Task-based performance in a second language

Downloaded from UvA-DARE, the institutional repository of the University of Amsterdam (UvA) http://dare.uva.nl/document/207304

File ID Filename

207304 Thesis

SOURCE (OR PART OF THE FOLLOWING SOURCE): Type Dissertation Title Cognitive and interactive aspects of task-based performance in Dutch as a second language Author M.C. Michel Faculty Faculty of Humanities Year 2011 Pages xiii, 213 ISBN 978-90-8891-232-0

FULL BIBLIOGRAPHIC DETAILS: http://dare.uva.nl/record/369312

Copyright It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other then for strictly personal, individual use.

UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

Cognitive and interactive aspects of task-based performance in Dutch as a second language

ISBN 978-90-8891-232-0 NUR 616 This book was typeset using LATEX. Printed by Proefschriftmaken.nl k Printyourthesis.com. Published by Uitgeverij BOXPress, Oisterwijk. ¨ Cover illustration and graphics: Dominika Plumpe und Luisa Heinrich c 2011 by Marije C. Michel. All rights reserved. Copyright

Cognitive and interactive aspects of task-based performance in Dutch as a second language

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. D.C. van den Boom ten overstaan van een door het college voor promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op donderdag 10 maart 2011, te 14:00 uur

door

Marije Cornelie Michel geboren te Lachen, Zwitserland

PROMOTIECOMMISSIE

Promotor:

Prof. dr. F. Kuiken

Co-promotor:

Dr. S.C. Vedder

Overige Leden:

Prof. dr. C.L.J. de Bot Prof. dr. P.C. Hengeveld Dr. A. Housen Prof. dr. J.H. Hulstijn Dr. J. Kormos Prof. dr. P. Robinson

Faculteit der Geesteswetenschappen

Voor Louise van den Berg-Gerlagh en Christina Roelofs-de Jonge twee vrouwen, geboren aan het begin van de vorige eeuw, ´ gestudeerd, de ander door het leven minstens net zo wijs. de e´ en Ik ben trots in hun voetstappen te staan.

Dankwoord

Writing and finishing a PhD-thesis depends on much more than the content of this book. I’m in debt to so many people in my surroundings whom I would like to dedicate some words of thank to in the following paragraphs. First of all, I am very grateful to my promotor prof. dr. Folkert Kuiken and my co-promotor dr. Ineke Vedder of the University of Amsterdam. Both of them were very conscientious in their role as supervisors and they took great care from the earliest discussions on until the last words written in this book. They introduced me generously to their network and shared opportunities for presenting and publishing. Most important, they established a cordial environment where even in difficult times we could laugh about Folkert’s jokes. I am in debt to the members of the scientific committee. I would like to thank prof. dr. Peter Robinson whose theory was the inspiration for this thesis but who more importantly shared his thoughts about task complexity by mail or in person when meeting at different ends of the world. I am also grateful to prof. dr. Jan Hulstijn for his critical comments and supporting words. I very much enjoyed being a member of the inspiring research group CASLA at the University of Amsterdam and the international network SLATE. Prof. dr. Kees Hengeveld I would like to thank, not only in his role as member of the reading committee but also for accompanying me during the whole process of this PhD-research at the ACLC as the supervisor of my progress meetings. Prof. dr. Kees de Bot, dr. Alex Housen, and dr. Judit Kormos I would like to thank for the reading and helpful comments but also for their openness to my work and the interesting discussions when meeting at conferences. The Dienst Maatschappelijke Ontwikkeling (DMO) of the city of Amsterdam I am thankful to as they financially supported this research. In Amsterdam, Utrecht and the surroundings I would like to thank all the L2-learners and native speakers who participated in my studies. Similarly, I thank the different NT2-teachers and language institutes who helped me in finding my participants (e.g., INTT, VASVU, Taal- en Schakelklassen HvA and INHolland, ROC Amsterdam and many more). I am thankful to the language institute of the University of Groningen for sharing their testing material. I thank the University of Amsterdam (UvA), most prominently the Amsterdam Center for Language and Communication (ACLC) for giving me the rich opportunities – in researching, teaching, traveling, studying, but also meeting so many interesting people – most PhD-students all over the world may only

viii

Dankwoord

dream of. In the course of conferences I benefited from critical audiences, e.g., Peter Skehan, Pauline ´ esz, ´ Foster, John Norris, Lourdes Ortega, Rod Ellis, and not least Gabriele Pallotti, Andrea Rev Roger Gilabert and Koen van Gorp. I hope there will be plenty of opportunities to keep seeing each other and go on in the field we are unraveling. I also thank the capacity group of Dutch linguistics of the UvA, led by prof. dr. Fred Weerman, that has been a fruitful environment to me for research, teaching, and for sharing lunch (or breakfast) with me: Jan, Suzanne, Nada, Loulou, Marian, Sven, Daniela, Robert, Petra, Ingeborg, Astrit, Leonie, and especially Hedde for his unbearable jokes and personal advices. I thank all my colleagues of the ACLC from Dirk Jan Vet and Ton Wempe (for technical help), to Bart and Raffael (for chats, jokes, and thoughts), from Jan de Jong (for giving me a good start) to Esther, Margot, Eva, Marcel, Josefien and all the other PhD-students for sharing good times. I will in particular miss the meetings with the research group CASLA because of the critical discussions in a cordial environment: Jan, Arjen, Sible, Lotte, Marjolein, Mirjam, Klaartje, and Margarita. Rob, Nomi, and Nivja I thank in particular for their advices on statistics. I thank my student assistants for helping me with all the transcription and coding: Kimberley Mulder, Dirk van der Meulen, Mascha Kuijlenburg, Olga Abell, and above all Rachel Jobels - my favorite intern. Thanks to Tikitu for proofreading the English and my colleagues who checked the manuscript for typos. ¨ ¨ zu Koln ¨ und dem DaZDank gebuhrt dem ZFL der TU Dortmund, den English Studies der Universitat ¨ Tubingen. ¨ ¨ ¨ die deutsche Geselligkeit in Studiengang der Universitat Ganz herzlich mochte ich mich fur ¨ ¨ Amsterdam bedanken bei Antje, Diana, Sebastian, Roland, Luisa und naturlich Maren und Irene fur ihren unschlagbaren Humor, Tanz und die konstruktiven Kommentare in allen Lebenslagen. Tot slot, dank je Catherine - omdat ik me geen beter kamergenoot en paranimf had kunnen wensen. En dan zo velen buiten de taalkunde maar wel even betrokken: Anne Marijke, Gerdi, Jacomijn, Noor, Michel, Carolien, Liza, Sandor, Laurens, Ersin, Jolein, Frouke, Natasja, en natuurlijk Esther en Stefan met Elisa, Ida en Julie voor de warmte en gezelligheid te allen tijde en Sandra, Theo, Ursula, Jeroen en Jeroen en de anderen van Fenix voor alle muziek. ¨ den Umschlag und all die Jahre) mit Meine Berliner, Dortmunder und Schweizer: Dominika (fur ¨ und Thomas mit Ida, Kilian, Benedikt, Milan und Habakuk (fur ¨ paddeln Carsten, Nina und Milan; Sulle ¨ Kontakt und so vieles mehr); Hanna, Antje, Ulrike, Eva, Ingo (und Eltern), Sebastian und Roland (fur ¨ ¨ ¨ uber Zeit und Grenzen) und naturlich Florian; Verena, Ocka, Vero und naturlich Anastasia, Ilka und ¨ Freude in DO und anderswo); die Schwiizer Madels ¨ Julius mit Ole (fur und Urs. Ik bedank mijn familie, in het bijzonder Louco en Joost, und meinem Bruder Diederik mit Patricia und ¨ die Herzlichkeit und guten Weine. My father and Susie I would like to thank for their support. Thierry fur Het is zo fijn om jouw steun te ervaren, Papa. Davide, non sapro` mai come ringraziati per tutto. Lieve Mamma en Aat, geen woorden zouden kunnen uitdrukken hoe veel dank ik voel voor jullie steun. Dit boek is aan jullie beider moeders opgedragen.

ix ¨ Und zu guter letzt, Nina: Ohne Dich waren Amsterdam und diese Diss sowieso undenkbar!

Contents

Dankwoord

vii

1 Task-based performance in a second language: cognitive and interactive aspects 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Outlook coming chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 The task-based approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 Task: definition and characteristics . . . . . . . . . . . . . . . . . . . . . . . 1.3.2 Principles underlying the task-based approach . . . . . . . . . . . . . . . . 1.3.3 Task-based research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.4 The Limited Attentional Capacity Model . . . . . . . . . . . . . . . . . . . . 1.4 The Cognition Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 Attentional allocation during task-based performance . . . . . . . . . . . . 1.4.2 Interference theory and the Multiple Attentional Resources Model . . . . . 1.4.3 Cognitive factors of task complexity . . . . . . . . . . . . . . . . . . . . . . 1.4.4 Interactive factors of task condition . . . . . . . . . . . . . . . . . . . . . . . 1.4.5 Learner factors of task difficulty . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.6 The Triadic Componential Framework . . . . . . . . . . . . . . . . . . . . . 1.4.7 Summarizing the theoretical claims of the Cognition Hypothesis . . . . . . 1.5 Measures of task-based performance . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1 Basic units of speech . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.2 Linguistic complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.3 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.4 Fluency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.5 Global versus specific measures of task performance . . . . . . . . . . . . 1.6 Earlier empirical investigations and open issues . . . . . . . . . . . . . . . . . . . . 1.6.1 Investigating cognitive task complexity . . . . . . . . . . . . . . . . . . . . . 1.6.2 Investigating interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.3 Investigating task specific measures . . . . . . . . . . . . . . . . . . . . . . 1.6.4 Investigating native speaker task-based performance . . . . . . . . . . . . 1.7 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 3 5 5 6 9 10 12 12 15 17 20 22 23 24 26 26 27 28 29 29 30 30 32 35 35 36

2 The studies in this book 2.1 Goal . . . . . . . . . . . . . . . . . . 2.2 The claims under investigation . . . 2.3 Research questions and hypotheses 2.4 Design . . . . . . . . . . . . . . . . . 2.5 Participants . . . . . . . . . . . . . . 2.6 Measures . . . . . . . . . . . . . . . 2.7 Concluding remarks . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

37 37 38 40 42 42 44 47

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

xii

Contents

3 Study 1: The influence of complexity in monologic versus dialogic tasks in Dutch L2 3.1 The Cognition Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 Task complexity: cognitive factors . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 Task condition: interactive factors . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3 Cognitively complex interactive tasks . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4 Previous research within the Cognition Hypothesis . . . . . . . . . . . . . . . . 3.2 Research questions, method, and design . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Production measures: accuracy, complexity, and fluency . . . . . . . . . . . . . 3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Fluency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.4 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Hypothesis 1: effects of task complexity . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Hypothesis 2: effects of task condition . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Hypothesis 3: combined effects of task complexity and task condition . . . . . 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Study 2a: Effects of task complexity and interaction on L2-performance 4.1 Attention and task performance . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Cognitive task complexity . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.3 Cognitive task complexity and interaction . . . . . . . . . . . . . 4.2 The present study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Effects on linguistic complexity . . . . . . . . . . . . . . . . . . . 4.3.2 Effects on accuracy . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Effects on fluency . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Effects of increased cognitive task complexity? . . . . . . . . . . 4.4.2 Interaction and L2-performance . . . . . . . . . . . . . . . . . . . 4.5 Conclusion and directions for future research and practice . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

