they meet with

remote colleagues face-to-face, for instance going out to dinner).

In a conceptual article, Lueg (2000) questions the transferability of the concept of

CoPs, “deeply rooted in the lived-in world”, to the online world. He analyses

newsgroup de.rec.bodyart, a German-language group focused on bodyart (tatoos, ear

and nose piercings, etc.). He notes there is true knowledge-sharing and problem-

solving, but concludes the experience and shared practice (tattooing, body piercing)

take place in the real world, thus discounting the group as a true “virtual” CoP. While it

is true that the online medium cannot capture the nuances of some “embodied”

practices (Cook and Yanow’s famous flute-making example comes to mind), there do

exist text-based professions, such as law, mathematics or computer programming,

whose specialised discourses can be accurately captured in the plain-text messages of

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
69
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Usenet. The article yields evidence of ENGAGEMENT and LEARNING (knowledge-

sharing and problem-solving are explicitly mentioned), as well as JOINT ENTERPRISE

(because of the very narrow topic of the newsgroup).

Baym (1995) studied rec.arts.tv.soaps, a newsgroup of soap-opera fans. She describes

online IDENTITY construction, as fans identify with this character or reject another, with

a great deal of self-disclosure involved. In a later book-length ethnography (Baym,

2000), she describes the newsgroup’s social customs and norms (such as netiquette),

the in-depth topic knowledge of regular members, and the interpersonal relationships

between members which sometimes led to personal online friendships. Baym argues

this community is constantly recreated in “habitual and recurrent ways of acting or

practices” (emphasis in the original), and labels it a community of practice. However,

Wenger (1998; 2000a) argues a CoP is more than a community of interest: a CoP is

task-oriented; members go beyond talk to get actual work performed collaboratively.

Having said this, Baym’s ethnography yields evidence of COMMUNITY (the tightly-knit

group of fans), JOINT ENTERPRISE (their narrow topical focus on specific soap operas)

and LEARNING (in the form of IDENTITY ACQUISITION). Lacking a practice though, there

is no evidence of real ENGAGEMENT or REPERTOIRE.

Baym’s research highlights an important issue: whether fan discussion of books, movies

or TV programmes can be considered MUTUAL ENGAGEMENT. According to Wenger et al

(2002) it cannot, for fans do not actually work with the books, movies or personalities

they enjoy talking about. However, it could be a different situation for a group of literary

or movie critics, whose professional method-based discussion of technical and artistic

merits could qualify as ENGAGEMENT. Thus it can be difficult to distinguish a practice-

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
70
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

forming domain from a serious hobby. Hence, this research will conservatively focus on

professional domains, while acknowledging that even the term “professional” admits

some ambiguity. For instance, the topic of the U.S. Civil War can constitute a

professional practice for a historian, while just a hobby for someone who likes History.

Wasko and Faraj (2000) conducted an e-mail survey of participants in three computer-

related newsgroups (comp.lang.c++, comp.objects and comp.database) which they

referred to as “virtual communities of practice”. Using open ended questions, they

asked identified posters why they participate in the newsgroup and help others.

Respondents mentioned past help they had received from the newsgroup and a desire to

reciprocate; the pleasure of helping someone at negligible cost to themselves; the

challenge of solving tricky computational problems and the learning and reputation-

enhancement they derived from this. Although Wenger’s constructs are not explicitly

named, respondent comments cited in the article yield evidence of MUTUAL

ENGAGEMENT (in the form of problem-solving), LEARNING (respondents report “gaining

knowledge” by participating in the newsgroups), SHARED REPERTOIRE (in the form of

jargon and symbolic language), JOINT ENTERPRISE (in the form of specialised domains

in each newsgroup) and COMMUNITY (respondents value the newsgroups as a

community of like-minded peers). One problem with this study’s very interesting

results is that the authors made no attempt to differentiate the replies of stable members

of the communities from ephemeral visitors, who are much more numerous (Smith,

1999). Thus, responses from this latter group were probably over-represented.

Teigland (2000) presents an analysis of patterns of individual-level knowledge flows at

an Internet consultancy and their impact on individual performance. Data collection

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
71
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

comprised 30 in-depth interviews with employees from all areas, later complemented

by a survey of all 242 employees. The author found a face-to-face CoP formed by

programmers who met after work to discuss problems and innovative solutions. In

addition, these programmers belonged (some for many years) to Internet-based

professional communities who discussed similar technical issues. In fact, the

programmers consulted their online communities more often than co-located

colleagues, because their knowledge base was broader and more easily consulted than

local colleagues. Teigland notes these electronic communities exhibit many of the

properties of traditional CoPs. Respondents reported MUTUAL ENGAGEMENT (through

online problem-solving), LEARNING (because people explicitly sought and received

knowledge from Internet communities), IDENTITY ACQUISITION (because belonging to

certain Internet communities conferred prestige) and COMMUNITY (because members of

Internet communities had formed stable friendships). Since specific Internet

communities are not identified in the article, evidence of specific JOINT ENTERPRISE or

SHARED REPERTOIRE is not mentioned. Although clearly impressed by these online

communities, Teigland argues against using the CoP label because explicit text-based

messages are not a conduit for sharing tacit knowledge as CoPs are known to do

through close interaction.

Kimble et al (2001) rhethorically ask if a CoP can be virtual, equating “virtual” with

“geographically distributed”. They describe a distributed CoP in a company, charged

with IT-support work, and formed by four co-located members in the UK, five in the

US and a lone member in Japan. Daily communication is held via e-media (e-mail,

voice mail, video link, telephone conferencing and Microsoft NetMeeting), and every

six months the entire group holds a face-to-face meeting. Though not exclusively

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
72
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Internet-based, the article provides a detailed case study of a geographically distributed

CoP which conducts a substantial part of its day-to-day work through electronic

networks. The group is described as having a sense of common purpose (hence JOINT

ENTERPRISE), a strong feeling of identity (hence IDENTITY) and its own specialised

language (hence SHARED REPERTOIRE). In addition, the case study description of

member interactions yields direct evidence of MUTUAL ENGAGEMENT, and emphasises

the importance members give to strong personal relationships, as they aided smooth

group functioning during the six-month periods without face-to-face contact. This was

taken as evidence of COMMUNITY.

Wasko and Teigland (2002) apply theories of public goods and collective action to

explain the knowledge collectively created by members of Electronic Networks of

Practice (ENoPs) in the form of a publicly available discussion archive. In this

empirical study they used Social Network Analysis to determine whether a successful

extra-organisational ENoP was sustained by a critical mass of participants, and

whether heterogeneity of resources and interests among participants contributed to

the success of the group’s collective action. The studied ENoP was a subscription

electronic service of a U.S. professional legal association, which used bulletin-board

technology similar to newsgroups. Using a two-month message sample, Social

Network Analysis was performed on 2460 messages posted by 526 individuals.

Results showed the network had a star shape with a critical mass of 23 participants

contributing most messages. The authors also sent an e-mail survey to participants to

determine individual resources and interests. Results showed network tenure,

individual expertise, a desire for reputation enhancement, and a lack of local peers

with whom to discuss professional issues were all significantly related to responding

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
73
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

to questions posted by others. The article provides evidence of SHARED REPERTOIRE

(i.e. online discussion archive, the technical language of the legal profession),

LEARNING (because participants asked for and received help) and JOINT ENTERPRISE

(because of the narrow focus of the list). Still, online exchanges are not described in

sufficient detail to yield evidence of MUTUAL ENGAGEMENT, nor is any evidence of

COMMUNITY reported.

Hara and Kling (2002) describe an online CoP based on a closed mailing list for licensed

state public defenders. The forum exhibited heavy traffic (average 107 weekly messages)

and about 250 suscribed members. Interviews with participants revealed they used the

mailing list for basic information sharing, more complex “brainstorming” and problem-

solving (hence MUTUAL ENGAGEMENT and LEARNING) and for learning about and

discussing current legal issues (hence JOINT ENTERPRISE). Evidence for SHARED

REPERTOIRE can be reasonably inferred from the technical language used by the legal

profession. No evidence of COMMUNITY was mentioned in the paper. The article reports

also some perceived drawbacks of the community: some attorneys attempted to “free-

ride” on the efforts of others by posting legal questions which they had not researched at

all, hoping to get quick and free answers. This led some experienced lawyers to describe

the mailing list as containing a “lot of junk”.

Wenger et al (2002) describe a “global community” of geologists involved in deep-

sea petroleum exploration at Shell. The community was intentionally designed and

supported, with the vision “to bring the world’s leading expertise, no matter where it

was located, to bear on problems and issues, no matter where they occurred” (2002:

114). This community eventually spanned 18 time zones, 20 independent Shell

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
74
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

companies, and over 1500 members. At the centre of this global community is a

network of local community coordinators, who see their role as brokering local and

global relationships, rather than knowledge. Although the authors (including Wenger)

describe this as a virtual or distributed CoP, it is more similar to a constellation of

communities (Wenger, 1998), or to a Network of Practice (Brown and Duguid, 2001).

The impossibility of direct ENGAGEMENT limits the size of particular communities,

and though ENGAGEMENT between geographically distributed members is entirely

possible, calling the entire global community a CoP stretches the original concept

beyond recognition (Roberts, 2006). Because of this, the study only yields evidence

of LEARNING (in the form of acquiring new knowledge), SHARED REPERTOIRE (in the

form of jargon and shared practices) and JOINT ENTERPRISE (in the form of a narrow

technical specialty).

Saint-Onge and Wallace (2002) provide a detailed description of the design and

successful launch of a web-based CoP for insurance agents at Clarica, a Canadian life

insurance company. The project was designed to support the company’s 3000

independent agents, most of whom work alone from various points of Canada’s huge

territory, conducting day-to-day business with Clarica through their laptops. The Agent

Network is an extension of Clarica’s proprietary intranet technology. It is a fairly

conventional KM project, but it exhibited unusual sensitivity to people issues, probably

due to the project designers attending a CoP workshop conducted by Etienne Wenger.

For instance, the completed technological infrastructure was populated by making a

voluntary invitation to 150 agents, who were expected to take ownership for the

community and become permanent hosts. This call had a 95% acceptance rate, where

developers expected no more than 50%, indicating high agent interest in electronically

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
75
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

meeting and sharing knowledge with other agents facing their same business

circumstances. The description of the community provides evidence of MUTUAL

ENGAGEMENT (in the form of problem-solving), COMMUNITY (the study notes that

almost from its launch the Agent Network developed a distinctive sense of

community), SHARED REPERTOIRE (in the form of online tools and knowledge

repositories), JOINT ENTERPRISE (because the agents shared the same professional

practice) and LEARNING (described as support for novice agents and continuous

renewal for experienced agents).

Schlager et al (2002) describe an online CoP aimed at K-12 teachers called TAPPED-IN.

Using MUD’s, they created virtual rooms which allow both public and private

discussion, with whiteboards and sharing of documents. They invited several education

organisations to participate as tenants, and provide professional development content.

The article reports three years of usage statistics, and several snippets of interactions

between teachers. From these, and from community descriptions, evidence for most of

Wenger’s constructs can be inferred. Specifically, there is evidence of MUTUAL

ENGAGEMENT (in the form of teachers working collaboratively), JOINT ENTERPRISE

(because the forum focuses on teacher professional development), SHARED REPERTOIRE

(in the form of online tools and practices) and LEARNING (in the form of knowledge

sharing). Schlager et al (2002) characterise TAPPED-IN as a successful online CoP, but

warn online facilitation and continuous support are essential to maintain the CoP.

Therefore, they see unaided Internet media such as listservs or newsgroups as insufficient

to “support the ebb and flow of discourse and collaboration that is characteristic of

professional practice” (2002: 152); concluding that a listserv or newsgroup, no matter

how well-trafficked, is not a CoP, a conclusion challenged in the next section.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
76
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Pan and Leidner (2003) present a case study of a firm’s global knowledge-sharing

efforts using a subscription-based networks (Compuserve). They argue their study

provides empirical evidence of KMS used to successfully provide bridges between

local CoPs. Still, their definition of CoP does not include co-location or even guarantee

MUTUAL ENGAGEMENT actually takes place in a sustained manner. Thus, the weaker

concept of NoP is more consistent with the authors’ operational definition. This is

further indicated by the description of interactions they provide, which do not seem

frequent or intensive enough to justify calling the aggregate a CoP, although there is

some evidence of LEARNING (because requests to the network for information or

knowledge did not go unattended) and REPERTOIRE (in the form of online tools and

knowledge repositories).

Ardichvili, Page and Wentling (2003) report a qualitative study of “communities of

knowledge-sharing” at Caterpillar Inc. These communities interact through the

Knowledge Network, an intranet-based system designed to provide an infrastructure for

community functioning supported by a group of KM technology experts as well as

members of Caterpillar’s corporate university. Qualitative interviews with 30 members

of three distinct communities were performed. The aim of the study was to determine

motivations and barriers to employee participation in these communities. As in the Pan

and Leidner (2003) study, local CoPs are much more in evidence than virtual ones.

Tellingly, participants in some interviews argued the Knowledge Network should be

designed to support local CoPs rather than absorb them. The article yielded evidence of

MUTUAL ENGAGEMENT and LEARNING (employees used the communities for problem-

solving) as well as SHARED REPERTOIRE (in the form of network access to Best

Practices and a Lessons Learned database).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
77
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

DeSanctis et al (2003) conducted an ambitious and original study in which they make an

explicit comparison of learning outcomes across various kinds of electronic communities.

Specifically, they investigated virtual learning networks supported by three different e-

based venues: video-conferenced MBA classrooms, team electronic discussion spaces,

and Internet-based communities. In addition, every group was assessed on three different

types of group learning: declarative and procedural information exchange, transactive

learning and sense-making. These reflect increasingly sophisticated forms of electronic

interaction. The authors expected sense-making to be more difficult to achieve in Internet

communities, and their results generally confirmed this. Specifically, they studied a

sample of 40 Internet communities, which they classified into three main categories:

information kiosks, associations and communities of practice. The authors provide a case

description of one such community of practice, whose activity they studied between

December 1999 (when it first started) to April 2001. This community was launched with

an explicit message by its founder and coordinator inviting interested parties to discuss

and learn about the specific topic of public sector Knowledge Management. Over the first

months, discussions, though lively, stayed at the level of declarative and procedural

information exchange. However, after ten months, discussions became more complex

and reflective and the community achieved all three forms of learning, while a stable 20-

member core formed. DeSanctis et al, assessed this community as a CoP because it

achieved sense-making, which they define as the process of developing shared mental

models, and which they believe contributes to the formation of a CoP. The case

description provides evidence of MUTUAL ENGAGEMENT and LEARNING (in the form of

sense-making knowledge exchange through deep and reflective discussions), a JOINT

ENTERPRISE (the public sector KM domain), SHARED REPERTOIRE (jargon, group charter),

and COMMUNITY (exchange of personal information; a core group with longer tenure).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
78
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Thus, the article makes a strong case for the existence of a successful virtual CoP, only

limited by its non-reliance on Wenger’s theoretical framework.

Lee and Cole (2003) performed a case study of the online Linux community, which

many consider an Internet-based CoP (e.g. Markus, Manville and Agres, 2000; Kogut

and Metiu, 2001; Snyder, 2005). Their aim was to develop an extra-organisational

community-based model of knowledge creation as opposed to a firm-based model.

Their data sources were the Linux kernel mailing list from 1995-2000, a survey of list

members, and the artifacts produced by the community, notably the Linux kernel

source code. The authors report the community is self-organised into a 122-member

core and a much larger periphery of over 4 thousand. The periphery performs the dual

tasks of writing code patches and reporting problems, while core members specialise in

the various subsystems (132 in all) making up the kernel, and focus on selecting the

best submitted patches for inclusion in the official release. The article provides explicit

evidence of JOINT ENTERPRISE (the Linux Operating System), SHARED REPERTOIRE

(community artifacts such as the source code and archived discussions), MUTUAL

ENGAGEMENT (in the form of collaborative code development, testing and critique),

COMMUNITY (because frequent exchanges and shared responsibilities make at least core

members known to each other), and LEARNING (through the process of continually

improving the product). In sum, Lee and Cole (2003) provide an excellent description

of the Linux kernel community which they characterise as an extra-organisational CoP.