49 49 50 51 52 52 53 54 55 56 58 59 59 60 62 62 63 63 64

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

67 68 68 70 73 75 76 81 81 82 83 85 85 89 91

5 Study 2b: The use of conjunctions in cognitively simple versus complex L2-tasks 5.1 Cognitive task complexity and the Cognition Hypothesis . . . . . . . . . . . . . . . 5.2 Global versus specific measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 The use of conjunctions as a task specific measure . . . . . . . . . . . . . 5.3 The present study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Research questions and hypotheses . . . . . . . . . . . . . . . . . . . . . 5.3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 The use of conjunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Focusing on specifically task relevant conjunctions . . . . . . . . . . . . . . 5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1 The use of conjunctions in cognitively simple versus complex tasks . . . . . 5.5.2 The factor ± few elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.3 Good measures of task performance . . . . . . . . . . . . . . . . . . . . . . 5.6 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

93 93 95 97 98 98 99 103 104 105 108 108 109 110 111

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

Contents

xiii

6 Summary of the findings, discussion, and implications 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Theoretical basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 The claims of the Cognition Hypothesis . . . . . . . . . . . . . . . . 6.2.2 Alternative accounts on cognitive task complexity and interaction . . 6.2.3 The hypotheses under investigation . . . . . . . . . . . . . . . . . . 6.3 Empirically investigating effects of cognitive task complexity and interaction 6.3.1 Experimental studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Effects of cognitive task complexity . . . . . . . . . . . . . . . . . . . 6.3.3 Effects of interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.4 Combined effects of cognitive task complexity and interaction . . . . 6.3.5 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.1 Cognitive task complexity . . . . . . . . . . . . . . . . . . . . . . . . 6.4.2 Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.3 Cognitive task complexity and interaction in combination . . . . . . . 6.4.4 Global versus specific measures of task performance . . . . . . . . 6.4.5 The benefits of a native speaker baseline . . . . . . . . . . . . . . . 6.5 Theoretical implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.1 Cognitive and interactive factors of task design . . . . . . . . . . . . 6.5.2 Manipulating cognitive task complexity . . . . . . . . . . . . . . . . . 6.5.3 Manipulating interaction . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.4 Manipulating cognitive task complexity by interaction . . . . . . . . . 6.5.5 Measuring task-based L2-performance . . . . . . . . . . . . . . . . 6.6 Practical implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.1 Challenging tasks may include many elements . . . . . . . . . . . . 6.6.2 Learning and testing by monologic and dialogic tasks . . . . . . . . 6.6.3 Sequencing tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

113 113 113 114 115 116 117 117 119 120 121 122 124 124 129 131 132 133 134 134 136 136 138 138 140 140 141 141 143

A Appendices to study 1 A.1 Proficiency task . . . . . . . . . . . A.2 Score sheet for the proficiency task A.3 Background information sheet . . . A.4 Tasks and instructions . . . . . . . A.5 Examples of transcripts . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

145 146 148 150 151 155

B Appendices to study 2 B.1 Proficiency task . . . . . . . . . . . . . . . . . . . . . . . . . . . B.2 Background information sheet for L2-learners and L1-speakers B.3 Perceived task difficulty assessment sheet . . . . . . . . . . . . B.4 Tasks and instructions . . . . . . . . . . . . . . . . . . . . . . . B.5 Examples of transcripts . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

161 162 163 165 166 171

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

English summary

187

Nederlandse samenvatting

195

Bibliography

205

List of Tables

1.1 The Triadic Componential Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.2 Predicted effects of task complexity and interaction based on the Cognition Hypothesis . 25 2.1 Predicted effects of task complexity and interaction of the present studies . . . . . . . . . 39 2.2 Experimental design of the present studies . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.3 Background information for all participants of the present studies . . . . . . . . . . . . . . 43 3.1 3.2 3.3 3.4 3.5 3.6 3.7

Predicted effects of task condition and task complexity . . . Manipulated factors . . . . . . . . . . . . . . . . . . . . . . Measures of accuracy, complexity, and fluency . . . . . . . Descriptives of all measures . . . . . . . . . . . . . . . . . . Results of the repeated measures MANOVA on accuracy . Results of the repeated measures MANOVA on complexity Results of the repeated measures MANOVA on fluency . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

52 54 55 57 58 60 61

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12

Previous work on cognitive task complexity and interaction . . . Predicted effects of task complexity and interaction . . . . . . . Experimental design . . . . . . . . . . . . . . . . . . . . . . . . Background information for all participants . . . . . . . . . . . . Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Descriptives of the measures of linguistic complexity . . . . . . Statistics of the measures of linguistic complexity . . . . . . . . Descriptives of the measures of accuracy . . . . . . . . . . . . Statistics of the measures of accuracy . . . . . . . . . . . . . . Descriptives of the measures of fluency . . . . . . . . . . . . . Statistics of the measures of fluency . . . . . . . . . . . . . . . Manifested effects of cognitive task complexity and interaction .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

71 75 76 77 79 80 81 82 83 84 85 86

5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8

Background information for all participants . . . . . . . . . . . . . . . . Dutch conjunctions under investigation . . . . . . . . . . . . . . . . . . Absolute numbers for the frequency of conjunctions for L2-learners . . Absolute numbers for the occurrence of conjunctions for L2-learners . Descriptives on frequency and occurrence of all conjunctions . . . . . Statistics on frequency and occurrence of all conjunctions . . . . . . . Descriptives of specifically task relevant conjunctions for L2-learners . Descriptives of specifically task relevant conjunctions for L1-speakers

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

100 102 103 104 106 106 107 107

6.1 Summary of results of the present studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.2 Summarizing effects of the factors ± few elements and ± monologic . . . . . . . . . . . . 135 6.3 Example of task sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

List of Figures

1.1 1.2 1.3 1.4 1.5

Which two girls would make the best studying couple? . . . Predicted trade-off effects . . . . . . . . . . . . . . . . . . . Robinson’s model of memory and attention . . . . . . . . . The Multiple Attentional Resources Model . . . . . . . . . . Resource-directing versus resource-dispersing dimensions

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

2 11 13 16 19

Chapter

1

Task-based performance in a second language: cognitive and interactive aspects

1.1

Introduction

The task-based approach to second language acquisition (SLA) promotes learning a language by means of tasks. In box 1.1 on page 2 you find an example of a typical task that second language learners may be confronted with in a task-based textbook. When performing this task a whole set of linguistic and cognitive skills is needed – like in real-life communication. First, task performers need to understand the written information that is given in order to know what they need to do. Second, to reach the extra-linguistic goal of the task they have to further process this information. In concreto, in order to convince their friend they need to balance reasons for or against possible task outcomes. During the two minutes preparation time, they will think about a possible solution but on top they are likely to focus their attention towards linguistic aspects of performance. For example, they think about what phrases and words they may use in order to express their opinion. Third, the actual task performance asks for a demanding oral activity in their second language (L2): They have to discuss their choice with an interactant during a phone call. For L2-learners talking on the phone is especially challenging because they can only rely on aural input that is not accompanied by visual cues, like facial expressions of the interlocutor. Last but not least, this task requires L2-learners to process the target language not only in production but also in perception. After all, they read the instruction, they formulate their own messages, and they listen to their speaking partner. In sum, this example presents a challenging real-life communicative task that requires language use during a holistic activity – which is the aim of the task-based approach.

2


Box 1.1: Traveling to Antwerp task1

STUDIYING IN THE NETHERLANDS Petra age: from: study: reading: L2-classes since: taking final exam:

Sofie 25 years Poland pedagogy regularly 8 months maybe

age: from: study: reading: L2-classes since: taking final exam:

23 years France pharmacy often 12 months yes

STUDIYING IN BELGIUM Dzifa age: from: study: reading: L2-classes since: taking final exam:

Marta 24 years Ghana French never 5 months no

age: from: study: reading: L2-classes since: taking final exam:

22 years Germany history often 14 months maybe

Figure 1.1: Which two girls would make the best studying couple?

Instruction Together with a friend you are organizing a study exchange program. Students of Dutch as a second language in the Netherlands and Belgium will be paired into studying couples. Together they will practice the Dutch language by using e-mail, chat and telephone. Next weekend all the participants of the program will get to know each other in Antwerp. There are two more places left but you received applications of the four girls you see in Figure 1.1. Together with your friend you need to decide who will be accepted for the program. In two minutes from now you will call your friend and discuss on the phone which couple is your favorite. Look at the descriptions of the students. Take a decision about which two (one from the Netherlands and one from Belgium) would probably make a good couple and who is likely to work happily together. Prepare yourself to explain in detail who you would choose. Describe not only the benefits of the best choice but include also why other couples are less likely to be a good match. Note, try to be convincing because you and your friend will have to agree in the end. There is only one more couple joining you to Antwerp.

1 The depicted people agreed on using their photograph for this research project. The listed names and characteristics are fictitious. Any resemblance to their real identity is purely coincidental.

1.2 Outlook coming chapters

3

The task-based approach ‘is an educational framework for the theory and practice of teaching second or foreign languages’ (van den Branden, Bygate, and Norris 2009: X). In order to learn the second language a task-based curriculum asks students to perform on carefully designed tasks. As the performance on these tasks requires target language use, this generates possibilities for interlanguage development and L2-learning. This raises the question, what is a carefully designed task? The work presented in this book adopts a cognitive perspective on task-based research when addressing this issue. In general, it is interested in how tasks may be manipulated such that they give the most beneficial output with respect to L2development. From a more research oriented point of view this book is interested in what and how task characteristics influence the cognitive processes underlying L2-task performance. With respect to these questions Robinson (1995b, 2001a, b, 2003b, 2005) proposed a theory that makes specific claims about how manipulations of task characteristics affect task-based L2-performance. His proposal, that became known as the Cognition Hypothesis, is the theory under investigation in the present work. More specifically, this book focuses on the predictions of the Cognition Hypothesis with respect to effects of cognitive task complexity and effects of interaction on L2-learners’ oral task performance. These factors are investigated both on their own and in combination.

1.2

Outlook coming chapters

Adopting a cognitive perspective on task-based research, this chapter gives the theoretical background of the studies presented in this book. Section 1.3 elaborates on the basis of task-based research as it defines the central unit of investigation, i.e. a ‘task’, and discusses how task-based language performance may foster L2-development. Furthermore, this section explains how the present work may contribute to four different strands of task-based research and elaborates on one alternative cognitive model of task-based L2-production, the Limited Attentional Capacity Model (Skehan 1996, 2001, Skehan and Foster 2005). Section 1.4 discusses in detail the Cognition Hypothesis and presents Robinson’s claims that are investigated in this book. The focus of this section is on predicted effects of increased cognitive task complexity and effects of interaction on task-based L2-performance. In section 1.5 the dependent variables used in the work at hand are discussed, i.e., global and specific measures of task-based L2-performance. Finally, section 1.6 reviews earlier work investigating the Cognition Hypothesis by addressing open issues that ask for more empirical work. Chapter 2 introduces the empirical work presented in this book. It formulates the general research questions and hypotheses and gives a description of the design and participants of the empirical studies that are presented in chapters 3, 4, and 5. It is important to consider that these empirical chapters are each intended as individual papers. That is, they have been or will be published outside the context of this book. As a result there is some overlap in their content – in particular when reviewing the theoretical

4


basis of the work. As far as there seem to be differences in, for example, the predictions between the studies, these reflect the evolution of the work presented here. In other words, the findings of study 1 (chapter 3) were the base for study 2 presented in chapter 4. The study presented in chapter 5 in turn is built on the knowledge gained from the two earlier investigations. Consequently, differences mirror the chronological growth of the research in this book. Chapter 3 elaborates on study 1, a first empirical investigation among 44 Turkish and Moroccan learners of Dutch as an L2. Participants in this study act on cognitively simple and complex tasks either alone (monologic) or in pairs (dialogic). The simple and complex version of an oral argumentative reasoning task include a different number of elements participants have to take into account for successful task completion. In the simple task (+ few elements) participants are asked to help a friend with the choice between two electronic devices. In the complex version (– few elements) there are six different options. The analysis of the L2-learners’ task performances focuses on global measures of linguistic complexity, accuracy, and fluency. The discussion highlights effects of cognitive task complexity and interaction both on their own and in combination. Chapter 4 gives the details of study 2, the second empirical investigation. It elaborates on the results of chapter 3 but overcame some methodological problems and extends the groups of participants. It examines the task performance of 64 L2-learners of Turkish and Moroccan background on oral argumentative reasoning tasks. In addition, 44 L1-speakers of Dutch serve as a native speaker control group. Again, participants act on cognitively simple and complex tasks in either a monologic or a dialogic setting. This time, the tasks require participants to combine people into pairs. The simple version presents four, the complex version nine possibilities, respectively. The discussion focuses on (combined) effects of cognitive task complexity and interaction and interprets the L2-results established by means of global measures of linguistic complexity, accuracy, and fluency in light of the L1-speaker’s baseline data. Chapter 5 presents a more elaborate analysis of the data of study 2. As an extension by means of a task specific measure this investigation examines the L2-learners’ and L1-speakers’ task performances with respect to the use of conjunctions. The data analysis evaluates the frequency and occurrence of conjunctions in the L2- and L1-speaking performances on cognitively simple versus complex argumentative reasoning tasks. The discussion focuses on the use of global versus specific measures of task performance. Chapter 6 discusses the general findings, implications, and conclusion of this research. After presenting the claims under investigation it summarizes the outcomes of the empirical work presented in chapters 3, 4, and 5. Results are discussed in light of the Cognition Hypothesis and the hypotheses formulated in chapter 2. Furthermore, this chapter addresses the concept of cognitive task complexity, relates the data to the role of interaction, discusses the use of global versus specific measures of task performance, and elaborates on the differences between native and non-native task performance. In

1.3 The task-based approach

5

the end directions for future research and practical implications for L2-pedagogy based on the findings presented in this book are given.

1.3

The task-based approach

In the last decades many scholars have devoted themselves to research into task-based second language teaching (TBLT). This book addresses task-based language research (TBLR) and the explanations here will be limited to those aspects of the task-based approach that are relevant for the present work.2 The guiding idea of the task-based approach is that tasks are the central unit of any pedagogical intervention. From a theoretical point of view many researchers have defined what exactly they mean by a ‘task’ (a.o., Bygate, Skehan, and Swain 2001, Candlin 1987, Crookes 1986, Long 1985, Nunan 1989, Prahbu 1987). The next section reviews those definitions that emphasize the cognitive aspects of a task and therefore are important for the studies in this book.

1.3.1

Task: definition and characteristics

For Prahbu a task is ‘[a]n activity which required learners to arrive at an outcome from given information through some process of thought and which allowed teachers to control and regulate that process’ (Prahbu 1987: 24). This definition highlights the cognitive processing (‘process of thought’) of information during task performance. Also Ellis characterizes tasks as ‘external means by which we can influence the mental computations that learners make’ (Ellis 2000: 198). However, he defines tasks as a ‘workplan’ consisting of ‘(1) some input (i.e., information that learners are required to process and use); and (2) some instructions relating to what outcome the learners are supposed to achieve’ (Ellis 2000: 195). Ellis’ second point stresses the importance of the outcome of the task. Also Skehan (1998) highlights the primacy of an extra-linguistic goal that needs to be reached by means of language use. He explains that the activity induced by a task should have a communicative problem to solve and that task performance will be evaluated with respect to this goal. Similarly, van den Branden emphasizes the communicative goal: ‘A task is an activity in which a person engages in order to attain an objective, and which necessitates the use of language’ (van den Branden 2006: 4). Ellis adds that tasks are ‘activities that call for primarily form-focused language use’ (Ellis 2003: 3). And also Long and Robinson (1998) argue that although tasks may be primarily meaning oriented, they should have some concern for form as a potential for L2-development. 2 For more extensive discussions I refer to the following texts: Overview articles on the theoretical rationale behind the taskbased approach are provided by Ellis (2000) and Skehan (2003). More elaborately, Samuda and Bygate (2008) and Eckerth and Siekmann (2008) present the theory and research into TBLT together with empirical studies. The focus of these volumes lies on implications for the classroom. van den Branden et al. (2009) reprinted twenty seminal papers in the field of TBLT in order to summarize the most prominent ideas and themes.

6

Task-based performance in a second language: cognitive and interactive aspects In sum, even though they stress different aspects, most researchers seem to agree on the following

points: (a) A task is an activity that promotes holistic language use. (b) The main aim of a task performance is to achieve the communicative (extra-linguistic) goal. (c) In order to reach this goal, the use of language is required. (d) The imposed language use is linguistically challenging such that the L2-learner is pushed to pay attention to language form. (e) The cognitive processes that are involved in meeting the (extra-)linguistic task demands promote interlanguage development. Samuda and Bygate give a conclusive summary when they define a task as ‘a holistic activity, which engages language use in order to achieve some non-linguistic outcome while meeting a linguistic challenge, with the overall aim of promoting language learning, through process or product of both’ (Samuda and Bygate 2008: 69). The present book will work with this last definition when referring to a task.

1.3.2

Principles underlying the task-based approach

The task-based account promotes language learning by means of tasks. The idea is that in the taskbased classroom students make use of the same communicative acts as in the real world outside the classroom. Therefore, task-based activities are said to have (at least) a two-fold advantage over traditional language teaching methods: First, during task performance L2-learners practice those skills (e.g., fluency) and that knowledge (e.g., vocabulary and syntax) they need for actual use of the second language outside the classroom. When learners perform on a task in their L2, next to learning how to convey their message (= meaning), they will also be triggered to use the appropriate words and linguistic structures (= form) for that task. Carefully designed tasks therefore may promote the use of specific linguistic structures and forms that are necessary for successful task performance. Second, performing on communicative tasks in authentic contexts most likely is more motivating for L2-learners than doing traditional exercises. As explained in the introduction (see p. 1 to 2), the ‘Traveling to Antwerp’ example task largely follows the requirements (a) to (e) above. It asks for a communicative activity that is guided by achieving an extra-linguistic goal, i.e., convincing the friend about ones choice. In order to reach this goal it presents a challenging linguistic task. L2-learners may formulate in a clear way a convincing line of argumentation. As speakers express and accordingly lexically mark their choices, this task therefore possibly will promote the use of linguistic structures that are related to argumentation (e.g., ‘I think


7

that. . . ’ or ‘because’ and ‘so’). A typical performance sequence on this task between two speaking partners A and B may have the following form:3 A: Maybe it is nice to combine Dzifa and Sofie because uh Sofie is from eh France. B: yes A: And uh Dzifa, she studies French. So Dzifa could learn a lot from it, uh I guess. B: But they will speak French with each other. And that is not, not really the idea of uh . . . A: No, because the idea is that they will learn Dutch. B: I thought of Sofie and Marta, because they uh yes both are European. Performance on the ‘Traveling to Antwerp’ example may ask for a communicative activity that encourages the use of linguistic means that mark the balancing of reasons. The example shows some repeated instances of the conjunction ‘because’ and lexical items like ‘guess’, ‘thought’, and ‘maybe’. A picture description task with the same input material most likely would not induce these forms and could be satisfactorily completed by using ‘and’, ‘and’, ‘and’. Successful performance on the actual example, however, may foster the use of causal conjunctions like ‘because’. In contrast to a more traditional language exercise, e.g., a fill in the gaps sheet with blanks on every conjunction, this task gives L2-learners the chance to use these forms during the natural context of an argumentative activity. As task-based performance accordingly may generate opportunities for uptake and intake of linguistic forms that are required for successful task performance it has been related to interlanguage development. Different seminal hypotheses in second language acquisition (SLA) corroborate the task-based approach as a pedagogical intervention for learning a second language. Krashen’s Input Hypothesis states that language acquisition occurs if and only if L2-learners are confronted with ‘comprehensible input’. Comprehensible input is information, that is slightly too difficult, i.e., input that is one step beyond the current level of the processing ability of L2-learners. Krashen (1985) termed this ‘i+1’. The input is understood – and eventually learnt – because the unknown aspects of the input are embedded in a rich linguistic and non-linguistic context that is comprehensible. The Output Hypothesis puts forward that input processing alone is not enough for L2-development because comprehension can be successful without fully understanding the linguistic input. Swain and Lapkin therefore argue that only through the production of output L2-learners will focus their attention on formal aspects of language, which generates possibilities for learning (Swain 1985, 1998, Swain and Lapkin 2001). Pushed output forces L2-learners to compare their own production to the target forms. This cognitive comparison will make them aware of differences between their own actual interlanguage 3 English translation of two Dutch native speakers performing the ‘Travelling to Antwerp’-task. See Appendix B.5 for the complete original in Dutch.

8


and the target performance of native speakers that will be classified as a gap in the interlanguage system. ‘Noticing the gap’ raises the awareness about which gaps still need to be ‘filled’ through L2learning (Gass, Mackey, and Pica 1998). Also Schmidt (1990, 2001) states in his Noticing Hypothesis that L2-learners need to pay attention to details of the L2 as only focused attention may affect L2-learning. Noticing raises the awareness about the own interlanguage system. As learner’s will notice what forms, words, and structures exactly they need to learn in order to perform successfully in their L2, Long (1985) sees noticing as a necessary condition for uptake and intake of new information. Not least importantly, heightened interlanguage awareness presumably has a positive effect on the learner’s motivation and creates opportunities for language learning. Finally, Long’s (1985, 1989) Interaction Hypothesis explains that in particular interactive tasks are important for SLA. For successful interaction it is crucial to understand and to be understood. L2learners are confronted with input they need to analyze and comprehend as a hearer while in addition, they have to make meaningful use of their L2-knowledge as a speaker. That is, they are pushed to perceive input and to produce output. When a speaker fails at being comprehensible, he will receive negative feedback from the speaking partner (Pica 1994). As the hearer may ask for clarification or the speaker checks for comprehension, interlocutors start to negotiate about the meaning of an utterance. Especially, if errors affecting the linguistic code obscure meaning, the hearer presumably pushes the speaker to reformulate his initial communicative attempts. It is likely that this induces so-called ‘language related episodes’ (LREs). ‘[W]hen learners negotiate meaning by means of requests for clarification or confirmation checks, they can obtain interactionally modified input that both helps them to comprehend the input and focuses their attention on new or partially learned linguistic forms, thus enabling their acquisition’ (Ellis and He 1999: 286). Importantly, as the speaker receives this feedback at ‘the most propitious moment [. . . ] when the meaning is problematic and when the learner is thought to be most receptive’ (Skehan 2003: 3), this process may also induce a restructuring of the underlying system. Accordingly, during LREs the attention is briefly pointed towards form without losing the primary focus on meaning (Long 1989, Long and Robinson 1998). This establishes what Long calls ‘Focus on Form’ (FonF): the attention of both interlocutors is drawn to the linguistic code. FonF is a prerequisite for language learning as only tasks that generate (some kind of) awareness to language form will contribute to L2-development (Long 1983, 1985, 1989). During meaning-oriented interaction ‘[i]t is the realization of divergence between L2 forms and target language (TL) forms that becomes the catalyst for learning’ (Gass et al. 1998: 301). FonF and LREs during interaction thus enhance noticing, uptake and intake of new information (Pica 1994, Schmidt 1990). In short, interaction has the potential to trigger cognitive processes that generate L2-learning


9

opportunities and possibly promote L2-development.4 As a whole, these accounts explain how a task-based approach may promote SLA. After all, performing a task (especially in interaction) is a holistic activity that requires input and output processing in the rich context of an authentic task (Krashen 1985, Swain and Lapkin 1995). As L2-learners are primarily driven by reaching the communicative goal (for example, convincing their speaking partner about the best studying couple in the ‘Traveling to Antwerp’-task) they make meaningful use of their target language knowledge. Consequently, task-based L2-performance generates many opportunities for noticing and FonF (Long 1989, Schmidt 1990). As the linguistic challenge of the task draws the learner’s attention towards formal aspects, which are needed for successful task completion (e.g., the use of causal conjunctions in argumentation), task-based performance can push meaning and form in parallel (Robinson 2003b). As a result, it may (implicitly) enhance uptake and memory for correct language forms. It follows that performance on carefully designed L2-tasks can contribute to SLA. The question how exactly task-based activities affect L2-performance and may promote language learning is the subject of task-based research, which will be discussed in the next section.