Even though they do not use Wenger’s full theoretical framework, sufficient evidence

of the constructs is provided to make this a plausible assessment for the core; yet it is

unlikely the entire 4,000-strong community can be assessed as a CoP because of the

limits imposed by the criterion of direct ENGAGEMENT.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
79
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Teigland and Wasko (2004) studied intra-organisational knowledge-sharing networks

at a division of Cap Gemini, a large European consultancy. They report the results of a

survey of participants in a listserv-based ENoP focused on the application of Microsoft

products at a consultancy. Results of this survey showed a positive association between

participation activity and tenure in the ENoP with acquiring knowledge from the

network as well as contributing to others. Additionally, knowledge acquisition from the

network and knowledge contribution to the network are positively related to individual

performance. The study explicitly mentions Wenger’s construct of MUTUAL

ENGAGEMENT as the cause of the ENoP, although no evidence of it is actually provided

in community descriptions or participant responses. The article does provide evidence

of LEARNING (the reported principal reason of ENoP members for participating),

SHARED REPERTOIRE (the online discussion archive) and JOINT ENTERPRISE (in the form

of the specialised topic of the mailing list).

At this point, two broad sets of conclusions emerge from this admittedly heterogenous

collection of papers. First there are the lessons learned from the collection as a whole,

specifically, what virtual learning communities are capable of doing, and how this

contributes to the thesis. Second, there are the limitations and gaps detected in previous

studies of “virtual” CoPs, or what the thesis can contribute to the current state of

knowledge.

With respect to the capabilities of virtual learning communities, the articles show:

• They can achieve a professional orientation (e.g. Murray, 1996; Hara and

Kling, 2002)

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
80
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• They can maintain focused discussions (e.g. Thomsen, 1996; Millen and

Dray, 2000)

• They enable powerful collaboration between widely distributed members

(e.g. Robey et al, 2000; Kimble et al, 2001; Lee and Cole, 2003)

• They enable information sharing (e.g. Hara and Kling, 2002; Pan and

Leidner, 2003)

• They enable collective problem-solving (e.g. Wasko and Faraj, 2000; Hara

and Kling, 2002; Lee and Cole, 2003)

• They enable sharing of knowledge and best practices (e.g. Lueg, 2000;

Wenger et al, 2002; Saint-Onge and Wallace, 2002)

• They can develop knowledge repositories and online tools (e.g. Schlager et

al, 2002; Ardichvili et al, 2003; Pan and Leidner, 2003)

• They enable formation of stable and persistent virtual communities (e.g.

Thomsen, 1996; Baym, 2000)

• They enable building strong personal relationships between members who

have never met personally (e.g. Teigland, 2000; Baym, 2000)

• They are frequently led by an energetic core of members (e.g. Thomsen,

1996; DeSanctis et al, 2003; Lee and Cole, 2003)

• They are valued by participants as learning forums (e.g. Murray, 1996;

Teigland and Wasko, 2004)

• They are valued by participants as communities of like-minded peers (e.g.

Wasko and Teigland, 2002; Saint-Onge and Wallace, 2002)

• They provide a forum for enacting and validating desired professional

identities (e.g. Thomsen, 1996; Bowers, 1997; Teigland, 2000)

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
81
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Thus virtual learning communities exhibit an impressive range of capabilities which

taken together make a strong case for the feasibility of true virtual CoPs. To this should

be added the evidence summarised in Table 3.2 about the presence of Wenger’s

constructs in reviewed communities. The sheer number of occurrences, and the fact that

several communities reportedly displayed the full range of Wenger constructs,

strengthens the theoretical case for the existence of virtual CoPs.

Turning now to the gaps and limitations of the reviewed references, it should be noted

first that for most studies, the feasibility of “virtual” CoPs was unproblematically

assumed. Only three studies –Baym (2000), Kimble et al (2001) and DeSanctis et al

(2003)– include a critical discussion of the issue, but without using Wenger’s full theory.

A second observation, not necessarily a limitation, concerns a divide Schwen and Hara

(2003) have noted in CoP literature (the reviewed articles are no exception) between

studies whose intent is design, i.e. that provide a prescription for the design and launch

of a virtual CoP (e.g. Schlager et al, 2002; Saint Onge and Wallace, 2002; Dubé et al,

2005) and studies whose intent is description, i.e. to examine a pre-existing group and

looks for traits known to be associated with CoPs (e.g. Baym, 2000; Hara and Kling,

2002). Schwen and Hara further argue Wenger’s (1998) book is a rich theoretical

description of a fully mature and constructive co-located CoP, but not a prescription for

designing a successful CoP, co-located or virtual. That this thesis does not focus on

design is by now fairly obvious. But neither is description its immediate objective.

Rather it starts with a broad discovery stage, where it will attempt to develop empirical

Internet-based indicators of CoP activity and maturity, and will then set out to locate

mature Internet-based CoPs.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
82
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Thirdly, the 20 reviewed articles exhibit an important gap. Of the 16 that explicitly

address “virtual” CoPs, 15 define CoPs by citing Wenger (only Baym doesn’t).

However, none applies Wenger’s complete theoretical framework, specifically the

complete set of constructs (although COMMUNITY and LEARNING were mentioned by

all). For instance, no study attempted to document MUTUAL ENGAGEMENT taking place

over the network, or looked for evidence of a SHARED REPERTOIRE developed and based

on the network. This non-reliance on Wenger’s theory makes it difficult to assess to

what extent the virtual communities described in these studies are CoPs as defined and

described by Wenger (1998). The thesis will address this gap by using the complete set

of constructs as guidelines for the development of the virtual CoP model and the

Research Questions. The identification of a gap in the “virtual” CoP literature is the key

result of this section. Therefore:

The identified research need is to determine whether CoPs, rigorously

defined according to Wenger’s theoretical framework, can be fully

“virtual” or Internet-based.

3.3 – Challenges to the concept of “virtual” Communities of practice

Wenger’s (1998) theory does not make member co-location a requirement for a CoP,

only regular ENGAGEMENT. The theory is thus open to the possibility of virtual CoPs, as

long as the electronic networks over which ENGAGEMENT takes place can adequately

support it, with “adequacy” defined not a priori but solely by CoP members.

Nevertheless, some researchers reject the possibility of true CoPs forming over

electronic networks, particularly the Internet. This section will discuss their critiques.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
83
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Jones (1997) argues that for an online collective to qualify as a virtual community

requires the presence of affective bonds between members, not just a high level of

interaction or a continuity of participation. Similarly, Kling and Courtright (2003)

believe not just the notion of “virtual CoP” is problematic, but more generally the

casual use of the term “community” to characterise groups that participate in electronic

forums. They critique the “aspirational” definition of virtual community proposed by

Rheingold (1993: 5):

Virtual communities are social aggregations that emerge from the Net when

enough people carry on those public discussions long enough, with sufficient

human feeling, to form webs of personal relationships in cyberspace.

Kling and Courtright argue the definition does not distinguish among various kinds of

social organisation that can use the Internet for interaction, such as “hangouts”, fan

clubs, associations, and communities. More generally, they take issue with published

descriptions of online groups which characterise the forum in question as a virtual

community only by virtue of the fact that its members interact online. This and Jones’

(1997) earlier point are valid concerns, and would discourage the usage of the term

“community” for the online collectives selected mostly through quantitative methods in

Stage III of the Research Strategy. However, succeeding Stages will guarantee that

selected collectives are characterised by much more than high volumes of participant

interaction.

Based on three years of research designing and supporting an Internet-based CoP for

teachers, Schlager et al (2002: 152) have concluded that unaided Internet media such as

listservs or newsgroups are insufficient to “support the ebb and flow of discourse and

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
84
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

collaboration that is characteristic of professional practice [therefore] a listserv or

newsgroup, no matter how well-trafficked, is not a community of practice”. The

argument seems to suggest that the main obstacle lies in media richness. Yet, as

previously noted, the research of Lee (1994), Carlson and Zmud (1999), and Burke et

al (2001) supports the notion that a lean medium is not an insurmountable obstacle.

Thus even if the success of Schlager et al’s designed virtual CoP is heavily dependent

on online facilitation and continuous support, it does not follow that a newsgroup or

listserv cannot succeed as a CoP on its own.

Brown and Duguid’s construct of NoP is a more sophisticated challenge to the

possibility of “virtual” CoPs. They quite agree that Internet-based groups can

interact and share a practice, but they propose the weaker concept of NoP for these,

and reserve the strong concept of CoP only for co-located groups (Brown and

Duguid, 2000b).

Networks of practice are made up of people that engage in the same or very

similar practice, but unlike in a community of practice, these people don’t

necessarily work together [yet] such a network shares a great deal of common

practice. Consequently, its members share a great deal of insight and implicit

understanding. And in these conditions, new ideas can circulate. These do not

circulate as in a community of practice, through collaborative, coordinated

practice and direct communication. Instead, they circulate on the back of

similar practice (people doing similar things but independently) and indirect

communications (professional newsletters, listservs, journals, and conferences,

for example) (Brown and Duguid, 2000a: 28).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
85
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

An example they give is the 24,000-member cadre of Xerox photocopier repair

technicians (Brown and Duguid, 2001). They are geographically dispersed and too

numerous to know and engage with each other, so they cannot be a CoP, but they share

a common practice and are thus capable of sharing a great deal of knowledge. Another

example of NoPs are professions, “where again similar practitioners, by virtue of their

practice, are able to share professional knowledge through conferences, workshops,

newsletters, listservs, Web pages and the like” (Brown and Duguid, 2001: 206).

Brown and Duguid (2001) think of NoPs as equivalent to Strauss’ (1978) social worlds,

and Knorr Cetina’s (1999) epistemic cultures. In fact, they acknowledge their concept is

close to van Maanen and Barley’s (1984) occupational community, using at one time the

metaphor of a “virtual guild” (Brown and Duguid, 2000a: 29), but they wish to re-direct

attention from the “community” aspect of such groups to the shared practice of these

networks (ibid). They see NoPs as extended epistemic networks where practice provides

a common substrate which makes them capable of effectively sharing a great deal of

knowledge, even if most of their members “will never know, know of, or come across

one another” (2001: 205). They do not limit the definition or size of a NoP to members

who know each other, or have actually exchanged some knowledge. Two hematologists

who have never met would be part of the same NoP because of the highly specialised

practice they both belong to (Brown and Duguid, 2000a). Hence, the membership

criterion is not actual but only potential knowledge-sharing. Herein lies the key

difference with Wenger’s CoP, where the criterion is direct and sustained ENGAGEMENT.

Using the fact that shared practice is a common denominator of CoPs and NoPs,

Brown and Duguid propose a theoretical framework that explains why knowledge is

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
86
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

sometimes “sticky” while at other times it is “leaky” (Brown and Duguid, 2000a,

2001). They draw upon well-known examples from the literature about instances

where knowledge could not flow within the organization (such as the GUI

breakthrough at Xerox), but could easily leak between different organizations

(specifically from Xerox to Apple). Brown and Duguid argue that knowledge cannot

easily flow between different CoPs, i.e. between people who have different practices.

By contrast, knowledge can easily flow between people who share a practice, be they

within the same CoP at a single organization, or within similar CoPs at different

organizations. They are like-minded professionals, among whom there exist such

weak ties as attending the same professional conferences, reading trade magazines, or

contributing to the same online newsgroups or listservs. These weak ties give rise to

an extensive NoP which connects the different CoPs, and provides conduits for

knowledge to leak from one organisation to another (even if the actual leakage is a

contingency; as was the famous visit of Steve Jobs to Xerox’s PARC). By contrast,

CoPs with different practices will have trouble sharing knowledge with each other,

even if they both belong to the same organisation, thus accounting for the stickiness

of knowledge.

Within this framework, CoPs are local high-density sections of larger NoPs. Members

of CoPs work together face-to-face and thus develop much stronger ties than those

prevalent over the NoP (Brown and Duguid, 2000b). This suggests that a viable

search strategy for CoPs (either virtual or co-located), is to examine the social

network structure of a known NoP for areas of high-density, a lesson that will be used

during the construction of the model of Internet-based CoP.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
87
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Brown and Duguid proposed the concept of NoP with embedded CoPs in order to build

a single unified theory to explain the sticky and leaky properties of knowledge. Yet,

their use of the CoP concept is different from Wenger’s. First, they assert members of a

CoP work together whereas members of a NoP don’t (Brown and Duguid, 2000b).

However, in Wenger’s definition of CoP, “working together” both in the sense of

belonging to the same workgroup (as did Orr’s service reps, even though they normally

worked on their own), or in the sense of having a common employer or a shared

physical workplace (as did Wenger’s claims processors), is not a requirement for a CoP

to exist (Wenger, 1998, 2000a). What is absolutely necessary is that members interact

on a regular basis so that they can have the MUTUAL ENGAGEMENT that will build up

both the community and its indigenous practice. With regular interaction as a premise,

members of CoPs need not work together every day, nor do they need to be co-located.

Second, Brown and Duguid (2000b) see CoPs as face-to-face intra-organisational

groups, and thus discount the possibility of extra-organisational or Internet-based CoPs.

However, restricting the CoP concept only to co-located intra-organisational groups is

not really necessary to their theoretical argument about knowledge leaking or sticking.

It rather looks as an added restriction to make a stronger contrast between NoPs (which

can be virtual) and CoPs (which allegedly can’t).

Brown and Duguid’s work on NoPs has not gone beyond the theorising stage. By

contrast, Wasko and Teigland have applied this construct on several empirical studies,

which were reviewed in the previous section as virtual learning communities. The

examined communities scored well on the presence of Wenger’s constructs, as shown

previously in Table 3.2. Nevertheless, it is Wasko and Teigland’s reasoned position that

virtual communities based on electronic networks can at most achieve the status of

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
88
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

NoPs, but not true CoPs (Wasko and Teigland, 2004). Hence, they pose a strong

theoretical challenge to the position of this thesis, which warrants a careful review of

their work.

After several years of empirical studies of NoPs, including their respective doctoral

dissertations, Wasko and Teigland (2004) recently turned to formal theorising, and

proposed a model for electronic NoPs or ENoPs, for which they provide a four-part

definition:

[W]e define an electronic network of practice as a self-organising, open activity

system, focused on a shared practice that exists through computer-mediated

communication (2004: 27).

First, ENoPs are self-organising because participation is voluntary and without

assigned deliverables, thus setting these structures appart from virtual project teams.

Second, ENoPs are open because they conduct discussions in public electronic spaces

where anyone can read them, or contribute messages. This sets them appart from

collectives that limit membership.

Third, members share a common practice and interact with each other to solve practice-

related problems, in effect conducting online MUTUAL ENGAGEMENT.

Fourth, ENoPs only exist over the network, they have no physical presence. This means

the knowledge exchanged through the network is codified as text messages, and is

publicly preserved online. It also means they can have any number of participants who

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
89
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

need not know each other, or coordinate in any way. This sets them appart from

collectives where people know each other and good behaviour is enforceable through

social sanctions.

In order to explain the success of these online collectives, a success based on active

participation and seemingly selfless knowledge contribution, Wasko and Teigland

(2004) apply theories of public goods, social dilemmas and collective action.

Specifically, they consider the archive of the ENoP (typically a newsgroup) as a public

good collectively created by the community, echoing an earlier suggestion by Millen

(2000). This raises the issue of free-riding by lurkers, who have access to discussions

but do not post any messages. Wasko and Teigland see this as as a social dilemma,

because if all participants attempt to free-ride, the public good is not produced. They

suggest several factors can make participants overcome self-interested behaviour and

actively participate; among them:

a) the more heterogeneous the network is, the more likely that a critical mass of

dedicated and competent “organisers”, able to create the collective good on their

own, will be achieved (Oliver, Marwell and Teixeira, 1985).

b) individuals who have higher levels of professional expertise and

organizational tenure are more likely to participate (Constant, Sproull and

Kiesler, 1996).

c) individuals who lack access to local discussion alternatives, such as co-

workers or colleagues are more likely to participate (Wasko and Teigland, 2002).

d) affective factors, such as moral obligation from previously receiving help from

the network (Wasko and Faraj, 2000) or a sense of identification with the

network and its goals (Lakhani and von Hippel, 2000).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
90
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

In a previous study, the authors found empirical support for propositions b) and c)

above (Wasko and Teigland, 2002).

The ENoP model has two important coincidences with Wenger’s CoP model: the use of

the MUTUAL ENGAGEMENT construct, and the emphasis on a critical mass, which is not

a property of NoPs, but is common in CoPs which through their intensive ENGAGEMENT

develop core-periphery structures. However, Wasko and Teigland are trying to explain

all CoP-like online behaviour through the alternative NoP concept, because they share

Brown and Duguid’s position about CoPs being restricted to face-to-face groups. Still,

if some CoP-like behaviour in the Internet is caused by virtual CoPs (the position of

this research), then Wenger’s theory can provide an alternative explanation to the one

offered by the ENoP model.

Specifically, the lurker, or free-rider issue, can be explained by recalling Wenger makes

LEARNING a function of MUTUAL ENGAGEMENT. Therefore, the knowledge acquired by

a participant who actively ENGAGES in discussion, problem-solving and responding to

questions (even without any assurance of reciprocity), will not in general be the same

as the knowledge acquired by a lurker who merely reads through the exchange.

Through their ENGAGEMENT, active participants acquire some knowledge which is not

explicitly written on the textual messages. This is because, by hypothesis, participants

already share a practice (whether the group is a CoP or only a NoP) and thus share a

great deal of knowledge they do not need to make explicit in their textual exchange.

This tacit knowledge is put to work through ENGAGEMENT, using the exchanged

messages partially as a medium, partially as a catalyst or affordance (Cook and Brown,

1999). In other words, participants involved in MUTUAL ENGAGEMENT get more

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
91
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

meaning from their textual exchange than they can possibly put into words; a variant of

Polanyi’s famous argument that “we can know more than we can tell” (Polanyi, 1966).

Hence, the aim of active participation is the LEARNING that each participant derives

from ENGAGEMENT, rather than the creation of the public archive, which would be more

of an instrumental aid (hence, part of SHARED REPERTOIRE). Therefore, the problem of

“free-riding on the efforts of others” is a mirage, because not all knowledge created

through ENGAGEMENT is captured in the public archive. The ENGAGEMENT of active

participants results in greater “private” LEARNING for them, which is undisturbed by the

existence of lurkers. Even the answering of questions without receiving anything back

is rewarded through LEARNING. For instance, it was found during this research that

newsgroups CPLUS and TAXES explicitly encourage questions from novices (see

Chapter Nine), because the regulars believe many of the best discussions have started

from seemingly innocent novice questions.

Having said this, the NoP concept is still the strongest theoretical challenge to virtual

CoPs in current literature. Both Brown-Duguid and Wasko-Teigland consider true

CoPs to be small tight-knit groups of members who know each other, work together

and communicate face-to-face. Both consider practice-sharing Internet-based groups

can at most reach the status of NoPs, but not true CoPs. Thus they alternatively make

members’ knowledge of each other, or their working together, or their interactions

being face-to-face, requirements for the existence of true CoPs. Wenger, on the other

hand, makes sustained MUTUAL ENGAGEMENT, by whatever means, the sole

requirement. This thesis will follow Wenger’s more developed theory, hypothesise

virtual CoPs are feasible, and seek actual working examples in the Internet. Of course,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
92
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

even if it succeeds, the thesis drew important lessons and insights from previous

research by Brown-Duguid and Wasko-Teigland, specifically:

• NoP theory gives no real argument that restricts CoPs to co-located groups.

• Professionally-focused newsgroups are the Internet-based portions of wider

NoPs.

• NoPs contain embedded CoPs, which are areas of direct and intensive

interaction, resulting in higher network density.

3.4 – Summary

This chapter has reviewed the theoretical arguments regarding the feasibility of

“virtual” CoPs and examined various kinds of “virtual” CoPs described in previous

studies. On balance, the literature tends to be fairly optimistic about the possibility and

advantages of virtual CoPs. However, in most cases this optimism is grounded on an

intuitive grasp of the CoP concept, and in no case were Wenger’s theoretical constructs

of MUTUAL ENGAGEMENT, COMMUNITY, SHARED REPERTOIRE, JOINT ENTERPRISE, and

LEARNING/IDENTITY ACQUISITION applied to rigorously assess whether a particular

online group functions as a CoP. This is the principal gap and research need the thesis

will concentrate on.

Previous articles about virtual learning communities yielded considerable evidence

about the viability of the Wenger constructs in online environments, which makes

virtual CoPs at least plausible. In addition, some of the capabilities revealed by these

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
93
Virtual Communities of Practice
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

communities will be used in the construction of the model of Internet-based CoP,

forthcoming in Chapter Four.

Finally, theory and empirical studies about NoPs were reviewed, because they formally

challenge the notion of a “virtual” CoP. They were found not to provide solid

arguments against the possibility of virtual CoPs. Yet they yielded the useful insight of

studying professionally-oriented newsgroups, which are an example of NoP, as a

promising place to search for virtual CoPs, revealed as high-density areas within the

network. The Research Strategy, also forthcoming, will capitalise on this fact.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
94
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Part Two

Research Strategy: Design

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
95
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CHAPTER FOUR

Research Design

Previous chapters reviewed theories of CoPs and virtual CoPs. The time has come to

build a theory-informed model of Internet-based CoP to provide direction to this

research. This is the key task of Chapter Four.

The chapter is organised in seven sections. The First develops and discusses a model of

Exemplary Internet-based CoPs that will provide the guidelines for the research design,

including the formulation of the Research Questions and Research Strategy. Section

Two then formally presents the aims and specific objectives of this research. Section

Three formulates the Research Questions that provide direction and focus to this study.

Section Four then makes an important decision by narrowing the scope of the thesis to

the Usenet discussion network. The rationale for this decision is discussed, and

previous research on Usenet-based communities, reviewed. Section Five addresses

thesis philosophy, including ontological and epistemological assumptions, as well as

the rationale for using triangulation of methods. Finally, Section Six uses the proposed

model as a blueprint for the development of an original Research Strategy, which is

structured around the stated Objectives, and carefully targets the Research Questions.

The last section is a chapter summary.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
96
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.1 – The model of “virtual” Community of Practice used in this research

This section proposes a theory-grounded model of Internet-based CoP that describes the

hypothesised target this study will search for. The model is conjectural, since the

existence of Internet-based CoPs is yet to be established. Still, by describing a specific

target, the model narrows and brings into focus the requirements the research design must

fulfil, and provides a detailed blueprint for the formulation of the Research Strategy.

4.1.1 – Model specification

The model intentionally specifies a particular type of community, referred to as an

“exemplary” virtual CoP. It includes two separate sets of visible attributes or traits.

First and foremost, labelled as Essential Traits, are Wenger’s constructs. According to

the operational definition of CoP proposed in Chapter Two, the presence of the

constructs is a necessary and sufficient condition for characterising a group as a CoP.

Second, the model includes an additional set, labelled Exemplary Traits. These include

contingent attributes that are frequently observed in CoPs and other communities, and

render a community more focused, productive and energetic.

Each set plays a different role. The role of the Essential Traits is to rigorously identify a

group as a CoP. However, assessing the presence of the Essential Traits requires

qualitative analysis of member interactions, which is time-consuming. Hence, the

Essential Traits are not suitable for conducting extensive searches. That is the role of

the Exemplary Traits. These cannot uniquely identify CoPs, because they appear in all

kinds of virtual communities. Yet, they are highly visible, and some can be quantified,

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
97
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

making empirical detection easier. Thus, they are well-suited for extensive searches,

and for detecting productive, focused and energetic online communities, “exemplary”

communities, as it were. Although the two sets of Traits play different roles, they are

not independent entities. Rather, the Essential Traits are intensified or enriched by the

Exemplary Traits, as explained presently.

The rationale behind the model is that if a suitable number of virtual communities are

located that display all Exemplary Traits, there is a very good probability that some will

also display all Essential Traits, hence qualifying as “exemplary” Internet-based CoPs.

The virtual CoP or VCoP model, as it will be called, comprises the twelve visible Traits

displayed in Figure 4.1, and subsequently discussed.

A Model of “Exemplary” Internet-based CoP

Exemplary Traits Essential Traits

Mid-sized stable group MUTUAL ENGAGEMENT


High-volume of participant interaction SHARED REPERTOIRE
Core-periphery structure JOINT ENTERPRISE
Highly focused discussions COMMUNITY
Topic is identifiable profession LEARNING / IDENTITY ACQUISITION
High-quality institutional documents
Non-conflictive

Figure 4.1 – The VCoP Model

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
98
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.1.2 – Exemplary Traits

4.1.2.1 – Mid-sized stable group

The model first requires an identifiable online group of people. This Exemplary Trait is

related to the Essential trait of COMMUNITY, and restricts it to “mid-sized”. An exemplary

CoP is not so small that it fails to achieve a critical mass that supports energetic

ENGAGEMENT (Oliver and Marwell, 1988), nor so large that direct ENGAGEMENT becomes

impossible (Wenger, 1998). Wenger et al, (2002) note conventional CoPs are groups of

between 15 and 50 stable participants, possibly with a larger periphery of onlookers and

visitors, who are not active participants. These numbers will be used as guidelines for the

search, making an implicit assumption that successful Internet-based CoPs are about the

same size as conventional CoPs.

4.1.2.2 – High-volume of participant interaction

The stronger ENGAGEMENT is within a group the more it will develop an indigenous

practice and function as a CoP (Wenger, 1998). Therefore, an exemplary Internet-based

CoP exhibits energetic participants who exchange a high-volume of messages, thus

building strong online ties between them. This Exemplary Trait is linked to the Essential

Trait of ENGAGEMENT, which it intensifies, rendering the community more visible.

4.1.2.3 – Core-periphery structure

An exemplary CoP exhibits a core-periphery structure, with core members contributing

most messages, thus setting the tone and direction of discussion (Wenger, 2000a).

Everett and Borgatti (1999: 397) define such a structure as follows:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
99
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

A network has a core-periphery structure if the network can be partitioned into

two sets: a core whose members are densely tied to each other, and a periphery

whose members have more ties to core members than to each other.

This Exemplary Trait “enriches” the Essential Trait of MUTUAL ENGAGEMENT. A core-

periphery structure reflects various degrees of ENGAGEMENT, which opens the

possibility of active participation to people other than core members (Wenger, 1998).

This increased membership makes for a more energetic and diverse community.

4.1.2.4 – Highly-focused discussions

This Exemplary Trait is linked to the Essential Traits of JOINT ENTERPRISE and

MUTUAL ENGAGEMENT, requiring that the latter remain centred on the former. In

essence, it requires that the community is successful in preserving its topical focus.

This is generally a significant indicator of quality in Internet-based communities, which

by being public and open must deal with off-topic messages from non-members and

commercial advertisers (Kollock and Smith, 1996; Smith, 1999).

4.1.2.5 – Topic is an identifiable profession

An exemplary CoP is defined as centred on a profession. Of all the Exemplary Traits,

this is the one most closely associated to a practice, because all professions are

practices (Brown and Duguid, 2001). Hence, this Trait probably exerts the greatest

influence on the “CoP-potential” of a virtual community. All other Exemplary Traits

can be present, and to a high degree, in a hobby community (e.g. alt.fan.harry-potter);

the profession Trait, by definition, cannot.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
100
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

This Exemplary Trait is linked to JOINT ENTERPRISE, whose theoretical purview it

narrows. Theory does not limit CoPs to professions; Lave and Wenger (1991)

examine the case of rehabilitated alcoholics. However, Wenger et al (2002: 30)

argue it is easier for a CoP to define its ENTERPRISE “when there is already an

established discourse, as is the case with a professional discipline.” Furthermore,

professionally-focused CoPs dominate the literature (e.g. Orr, 1990; Brown and

Gray, 1995; Gherardi et al, 1998; Wenger, 1998), suggesting these are more

relevant, and indeed more “exemplary”, to organisational studies, than CoPs pursuing

other interests.

4.1.2.6 – High-Quality institutional documents

Mature Internet communities develop permanent documents that fulfil a practical

function, such as a charter to tell visitors what the community is about, or a FAQ to

avoid endless repetition of routine questions (Hahn, 2000). These will be referred to as

institutional documents, and are a good indicator of the capacity for collective action of

an online community and of the commitment of its members. This Exemplary Trait

increases the quality of SHARED REPERTOIRE.

4.1.2.7 – Non-conflictive

In an exemplary CoP members enjoy working collectively on domain-related problems,

which presuposes a measure of mutual respect. Wenger (1998) mentions interpersonal

conflict as a day-to-day reality of some CoPs, even as part of the practice. However, in

the Internet social environment it is easy for argument and conflict to escalate, and

some newsgroups have actually been destroyed by “flame wars”. Thus an Exemplary

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
101
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

virtual CoP will display little conflict, as manifested by Subject headers and message

tone. This Exemplary Trait intensifies the Essential Trait of COMMUNITY.

4.1.3 – Essential Traits

The Essential Traits of the model are Wenger’s constructs, discussed in Chapter Two.

Here, they are operationally defined in terms of several representative and visible

manifestations.

4.1.3.1 – MUTUAL ENGAGEMENT

Wenger (1998) emphasises sustained ENGAGEMENT is the root cause that makes a CoP

cohere and become a COMMUNITY. The combined Exemplary Traits of high-volume of

participant interaction and focus on a profession will practically guarantee the existence

of MUTUAL ENGAGEMENT, but actual confirmation requires qualitative analysis of

interactions. Real CoPs get real work done, they are not just communities of shared

interests (Wenger et al, 2002). Thus, member interactions should reflect collaborative

work (i.e. not just talk) that sustains and advances the practice. Typical manifestations

would be collaborative problem-solving, debating professional issues or sharing useful

information.

4.1.3.2 – SHARED REPERTOIRE

The presence of high-quality institutional documents is an indicator of SHARED

REPERTOIRE, but again it is not conclusive. To establish the existence of a REPERTOIRE,

the instrumental dimension of practice, it is necessary to qualitatively analyse the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
102
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

products of, and the tools deployed in, on-line interactions. Typical instances of a

SHARED REPERTOIRE include shared artifacts, shared criteria and shared practices.

4.1.3.3 – JOINT ENTERPRISE

The Exemplary Trait of professional topic goes a long way toward establishing a JOINT

ENTERPRISE, but again is insufficient by itself. In fact, JOINT ENTERPRISE is a more

elusive construct, because even though members of a CoP collectively negotiate the

ENTERPRISE and hold each other accountable, they do not need an explicit exchange or

a very frequent one to do it (Wenger, 1998). Again, the judgement of its presence is

qualitative, and a typical manifestation is that the group cares about an identifiable

domain of knowledge.

4.1.3.4 – COMMUNITY

The combined Exemplary Traits of a mid-sized stable cohesive and non-conflictual

group provide a foundation for the existence of COMMUNITY, but this requires further

qualitative assessment of interactions. Manifestations of COMMUNITY include

members’ knowledge of each other and a shared sense of community.

4.1.3.5 – LEARNING/IDENTITY ACQUISITION

Wenger’s (1998) construct of LEARNING has no antecedent in the Exemplary Trait set;

it has to be fully ascertained by participant-provided information or by examining the

content of interactions. Typical manifestations of LEARNING include new knowledge or

new skills acquired from participation in the community. Although these cannot be

directly observed, members can be expected to express satisfaction when they learn

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
103
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

something, or to report LEARNING when expressly asked. IDENTITY ACQUISITION is an

even more elusive construct, as member messages addressing issues of IDENTITY are

probably not very common. Discussion of career issues will be used as a visible

manifestation of IDENTITY ACQUISITION.