1.3.3

Task-based research

Research into task-based language pedagogy is interested in exploring a wide range of issues. Skehan (2003) identified four different strands of task-based research. The first strand of research is in line with Long (1989). It advocates a psycholinguistic point of view and emphasizes the role of interaction. Following the Interaction Hypothesis this approach argues that tasks should be designed such that they induce the most interaction possible. This type of task-based research explores how tasks can be designed such that they trigger fruitful interaction. The second strand adopts a sociocultural view on interaction and focuses on the co-construction of meaning. According to Lantolf (2000) the collaborative L2-performance creates an output that is beneficial for both speaking partners. As both provide the relevant structures they know, the joint knowledge potentially stretches interlanguage and pushes L2-development (Swain and Lapkin 2001). This perspective accordingly explores the type of interaction that evolves upon different kind of tasks. Thirdly, the structure-focused account advocates designing tasks that make learners use a specific linguistic structure, which is practiced and eventually learnt by performing on that task (Loschky and Bley-Vroman 1993). For example, a task where you tell about your nicest holiday experience asks for past tense while a task with the title ‘What will you do the coming holidays?’ requires the use of future tense. Both of them will ask for specific lexical items for time reference. Research within this strand tries 4 Also for L1-development interaction is essential for linguistic growth: The Vygotskian theory claims that only through interaction with adults or peers children are pushed to higher levels of performance such that higher order cognitive functions, including linguistic skills, are developed (Watson-Gegeo and Nielsen 2003).

10


to identify characteristics and effects of structure-focused tasks. The cognitive strand of task-based research aims at understanding the cognitive and attentional processes during L2-task performance. It investigates how manipulations of different task characteristics may affect attentional allocation because only focused attention has the potential to promote L2-development (Schmidt 1990). Especially the cognitive load a task puts forward, i.e., cognitive task complexity, has received a lot of attention as it is crucial in guiding the focus of attention during taskbased performance (Robinson 1995b, 2001b, 2005, Skehan 1996, Skehan and Foster 2001). Within the cognitive strand there exist contrasting hypotheses concerning factors of cognitive task complexity. Robinson’s (2005) theory, that has become known as the Cognition Hypothesis, is based on the idea of multiple attentional resources. As the research presented in this book takes Robinson’s claims as object of investigation section 1.4 will give a detailed description of the Cognition Hypothesis. The following section 1.3.4 will discuss an alternative cognitive account on task-based L2-performance that advocates the idea of limited attentional capacity (Skehan 1996, Skehan and Foster 2001).

1.3.4

The Limited Attentional Capacity Model

The Limited Attentional Capacity Model, as its name states, claims that attentional capacity is limited (Skehan 1996, 1998, 2003, Skehan and Foster 1997, 1999, 2001, 2005). Assuming capacity limits means that there is some maximum in the amount of information one can keep active or pay attention to. Also the number or size of attentionally controlled processes that can take place in parallel is limited (Schmidt 2001).5 Accordingly, when performing a task the available resources have to be shared between all the processes a task asks for, e.g., input selection, goal oriented information processing and response actions (Baddeley 2003). If various task demands exceed the total of available resources, the different processes come into competition and the control function of attention will decide where to allocate attention to. For example, it will prioritize important over unimportant task aspects. The central point of Skehan and Foster’s account is that also during task-based L2-performance the ‘attentional limitations for the L2-learner and -user are such that different areas of performance [linguistic complexity, accuracy, and fluency] compete for one another for the resources that are available’ (Skehan and Foster 2001: 205). In other words, the different ongoing processes during task performance are in competition with each other for attentional resources. As only those aspects that receive enough attention will reach optimal performance, processes which receive no or limited attention will fail or become erroneous such that overall performance declines. Skehan and Foster argue that during L2tasks learners first and foremost want to reach the communicative goal such that they will prioritize meaning over form (VanPatten 1990). 5 The exact nature and size of the attentional capacity limits are still under debate but it may be restricted to four elements or chunks of related information (Halford, Cowan, and Andrews 2007). See Baddeley (2003), Baddeley and Hitch (1974), Cowan (1988, 1993, 2000), Repovs and Baddeley (2006), Robinson (2003a), Schmidt (2001), Talmy (2008), Ullman (2004) for more information about the human cognitive systems of memory and attention.


11

Figure 1.2: Predicted trade-off effects (adapted from Skehan and Foster 2001)

dimensions of performance

fluency

form

accuracy

linguistic complexity

According to the limited attentional capacity perspective L2-tasks thus should not be too heavy in terms of cognitive complexity. After all, when performing cognitively too complex tasks most attention may be ‘swallowed’ by the communicative aim and only the residual is left for formal aspects. As a result the different dimensions of L2-task performance come into competition with each other, which is likely to generate trade-off effects between the dimensions of performance.6 Skehan and Foster (2001) present a model that illustrates these trade-offs during L2-language production (see Figure 1.2). In a cognitively complex task, first, fluency and form compete with each other, i.e., speakers prioritize fluency at the cost of form or vice versa. At the next level the formal notions – linguistic complexity and accuracy – compete for the attentional capacity that remains after ‘subtracting’ the fluency dimension. Crucially, in this model limitations in cognitive capacity may be most prominent as trade-off effects between linguistic complexity and accuracy. The authors argue that when performing a cognitively complex task, L2-learners will focus their attention (either consciously or not) to one of the three dimensions linguistic complexity, accuracy, or fluency. That is, if they focus on linguistic complexity this will reduce the accuracy of task performance. For example, when L2-learners explore more complex structures and unknown vocabulary, they will produce more errors. Skehan and Foster termed this the ‘accuracy last’ approach. If in contrast L2learners choose to have control over their interlanguage, they adopt a ‘safety first’ approach: They avoid uncontrolled explorations of the target language because they focus on accuracy. The speech performance is linguistically less complex but more accurate because L2-learners rely on simpler and known language forms (Skehan and Foster 2001). 6 Traditionally, L2-task performance is evaluated by measures of linguistic complexity, accuracy, and fluency. Section 1.5 will discuss these constructs in more detail. For now it will suffice to acknowledge the three dimensions Complexity, Accuracy, and Fluency, and remember them as so-called CAF-measures.

12

Task-based performance in a second language: cognitive and interactive aspects The Limited Attentional Capacity Model predicts that cognitively simpler tasks are more likely to let

L2-learners focus on both linguistically complex and accurate performance in parallel, because then L2learners do not perceive limitations of their attentional capacity. Cognitively complex tasks will inevitably yield an L2-performance of lower linguistic quality than cognitively simple tasks because attention is a resource of limited capacity. As complex tasks put L2-learners under pressure, trade-off effects between the dimensions of performance occur which manifest themselves most obviously between linguistic complexity and accuracy (Skehan 1996, Skehan and Foster 2001). In contrast to the Limited Attentional Capacity Model the Cognition Hypothesis by Robinson (2005) claims that L2-learners can rely on multiple attentional pools during task performance such that complex tasks may not induce trade-off effects. As the Cognition Hypothesis is the theoretical framework under investigation in the present book, the next section will give a detailed description of Robinson’s ideas.

1.4

The Cognition Hypothesis

Since the early 1990’s Robinson has been developing a theory of task sequencing and L2-development that has become known as the Cognition Hypothesis (Robinson 1995a, b, 2001a, b, 2003a, b, 2005, 2007a, b, Robinson, Cadierno, and Shirai 2009, Robinson and Gilabert 2007). This section elaborates on Robinson’s view on attention and cognitive task complexity to give the rationale behind the Cognition Hypothesis. Finally, it presents Robinson’s (2005) taxonomy of factors that influence task performance because the studies in this book investigate some of his claims related to this so-called ‘Triadic Componential Framework’.

1.4.1

Attentional allocation during task-based performance

According to the Cognition Hypothesis attention is crucial for processing linguistic information as it ‘is the process that encodes language input, keeps it active in working and short-term memory and retrieves it from long-term memory’ (Robinson 2003a: 631). Robinson’s perspective is based on Cowan’s (1988, 1993) idea of hierarchical subset relations between memory and attention: Short-term memory is the activated part of long-term memory, while working memory in turn is the activated part of short-term memory, i.e., that part of memory that is in the current focus of attention (see Figure 1.3). Robinson extends Cowans’ idea by highlighting two attentional processes that are important for SLA: detection and noticing. Detection refers to a mostly unaware automatic recognition process in short-term memory while noticing takes place in the focus of attention. If a specific bit of information receives focal attention it is processed in working memory. As only information that is in the focus of attention can be activated for further processing, noticing, in contrast to detection, does presuppose some kind of awareness

1.4 The Cognition Hypothesis

13

Figure 1.3: Robinson’s model of memory and attention (adapted from Robinson 2003a)

(Schmidt 1990). Input that is automatically recognized (i.e., detected) but not noticed stays in the peripheral scope of attention. That is, it stays in short-term memory and is not processed by working memory. To give an example, when talking to someone on the phone who stands on a windy spot one may detect the acoustic signals of both the voice and the wind. Focused attention will help to filter the voice signals for further processing (e.g., try to understand the words) while it will aim at not attending to the noise of the wind. Robinson distinguishes furthermore ‘data driven’ from ‘conceptually driven’ processes of learning. These processes link the noticed information to the knowledge store in long-term memory. Robinson calls bottom-up, automatic processes that are activated by the data itself data driven. In contrast, conceptually driven processes are participant initiated, top-down, and attentionally controlled. For example, the task to describe a picture with six cars of different colors requires more various color names than describing a picture with two blue cars. Here, data driven processes of the task itself ask for the use of specific lexical color items. Performance on the same picture description task, however, can be guided by top-down processes. If the task instruction on the one hand asks to describe the picture to a police officer as if it was a street scene of an accident, then the performer probably will try to give an accurate and detailed description. As a result, speech performance will slow down while more attention is given to accuracy and lexical precision. If on the other hand the instruction gives the