4.1.4 – Model discussion

The VCoP model is intentionally designed to address the distinct problems of search

and assessment. The Exemplary Traits address the search problem, mainly through

extensive methods. The Essential Traits address the assessment problem, mainly

through intensive methods.

The model derives its theoretical validity from being transparently linked to Wenger’s

theory through the inclusion of the Essential Traits. Although the theoretical range of the

concept of CoP is narrowed by the inclusion of the Exemplary Traits, validity is not

compromised, because Exemplary Traits do not change the nature of the Essential Traits,

they only render them more pronounced and visible. For instance, the Exemplary Trait of

High-volume of participant interactions will be present if a CoP displays strong MUTUAL

ENGAGEMENT. Likewise, the Exemplary Trait of High-quality institutional documents is a

logical outcome of a well-developed REPERTOIRE.

Imposing on the model the additional requirements of the Exemplary Traits implies a

tradeoff, since they do restrict the greater theoretical range of CoPs defined solely by

Wenger’s constructs. Thus, an acknowledged limitation of the VCoP model is that it

may well discard “un-exemplary” and yet entirely valid Internet-based CoPs, simply

because they lack one or more of the Exemplary Traits. In exchange, by demanding

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
104
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

highly-visible traits that simplify empirical detection, the model enables efficient

search and selection of communities with better-than-average CoP-potential. In

addition, by enabling extensive Internet searches, the model opens the door to a huge

search range, allowing the size and diversity of Internet to directly contribute to this

study’s success. Finally, the Exemplary Traits allow a clearer characterisation of

virtual CoPs detected by the model, because they will be more focused and energetic,

“exemplary” as it were.

The VCoP model can search and assess all kinds of Internet-based collectives. It is not

limited to a particular area of the Internet or a particular communication technology.

The only assumption is that the technology allows participants to sustain online

interactions. Neither is the model tied to particular research methods; in selecting these,

both theory and technological considerations should be weighed to make an informed

choice. In Section 4.4, the study becomes focused on a specific Internet area, but this is

a pragmatic choice, rather than a model limitation.

4.2 – Research aim and objectives

The development of the VCoP model is an important milestone since, together with the

Research Objectives, it will guide the formulation of Research Questions and Research

Strategy. This section will revisit and discuss the aims and Research Objectives initially

outlined in the Introduction; the next section will focus on the formulation of the

Research Questions. Beginning then with the aim:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
105
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The aim of this research is to extend Wenger’s theory of CoPs to the social

areas of the Internet by undertaking a systematic search for, and a rigorous

assessment of, working exemplars of “virtual” or Internet-based CoPs.

This general aim was introduced in Chapter One, but explaining why it is a worthwhile

and relevant endeavour corresponded to Chapters Two and Three. The importance of

CoPs to business organisations and the increasing turbulence in the workplace make the

study of distributed CoPs a priority. Yet, there are few rigorous studies of virtual CoPs,

and even fewer that have fully applied Wenger’s (1998) theory.

The general aim can be broken down into six specific and cumulative Research

Objectives, which are formulated and discussed next.

In order to extend Wenger’s theory of CoPs to the Internet, the first step is the

development of a theory-informed model that describes in full detail how an Internet-

based CoP should be like. Such a model must provide a hypothetical, and yet, very

concrete target for the subsequent development of an effective Research Strategy.

Hence:

Research Objective One: To build a model of Internet-based CoP that is

consistent with Wenger’s theory of CoPs, and includes theory-informed criteria

that enable selection of virtual communities with high “CoP-potential”.

This Objective has been achieved already through the development of the VCoP model,

in Section 4.1.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
106
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The second step is to use the model to devise a Research Strategy comprising the two

major tasks that structure this study: to search the Internet for virtual communities

exhibiting certain characteristics, and to theoretically assess CoP attributes of detected

communities. Therefore:

Research Objective Two: To design, under the guidance of the model, a

Research Strategy that will, first, enable detection of Internet-based

communities exhibiting high CoP-potential, and second, enable theory-

grounded analysis of selected communities to rigorously assess whether or not

they truly are CoPs.

The next Objective is the first to address the field research proper. This begins with a

broad search of the Internet aimed at locating the virtual communities that will

participate in this study on the basis of their strong affinity to the VCoP model. The

Objective can be framed thus:

Research Objective Three: To conduct a highly comprehensive search of the

Internet, and to locate the virtual communities that display the strongest affinity

to the model of Internet-based CoP.

Affinity to the model gives high CoP-potential to selected communities, but

considerably more evidence is required to make a valid CoP assessment. Therefore,

two independent and triangulating methods will be deployed to obtain quantitative and

qualitative evidence of the presence or absence of the Essential Traits in each

community. The corresponding Objectives are:

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
107
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Research Objective Four: To question regular participants of selected

communities about the presence or absence of visible manifestations of

Wenger’s constructs.

Research Objective Five: To directly examine online interactions of regular

participants in selected communities to determine the presence or absence of

visible manifestations of Wenger’s constructs.

The final Research Objective calls for integrating all available evidence to make an

assessment of which, if any, of selected communities can be characterised as an

Internet-based CoP lacking none of the attributes identified in Wenger’s CoP theory.

Research Objective Six: To assess the presence or absence of Wenger’s constructs

in each community, using concurring evidence from two independent methods, and

thereby establish whether any among selected virtual communities display all

theoretical properties of CoPs, and can therefore be characterised as such.

The Research Objectives describe the specific tasks the study must accomplish. The

forthcoming Research Strategy will address the techniques and procedures for

achieving the Objectives. First, though, it is important to carefully think through the

Research Questions the study must answer to fully achieve the Research Objectives, so

that the Strategy can be designed accordingly.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
108
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.3 – Research Questions

The Research Questions of the thesis are derived directly from the VCoP model. The

various sets of questions mirror the logic of the model in that some questions address

issues of community pre-selection, while others address the ultimately decisive issues

of CoP-assessment. This section will formulate explicit research questions to cover

every aspect of the study, in order to make the research logic more transparent, and

render implicit assumptions visible.

The approach followed by proposing the VCoP model is to describe conjectural

Internet CoPs, devise instruments for detecting them, and assess them rigorously

according to Wenger’s theory. This is a postpositivistic approach, as discussed in

Section 4.5. Therefore, the Research Questions will be framed as testable statements,

even if they are not statistical hypothesis, in order to better reflect this research

approach and philosophy.

4.3.1 – Principal Research Question

The overarching Research Question this thesis addresses, which logically follows from

the previously stated aim, can be framed as follows:

Principal Research Question. There exist stable Internet-based groups that

function as Communities of Practice.

The scope of this central question is very broad because Wenger’s concept of CoP

is complex, comprising the interrelated constructs described in Chapter Two.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
109
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Therefore, the Principal Question will be addressed indirectly, through a set of

subsidiary research questions centred on the Essential Traits of the VCoP model, as

described later in this section. First, though, some preliminary issues must be

addressed.

4.3.2 – Preliminary Research Questions

The Principal Question implies suitable groups must first be located in the Internet

before they can be examined to determine whether or not they function as CoPs.

Therefore, as a necessary antecedent, several methodological questions are

formulated addressing the feasibility of an Internet search for hypothesised virtual

CoPs. The intent of these questions is to make explicit the various assumptions

underlying this research, thus enabling critical discussion. Preliminary Research

Questions are framed thus:

Preliminary1. There exist large numbers of stable Internet-based groups, and

efficient techniques can be devised to locate them.

Preliminary2. Internet-based groups exhibit considerable variation along the

following dimensions:

a) the size of the group

b) the volume of participant interaction

c) the extent to which they adopt a core-periphery structure

d) their success in maintaining a focused discussion

e) the topics they focus on

f) the institutional documents they develop

g) the level of conflict in participant interactions

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
110
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Preliminary3. Techniques can be devised to assess the affinity of Internet-

based groups along each dimension to the corresponding Exemplary Trait of the

VCoP model.

Preliminary4. Internet-based groups exhibiting all Exemplary Traits to a high

degree are more likely to be CoPs.

The assumptions reflected in the Preliminary Questions are plausible, but they cannot

be proved before launching the research, because there is no ex-ante way of

determining the complete population or the character of stable Internet-based groups.

Hence, only the success of the study can provide an affirmative, though limited,

response to these questions. Until then, the study must take for granted these

assumptions, which are discussed next.

The existence of Internet-based communities is well-established (e.g. Roberts, 1998;

Wellman and Gulia, 1999), making plausible the premise of Preliminary1. It is up to the

Research Strategy to propose techniques that can efficiently locate these communities;

their success or failure in the field will provide the basis for addressing the question.

The second Preliminary question assumes large numbers of Internet communities will

be sufficiently heterogeneous to achieve considerable variation along seven observable

group characteristics which match the Exemplary Traits of the VCoP model. Again,

confirmation depends on field results.

Assuming Preliminary2 is in fact correct, Preliminary3 posits techniques can be developed

to rank groups along each dimension according to their greater or lesser affinity to the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
111
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

criteria specified by the Exemplary Traits. Again, it is up to the Strategy to propose such

techniques, and their success in the field will give grounds for answering the question.

Finally, Preliminary4 states the strongest assumption of all: that the subset of Internet

groups exhibiting greatest affinity to the Exemplary Traits of the VCoP model is more

likely to contain true CoPs. It is a strong assumption because if proven correct, then it

would mean CoPs are detectable over the Internet, at least to a certain extent. The

plausibility of this assumption rests on the following rationale: In general, the presence

of the Exemplary Traits will tend to make selected virtual communities more focused,

productive and energetic. More particularly, though, if the community is centred on a

profession, and a practice is thus assured, its very focus, productivity and energy will

give it better odds of achieving the status of CoP. This is why the Exemplary Trait of

professional topic was previously judged the one most responsible for the “CoP-

potential” of a virtual community. As before, though, confirmation of Preliminary4 will

have to await field results.

4.3.3 – Exemplary Trait Research Questions

Having explicitly addressed the feasibility of Internet searches, the next set of questions

addresses the criteria which should guide those searches in order to increase the

probability of eventual success. These criteria are none other than the Exemplary Traits

of the VCoP model, hence the name of this set of questions. They are:

Exemplary1. An Exemplary Internet-based CoP is mid-sized.

Exemplary2. An Exemplary Internet-based CoP exhibits a high volume of

participant interaction.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
112
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Exemplary3. An Exemplary Internet-based CoP exhibits a core-periphery

structure.

Exemplary4. An Exemplary Internet-based CoP exhibits highly-focused

discussions.

Exemplary5. The topic of an Exemplary Internet-based CoP is an identifiable

profession.

Exemplary6. An Exemplary Internet-based CoP exhibits high-quality

institutional documents.

Exemplary7. An Exemplary Internet-based CoP is non-conflictive.

Discussion of these questions is omitted since it would parallel previous discussion of

the Exemplary Traits in Section 4.1.

4.3.4 – Essential Trait Research Questions

Assuming then that a number of Internet groups with high affinity to the Exemplary

Traits (and therefore high CoP-potential) can be found, the Principal Research Question

can be addressed. This is done by formulating an additional set of interrelated questions

focused on the Essential Traits, hence the name of the set. They are:

Essential1. There exist groups that sustain MUTUAL ENGAGEMENT through

Internet-based interaction.

Essential2. There exist groups that negotiate a JOINT ENTERPRISE through

Internet-based ENGAGEMENT.

Essential3. There exist groups that develop a SHARED REPERTOIRE through

Internet-based ENGAGEMENT.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
113
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Essential4. There exist groups that cohere as a COMMUNITY through Internet-

based ENGAGEMENT.

Essential5. There exist groups whose members achieve LEARNING through

Internet-based ENGAGEMENT.

Essential6. There exist groups whose members ACQUIRE AN IDENTITY through

Internet-based ENGAGEMENT.

These questions can only be said to provide an affirmative answer to the Principal

Research Question when they are all answered affirmatively. This is because each of

them enquires after an essential element of a CoP, even if there is some degree of

overlap. It is simply a different way of stating that, as defined by this study, assessing a

group as a CoP requires the presence of all Wenger constructs.

4.4 – Focusing the study on the Usenet discussion network

Having arrived at a set of Research Questions that will provide direction and focus to this

study, and before designing the Research Strategy, a prior decision must be made: the

informed choice of Internet area this study will focus on. The Internet is a complex

network that encompasses many different communication technologies. Several have

been designed to support one-to-many communication, in order to facilitate online

discussion. It is in these “social areas” of the Internet that different types of virtual

communities exist (for typologies, see Kling and Courtright, 2003; Porter, 2004). The

most common technologies are:

• Name-brand discussion groups, such as Yahoo Groups (DeSanctis et al, 2003).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
114
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• Private online discussion forums set up by professional associations

(Romano, 1998; Wasko and Teigland, 2002).

• Multi-user Dungeons, also known as MUD’s (Schlager et al, 2002;

Davenport, 2004).

• Blogs, which are personal journals published on the Web and commented on by

readers (Baker and Green, 2005). The name is a shortened form of “Web Log”.

• Chat rooms, which are spaces for synchronous (i.e. live) computer-mediated

conversation.

• Mailing lists, also known as listservs (Thomsen, 1996; Millen and Dray,

2000).

• Usenet newsgroups (Wasko and Faraj, 2000; Baym, 2000).

The communication dynamics characterising each of these technologies are quite

different: some are synchronous (e.g. chat rooms), most are asynchronous (e.g. listservs

and newsgroups); some are open to any participant (e.g. most newsgroups), while

others erect significant barriers to entry (e.g. private discussion groups). These

differences impact the quality and character of the social spaces that develop around the

technologies (Hahn, 2000). Moreover, despite their being all Internet-based

technologies, each poses different technical problems regarding the search for virtual

communities. Hence the need to tailor the Research Strategy to a particular area. The

informed choice was the Usenet discussion network, for several reasons.

First, Usenet can be efficiently searched using Smith’s (1999) newsgroup analysis tool,

Netscan. Several Traits of the Exemplary set can be measured with Netscan, and

analysed newsgroups can be sorted according to their affinity to them.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
115
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Second, Social Network Analysis (SNA) can be applied to Usenet newsgroups to reveal

stable cohesive subgroups (Murillo, 2002). In particular, a core-periphery model can be

fit to individual newsgroups, thus measuring another Exemplary Trait.

Third, Usenet is big: it is the largest and oldest discussion network on Earth (Hahn, 2000).

Size makes the search more challenging, yet it is a resource as well, because large numbers

of heterogeneous participants improve the odds of achieving collective action. Oliver and

Marwell’s (1988) theory of the critical mass predicts that large networks are more likely to

succeed in collective action because they have a better chance of containing a number of

dedicated and competent “organisers” that can band together and mobilize others.

Fourth, the asynchronous character of Usenet communication may foster deeper analysis

and reflection of issues. Benbunan-Fich and Hiltz (1999) found student teams using

asynchronous discussion software to solve a case study problem produced better and longer

reports than teams solving the same case through face-to-face discussion.

Fifth, Usenet represents an interesting limit case given its low media-richness. If complex

MUTUAL ENGAGEMENT can be effectively achieved within the limitations of plain-text

messages, giving rise to true Usenet-based CoPs “in the wild”, then it should be easier to

achieve in the richer media provided by modern intra-organisational networks.

Finally, there is considerable previous research about Usenet virtual communities,

providing useful lessons for this thesis. Relevant studies are reviewed in this section,

immediately following some background about Usenet’s structure.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
116
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Usenet is self-organised by discussion topic into thousands of distinct newsgroups,

providing a practical way for people with shared interests to find each other and talk. With

the number of newsgroups exceeding 100,000 (Smith, 2002), every topic of human interest

is discussed somewhere, from wildlife photography to medieval genealogies. Early in the

history of Usenet, a classification scheme was established to help people find the topics

they were interested in. A system of categories called hierarchies was launched between the

summer of 1986 and March of 1987, with seven initial categories (Hahn, 2000):

comp Computers

sci science and technology

news Usenet itself

rec recreation, hobbies, arts

soc social and cultural issues

talk debate, controversial topics

misc miscellaneous

Today, each hierarchy contains thousands of newsgroups focused on the specific

topic announced in their name. Newsgroup names consist of two or more parts,

separated by dots. The first part states the hierarchy, and the others describe with

increasing specificity the topic of the newsgroup. For example, the newsgroup

comp.lang.java discusses the JAVA computer language. As discussion becomes more

and more specialised, new groups are launched with increasingly specific names,

such as comp.lang.java.security. This hierarchical organisation, coupled with the

huge number of participants reached worldwide, has supported the formation of many

topically-focused communities. By conducting discussions in public, newsgroups

signal worldwide their topic, level of expertise and the often colourful personalities of

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
117
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

members. Hence, they attract new participants, either as occasional question-askers or

as new stable members.

Previous studies about Usenet-based communities provide evidence of the presence of all

Essential Traits and some Exemplary Traits. Therefore, though not the intended aim of

the studies, they do provide indirect support for the thesis that Usenet CoPs are feasible.

Despite newsgroup posters’ supposed anonymity, which would arguably render

community impossible, Donath (1998) found identity cues in Usenet are closely

monitored by participants, who become attuned to nuances of e-mail addresses and

signature styles. In time, persistent contributors to newsgroups make a strong impression

on other members, and build up a fairly accurate reputation. Hence it is very difficult for

a regular participant to achieve perfect anonymity. Many simply opt for pseudonymity,

posting regularly under an assumed name that is untraceable to a real world person.

Having made this qualification, COMMUNITY is the Essential Trait most often reported in

Usenet research. North (1994), an early researcher of Usenet, performed an ethnography

of newsgroups, actively participating in several, and conducting some online interviews.

He reports a very social environment and a strong sense of online community, despite

what he calls Usenet’s “faceless textual interface”. Patterson (1996) examined newsgroup

alt.good.morning, and found a rich culture, local institutions, language play, and the firm

conviction of members of belonging to a virtual community. In addition, online surveys

conducted by several studies have found Usenet-based communication supports the

formation of true personal relationships (Parks and Floyd, 1996; Roberts, 1998;

McKenna, Green and Gleason, 2002), thus yielding evidence for the sub-construct of

‘Members’ knowledge of each other.’ More to the point, respondents report a “sense of

community” in specific newsgroups (Parks and Floyd, 1996; Roberts, 1998).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
118
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Evidence of SHARED REPERTOIRE comes in the form of communities developing and

enforcing local rules and conventions, usually to protect discussion focus, an Exemplary

Trait. Kollock and Smith (1996) argue the difference between successful and

unsuccessful newsgroups lies on the former’s ability to enforce local norms and customs

aimed at solving social dilemmas such as posting off-topic messages. McLaughlin,

Osborne and Smith (1995) performed a discourse analysis of all messages posted to five

newsgroups during a three week period to detect conduct-correcting episodes. They

derived a taxonomy of Usenet offences, and concluded regular participants enforce

respect of local norms and idiosyncrasies to protect the identity and smooth functioning

of their virtual communities. Tepper (1997) describes how members of alt.urban.folklore,

who call themselves ‘Old Hats’, adhere to subtle but effective local customs to reinforce

the identity and subculture of the community and clearly mark its boundaries.

The strong topical focus displayed by some newsgroups can be interpreted as evidence of

JOINT ENTERPRISE. Sivadas, Grewal and Kellaris (1998) surveyed six newsgroups

focused on different types of music, and found posters to be highly involved with their

newsgroup, and with the type of music it espouses (e.g. buying recorded music, attending

concerts, etc.), but minimally so with respect to other music newsgroups. More in-depth

case studies of fan communities are provided by Watson (1997) and Baym (2000); the

first about a rock group, the second about soap operas. More evidence of JOINT

ENTERPRISE is afforded by support newsgroups, which not only discuss a topic, but

actively provide social and emotional support to their members. Winzelberg (1997)

applied content analysis to a sample of messages from a newsgroup oriented to

individuals with eating disorders and discovered strong emotional support of experienced

newsgroup participants toward newbies. Preece (1999) described how empathy was

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
119
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

delivered in a group focusing on knee-injuries, where, unlike most other newsgroups,

elementary oft-repeated questions from newbies did not elicit reproachful or ironic replies

(flames). Newsgroup veterans were extremely patient and understanding with newbies,

striking a thoughtful balance in their provision of factual information and sympathy.

Evidence of MUTUAL ENGAGEMENT, and indirectly of LEARNING, is anecdotal but

comes from numerous studies. Usenet communities provide sophisticated technical

support to practitioners from diverse professional fields such as acounting (Alexander,

2000), computer programming (Wasko and Faraj, 2000), cellular biology (Hengen,

1997), taxes (Fuller, 1999), veterinary diagnosis (Lofflin, 2000), handcrafts (Lovelace,

1998), and law (Samborn, 1999; Ricci, 2000). As Wellman (1999: 15) aptly sums up:

Usenet “supports emotional, nuanced and complex interactions, belying early fears that

it would be useful only for simple, instrumental exchanges”.

Two studies provide evidence of IDENTITY ACQUISITION. Lakhani and von Hippel

(2000) conducted a survey of newsgroup comp.infosystems.www.servers.unix and

found strong identification of participants with their online community. Blanchard and

Markus (2004) performed participant observation and member interviews in a Usenet

community focused on triathlons, and found participants create through their posts an

online identity closely aligned to the goals of the community, and also display

recognition of the online identities constructed by other members.

Finally, evidence for the Exemplary Trait of core-periphery structure comes from several

studies that report a clear differentiation of participants with respect to their involvement

in the community. Some are stable and prolific participants, others post occasionally, and

some post only once. They use the term “core” to refer to the subset of stable and active

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
120
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

participants, who dominate the discussion and heavily influence the identity, culture and

local customs of the community (McLaughlin et al, 1995; Thomsen, 1996; Tepper, 1997;

Smith, 1999; Millen and Dray, 2000; Blanchard and Markus, 2004). Previous research by

this author should also be mentioned, since it provides an important methodological

foundation for this thesis. Three SNA models of cohesive sub-group –cliques, k-plexes

and continuous core-periphery structure– were fitted to 30-week message samples from

five Usenet newsgroups: comp.lang.c++.moderated, humanities.philosophy.objectivism,

misc.invest.mutual-funds, misc.writing and sci.physics. The cohesive sub-groups were

tested for activity (the number of exchanged exchanged) and stability (by separately

testing for activity in each 10-week sub-sample). The study found stable and active

cliques and k-plexes in four out of five newsgroups, and stable and active core-periphery

structures, also in four out of five newsgroups (Murillo, 2002). The SNA methods used

there have been adapted for this doctoral research.

4.5 – Research philosophy

The researcher followed a particular research logic from the start; only later was the

approach reflectively identified as postpositivism or Critical Rationalism. Accepting

Wenger’s (1999) argument that CoPs have existed wherever and whenever human beings

congregate (ancient tribes and medieval guilds are given as examples); the researcher

hypothesised they would also spontaneously convene in the social spaces of Usenet;

hence the early focus on designing detection instruments. The attempt to apply a well-

established theory to a new domain, and the described hypothesis-testing logic,

correspond to what Blaikie (1993) calls a deductive research strategy. This fits the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
121
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

researcher’s realist ontology, and an epistemology that while acknowledging the theory-

ladenness of observation, and the approximative and perfectible condition of scientific

explanations (Artigas, 1992; 1999), is still committed to the notions of objectivity and

truth as the guiding ideals of the research enterprise (Phillips, 1990; Searle, 1993). As

Miles and Huberman (1994: 4) aptly sum up:

Human relationships and societies have peculiarities that make a realist approach

to understanding them more complex–but not impossible. Unlike researchers in

physics, we must contend with institutions, structures, practices, and conventions

that people reproduce and transform. Human meanings and intentions are worked

out within the framework of these social structures–structures that are invisible but

nonetheless real. In other words, social phenomena, such as language, decisions,

conflicts, and hierarchies, exist objectively in the world and exert strong influences

over human activities because people construe them in common ways. Things that

are believed become real and can be inquired into.

Although more frequently associated with quantitative methods, the deductive research

strategy also admits qualitative methods as well as combinations of quantitative and

qualitative methods, provided they are used with the same ontological assumptions

(Blaikie, 2000). In this case, the ontology is realist throughout, and the hypothesised virtual

CoPs are the assumed ‘objective’ target of the study. Therefore, a compatibilist, pragmatist

position will be adopted, allowing the researcher to mix and match design components to

achieve the best fit with the research problem and logic (Johnson and Onwuegbuzie, 2004).

Specifically, quantitative and qualitative methods will be combined to develop a more

reliable composite assessment of hypothesised virtual CoPs, and to build triangulation into

the study, thus increasing validity (Rocco, Bliss, Gallagher and Pérez-Prado 2003).

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
122
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The first problem the Research Strategy must address, previously referred to as the

search problem, is the detection of virtual communities in Usenet. Not all newsgroups

contain virtual communities, but those that do exhibit a dense and persistent cluster of

communication between a subset of stable participants, and suitable methods must be

devised to locate such clusters. The second problem, earlier referred to as the

assessment problem, involves in-depth examination of detected communities to

determine whether or not the Essential Traits of the VCoP model are present.

The search problem will be addressed by the combined deployment of two quantitative

tools that are particularly suited to the detection and measurement of dense clusters in

Usenet, the Netscan analyser and Social Network Analysis. The assessment problem

will be addressed through the combined use in selected communities of quantitative

and qualitative methods, in order to build robust triangulation into the study (Rocco et

al, 2003). The process is illustrated in Figure 4.2, which displays the social network of

a hypothetical virtual community exhibiting a core-periphery structure, as revealed by

strong ties between core members. The community will be assessed by two

independent instruments. A quantitative Survey will be applied to a sample of

community members, enquiring about visible manifestations of Wenger’s constructs,

i.e. the sub-constructs. This will be complemented by a qualitative Content Analysis of

the interactions between core-members, again targeting the sub-constructs. Both sets of

results will be triangulated into a single composite assessment regarding the presence in

the community of the Essential Traits. Specifically, the assessment will require that the

presence of each Essential Trait be corroborated by concurring results from both

instruments.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
123
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Essential Trait assessment


through concurring survey
MUTUAL and content analysis results
ENGAGEMENT

Essential Traits
manifested in
online interactions

Essential Traits
reported by survey
respondents

Virtual community with


core-periphery structure

Figure 4.2 – Triangulation of Survey and Content Analysis results

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
124
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Like the previous decision making Exemplary Traits part of the model, the requirement

that each construct be independently detected by two instruments involves a tradeoff.

Making the assessment more rigorous reduces the field of candidates, but increases

confidence in the validity of results. In addition, the use of two separate instruments

will provide a richer picture of the examined communities than either method could

achieve separately.

Yauch and Steudel (2003) used a similar mixed methods approach to assess the

organisational cultures of two small manufacturers. The authors first used qualitative

document review, participant observation and group interviews to identify organisational

values, assumptions and behavioural norms. They followed up with a quantitative survey,

the Organizational Culture Inventory (Cook and Lafferty, 1987), a validated instrument

which measures 12 sets of behavioural norms using 10-item scales. Robust triangulation

was achieved when both qualitative and quantitative methods corroborated the presence

of particular cultural factors in the studied firms. Specifically, both methods identified

Avoidance and Complancency behaviours at one of the firms. In addition, the use of

mixed methods resulted in deeper understanding of the organisation than either method

on its own could have accomplished (Yauch and Steudel, 2003).

The search and assessment problems are both addressed with extensive, rather than

intensive methods. An alternative strategy would be to perform an ethnography of one

or several potential CoPs, as done in most previous studies (Orr, 1990; Lave and

Wenger, 1991; Wenger, 1998). However, ethnography is extremely time-consuming;

examining more than two or three communities would exceed the scope of a doctoral

project. In addition, the problem would remain of how to first locate communities with

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
125
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

high CoP-potential, so that the ethnography is well-targeted. Hence, extensive methods

were preferred, with an emphasis on discovery prior to in-depth analysis.

Having discussed the research philosophy, a more detailed description of the Research

Strategy follows, though methods are discussed separately, in Chapter Five.

4.6 – Research Strategy

The VCoP model will be the blueprint for developing the Research Strategy, a multi-stage

process structured around the Research Objectives, and targeting the Research Questions.

4.6.1 – Stage I: Model Specification

Stage I of the Strategy, corresponding to Objective One, is to build a theoretically-

grounded model of Internet-based CoP as a hypothetical but concrete target for the

Usenet search. Model specification is not only a major output of the Research Strategy

but also a key input, because by specifying a target for this research, the model

substantially determines ulterior methodological choices. For this reason, model

specification was actually addressed in Section 4.1, and has since guided the

formulation of Research Questions and Strategy.

4.6.2 – Stage II: Strategy Formulation

Each Stage of the Research Strategy is an operational response to a specific Research

Objective. Moreover, the Strategy follows the design and logic of the VCoP model.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
126
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Therefore, it is a two-part process: a first part consisting of an extensive search for

suitable Usenet communities, conducted under the guidance of the Exemplary Traits,

and a second part consisting of intensive analyses of selected communities, under the

guidance of the Essential Traits. The methods and techniques deployed in each part

constitute the remaining Stages of the Strategy as described next.

4.6.3 – Stage III: Usenet Search

The Third Stage, which addresses Objective Three, involves the selection of

participating newsgroups, a critical task on which hinges the eventual success of the

entire research. It will rely on the Netscan newsgroup analyser to perform

comprehensive Usenet searches. Operational quantitative and qualitative selection

criteria will be derived from the Exemplary Traits. Among quantitative criteria are

Netscan statistics of posting activity and the poster-to-post ratio, which will aid in

identifying newsgroups with dense cores. Qualitative criteria include professional

topics and high-quality institutional documents. The criteria will be progressively

applied to, and a ranking performed of, the initial population of newsgroups,

beginning with extensive quantitative criteria, followed by more time-intensive

qualitative criteria, and finally, the core-periphery analysis, which being the most

time-intensive will be applied to a much reduced newsgroup subset. Given these

stringent selection criteria, and the large size of the initial sampling universe, finalist

communities will exhibit the highest available affinity to the Exemplary Traits of the

VCoP model, and therefore, highest possible CoP-potential. The results of this Stage

will provide elements to addresses both Preliminary and Exemplary Trait Research

Questions.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
127
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.6.1 – Stage IV: Participant Survey

Stage IV, corresponding to Objective Four, involves questioning regular participants of

selected communities about the presence or absence of the Essential Traits of the VCoP

model. In order to maximise convenience to respondents, and likely response rates, a

web-based Survey will be used, based on point-and-click Likert scales. The design of

scales will closely follow the Essential Trait Research Questions. Specifically, the

Survey will ask high-coreness community members about visible indicators of

Wenger’s constructs, such as collective problem-solving in the case of MUTUAL

ENGAGEMENT. The Survey will thus gather quantitative evidence about the Essential

Trait Research Questions.

4.6.5 – Stage V: Content Analysis of Discussions

Research Objective Five calls for direct examination of newsgroup discussions, under

the premise that if selected communities are CoPs, then online interactions should yield

evidence of Wenger’s constructs. This will be accomplished through a Content

Analysis of selected discussion threads. Because of the huge amount of textual

discussion each community contains, a theoretical sample of threads will be selected,

using theory-informed criteria from the VCoP model. Specifically, threads dominated

by core members and focused on professional topics will be selected. The coding

categories for the Content Analysis will be derived from the Essential Traits, using the

same indicators previously used to design the Survey. The Content Analysis will thus

gather qualitative evidence about the Essential Trait Research Questions.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
128
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.6.6 – Stage VI: CoP Assessment

The final stage in the Research Strategy is making a critical synthesis of the quantitative

and qualitative data obtained about each community, and assessing to what extent each of

them can be considered a true CoP. This is the agenda of Research Objective Six. Each

community will be tested for the presence of the complete set of Wenger constructs. To

maximise validity, the presence of each construct must be confirmed by concurring

results by both the Survey and the Content Analysis. Communities exhibiting all of

Wenger’s constructs, and therefore lacking none of the theoretical attributes of CoPs,

will be assessed as Exemplary Usenet-based CoPs.