14


speaker maximally 60 seconds to describe all the cars, the performer most likely will speak faster, will use easily activated vocabulary items and structures, and probably will allow some errors. Generally, humans are aware of the information that is in the focus of attention (Robinson 1995a). Even so, as shown by the example in the picture description task with the cars, not all information that is noticed is a result of conceptually driven top-down processes. Some detection (followed by noticing) may be triggered by bottom-up data driven processes. It follows that a task itself and its characteristics have the potential to affect attentional allocation during task performance. Different aspects of task performance may receive attention guided by on the one hand top-down processes, e.g., induced by a task instruction or by the aim of the task performer to reach a high level of performance. On the other hand data driven processes triggered by task inherent characteristics may draw the attention on the use of, e.g., specific linguistic structures and lexical forms. Both ways, detection in the focus of attention can be followed by noticing. As noticing can be seen as a result of focused attention, attentional allocation during task-based L2-performance is crucial for SLA. After all, only information that is in the focus of attention and is processed by working memory, is noticed and only what is noticed may be learnt (cf. the Noticing Hypothesis in section 1.3.2). Consequently, attentional allocation during L2-task performance to a great extent determines what can be learnt from a task (Robinson 1995a, 2003a, Schmidt 1990). Task aspects that receive more attention have more chances to be noticed (Robinson 2003a). However, what aspect of a task receives most attention is determined by how successful it is at attracting attention (Navon and Gopher 1979). The amount of attention a task may ask for is termed cognitive task complexity (Robinson 2003a). A central claim of the Cognition Hypothesis is that especially cognitively complex tasks are successful in attracting attentional allocation during task-based L2-performance (Robinson 1995a). Robinson ´ (1995) children refers to first language acquisition research to support this claim. According to Givon develop from a ‘pragmatic’ mode to a ‘syntactic’ mode of communication. They start in the pragmatic mode, where they talk within the here-and-now and refer to the immediate environment. Only at later stages of their cognitive development children may use the syntactic mode. Once they have learnt to communicate with complex syntactic and lexical structures more efficiently, they may also talk about the there-and-then. Unlike children, L2-learners do have the cognitive ability to operate in the syntactic mode. Due to their incomplete L2-knowledge, however, they tend to rely on the pragmatic mode. This speech may be characterized by simple morphosyntactic structures and a less varied lexis. Even so, if a task is communicatively or conceptually complex it can prompt and elicit the syntactic mode of production. L2learners performing in the syntactic mode show ‘higher’ levels of linguistic production (e.g., their speech is characterized by a greater use of morphology, greater syntactic subordination, and a higher noun to verb ratio as well as a more varied lexicon, Robinson and Gilabert 2007). In other words, when performing a task in the second language top-down processes guide the


15

attention to the accomplishment of the communicative goals (VanPatten 1990, 1996). Presumably L2learners choose the safe way by using the language they know. However, sometimes, simple linguistic features do not suffice to meet the communicative goals of a task. Some task factors may put up conceptual or communicative demands that require structures and lexical items that are quite complex. As complex linguistic means then are needed in order to fulfill the task correctly, data driven processes push the L2-learner to use more complex language than a simple task would do. For example, tasks that ask learners to tell a story in the there-and-then require them to use verbs in the past tense and to use vocabulary items for time reference. It is obvious that performing such a task creates more opportunities to notice and eventually learn the past tense than when the story can be told in the here-and-now. This rationale explains how cognitively complex tasks, in which the conceptual demands require a more elaborate use of linguistic structures and items, are able to push L2-learners’ performance. Importantly, L2-learners conceptually focus on pursuing the communicative aim of a task while data driven processes induced by task inherent characteristics draw the learner’s attention to formal aspects of the L2 and accordingly may stretch the interlanguage.

1.4.2

Interference theory and the Multiple Attentional Resources Model

As explained in section 1.3.4 performing a cognitively complex task in a second language potentially exceeds the available resources of the L2-learner’s attention such that task performance may suffer. However, other than the Limited Attentional Resources Model (Skehan and Foster 2001), the Cognition Hypothesis does not predict trade-off effects. Instead Robinson (2001a, 2003b, 2005) incorporates the ideas of interference and multiple attentional resources into his theory (Navon 1989, Navon and Gopher 1979, Neumann 1996, Wickens 1991, 2002, 2007). Interference theory predicts that if the cognitive complexity of a task is higher than the available resources, the cognitive system itself loses the control over selective processing (Navon 1989, Navon and Gopher 1979, Neumann 1996). Performance problems then occur due to involuntary attentional shifts that result from a loss of control over attentional allocation. Irrelevant information is processed although it is not a primary task goal because attention focuses on other aspects of performance than the task performer may want to. A drop of performance is the result. For example, people tend to ‘waste’ attention to more automatic processes, which interfere conscious selective processing. Sometimes automatic L1-processes interfere L2-production because there are not enough attentional resources to suppress the automatic activation of the mother tongue (Poulisse and Bongaerts 1994).7 7 Also the Stroop (1935) task is a good example. Even though people are focused on the task (i.e., naming the color of the ink a word is written in) they cannot inhibit the automatic process of semantically processing a word. If the stimulus is a color name, e.g., ‘blue’ written in red ink, task performers tend to answer ‘blue’ rather than ‘red’. The automatic process of reading and semantically recognizing the word ‘blue’ cannot be inhibited. It accordingly interferes the controlled process of reaching the task goal, that is, naming the ink color.

16


Figure 1.4: The Multiple Attentional Resources Model (adapted from Wickens 2007) processing stages perception cognition

responding

spatial input modalities

visual verbal

manual spatial

re sp on se vocal s verbal

auditory

s es oc s pr ode c

spatial

g

in

verbal

The Cognition Hypothesis acknowledges that upon certain increases of cognitive task complexity attention fails at controlling the parallel processes, which are needed for successful L2-task performance. As a result L2-performance problems occur. Yet, the fundamental claim of Robinson (2005) is that not every increase in cognitive task complexity generates interference. The Cognition Hypothesis argues that L2-learners can rely on different attentional resource pools (Navon 1989, Navon and Gopher 1979, Wickens 1989). As task performance depends on the capacity of each of them and various task demands may draw on different pools of attention increasing the cognitive task complexity does not inevitably harm overall task performance. Crucially, assuming multiple attentional resources does not deny the existence of capacity limits but this view claims that not every increase in cognitive task complexity will automatically reach these limits. As long as a task addresses different resource pools no problems during task performance are expected. Wickens (1991, 2002, 2007) developed a model that explains in detail what different task demands may address what kind of attentional pools. His Multiple Attentional Resources Model (see Figure 1.4) distinguishes three different dimensions of human information processing: input modalities, processing codes, and processing stages. First, there are two input modalities: auditory (e.g., spoken word or cue tone) versus visual (e.g., a graphical sign or a blinking lamp). Second, information is processed in different codes, i.e., the spatial code for spatial and analog information or the verbal code for verbal and linguistic material. The third dimension proposes perception, cognition, and responding as different stages during information processing. At every stage different pools of resources are responsible for a different code of information: in perception (speech and print versus graphics and motions), in cognitive processing (linguistic information versus spatial memory processes), and in responding (a verbal reaction like a spoken answer versus a spatially guided manual action like pushing a button).


17

The central claim of Wickens’ model is that when a task requires processes at the same dimension parallel processing of information generates performance problems. For example, one cannot give two spoken answers at the same time. However, if task performance addresses different dimensions, parallel processing is possible without competition for attentional capacity, e.g., one can read a word while pushing a button. As information of a different code, modality or at a different stage may draw on different pools of attention no competition will occur. Only if two tasks both address the same pool of resources performance problems become visible.8 Wickens’ (2002) model is based on empirical research in the field of ergonomics on multi-tasking by pilots or drivers in heavy traffic. The results demonstrate that indeed task performance is hardly affected by parallel processing of different input signals (e.g., visual and auditory), whereas parallel processing of similar information (e.g., two tasks of visual perception) does cause interference. For example, a car driver can on the one hand listen to the radio without being distracted from driving because this is auditory, verbal and linguistic input that is not harming the visual input and manual response processes of driving. On the other hand, searching for a radio station affects driving skills because it withdraws resources from the same attentional pools as driving (visual input and manual response action). Considering task-based performance in a second language Wickens’ model may not be specific enough because it presents only a broad distinction between verbal/linguistic and visual/spatial information processing. Furthermore, other than predicting a drop in performance Wickens does not go into details about the processing of information of the same kind. Still, the idea of multiple resource pools has been appealing for theories on attentional aspects of task-based L2-performance – mere verbal/linguistic information – like Robinson’s Cognition Hypothesis.

1.4.3

Cognitive factors of task complexity

The Cognition Hypothesis argues that some task characteristics may require a more elaborate use of linguistic structures and forms than others. Due to data driven processes induced by their characteristics tasks may attract the L2-learner’s attention and focus it to form. Consequently, the characteristics of a task – especially those that attract attention – are crucial factors for task design. Robinson (2005) termed task characteristics that affect the L2-learner’s attention during task performance ‘cognitive factors of task complexity’.

8 Wickens furthermore predicts that ‘to the extent that any two tasks share common levels along more dimensions (0, 1, 2, or 3), interference will be greater’ (Wickens 2007: 187). Moreover, as revealed by empirical investigations of the model the resources for the first two stages (perception and cognition) are the same and functionally different from those for responding (Wickens 2002). In other words, more interference is expected between perception and cognition but not between either of them and response actions.

18


Resource-directing versus resource-dispersing cognitive factors Robinson (2003a) states that tasks with more complex demands by means of a resource-directing factor focus the L2-learner’s attention towards the linguistic form because the cognitively complex conceptual and performative demands require complex linguistic means. Figure 1.5 names three resource-directing factors: ± here-and-now, ± few elements, and ± no reasoning demands. For example, a resourcedirecting task with many elements rather than a few elements is expected to ask for a more specific lexis and to induce more complex syntactic structures because all the different elements need to be named and distinguished. Similarly, a task that takes place in the there-and-then rather than in the here-and-now will lead to more complex use of vocabulary (e.g., references to time) and structures (e.g., past tense). Alike, complex reasoning tasks will generate more complex language because the line of argumentation may be lexically marked (e.g., by the verbs ‘claim’, ‘propose’, ‘argue’), and syntactically expressed by means of complex sentence structures of argumentation (e.g., ‘if. . . then’ clauses). Simple tasks, that do not involve reasoning, can stick to simpler structures (e.g., clauses coordinated by ‘and then. . . ’).9 Accordingly, resource-directing cognitively complex tasks may result in an L2-output that shows a higher structural and lexical complexity induced by the higher cognitive demands. Due to the increased attention for language also the formal dimension of accuracy will be pushed. So, the increased ´ cognitive complexity may encourage L2-learners to perform in a syntactic mode of processing (Givon 1995). Consequently, resource-directing cognitively complex tasks can promote meaning and form of L2-performance in parallel, that is, linguistic complexity and accuracy may increase both. As L2-learners can rely on different attentional resource pools no competition for attentional resources may be expected (Robinson 2001a, Wickens 2002, 2007). Only fluency possibly suffers from increased cognitive task demands, i.e., complex tasks may induce slower speech with more hesitations, pauses and self-repairs, because fluency is of a more performative nature and therefore may suffer of the high processing effort (Levelt 1989, Riggenbach 2000). At this point, it may be worth noticing that Robinson (2001a, 2003b, 2005) relates his theory to Wickens’ idea of multiple attentional resource pools but does not overtly state how exactly task performers in a second language may address different attentional pools. However, as put forward in section 1.4.2, Wickens distinguishes between verbal/linguistic information and visual/spatial information, while taskbased L2-performance is based on processing information of the linguistic/verbal kind at all stages of processing: perception, cognitive processing, and responding. Wickens’ Multiple Attentional Resources Model therefore presumably would predict interference between the different task demands in a complex task that is manipulated e.g., on the factor ± reasoning or ± few elements, unless one may increase 9 N.B. In this model task complexity factors are presented as a dichotomy (as shown by the + and – symbols). Robinson (2007b), however, explains that most of the factors may be seen on a continuum from simple to highly complex tasks.