4.6.7 – The Funnel Research Strategy

The Research Strategy outlined above can be visually represented, using the

conventional image of a Funnel, as displayed in Figure 4.3. The logic of the Funnel

Strategy involves taking in as large an input as possible of Usenet newsgroups, through

the upper opening of the Funnel, and making a first selection using quantitative criteria

from the Exemplary Traits that can be extensively applied through Netscan’s wide

scope. This results in a considerably reduced newsgroup subset which can then be

examined, through direct observation, for the qualitative criteria of the Exemplary

Traits; a more time-intensive method. This once again results in a substantial reduction

of elligible newsgroups, which can then be analysed through the computationally-

intensive core-periphery analysis. Virtual communities exhibiting the highest affinity to

the Exemplary Traits are then allowed to proceed to the next level of testing, where the

Essential Traits come into play. Two methods will independently test for the presence

or absence of Wenger’s constructs, the Survey of high-coreness participants, and the

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
129
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

USENET

Quantitative Exemplary Traits


• High-volume participant interaction
• Mid-sized subgroup
• Highly focused discussions Stage III

Qualitative Exemplary Traits Netscan


• Topic a profession
• High-quality documents Qualitative
• Non-conflictive analysis

Core-periphery
• core-periphery structure analysis

Essential Traits
Stage IV – Survey
Wenger’s
constructs
Stage V – Content Analysis
Stage VI – Assessment and
identification of VCoPs

● ●
● ● Hypothesised VCoPs

Figure 4.3 – The Funnel Research Strategy

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
130
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Content Analysis of core-member discussions. The final Stage will critically assess

the quantitative and qualitative data obtained by all previous Stages, and identify

communities exhibiting all Essential Traits of the VCoP model. The planned outcome

of this study is for at least some virtual communities to successfully navigate the

multiple filters of the Funnel by demonstrated possession of all Essential and

Exemplary Traits of the VCoP model; thus qualifying as Exemplary Usenet-based

CoPs. Throughout, the logic of the Funnel design favours deployment of more

extensive methods near the top (notably, Netscan) and more intensive methods near

the bottom (notably, the Content Analysis). This makes it possible to conduct a very

extensive Usenet search, and still make it compatible with very time-intensive

assessment methods.

For illustrative purposes, the heavily process-oriented Funnel diagram of Figure 4.3 can

be complemented with Figure 4.4, which rather emphasises the correspondence

between Research Objectives and Strategy Stages, and the latter’s position within core

thesis chapters. This Figure was previously displayed, for summary purposes, at the end

of the Introductory chapter.

Finally, an approximate timeline of the execution of the Research Strategy is provided

in Appendix A. It should be noted that the PhD was undertaken on a part-time basis,

which is the reason the field research eventually spanned over two years.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
131
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Research Objectives Research Strategy Thesis Chapter

Objective One Stage I

Build a theoretically-grounded Model Specification Chapter Four


model of Internet-based CoP

Objective Two Stage II

Develop an effective strategy for Strategy Formulation Chapter Four


locating and assessing Internet CoPs

Objective Three Stage III

Comprehensively search the Internet Usenet Search Chapter Five: design


and identify virtual communities with Chapter Six: results
the strongest affinity to the model

Objective Four Stage IV

Ask community participants about the Participant Survey Chapter Five: design
presence of Wenger constructs Chapter Seven: results

Objective Five Stage V

Directly examine participant interaction Content Analysis Chapter Five: design


for the presence of Wenger constructs Chapter Eight: results

Objective Six Stage VI

Assess the presence of Wenger CoP Assessment Chapter Nine


constructs, and identify communities
displaying all the properties of CoPs

Figure 4.4 – Research Objectives and Strategy Stages

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
132
Research Design
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

4.7 – Summary

This chapter has addressed the issues involved in the research design. It began by

using the theories presented in Chapters Two and Three to build a consistent model of

Internet-based CoP to act as a blueprint for the research design. In light of this model,

the aim and Research Objectives of the thesis were formulated and discussed.

Working from these goals, and making explicit the assumptions of the VCoP model,

the Research Questions were then formulated and discussed. The important decision

to focus the study on the Usenet network was explained, and previous studies on

Usenet-based communities were reviewed. The next section discussed the ontological

and epistemological assumptions underlying the study, which were most closely

identified with Critical Rationalism, and also discussed the use of mixed methods to

achieve triangulation. Finally, Section Six develops a Research Strategy uniquely

tailored to the search and assessment problems defined by the Research Objectives

and Research Questions. The Strategy can be represented by the image of a Funnel,

and will be so named.

The next chapter provides a detailed discussion of the methods deployed in each Stage

of the Strategy.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
133
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

CHAPTER FIVE

Methods

The previous chapter formulated the Research Objectives and Research Questions, and

tailored a Research Strategy to address them. Several methods are deployed within this

Strategy; the aim of Chapter Five is to discuss each. The methods are independent from

one another, and their collective coherence stems from their coordinated deployment

within the Funnel Strategy. The chapter is therefore organised into five sections

addressing respectively the Netscan newsgroup analyser, the Borgatti-Everett core-

periphery model, the web-based Survey, and the Content Analysis of threads, plus a

final summary section.

5.1 – The Netscan analyser

The original research problem posed a substantial challenge: to devise a procedure

capable of comprehensively searching the Usenet network, and capable of detecting

and ranking virtual communities according to their measured affinity to the VCoP

model. This section describes the Netscan newsgroup analyser, and how it will be

deployed to address these distinct search and assessment problems.

Netscan was developed by Smith (1999) to measure and record the flow of messages

posted to over 14 thousand newsgroups during a ten-week period from November 1996

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
134
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

to mid-January 1997. His aim was to map activity patterns over the entire Usenet

hierarchy. He found that cross-posted messages establish dense ties between topically-

related newsgroups; he also noted many newsgroups adopted core-periphery patterns.

Smith proposed three measures of social interaction within a newsgroup. The poster-to-

post ratio (PPRatio) is the reciprocal of the average number of messages posted by each

participant, and ranges from zero to one. A poster-to-post ratio close to zero indicates a

small number of participants contribute a large proportion of the newsgroup’s messages.

The thread-to-post ratio (TPRatio) is a more direct measure of social interaction; it can be

interpreted as the proportion of undirected messages or thread-heads in the newsgroup. A

value close to zero indicates a greater proportion of directed messages, and therefore, of

social interaction between participants. Finally, the percentage of cross-posting (%Cross-

Post) refers to the proportion of messages posted to more than one newsgroup. Echoing an

earlier suggestion by McLaughlin et al (1995), Smith (1999) argues newsgroups with a low

percentage of cross-posting are more likely to contain clearly defined topical communities,

while newsgroups with a large percentage of cross-posting evince a lack of focus.

Nowadays, Netscan is publicly available at http://netscan.research.microsoft.com. It

can comprehensively scan complete Usenet hierarchies (e.g. comp.*) containing

thousands of newsgroups, and calculate Smith’s measures of interaction, which can be

used to operationalise the quantitative Exemplary Traits of the VCoP model. This is

best illustrated through a worked example: Table 5.1 displays the results of a limited

search: all newsgroups whose name includes the word “invest”, for the period

September 1-30, 2002, sorted by descending number of posts. Definitions of raw

Netscan results are provided in Figure 5.1.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
135
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Table 5.1 – Netscan search for “invest” sorted by Posts (September 1-30, 2002)
UnR
No Newsgroups Ordered by Posts, descending Posts Posters PPRatio Returnees Replies Singles
MSGS
XPosts
1 misc.invest.stocks 7063 811 .11 274 5446 385 688 930
2 misc.invest.mutual-funds 1614 188 .12 63 1426 92 84 144
3 aus.invest 1087 266 .24 84 818 156 125 384
4 microsoft.public.investor.discussions 649 65 .10 15 518 31 68 63
5 alt.private.investigator 617 143 .23 27 385 76 144 157
6 misc.invest.futures 529 200 .38 71 315 122 136 147
7 misc.invest.financial-plan 496 84 .17 35 402 36 18 7
8 misc.invest.canada 423 161 .38 30 257 94 122 171
9 misc.invest.real-estate 420 208 .50 45 125 143 248 133
10 it.economia.investire 396 141 .36 29 257 76 87 25
11 misc.invest.technical 322 168 .52 47 136 108 134 146
12 alt.invest.real-estate 281 148 .53 29 76 100 170 74
13 misc.invest.misc 246 148 .60 30 81 116 136 120
14 alt.invest.penny-stocks 243 121 .50 20 115 83 102 79
15 misc.invest.options 223 111 .50 34 110 70 79 70
16 alt.invest 169 102 .60 21 49 67 102 77
17 misc.invest 142 98 .69 11 44 68 88 70
18 alt.invest.market.crash 137 59 .43 7 78 37 49 88
19 misc.invest.marketplace 115 80 .70 18 24 59 77 40
20 sg.invest.stocks 110 68 .62 13 35 45 61 80
21 dk.penge.investering 106 52 .49 18 72 32 17 6
22 alt.invest.penny-stock 69 41 .59 6 26 26 35 37
23 alt.invest.real-estate.methods 51 32 .63 4 0 21 50 7
24 alt.investors 43 34 .79 4 3 28 37 3
25 biz.marketplace.investors 39 28 .72 4 4 21 33 1
26 misc.invest.stocks.penny 33 27 .82 5 6 23 24 9
27 japan.finance.invest 29 17 .59 1 1 14 27 13
28 realtynet.invest 27 22 .81 4 1 19 25 2
29 acc.sbell.usa-today.invest 26 14 .54 2 0 9 26 1
30 alt.invest.technical-analysis.omega 26 20 .77 3 3 16 20 9
31 alt.invest.fsbo 24 18 .75 1 4 13 18 7
32 asu.investigates.info-highway 21 12 .57 2 0 7 21 3
33 uw.clubs.invest 21 14 .67 2 1 9 20 4
34 misc.invest.funds 15 8 .53 0 0 5 15 0
35 alt.invest.technical-analysis 15 11 .73 1 0 9 15 1
36 scnr.isp-investors 13 7 .54 1 8 6 1 0
37 misc.invest.stock 13 9 .69 0 1 7 10 2
38 misc.invest.precious-metals 6 6 1.00 1 0 6 6 1
39 jaring.invest 5 1 .20 1 0 0 5 5
40 microsoft.public.msinvester 5 4 .80 0 4 3 0 5
41 misc.invest.index-futures 5 5 1.00 0 0 5 5 1
42 alt.invest.market 5 5 1.00 0 1 5 4 2
43 alt.invest.penny 5 5 1.00 0 1 5 4 2
44 misc.invest.commodities 4 4 1.00 0 0 4 4 0
45 misc.invest.penny-stocks 3 2 .67 0 0 1 3 1
46 misc.invest.fixed-income 3 3 1.00 0 0 3 3 0
47 misc.invest.forex 3 3 1.00 0 0 3 3 0
48 usa-today.invest 3 3 1.00 1 0 3 3 1
49 demos.invest.russia 2 1 .50 0 0 0 2 1
50 misc.invest.asset-management 2 1 .50 1 0 0 2 0
51 news.misc.invest.stocks 2 2 1.00 0 0 2 2 2
52 alt.business.invest 2 2 1.00 0 0 2 2 2
53 alt.invest.canada 2 2 1.00 0 0 2 2 2
54 news.admin.invest.futures 2 2 1.00 0 0 2 2 2
55 misc.invest.stocks.ipo 2 2 1.00 0 0 2 2 0
56 news.groups.invest.real-estate 1 1 1.00 0 0 1 1 1
57 misc.invest.emerging 1 1 1.00 0 0 1 1 0
58 misc.invest.precious 1 1 1.00 0 0 1 1 1
59 gaia.fido.invest 1 1 1.00 0 0 1 1 0
60 microsoft.public.investor 1 1 1.00 0 0 1 1 0

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
136
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Post: An individual Usenet message.

Posters: This is the count of the number of people who contributed at least one
message to the newsgroup in the time period selected.

PPRatio: This is the ratio of posters to posts contributed to the newsgroup in the
time period selected. A P:P of 1 means that on average each poster contributed
one and only one message. A P:P of .1 means each poster contributed 10
messages.

Returnees: This is the count of the number of people who contributed at least
one message to the newsgroup in the time period selected, and who also
contributed a message in the previous time period; i.e. in the prior month.

Replies: This is the count of the number of message that were replies to
another message (in contrast to a message that is a thread head that starts a
new conversation).

Singles: This is the count of the number of people who contributed one and only
one message to the newsgroup in the time period selected. Sometimes referred
to as drive-by posters.

UnRMSGs: This is the count of the number of messages in the newsgroup that
did not receive a reply in the newsgroup in the time period selected.

XPosts: This is the count of the number of messages that were shared (cross-
posted) to at least one other newsgroup in the time period selected.

Source: Netscan Glossary of Terms (http://netscan.research.microsoft.com/help/glossary.asp

Figure 5.1 – Definitions of Netscan quantitative results

The results illustrate two important facts that will be turned to advantage during the

Usenet search. First, they confirm Smith’s (1999) observation that the majority of

Usenet newsgroups display little or no activity. Hence, over half of examined

newsgroups can be summarily discarded from the study since they exhibit less than one

post per day on average, thus failing to meet the Exemplary-Trait criterion of high-

volume of participant interaction. Second, a larger number of newsgroups (specifically

50) have high PPRatios, greater than 0.40, in this case. This is not consistent with a

core-periphery structure, another Exemplary-Trait criterion, because by definition, such

a structure implies that a small sub-set of participants (the core) contribute a

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
137
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

disproportionately large number of posts, driving the PPRatio close to zero (operational

thresholds are set in Chapter Six). Therefore, the search for newsgroups displaying

potential affinity with the VCoP model quickly narrows to just 10 of the original 60

newsgroups, or really just 9, because one of the groups with low PPRatio, no. 39, had

already been discarded for low number of posts.

Raw Netscan results do not display the TPRatio or the %Cross-Post, but they are easily

derived. Specifically:

TPRatio = 1 – ( Replies / Posts )

%Cross-Post = XPosts / Posts

Continuing with the example, Table 5.2 is a modified version of Table 5.1, which

includes new columns for the TPRatio and %Cross-Post, omits discarded newsgroups,

and sorts survivors by ascending PPRatio (original identifying numbers were retained).

Compatibility with a core-periphery structure requires a low PPRatio. A low TPRatio,

indicating strong social interaction, is also required, to filter out newsgroups dedicated to

public announcements (e.g. job offers) which have extremely low PPRatios but little

interaction (Smith, 1999). Finally, a low %Cross-Post is more compatible with a focused

discussion, an Exempary-Trait criterion (McLaughlin et al, 1995; Smith, 1999). Only the

topmost four newsgroups of Table 5.2 exhibit low values for all three measures.