19

Figure 1.5: Resource-directing versus resource-dispersing dimensions of cognitive task complexity (adapted from Robinson 2003a)

cognitive task complexity by changing from visual to verbal cues.10 With respect to the resource-directing factors named in Figure 1.5 the Cognition Hypothesis predicts no interference due to focused attention to the linguistic code. In contrast, Robinson (2005) identifies other task characteristics that do increase the cognitive complexity of a task but fail at focusing the L2learner’s attention towards language. He states that these resource-dispersing cognitive factors divert attention away from the formal aspects of task-based L2-performance and spread it over processes that are not relevant for the linguistic performance of L2-learners. Cognitively complex tasks manipulated by means of resource-dispersing factors focus the L2-learner’s attention towards non-linguistic aspects of a task and higher task demands may cause a drop in performance because of interference and attentional shifts. Considering the resource-dispersing variables given in Figure 1.5 a higher cognitive demand by means of the factors ± planning time, ± prior knowledge, and ± single task will affect learner speech negatively. For example, research gathered in Ellis (2005) has shown that increasing cognitive task complexity by means of the factor ± planning time leads to trade-off effects of linguistic complexity and accuracy. In a task with planning time learners can conceptualize their propositional message prior to the actual speaking act. In a no-planning-time condition, learners will need attentional capacity for the conceptualization of the message while they are formulating it (Levelt 1989). As attention is needed for online planning of the propositional content the learner’s attention is drawn away from the formal aspects of the linguistic message (VanPatten 1990). Consequently, cognitively complex tasks on the factor ± planning time are considered to disperse the attentional resources. The Cognition Hypothesis distinguishes so-called ‘resource-directing’ from ‘resource-dispersing’ 10 For example, a picture naming task where the input is visual should leave more attention for linguistic task performance, than an aural-oral word translation task, where the involved cognitive processes are all linguistic in nature. Wickens’ model would predict more interference and accordingly more problems in the spoken performance on the translation task than in the picture naming task.

20


cognitive factors of task complexity. Figure 1.5 depicts this distinction (see also Table 1.1 on page 23). Not providing prior knowledge or giving participants dual rather than single tasks is expected to equally disperse task performer’s attention. The resulting loss of control over attentional allocation creates an inefficient and effortful processing of information. In the end, linguistic complexity, accuracy, and fluency probably all suffer from increased cognitive task complexity on a resource-dispersing factor. To sum up, the Cognition Hypothesis predicts speech on cognitively complex tasks manipulated by means of a resource-directing factor to be more complex and accurate but less fluent, that is, resource-directing factors have the potential to stretch interlanguage use and create opportunities for L2-development (Schmidt 1990). On the contrary, cognitively complex tasks manipulated on a resourcedispersing factor presumably will generate interference and possibly create a drop in performance with respect to all three dimensions of task-based L2-performance. With respect to these claims, the research in this book is interested in the differential effects of cognitively simple versus complex tasks manipulated by means of the resource-directing factor ± few elements. In addition it focuses on what Robinson (2005) calls ‘interactive factors of task condition’ because they have predictive value for task outcome too. The next section will elaborate on task-based interaction.

1.4.4

Interactive factors of task condition

Guided by Long’s (1985) Interaction Hypothesis (cf. section 1.3.2) investigating L2-interaction has been a productive strand in the task-based approach. According to Ellis (2000) this research asked questions like: What task characteristics provide the most interaction? What kind of task generates what kind of interaction? What task manipulations may push L2-development as they promote the key interactional feature, i.e., negotiation of meaning? A body of research focused on interactional modifications during task-based L2-production (e.g., Gass and Varonis 1994, Kuiken and Vedder 2002, Mackey 1999, ´ esz ´ 2007, Shehadeh 2004, Swain and Lapkin 2000). Nassaji 2007, Nuevo 2006, Pica 1994, Rev Just as for cognitive factors of task complexity, the Cognition Hypothesis formulates predictions with respect to how different interactive factors of task condition may affect task-based L2-performance (see again Table 1.1 on page 23). Robinson (2005) distinguishes two types of interactive factors: (a) participation and (b) participant variables. The latter can be seen as grouping variables based on gender, familiarity, and power relations that may be or be not shared between participants.11 The first category, participation variables, are specified in terms of information-flow (e.g., one–way / two–way) and task outcomes (open / closed; convergent / divergent). Interactive factors determine the 11 In the empirical studies presented in this book, the participant factors, e.g., whether people of the same or different gender talk to each other, were controlled over and within conditions. Therefore, no further elaboration of these factors will be discussed here.


21

amount and nature of interaction that a certain task condition will promote. For example, closed and divergent task outcomes are related to more interaction than tasks with an open or convergent outcome (Ellis 2000). Differential effects based on the participation variable one–way versus two–way flow of information forms a major topic of the studies in this book. The distinction one–way / two–way refers to how task relevant information is distributed over participants and to what extent successful task performance involves the exchange of that information (Long 1990). A classical way to manipulate this factor is the use of split information tasks. For example, an interactive task, where one participant has to describe a picture to an interlocutor, will generate an interactional setting where one holds the role of speaker (the information giver) and the other one the role of hearer (the information receiver). As speaking in these kind of tasks is relatively one-sided, they are characterized as tasks with a one–way flow of information. In an interactive split information task, where two interlocutors exchange information and where both participants hold half of the task-relevant information, the interaction will be more balanced because both participants hold both the roles of hearer and speaker. As interlocutors need to give and receive parts of the information, both of them are active participants of a two–way interaction. These kind of tasks are said to ‘produce more negotiation work and more useful negotiation work than one–way tasks’ (Long 1990: 41). In short, two-way tasks may focus the L2-learners’ attention to form. The studies presented in this book adopted a radical form of manipulating the flow of information by investigating the factor ± monologic.

Monologue versus dialogue In a dialogic task condition L2-learners will interact – with all the beneficial aspects discussed in section 1.3.2. During interactive tasks L2-learners have many opportunities to test their hypotheses about the target language. If a speaker fails to be comprehensive the interlocutor gives feedback. Negative feedback causes interlocutors to review their hypotheses and linguistic processes. Furthermore, it is likely that such tasks induce negotiations about meaning and form, clarification requests, modified input and output, and other LREs. All these processes focus the attention of both interactants towards form (FonF) and thus push L2-accuracy (Long 1985, 1989, Pica 1994). In contrast, L2-learners act on their own in a monologic task condition where the flow of information by default is one–way. Accordingly, L2-learners have to rely on their own knowledge and resources. They do not receive other feedback and no interactional modifications will focus their attention to form or meaning. The only way to generate modified output is by monitoring the own speech. However, self-repair is an effortful process that needs time and attentional capacity especially in the L2 (Kormos 1999, 2000a, b). Even though the Cognition Hypothesis makes no spelled out predictions about the factor ± mono-

22


logic, the explanations given above and the claims with respect to effects of the interactive variable one–way / two–way flow of information may allow the following summary: Contrary to monologic tasks, dialogic tasks possibly have a beneficial effect on the accuracy of L2-performance. Mutual understanding is crucial so that the attention of both the L2-learners is focused on form. The turn-taking behavior, in contrast, may prevent L2-learners from producing elaborate linguistic structures as frequent interactional moves and interruptions presumably inhibit speakers to produce complex sentences (Robinson 2001a, 2005). Dialogic tasks therefore are expected to yield a lower linguistic complexity than monologues. In analogy to effects of a higher cognitive task complexity, the focused attention to accuracy may reduce the speed of production so that L2-learners may produce less fluent speech upon dialogic than monologic tasks. Besides the discussed cognitive factors of task complexity and interactive factors of task condition, Robinson (2005) distinguishes a third category of task characteristics, i.e., ‘learner factors of task difficulty’.

1.4.5

Learner factors of task difficulty

According to Robinson (2005) task difficulty is the perceived amount of cognitive effort needed to perform a task (see Table 1.1 on page 23).12 Robinson distinguishes (a) affective variables (e.g., anxiety and motivation) from (b) ability variables (e.g., working memory capacity and aptitude). In combination with the interactive factors of task condition under which a task is performed and the task inherent cognitive factors of complexity, learner factors determine how difficult a task was perceived to be by a learner. As such, the difficulty of a task of a stable cognitive complexity may vary depending on the task performer and the task conditions. Task difficulty is defined by the triple learner–setting–task and emerges during task performance. Robinson argues that it is important to take into account task difficulty factors when trying to interpret task performances of L2-learners. Learner factors possibly interact with factors of task complexity or task condition (Robinson 2001b, 2003b, Robinson and Gilabert 2007). However, the large variety of affective and ability factors that a group of L2-learners brings to the task is less important in task design because researchers cannot manipulate it. After all, task difficulty depends on learner internal factors and therefore, factors of task difficulty may be less important for task-based research interested in how task factors influence task performance. After all, they are not inherent characteristics of a task itself and they do not define the inherent cognitive complexity of a task. The empirical studies presented in this book measure some learner factors in order to be able to explain possible differences in task performance post-hoc. Even so, as cognitive factors of task complexity and interactive factors of task condition are in the focus of investigation, this short explanation 12 Other researchers (e.g., Skehan and Foster 1999) use the words ‘cognitive difficulty’ to refer to the cognitive demands of a task, i.e., what Robinson calls cognitive task complexity.


23

Table 1.1: The Triadic Componential Framework (adapted from Robinson 2005) TASK COMPLEXITY

TASK CONDITION

TASK DIFFICULTY

cognitive factors

interactive factors

learner factors

(a) resource-directing variables

(a) participation variables

(a) affective variables

± here-and-now

one-way/two-way

anxiety

± few elements

open/closed

motivation

± no reasoning demands

convergent/divergent

confidence

(b) resource-dispersing variables

(b) participant variables

(b) ability variables

± planning time

same/different gender

working memory

± prior knowledge

power/solidarity

aptitude

± single task

familiar/unfamiliar

intelligence

Sequencing criteria:

Methodological criteria:

Dependent factors:

Prospective decisions

On-line decisions

Take into account for

for syllabus design

about pairs and groups

post-hoc interpretations

of task difficulty factors may suffice here.

1.4.6

The Triadic Componential Framework

The claims of the Cognition Hypothesis presented in section 1.4.3, 1.4.4, and 1.4.5 are summarized in a classification system of factors of task design that Robinson (2005) termed the ‘Triadic Componential Framework’. This framework is given in Table 1.1 which shows the Triadic Componential Framework based on Robinson (2005).13 It provides a taxonomy of 18 different task design factors. Although, the taxonomic structure never changed, Robinson expanded the Triadic Componential Framework over the past years according to findings of empirical research (Robinson 2007b). Even so, the research at hand adopts the framework from 2005 because the newer version may be considered to be less feasible (Kuiken and Vedder 2007b). As points of critique Kuiken and Vedder named the following issues: First, the 2007 version of the Triadic Componential Framework consists of 36 variables (in contrast to the 18 in Figure 1.1), which leads Kuiken and Vedder to ‘wonder how all these variables can be operationalised and differentiated and how for instance the supposedly different kinds of reasoning should be tested in an experimental setting’ (Kuiken and Vedder 2007b: 265–266). As an example they mention problems with the factor ± few elements because a larger number of elements implies almost automatically an increase in the number of reasoning demands. The factor ± reasoning demands, however, is seen as a separate 13 In addition to the original, this table includes Robinson’s post-hoc considerations with respect to the dependent factors of task difficulty.