Therefore, only these four simultaneously meet the criteria of high-volume of participant

interaction, core-periphery structure and focused discussions; the remaining five can be

discarded. Hence, of 60 newsgroups analysed by Netscan, 56 are efficiently assessed as

displaying little affinity to the VCoP model, and thus discarded from the study.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
138
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Table 5.2 – Newsgroups pre-selected from Table 5.1 sorted by PPRatio


%Cross UnR
No Newsgroups Posts Posters PPRatio TPRatio Returnees Replies Singles XPosts
Post MSGS
4 microsoft.public.investor.discussions 649 65 0.1 0.20 0.10 15 518 31 68 63
1 misc.invest.stocks 7063 811 0.11 0.23 0.13 274 5446 385 688 930
2 misc.invest.mutual-funds 1614 188 0.12 0.12 0.09 63 1426 92 84 144
7 misc.invest.financial-plan 496 84 0.17 0.19 0.01 35 402 36 18 7
5 alt.private.investigator 617 143 0.23 0.38 0.25 27 385 76 144 157
3 aus.invest 1087 266 0.24 0.25 0.35 84 818 156 125 384
10 it.economia.investire 396 141 0.36 0.35 0.06 29 257 76 87 25
6 misc.invest.futures 529 200 0.38 0.40 0.28 71 315 122 136 147
8 misc.invest.canada 423 161 0.38 0.39 0.40 30 257 94 122 171

5.2 – Social Network Analysis and the core-periphery model

Social Network Analysis (SNA) provides a method, or rather a family of methods,

ideally suited for analysing the differentiated patterns of interaction that are evident in

many newsgroups (Wasserman and Faust, 1994; Garton, Haythornthwaite and

Wellman, 1997; Scott, 2000). Most social research methods work with attributional

data. They measure the individual actor attributes (such as age, gender, socio-economic

status, etc.), then try to group together individuals possessing similar profiles of

attributes, and explain social behaviour as the result of these common attributes

(Wellman, 1988). By contrast, SNA works with relational data. It adopts as unit of

analysis the ties or relations between actors, and explains social behaviour as a result of

the patterns of strong and weak ties between actors and the resulting constraints to

social behaviour (ibid). As long as ties can be measured no assumption is necessary

regarding spatial position or indeed any other attribute of actors. Thus, SNA has been

used to study networks of distributed actors, such as ham radio operators (Faust and

Romney, 1985), scientists communicating through e-mail (Freeman, 1984), or countries

conducting foreign trade (Wasserman and Faust, 1994).


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
139
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Even though Usenet provides a rich field of social interaction, with readily available

data, the literature review found relatively few SNA studies. Mostly they have focused

on mapping the links between newsgroups by defining ties as the number of messages

cross-posted between them (Smith, 1999; Donath, Karahalios and Viégas, 1999; Sack,

2001; and Choi and Danowski, 2002).

A different approach, arguably closer to the original spirit of SNA, is to treat individual

participants as actors, and the messages they post to each other as ties (Wasserman and

Faust, 1994). Such an approach can easily identify persistent, stable and cohesive sub-

groups of participants. Persistent here refers to a sub-group that is resilient over time,

while stable refers to a sub-group whose membership exhibits little turnover. For the

purposes of this research, a persistent sub-group whose members remain the same over

time and engage in intensive discussions between them, will be considered a virtual or

Usenet-based community. Note that this definition does not apply to the entire

newsgroup, since the majority of participants are usually ephemeral question-askers.

Note also that at this point, nothing need be said about the topic of such a community,

only about the intensity of interaction and the cohesiveness between fixed members.

The appropriateness of SNA methods for a particular study depends on the theoretical

validity of the operational definitions of actors and of the social ties between them

(Wasserman and Faust, 1994). This research will define actors as Usenet participants

who actually interact by posting a message to a newsgroup. Defining ties, though,

requires further elaboration.

Apart from their topical focus, newsgroups have an organising scheme that allows

multiple simultaneous discussions within a single newsgroup; specifically, they are

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
140
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

organized as threaded discussions. A thread is a set of messages which address the

question or discussion topic announced in the first message, called the thread head

(Hahn, 2000). Participants who wish to offer an answer or an opinion post follow-ups to

the initial message, that is, they send a message making explicit reference to the thread

head or to one of its subsequent follow-ups. At any given time, a newsgroup contains

dozens of ongoing threads or discussions, each with its own coherent set of messages.

Since every message posted to Usenet prominently displays the name (real or

pseudonimous) of its author, every follow-up message is in effect a directed message, from

a specific sender to a specific recipient. Since a directed message is intentionally (and

publicly) posted by one participant to another, it constitutes a personal communication.

Hence, this study defines a (directed) social tie between two Usenet actors as any directed

message posted from one to the other, within the threads of a given newsgroup.

Having thus defined actors and ties, the aim of the SNA runs will be to identify subsets

of actors exhibiting more frequent communication, since person-to-person

communication constitutes the basic building block of communities. To perform this

efficiently, the research will tap the technical information stored in message headers.

Specifically, header information will be imported into a database programme to allow

efficient handling of messages numbering in the tens of thousands. Headers identify the

author of each message, whether it is a thread head or a follow-up, and if the latter,

identifies the message (and therefore the author) it is a follow-up to. The database can

efficiently derive a listing of all directed poster-to-poster communications, which is in

effect the social network participants create through their interactions. This will then be

imported into SNA software and a core-periphery model fitted.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
141
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

At this point, a digression on the ethical handling of Usenet messages. Sudweeks and

Rafaeli (1996) consider the issue of informed consent does not apply to a quantitative

analysis in which only publicly available messages are studied (which is the case in

Usenet), as long as the names of message authors are kept confidential. Eysenbach and

Till (2001) consider the requirement of informed consent can be waived under certain

circumstances. They draw a parallel between passive analysis of messages on mailing

lists, and non-intrusive research in clinical studies based on retrospective use of

existing medical records, where informed consent can be waived if the material is

anonymised at the earliest posible stage, no inconvenience or hazard is caused to

subjects, and the institutional review board has approved the research protocol.

The researcher’s criterion, within the SNA phase of the research was to use relabelling

to anonimise participants’ names and e-mails. Informed consent for unobtrusive

analysis of newsgroup postings was not sought as participants know they are posting to

a public forum, and their messages, preserved in the Web, and accessible through the

Google search engine, are public domain. The forthcoming Content Analysis of

discussions will similarly respect poster anonimity, and disguise names when literal

messages are reproduced in the thesis.

The core-periphery structure targeted by Netscan searches is illustrated in Figure 5.2,

which plots on the z-axis the number of directed messages sent during 30 weeks (01-14-

2001 to 08-11-2001) by the 60 top posters in newsgroup soc.history.medieval sorted in

descending order by number of posts. The graph reveals a very active and cohesive

“core”, indicating top posters have frequent interaction, and thus strong social ties with

other top posters. Hence, the subset of participants close to the core can be characterised

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
142
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

as a cohesive sub-group. The newsgroup not only exhibits the sought-for core-periphery

pattern, but also, as shown on Table 5.3, the expected low values for the three interaction

measures.

FIGURE 5.2 - Posting activity of top 60 posters in soc.history.medieval

Table 5.3 – Smith’s interaction measures for soc.history.medieval


Newsgroup Posts Posters PPRatio TPRatio %Cross-Post

soc.history.medieval 3889 292 0.08 0.04 0.18

Note: Data for March 1-31, 2001

Borgatti and Everett (1999) have proposed a mathematical model for empirically

detecting such patterns, the continuous core-periphery model. Their assumption is that

network data consist of continuous values representing strengths of relationships, which

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
143
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

fits the study’s definition of ties as the number of directed messages exchanged between

actors. The algorithm, implemented in their own SNA software, Ucinet 6 for Windows

(Borgatti, Everett and Freeman, 2002), searches for a set of coreness scores which

maximise the correlation of the observed pattern with an ideal core-periphery pattern,

illustrated in Figure 5.3. The entries of the ideal-pattern matrix are assumed to be the

product of the coreness scores of each pair of actors, with coreness ranging from 0 to 1.

FIGURE 5.3 – Graph of ideal-pattern for core-periphery model

The usefulness of model for this research lies not just in the fact in the fact that CoPs

often exhibit a core-periphery pattern (Wenger, 2000a), but in the fact that by using

extended time-samples from the newsgroups (52 weeks), the persistence of detected

core-periphery structures and the stability of high-coreness members are assured.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
144
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

5.3 - The Survey questionnaire: objectives and design

The Survey is the first of two methods whose design follows the Essential-Trait

Research Questions. Furthermore, the design process closely follows the guidelines for

scale development proposed by DeVellis (2003); although Babbie (1990), Oppenheim

(1992) and DeVaus (2002) were also consulted. Since the former’s guidelines provide,

in addition to good practice, a clear and logical structure for organising this section,

they will be used as sub-section headings.

5.3.1 – What the Survey will measure

The instrument aims to detect the Essential Traits through valid and reliable scales. To

do this, the constructs must be broken down into more concrete sub-constructs or

indicators (DeVaus, 2002). Representative manifestations of each construct were

defined in Section 2.3; they will act as sub-constructs, and a scale will be developed for

each. These representative sub-constructs were derived through the item-generation

exercise described next.

5.3.2 – Generation of the item pool

The next step is to generate a large pool of items that at face value are related to the

construct, and are thus candidates for eventual inclusion into the scale (DeVellis, 2003).

As an aid for thinking up specific and relevant items, development began by deriving

broad, open, yet more specific questions from each of the Essential Trait Research

Questions, always working within the context of CoP theory. These broad questions were

later named thematic questions, and were concrete realisations of the Research

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
145
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Questions. In effect they were the kind of questions the researcher asked himself about

the virtual communities in the study. Yet, their concreteness, plus the restriction that

items must fit in a self-administered questionnaire with Likert-scales, were effective in

suggesting and discarding appropriate items. For instance, the thematic questions for

COMMUNITY were:

Is there evidence of stable relationships in the newsgroup?

Have members interacted with each other for some time?

Have members met personally off-line?

Are members comfortable to ask each other for help or advise?

Do members agree on who belongs?

Is there much turnover among regular members?

Is there evidence of leadership and community service?

These thematic questions in turn suggested specific closed items, such as:

I have had considerable interaction with the other members

I know well the views of the other members

I know well the expertise of the other members

I usually know beforehand how other members will react

My own views are well known by the other members

My own expertise is well known by the other members

The development of the item bank extended over several months (see Appendix A),

iterating between thematic questions and closed items. Eventually, as a result of continued

work and reflection, there came a point where thematic questions were put aside, and

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
146
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

representative sub-constructs were introduced. These were concrete manifestations of

Wenger’s constructs, whose presence or absence a newsgroup participant can report if

clearly asked about (DeVaus, 2002). In all, 12 sub-constructs were chosen as representative

manifestations of the underlying constructs; they are as follows:

LEARNING / IDENTITY ACQUISITION

• Acquiring new knowledge

• Acquiring new skills

• Acquiring/enacting a professional identity

MUTUAL ENGAGEMENT

• Debating issues related to the profession

• Collective problem-solving

• Sharing useful information

SHARED REPERTOIRE

• Shared criteria

• Shared practices

• Shared artifacts

COMMUNITY

• Members’ knowledge of each other

• Shared sense of community

JOINT ENTERPRISE

• Caring for some domain of knowledge

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
147
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

As noted before, this set of visible manifestations of Wenger’s constructs cannot be

exhaustive, only representative. Questionnaire size limitations preclude using more

than a few obvious manifestations of each construct, because five items were deemed

as the minimum acceptable size of the scale for each sub-construct. Although a search

was made for previously published scales and existing indicators (e.g. Rubin,

Palmgreen and Sypher, 1994), none appeared to fit the constructs and sub-constructs

listed above.

The constructs of LEARNING and IDENTITY ACQUISITION merit additional comment. In

Wenger’s (1998) framework, they are not independent constructs, since he views

LEARNING as the acquisition of a new IDENTITY. Therefore, though a separate scale is

built for IDENTITY ACQUISITION, it is considered a sub-construct of LEARNING.

The introduction of sub-constructs improved clarity and focus in the item-generation

exercise, and the item bank eventually grew to 214 items, of which the pilot

questionnaire included 66. Hence, one out of three items was selected for the

instrument, which follows the guideline suggested by DeVellis (ibid).

Items were worded in as clear and direct manner as possible; negative wording was

mostly avoided in the interests of clarity (Babbie, 1990). Hinkin (1995) reports authors

are not unanimous as to the adviseability of introducing negatively worded items in a

survey. Even though this practice reduces the potential for agreement bias, it can

introduce systematic error. DeVellis (2003) advises against, particularly when the

questionnaire is long, and this counsel was followed.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
148
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Throughout the iterative process of item generation, the feedback and comment of the

thesis supervisor were periodically sought to refine problematic wording, suggest

additional items, and provide a theory-informed critique about the developing

questionnaire. Nine distinct questionnaire drafts were produced before reaching a

version deemed ready for piloting. A record of the process is displayed on Table 5.4.

Table 5.4 – A record of the item-generation process of the Survey instrument


Version Version Constructs Thematic Sub- Items Supervisor
Date questions constructs reviewed

Draft 1 20-05-03 7 28 none 72 yes

Draft 2 25-05-03 7 40 none 125 yes

Draft 3 12-06-03 7 41 none 182 no

Draft 4 18-06-03 7 41 none 214 no


Development

Draft 5 29-06-03 7 37 none 210 no

Draft 6 02-07-03 6 none 13 54 yes

Draft 7 13-07-03 5 none 14 55 no


yes:
Draft 8 20-07-03 5 none 12 66
pilot-ready
Draft 9 06-08-03 5 none 12 66 yes: pilot
launch
Pilot 1 19-08-03 5 none 12 66 yes,
Piloting

reliabiliy
Pilot 2 09-09-03 5 none 12 66 yes, main
launch
Deploy

Main 23-09-03 5 none 12 63 no


Survey

The piloted version contained five constructs, twelve sub-constructs, five-item scales

for each sub-construct, plus one open question and 5 socio-demographic questions at

the end, resulting in a 66-item Survey (provided in Appendix B). As a visual aid to

respondents, items were organised into four major thematic blocks labelled “Learning

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
149
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

through participation in the newsgroup”, “Member interaction in the newsgroup”,

“Newsgroup-specific tools and standards”, and “Community traits in the newsgroup”.

Learning was used as the opening block because it addresses a familiar experience, and

helps to gradually introduce respondents to the main issues of the Survey (Babbie,

1990; Oppenheim, 1992). Within each block, scale items were randomised to test

whether Factor Analysis groups items under the hypothesised scales. Randomising the

entire 66-item Survey was deemed unwise, as it would convey a chaotic and disorganised

impression to respondents, making their task more difficult (Babbie, 1990).

5.3.3 - Design of the web-based instrument

Internet surveys rarely obtain good response rates (Witmer, Colman and Katzman,

1999). Therefore, to maximise convenience to respondents, an early decision was made

to use a Internet-based questionnaire, with point-and-click Likert scales that

participants could fill out quickly and accurately.

In addition, considerable care was invested into the web design of the Survey. For

instance, the choice of provider is critical. It must offer good layouts and aesthetics,

hinting at an interesting and inviting experience for the respondent (DeVaus, 2002). In

addition, the web page must display in a speedy and reliable manner, as a slow-loading

page will probably cause a respondent to become impatient and move on. Above all, the

provider’s home page must reassure respondents about privacy and security concerns

which are ever-present in Internet. Conscious of this, the researcher sought a provider

who by a professional-sounding name, and a highly-visible privacy policy, reassured

respondents about the legitimacy, security and ethical guarantees of the Survey.

Obviously, this could all be achieved by contracting one of several high-priced providers

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
150
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

available, but financial restrictions led the researcher to leave this option as a last resort,

and to search the Internet for alternatives. Thus the review of web-based survey providers

(included as Appendix C) is biased toward economic or free services, although the size of

the questionnaire and the potential sample population automatically discarded several

smaller services (for a current review, see Wright (2005) ). It turned out some

professional providers offer a lower price, or even a free service, to academic projects.

To test the layouts and responsiveness of potential providers, the researcher registered

for several trial-accounts. The test assessed the clarity and aesthetics of the published

questionnaire, the convenience of the item/survey editor and the invitation mail engine.

Surveyworld.net, based in the Netherlands, was chosen as the main survey provider

because of its excellent layouts, its explicit privacy policy, and its generous support for

academic research projects. After reviewing the researcher’s credentials and research

project, the company agreed to provide free-of-charge a professional account including

all the features and options of their service. This support is gratefully acknowledged.

As already mentioned, the Survey was visibly organised into four thematic blocks, with

fifteen items in each, and socio-demographic items at the end. In addition, the Survey

displayed fully on a single page, with the Submit button at the end. This gave

respondents the opportunity to read over the entire Survey, skip forward, change

previous answers, and make a final review of responses before hitting the Submit

button. Actual web display of Pilot 1 is provided in Appendix B.