24


factor in the Triadic Componential Framework. Accordingly, ‘it is nearly impossible to make a clear-cut distinction between them’ (Kuiken and Vedder 2007b: 265). Moreover, ‘it is far from clear how [all] these variables have to be operationalised, which of them are predominant, how they interact and how fine-grained they should be’ (Kuiken and Vedder 2007b: 265–266). For example, the framework does not explain in detail based on which criteria the number of elements in a tasks should differ in order to determine substantial differences in cognitive task complexity. Similarly, Ellis criticizes that the framework does not ‘specify how to weight the different factors hypothesized to contribute to complexity’ (Ellis 2009: 492). In sum, although it may seem to be more precise, the newer version of the Triadic Componential Framework opens many new questions too. For the present research that is interested in how task design features influence L2-task performance the Triadic Componential Framework from 2005 was used. It provides a useful agenda because it relates predictions to specific task characteristics.14 So far, this section discussed Robinson’s predictions with respect to cognitive factors of task complexity and interactive factors of task condition. The next section summarizes Robinson’s claims and introduces his predictions with respect to a combination of the two types of factors, that is task performance in cognitively complex interactive tasks.

1.4.7

Summarizing the theoretical claims of the Cognition Hypothesis

Table 1.2 gives a schematic overview of the effects of increased cognitive task complexity and changes in interactive task condition on their own and in combination predicted by the Cognition Hypothesis. The Cognition Hypothesis predicts that of the factors distinguished in the Triadic Componential Framework cognitive factors of task complexity are able to substantially manipulate attentional allocation during task-based L2-performance (Robinson 1995a, 2001a, 2005, 2010, Robinson et al. 2009, Robinson and Gilabert 2007). Among the cognitive factors only the resource-directing ones, e.g., the factor ± few elements, have the potential to focus the learner’s attention to language while conceptual demands ask for complex linguistic means. As L2-learners may draw on multiple pools of attentional resources, task performance is pushed towards greater linguistic complexity and accuracy in parallel at the cost of fluency. Increasing the cognitive demands on resource-dispersing factors, e.g., the factor ± planning time, results instead in the learner’s attention being diverted – away from the language code. Due to lost control and interference L2-task performance suffers. Following Robinson’s (2001b, 2005) view on interactive factors of task condition (e.g., one–way versus two–way flow of information) and based on Long’s (1989) Interaction Hypothesis, the factor ± monologic is expected to push accuracy but reduce linguistic complexity and fluency of L2-performance. 14 N.B. Also the Limited Attentional Capacity Model distinguishes different task characteristics that affect L2-task performance (cf. Skehan and Foster 2001 and section 1.3.4). Also Skehan and Foster’s (2001) model gives a distinction between factors of ‘code complexity’ (e.g., vocabulary load), ‘cognitive complexity’ (e.g., clarity and sufficiency of information), and ‘communicative stress’ (e.g., time pressure). In many ways their classification addresses comparable factors as Robinson’s Triadic Componential Framework. See Kuiken and Vedder (2007b, 2008) who give in-depth comparisons of the two models and highlight overlaps as well as contrasts.


25

Table 1.2: Predicted effects of task complexity and interaction based on the Cognition Hypothesis

TASK COMPLEXITY

INTERACTION

TASK COMPLEXITY × INTERACTION

MEASURE L2-LEARNERS

complex

dialogue

complex dialogue

ling. complexity

↑

↓

↓

accuracy

↑

↑

↑

fluency

↓

↓

↓

Note. ↑ = increase; ↓ = decrease

The frequent turn-taking and interactional moves in interaction will prevent speakers from building complex linguistic structures while the joint attention to language and the need for mutual understanding will push both interlocutors to produce more accurate speech at the cost of fluency.

Combining cognitive task complexity and interaction Finally, Robinson (2001a, b, 2003b, 2005) predicts a combined effect of increased cognitive task complexity and interaction. He states that cognitively complex interactive tasks yield more interaction than simpler versions of the same task would do. The rationale is that cognitively complex tasks may need more clarification between interactants than simple tasks. As a result, more negotiation work emerges, which promotes the amount of interaction. Consequently, a dialogic task setting may increase the accuracy of cognitively complex task performance because the higher cognitive task complexity and the interactive task condition both focus the L2-learner’s attention to form. In contrast, linguistic complexity may decrease in cognitively complex interactive tasks because frequent interruptions, clarification work, and comprehension checks decomplexify the syntactic structures that are generated in interaction. The fluency of L2-production may decline because of the higher procedural load induced by the focused attention. Robinson (2001b) argues that the effect of an interactive setting possibly mitigates against effects of a higher cognitive task complexity such that the linguistic complexity of cognitively complex interactive L2-performance is reduced. In contrast, both, interaction and cognitive task complexity, push accuracy at the cost of fluency. After this in-depth description of the theoretical framework under investigation, i.e., the Cognition Hypothesis by Robinson (2001b, 2003b, 2005), with a focus on how task characteristics may affect task performance, the next section will discuss the dependent variables in task-based research into L2-performance, i.e., the constructs of linguistic complexity, accuracy, and fluency.

26


1.5

Measures of task-based performance

The studies presented in this book use both so-called ‘global measures’ of linguistic complexity (C), accuracy (A), and fluency (F) – in short CAF, and a task specific measure. The choice for global CAFmeasures was based foremost on the aim of the present work to relate itself to a tradition of task-based cognitive research into L2-production across tasks and populations, as well as over source and target ´ esz ´ 2007, languages (e.g., Ferrari 2009, Gilabert 2005, Kuiken and Vedder 2007b, Nuevo 2006, Rev Robinson 2001b, Skehan and Foster 2007). In addition, the results of the different studies within the present investigations (that is chapters 3 and 4) need to be compared over tasks and populations. Therefore, task-based performance is evaluated by means of global CAF-measures. More recently, Robinson proposed to complement global by what he calls ‘task-specific’ measures (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). As the focus of the studies presented here lies on investigating Robinson’s claims, chapter 5 evaluates the task-based performance by means of a task specific measure (the use of conjunctions). In order to interpret the data of the empirical investigations this section will give a short introduction to each measure. Furthermore, it discusses the use of global versus specific measures of task performance. This review includes some controversies about the definitions of each constructs because the empirical chapters address this issue only briefly. The actual measures used in the present book are presented in the individual chapters as well as in section 2.6 that explains the design of this work.

1.5.1

Basic units of speech

Several suggestions have been brought forward for basic units of L2-performance, e.g., T-units, C-units, idea-units, utterances, clauses, S-nodes or the number of words (see Crookes 1990, Wolfe-Quintero, Inagaki, and Kim 1998 for an elaborate discussion of the benefits and drawbacks of these units). Basic units are needed in order to calculate measures that are corrected for sample length. Many CAFmeasures are based on frequency counts, e.g., of errors, for accuracy. Drawing conclusions from raw scores though, risks to generate confounded interpretations. L2-learners differ in how much they may (have to) say in reaction to a given input. This results in task performances of different lengths in terms of utterances, clauses, and words. For example, a complex task addressing many elements presumably yields more utterances than a simple version with only a few elements. After all, in the complex task the mere description of all the items probably will need more words than in the simple task. This increases the chance of making an error. It is higher in the complex than in the simple version. Any comparison of raw error frequencies in the simple and complex tasks then would be biased for sample length, such that the results turn ‘meaningless in comparison with other populations or across different tasks’ (Wolfe-Quintero et al. 1998: 10). Therefore it is necessary to correct measures for the length of task performance.

1.5 Measures of task-based performance

27

Foster, Tonkyn, and Wigglesworth (2000) developed the Analysis of Speech unit (AS-unit) and define it as ‘a single speaker’s utterance consisting of an independent clausal or sub-clausal unit, together with any subordinate clause(s) associated with either’ (Foster et al. 2000: 365). The authors present the AS-unit as an alternative to the popular T-unit15 that works well for written language but may not give a reliable tool for spoken language. After all, in spoken language, pausing and intonational patterns suggest that speakers may produce utterances that are longer than a single clause. Furthermore, speech is characterized by elliptical or minor utterances, which would not be considered as independent T-units. For example, coordinated phrases may belong to the same unit if the conjunction combines clauses like in ‘Peter likes chips and Sophie chocolate.’ However, if pauses and intonation suggest a clear break, the ‘and’ may serve as a conversational marker: ‘Peter likes chips (falling tone and pause of 0.5 seconds) and Sophie chocolate.’ The AS-unit acknowledges this difference. Although it first and foremost is a syntactic unit it takes into account intonational and pausing information where syntactic information does not suffice. Foster et al. (2000) also give instructions how to deal with false starts, repetitions, self-corrections, topicalizations as well as interruptions and joint utterances resulting from scaffolding. They furthermore suggest three levels of application, that may be used depending on the aim of the analysis (fine-grained full analysis or investigation of complete utterances only). As a whole, the AS-unit is a workable tool, which gives credit to the specificities of oral task-based L2-performance.

1.5.2

Linguistic complexity

‘[Linguistic] complexity is certainly the most problematic construct of the CAF triad because of its polysemous nature’ (Pallotti 2009: 592). A first problem is that in task-based research into effects of task manipulations the independent input variable of ‘cognitive task complexity’ is easily confused with the ‘linguistic complexity’, which is the dependent variable that measures task output. The second problem lies in the complex composition of the construct itself. Ellis defines linguistic complexity as the ‘extent to which the language produced in performing a task is elaborate and varied’ Ellis (2003: 340). Linguistic complexity comprises of at least two sub-constructs: structural complexity and lexical complexity. Both of them may be further divided into ‘subsubconstructs’ themselves. For example, Norris and Ortega (2009a) propose three sub-types of structural complexity: (i) complexity via subordination; (ii) general complexity indicated by length measures; and (iii) subclausal complexity based on clause length (see there and Norris and Ortega 2003, Ortega 2003, Wolfe-Quintero et al. 1998 for detailed discussions about measures of structural complexity). Also lexical complexity may be devided into two different sub-types. Daller, van Hout, and Treffers15 The T-unit is defined as ‘one main clause plus whatever subordinate clauses happen to be attached or embedded within it’ cf. Hunt 1996, as cited in Crookes 1990: 184.

28


Daller (2003) differentiate (i) text-internal measures, which are calculated with the information of the linguistic sample itself (e.g., the type-token ratio) from (ii) text-external measures, which use an external database, e.g. general word frequency, as point of reference (for example, the lexical profiles proposed by Laufer and Nation 1995). For a more elaborate review of measures of lexical complexity and separate definitions for lexical diversity, density, richness, and vocabulary range consult among others Bulte´ (2007), Daller et al. (2003), Laufer and Nation (1995), Malvern and Richards (2002), Skehan (2009), Vermeer (2000), and Wolfe-Quintero et al. (1998).

1.5.3

Accuracy

According to Housen and Kuiken (2009) accuracy is ‘the ability to produce error-free speech’ (Housen and Kuiken 2009: 1). Of the three CAF-measures, accuracy may be the least controversial. After all, it is based on the concept of the degree to which non-target like forms are used, i.e., error frequencies. People may criticize this very fact because accuracy measures are based on the comparison of learner output with a target norm (often the written standard of native speakers). Errors therefore ignore that some grammatical norm violations may be communicatively adequate in spoken language. Furthermore, accuracy measures evaluate learner performance in a momentum of time such that they cannot say much about interlanguage development (Norris and Ortega 2003, Pallotti 2009, Wolfe-Quintero et al. 1998). Error frequencies generally are related to the sample length in units or words. Some task-based research uses ‘error-free’ clauses or units. However, in populations of intermediate proficiency (e.g., the studies at hand), there may be hardly any units without errors. In addition, these general accuracy measures do not distinguish between an utterance with one error and a hardly understandable utterance consisting of ‘errors only’ (Wolfe-Quintero et al. 1998). On top, they risk overlooking errors that are indicators for developmental steps. Kuiken and Vedder (2008) chose to differentiate their errors based on gravity, from type 1 errors of e.g., spelling, to type 3 errors that obscured meaning. Weighting errors, however, can be rather subjective as it presumably is based on a researcher’s intuition rather than on objective, external means (Wolfe-Quintero et al. 1998). Other researchers therefore proposed to use error types based on different linguistic categories (e.g., Ortega 1999, Wolfe-Quintero et al. 1998). The problem with taxonomies differentiating e.g., lexical, morphosyntactic, and/or other error types, is that it can be difficult to unambiguously assign an error to a specific type (Lennon 1991). For example, when using a non-native choice of preposition does this violate lexico-semantics or morphosyntax? For more detailed reviews of accuracy measures see Norris and Ortega (2003), Pallotti (2009), Polio (1997), and Wolfe-Quintero et al. (1998).