Invitations to take the Survey were sent by the e-mail engine of the provider. The text

of the invitation was slightly longer than usual, trying to convey an informative and
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
151
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

professional look. It was important to differentiate the invitation from junk-mail,

universally hated in Internet. The invitation is provided in Appendix B.

5.3.4 - Qualified review of the item pool

The researcher tested evolving drafts of the questionnaire on his supervisor to refine

item wording and sharpen sub-construct focus. Hence, two knowledgeable researchers

vetted all items, before attempting a first pilot of the questionnaire.

5.3.5 – Possible inclusion of validation items

DeVellis (ibid) advises searching for and trying to include previously developed scales

as control variables or as constructs theory predicts to be related to the constructs under

study. Although this was attempted by a careful study of various scales published in

Rubin et al (1994), none appeared to be a good fit within the evolving questionnaire,

possibly because of its novel purpose –CoP assessment– and exploratory nature.

5.3.6 – Pilot runs of the instrument

Newsgroup comp.lang.ada was chosen for Pilot-1 of the questionnaire (newsgroup

selection is described in Chapter Six), and the invitation was sent on Tuesday 19

August, 2003 to 136 participants exhibiting coreness scores greater than or equal to

0.005. An automatic reminder was sent a week later to those who had not taken the

Survey nor expressly declined. The Survey closed ten days after that.

Since many participants use more than one e-mail address to post to the newsgroup, the

actual list of e-mail addresses contained 194 distinct addresses. The first visible result

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
152
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

of the invitation was a large number of error messages returned to the sender caused by

false addresses. However, within the first hour, several participants completed the

Survey. In all, 31 newsgroup members took the Survey during those 19 days, and it was

encouraging to note that, despite its length, all submitted surveys were complete. Pilot

logistics are recorded on Table 5.5.

Table 5.5 – Response rates for the Survey pilots


Pilot Launch Reminder Close Invitations Repeated Distinct Invalid True Response Response
sent names names address sample rate

Pilot-1 19-08-03 26-08-03 06-09-03 194 58 136 10 126 31 24.6%

Pilot-2 09-09-03 16-09–03 20-09-03 281 46 235 29 206 40 19.4%

5.3.7 - Item evaluation

Raw results were downloaded as an Excel file from the Surveyworld server and

imported into SPSS where items were re-grouped back again into the five-item scales

of each sub-construct. Descriptive statistics and Cronbach alphas were calculated. As a

result of the reliability analysis, significant wording changes were made in 14 items and

minor changes in another 7. Coreness, an independent variable measured for each

participant was not found to significantly affect results, but individual expertise

apparently did. Therefore, three items were selected to build a new scale for Expertise.

The extent of changes to the questionnaire advised running a second pilot with the

revised version. In addition, since Pilot-1 had been applied to a computer-related

newsgroup, it was decided to apply Pilot-2 the second pilot on a non-computer group;

sci.physics.research was chosen. Pilot-2 was launched on Tuesday, 9 September, 2003,

with an automatic reminder a week after, and a total run of 12 days. This was a larger
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
153
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

newsgroup than comp.lang.ada, as shown on Table 5.5, and a total of 40 completed

questionnaires were received.

The reliability analysis for the Pilot-2 revealed three items whose item-total correlation

was low, and whose alpha-if-item-deleted was substantially higher than current scale

alpha. When closely examined, two items were found to differ, at face value, from the

other items in the set, while the other was found to be too broad and unspecific.

Therefore, these three items were discarded. Other than this, though, the analysis

upheld the changes made to the previous questionnaire. In fact, the results from

sci.physics.research, omitting the three discarded items, were judged valid for merging

with the dataset from the main Survey. Given the positive results of Pilot-2, the Survey

instrument was deemed ready (the deployed instrument is available in Appendix D),

and the main Survey was launched on ten additional newsgroups starting on Tuesday

23 September, 2003.

The evaluation of the instrument cannot rest only on the reliability analysis. DeVellis

(ibid) strongly advises using Factor Analysis to refine and further validate scales. Thus,

an Exploratory Factor Analysis (EFA) was performed on the results of the main Survey,

which confirmed the validity of the instrument; results are described in Chapter Seven.

5.3.8 - Possible optimisation of scale length

The trade-off between size and reliability of scales for each sub-construct was an ever

present consideration because of the large number of sub-constructs (12). Five items

was deemed the minimum acceptable size for each scale, keeping in mind that some

items would be discarded during the reliability analysis. DeVellis (ibid) points out that

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
154
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

optimisation of scale length is confined to situations where the researcher has

“reliability to spare”, obviously not the case here.

This concludes the discussion of questionnaire design. Survey deployment and results

are presented and discussed in Chapter Seven.

5.4 - The Content Analysis: objectives and design

Sharing the lower level of the Funnel with the Survey is a Content Analysis of

community discussions, which will be used in a deductive research strategy to test the

applicability of Wenger’s CoP theory (Blaikie, 2000). The working hypothesis is that

some or all participating communities are true CoPs. Hence, interactions between

members should manifest the Wenger constructs, i.e. the Essential Traits. Detecting the

text-based manifestation –or lack of it– of the Essential Traits, in the newsgroup

messages that record community interactions, is the objective of the Content Analysis.

There are two major aspects in the design of the Content Analysis, and this section will

be organised around them. The first is the selection of the sample from a textual corpus

too large to analyse exhaustively. The second is the development of the coding scheme.

5.4.1 – Rationale and criteria for purposive sample selection

The research design acknowledges that, with eleven newsgroups in the study, and a

one-year message sample from each, an exhaustive content analysis is not feasible.

Therefore, some sampling strategy is required.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
155
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

This study will adopt threads as the sampling units for the Content Analysis. The advantage

of threads over individual messages lies in the former representing a complete and coherent

discussion. Thus a full thread is easier to interpret and content analyse because the various

messages provide a context and oftentimes a critical review of each other.

However, the researcher felt from the beginning that arbitrarily picking threads that

exhibited the sought-for qualitative traits mentioned above would be falling into the

trap of anecdotalism (Silverman, 2000), thus undermining the credibility of this study.

Therefore, a decision was made to select threads according to previously-devised and

theoretically informed criteria, following Silverman’s (2000: 105) advice: “Sampling in

qualitative research is neither statistical nor purely personal: it is, or should be,

theoretically grounded.”

Threads are already captured into the database, and can be sorted and classified by

posters or number of messages. Given this capability to sort and classify threads, it is

an easy to see that for the purposes of this research, not every thread in a newsgroup is

equally useful. Threads are very diverse in size, topic, participants and duration. Just as

“exemplary” VCoPs were defined as the target of the study, “exemplary” threads will

be defined as the target of the Content Analysis sample, using once again the VCoP

model to derive three selection criteria. Specifically:

Criterion 1: An “exemplary” thread addresses a professional topic.

This proposition mirrors the Exemplary Trait requiring the newsgroup to have a

profession for its topic. Even in newsgroups narrowly focused on a profession, not all

talk is about the business of the newsgroup, particularly in unmoderated newsgroups.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
156
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

There is also, especially among regular members, a significant amount of what can be

described as relaxed chatting about current news, the weather, political issues, etc.

Wenger (1998) argues that such conversation among members of a CoP, even if it does

not directly concern work issues, helps to build the personal relationships that facilitate

MUTUAL ENGAGEMENT. However, mere conversation cannot clearly differentiate a

Usenet-based CoP from other virtual communities, whereas collective problem-solving

or debating professional issues, can.

Criterion 2: An “exemplary” thread is longer than average.

The logic of this criterion goes back to some of the Essential Traits which require some

discussion “space” to fully play out. Specifically, both collective problem-solving and

debating issues, which are manifestations of MUTUAL ENGAGEMENT, are more likely to

be found in long rather than in short threads.

Criterion 3: An “exemplary” thread involves mostly core members.

The logic of this criterion goes back to the set of Exemplary Traits of the VCoP model,

which implied that an exemplary Usenet-based CoP would be the cohesive sub-group

formed by the stable core members of a newsgroup. Therefore, given the ability to

choose among all discussions taking place in a newsgroup, it is clear that the discussions

that remain mostly within the core, where membership is strongest (by virtue of intensive

ENGAGEMENT), are more likely to yield the sought-for evidence. Since newsgroups are

open to participation, it will be a rare thread that only contains messages by core

members, but threads where discussion is dominated by these members are numerous.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
157
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Using these criteria as filters will result in a very substantial reduction of the thread

population of each newsgroup, and improved odds of finding the Essential Traits on

those groups that are Usenet-based CoPs. From the reduced thread population, four

threads per newsgroup will be selected, one long and three middle-sized threads.

Furthermore, to take the decision out of the researcher’s hands, as it were, the selected

threads will be those that best match the previous three criteria, and non-selected

threads and the reasons for their exclusion will be provided in Appendix M.

5.4.2 – The coding scheme

In this study, content analysis is being used deductively, with a priori categories

derived from Wenger’s theory. Hence, codes here will be used to denote a fact,

specifically that a particular textual passage contains an assessed example or instance

of a sub-construct. Thus, codes will be used as representations of phenomena. This

coding approach is fundamentally different to the use of codes in CAQDAS software as

heuristic devices to mark passages in order to build theory or hypothesis with an

interpretivist inductive logic (Seidel and Kelle, 1995).

Though deductive, this Content Analysis is qualitative, because it will be concerned

with the presence or lack of it of the Essential Traits in specific textual passages rather

than how many times they manifest themselves. Nevertheless, a summary of code

counts will be provided, because it is a useful reference even in a qualitative context

(Miles and Huberman, 1994; Silverman, 2000).

Most studies performing a quantitative type of content analysis develop from relevant

theory a list of keywords, and report the number of times these appear in a sample of

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
158
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

text (Neuendorf, 2002). It is easy to see that Wenger’s constructs, and even the more

specific sub-constructs, are not amenable to reduction to a list of keywords, even an

extensive one. Hill and Hughes (1997) report a similar problem in their content analysis

of political communication in Usenet newsgroups. Because political communication

includes such a range of topics, they discarded a keyword strategy. Rather, they used

threads as their primary unit of analysis, and relied on their professional interpretation

of message content to code for several pre-determined qualitative variables, such as:

ideology of the thread head (left, no dominant, right), overall ideological direction of

entire thread (left, no dominant, right), whether the thread is a debate, and whether the

thread is predominantly “antigovernment”.

Clearly, this use of “broad” coding categories and expert researcher judgement is an

appropriate strategy for capturing the complex social interactions that were the focus of

Hill and Hughes’ study. “Broad” coding categories to test theory or obtain evidence for

specific research questions have been used in other studies of Usenet newsgroups. For

instance, Preece (1999) and Preece and Ghozati (2001) used it to test for the presence or

absence of empathic communication in various kinds of communities, particularly those

whose aim was providing patient and emotional support. Other studies using broad

categories include Thomsen (1996), Winzelberg (1997) and Millen and Dray (2000).

In a similar vein, identifying and coding the complex and nuanced interactions that

make up Wenger’s constructs, will require the researcher’s professional interpretation

of CoP theory, as it manifests itself in discussion threads. Ideally, this thesis would

adopt a multi-researcher coding strategy, in which several competent researchers would

code randomly selected threads, thus minimising the potential for bias and improving

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
159
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

overall coding consistency. This is not possible for a lone PhD candidate, but some

precautions were adopted nonetheless. Specifically, the candidate acted as the main

coder, and the thesis supervisor performed an independent review of the researcher’s

coding consistency in a substantial random sample of coded threads.

Most methods of qualitative textual analysis develop coding categories inductively,

very much in the spirit of grounded theory (Ryan and Bernard, 2000). Not many

qualitative methods adopt a deductive code development strategy (Mayring, 2000).

Since the aim of the Content Analysis is to test the applicability of Wenger’s theory,

coding categories will be deducted from Wenger’s constructs. More specifically, this

research will follow the code-creation procedure for a priori or pre-defined codes

suggested by Miles and Huberman (1994), which can be summarised as follows:

• Use relevant theory and the research questions to create a provisional “start

list” of codes. This initial coding scheme must have a clear theory-supported

structure, and each code similarly requires an initial theory-informed

definition, which will be refined as the research progresses.

• Apply the coding scheme to the textual corpus, possibly starting with a

reduced pilot.

• Revise codes:

o discard empty codes or absorb them into more successful ones.

o break down “bloated” codes into smaller subcodes.

o improve and further focus code definitions.

o where apropriate, create new codes that “emerge” from the textual corpus.

• In revising codes, always maintain clear links to the theoretical framework

and the research questions.


––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
160
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

• Along the way, perform some consistency and reliability checks of code

definition and application.

Theoretical codes will thus be initially defined as the 12 sub-constructs which,

following Wenger’s theory, were derived during the development of the Survey. A 13th

code will be labelled as “Other”. These starting codes are shown in Table 5.6, along

with working definitions that will provide guidelines for coding within NUDIST.

Table 5.6 – Initial theory-derived codes for the Content Analysis


Wenger’s Sub-constructs Working NUDIST definitions
constructs (coding categories)
Debating issues related to the Debating pros and cons of an issue related to
profession the profession

MUTUAL Collective problem-solving Collectively building the solution to a problem


ENGAGEMENT or case

Sharing useful information Providing (spontaneously or by request) useful


information related to the profession

Acquiring new knowledge New knowledge acquired from discussion

Acquiring new skills New skills acquired from discussion


LEARNING
Acquiring/enacting a Developing/enacting a professional identity
professional identity

Shared criteria Invoking an accepted professional criterion

SHARED Shared practices Invoking an accepted procedure within the


REPERTOIRE profession

Shared artifacts Invoking an accepted tool, real or virtual

Members’ knowledge of Evidence of newsgroup participant's knowing


each other each other
COMMUNITY
Shared sense of community Acknowledgement and positive regard for the
online collective

JOINT Caring for some domain of Expressed interest for/involvement in the


ENTERPRISE knowledge profession

OTHER Textual passages that do not fit under any code

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
161
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The actual coding of textual passages will proceed in several stages. Before actually

tackling the coding of the main thread sample, one or two pilot coding exercises will be

conducted, in order to test the consistency and precision of the categories as defined

above. It is to be expected that the wording of these definitions will be refined first by

the pilots exercises, and then by the main coding exercise, without actually breaking

their connection to the theoretical concepts they represent. It should be remembered

that only 12 sub-constructs were defined because of the length restrictions of the

Survey instrument, not because those sub-constructs exhaust the range of possible

manifestations of the constructs. Since the categories of the Content Analysis are not

inherently restricted to the defined sub-constructs, but to Wenger’s broader constructs,

it is likely that new sub-constructs will emerge from the textual data, reducing the

number of textual units classified as “Other” and resulting, in effect, in a better

theoretical categorisation of the thread sample. In turn, this should provide clearer

evidence regarding the Essential Trait Research Questions, which is the ultimate aim of

the Content Analysis.

Further details of the Content Analysis Stage and results of the main coding exercise

are provided in Chapter Eight.

5.5 – Summary

The chapter described and discussed the research methods used in this thesis; an

heterogeneous collection whose coherence stems from their coordinated deployment at

various Stages of the Funnel Strategy.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
162
Methods
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

The first two methods, the Netscan analyser and the core-periphery model, will be used

during the Usenet search, or Stage III of the Research Strategy, to select a sample of

communities exhibiting high CoP-potential. They will be guided by the Exemplary

Traits of the VCoP model, and will address both Preliminary and Exemplary Trait

Research Questions.

The last two methods, the Survey and the Content Analysis, will be applied to selected

communities, during Stages IV and V of the Research Strategy. Their aim is to

independently test for the presence or absence of Wenger’s constructs. They will be

guided by the Essential Traits of the VCoP model, and will address the Essential Trait

Research Questions.

This chapter concludes discussion of thesis methodology; description of field research

begins with Chapter Six.

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
163

You might also like