1.5 Measures of task-based performance

1.5.4

29

Fluency

Speed and pauses in speech production may serve as the most prominent measures of fluency. It is ‘the extent to which the language produced in performing a task manifests pausing, hesitation, or reformulation’ (Ellis 2003: 342). This definition sees fluency as a skill. It is the speedy retrieval and production of speech that relies on knowledge. As L2-proficiency increases, the processes in the conceptualizer, formulator, and articulator become more and more automatized and may be processed in parallel (de Bot 1992, Levelt 1989). Advanced L2-speakers accordingly are skilled and fast users of their L2-knowledge. With growing automaticity, more cognitive resources may become available for the process of monitoring. Consequently, the amount of self-repairs may be seen as a fluency measure too ¨ (Dornyei and Kormos 1998, Kormos 2000a). Summing up these three aspects of fluency, the distinction made by Tavakoli and Skehan (2005) has become famous and widely used. The authors distinguish three sub-dimensions of fluency: ‘The first sub-dimension of fluency is silence [. . . ]. A second sub-dimension of fluency deals with the speed with which language is produced. [. . . ] The third sub-dimension of fluency is what is known as repair fluency’ (Tavakoli and Skehan 2005: 254–255). Elaborate discussions on these and various other aspects of fluency can be found in the volume edited by Riggenbach (2000) including the conclusive chapter by Freed (2000) who relates objective measures to expert ratings. One may also consider the work by Chambers (1997), de Jong, Steinel, ´ Florijn, Schoonen, and Hulstijn (2007), Kormos and Denes (2004), Mehnert (1998), and Riggenbach (1991).

1.5.5

Global versus specific measures of task performance

The use of global measures of linguistic complexity, accuracy, and fluency is currently under debate (e.g., Pallotti 2009). In order to complement the broad view on task performance by means of CAF Robinson and colleagues propose to use task specific measures (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). The rationale is: ‘Such specific measures should be more sensitive to conception, task complexity, and its linguistic demands than general measures’ (Robinson et al. 2009: 550). Especially, if tasks are designed in such a way that the demands can be met by the use of specific linguistic structures, task specific measures presumably complete the picture that global CAF-measures leave open. For example, the effect of an increase of cognitive task complexity by means of the manipulation on the factor ± here-and-now may serve as a focused elicitation of specific linguistic markers of the past, e.g., past tense and temporal conjunctions. Investigating past tense in such a task may complement the picture gained by global measures. As such, the use of specific measures next to global CAF may give more insights in the actual linguistic performance of an L2-learner. Not least importantly, it is worth

30


taking them into account because up to now not much earlier work investigated task specific measures and the related claims of the Cognition Hypothesis. So far, this chapter has reviewed theoretical approaches to the cognitive strand of research into task-based L2-performance in general and in particular addressed the Cognition Hypothesis (Robinson 2005). The following section will review earlier research, that has taken Robinson’s theory as its point of departure for collecting and analyzing experimental L2-data.

1.6

Earlier empirical investigations and open issues

A body of research has focused on the investigation of Robinson’s Cognition Hypothesis. Another line of scientific work took the Limited Attentional Resources Model (Skehan 1996) as its base. Together these studies give a valuable perspective on the cognitive approach to task-based L2-performance.16 This section reviews this work by restricting itself to those studies that address the Cognition Hypothesis. As the present studies are intended to contribute to task-based research by addressing some of the open issues that have not been fully answered yet, this review focuses on general findings and open questions of earlier work (see chapters 3, 4, and 5 for discussions of individual papers).

1.6.1

Investigating cognitive task complexity

Manipulations on the factor ± here-and-now have generated positive effects of increased cognitive task complexity as proposed by the Cognition Hypothesis (Gilabert 2005, Iwashihta et al. 2001, Rahimpour 1997, Robinson 1995b). Looking at the manipulation of this factor, the interpretations, however, may be problematic. In most studies, participants in the here-and-now condition had to tell a story while looking at a sequence of pictures. The complex counterpart of this task asked participants to tell the story in the past tense after the pictures were removed in order to induce the there-and-then setting. As in the complex version participants told the story by heart it may be that this complex setting put up a higher load on working memory than the simple here-and-now condition. Using this task manipulation therefore probably is confounded by the factor ± dual task because participants on the one hand need to ´ esz ´ 2007 for a comparable tell the story and on the other hand have to remember the story line (see Rev manipulation by means of the resource-dispersing factor ± dual task). Other empirical investigations focused on the factor ± few elements. In Gilabert (2007a) and Gilabert et al. (2009) participants acted on a map task. The complex task did not only increase the number of elements but also the way how difficult or easy these items were to distinguish from each other. 16 See Cadierno and Robinson (2009), van Daele (2007), Foster (2000, 2001), Foster and Skehan (1996, 1999), Foster and Tavakoli (2009), Gilabert (2005, 2007a, b), Gilabert, Baron, and Llanes (2009), Ishikawa (2007), Iwashihta, MacNamarra, and Elder (2001), Kim (2009), Kuiken, Mos, and Vedder (2005), Kuiken and Vedder (2004a, b, 2007a, b, c, 2008), Michel (2009), Niwa ´ esz ´ (2007, 2008, 2009, in press), Robinson (1995b, 2001b, 2003b, 2005, 2007a, (2000), Nuevo (2006), Rahimpour (1997), Rev b, 2010), Robinson et al. (2009), Skehan (2003, 2009), Skehan and Foster (1997, 1999, 2005, 2007), Tavakoli and Foster (2008), Tavakoli and Skehan (2005), Yuan and Ellis (2003).

1.6 Earlier empirical investigations and open issues

31

Similarly, Robinson (2001b) used, when testing university students, in a simple map task a well known route on the campus while the complex version of this task addressed an unknown and larger area of town. Kuiken et al. (2005) and Kuiken and Vedder (2007b) choose a different manipulation. They asked participants to take into account a different number of criteria when deciding between five options. As one can see, these studies did use different operationalizations of the factor ± few elements. Probably, this is one of the reasons why findings of these studies are mixed and do not give a conclusive picture. Kuiken et al. (2005) and Kuiken and Vedder (2007b) found partial support for the Cognition Hypothesis as cognitively complex tasks triggered a higher accuracy while linguistic complexity was influenced in opposite directions in different populations. In contrast, Robinson (2001b) attested the predicted outcome on lexical complexity but parallel increases on accuracy manifested itself as trend effects only. Finally, the studies presented by Gilabert (2007a) and Gilabert et al. (2009) found an increasing effect on accuracy but did not test for complexity. As a whole, these studies therefore may not serve as support for Robinson’s theory. As Skehan (2009) points out, support for the Cognition Hypothesis should always reveal an augmentation on linguistic complexity and accuracy measures in parallel. Otherwise data speak more for the existence of trade-off effects as proposed by his Limited Attentional Capacity Model (see section 1.3.4 and Skehan 1996, 1998, Skehan and Foster 2001). A puzzling finding of the work by Gilabert (and colleagues) was that it yielded different results for different task types. That is, a narrative task, that manipulated the factor ± here-and-now, revealed significant support for the Cognition Hypothesis on 8 out of 9 measures. However, the instruction giving and decision making tasks, that manipulated the number of elements, failed at giving substantial support, e.g., the decision making task showed significant effects on 2 out of 9 measures only (see chapter 3 for a more elaborate discussion of this study). Pallotti (2009), however, points out that one should be cautious in interpreting data as support if there are more measures that are not in line with the Cognition Hypothesis than that give support to Robinson’s predictions. Taken together, these studies manipulating the number of elements therefore may not allow conclusive interpretations with respect to the Cognition Hypothesis. These findings with respect to the factor ± few elements show one problem with the Cognition Hypothesis that was addressed in section 1.4.6. Robinson’s theory does not give clear instructions how to manipulate the different factors of the Triadic Componential Framework. Possibly, results point into different directions because researchers opted for different operationalizations. In addition, some of the factors may be intertwined and therefore easily get confounded with each other, as explained with the ± dual task / ± here-and-now above. With respect to the factor ± few elements Kuiken and Vedder (2007b) argue that an increase in the number of elements almost always induces an increase in the amount of reasoning demands, which would again result in the factors ± few elements and ± reasoning demands to be confounded. Possibly, this can explain the findings for different task types in Gilabert (2007a) and

32


Gilabert et al. (2009) (see these points discussed in section 1.4.6 of this chapter and chapter 4). Some other controversies with the empirical support for the Cognition Hypothesis are of a more methodological nature. For example, Robinson (2001b) used the token-type ratio as a measure of lexical complexity. This measure, however, may be unreliable because it is sensitive to sample length and complex tasks often yield longer speech samples than simple tasks (see section 1.5 and Bulte´ 2007, Vermeer 2000). As the work by Kuiken and colleagues investigates L2-writing tasks it possibly suffers from a methodological blemish concerning time on task (Kuiken et al. 2005, Kuiken and Vedder 2007a, b). In their investigations, participants did receive equal amount of time to work on a written text. Even so there was no instruction on how to use the overall time for the task. Most likely participants differed in the amount of time they spent on planning, writing, or reviewing. This suggests that there possibly is a confound of the resource-dispersing factor ± planning-time, that may yield trade-off effects between linguistic complexity and accuracy (see section 1.4.3 and Ellis 2005). A last point concerns the input material. As far as the information was presented in the publications discussed so far, one must conclude that the visual and written input material (e.g., pictures and instructions) was rather different over conditions, and sometimes addressed various topics. However, verbal instruction, visual stimuli and task topic to a large extent influence subsequent task performance (e.g., Schoonen 2005). In sum, these issues reveal that more empirical work is needed with respect to investigations of cognitive task complexity.

1.6.2

Investigating interaction

Concerning the interactive factor of task condition one–way / two–way flow of information, empirical work within the Cognition Hypothesis is rather limited. Most studies investigating this factor looked at effects of how information was distributed over participants. Accordingly, these studies evaluated the amount and type of interaction in dialogic settings (e.g., Doughty and Pica 1986, Gass and Varonis 1994, Gass et al. 1998, Shehadeh 2004). Other researchers investigated effects on interaction due to an increase on cognitive task complexity. While data by Nuevo (2006) contradict the claims of the Cognition Hypothesis (there were more learning ´ esz ´ opportunities in simple than in complex reasoning tasks), Rev (2008) and Gilabert et al. (2009) found supporting effects. In these studies complex tasks generated more interactional moves than simple tasks. Again, effects in the study of Gilabert et al. (2009) were not equally clear over task types (see above and chapter 4, Table 4.1). Robinson (2001b) and Kim (2009) did find supporting data. Participants displayed more interaction in complex than in simple interactive tasks. However, these studies used raw scores for measuring the amount of interaction such that results may be confounded by sample length.


33

Up to now, hardly any investigation looked at the difference between task performances manipulated on the factor ± monologic. In a meta-analysis Skehan and Foster (2007) compare their data on monologic tasks with data on dialogic tasks and find that dialogues push accuracy and complexity but decrease fluency. Even so, this meta-analysis compares L2-performances on various different tasks that manipulated all kinds of other task factors. Consequently, there is a lack of systematic comparisons of monologic versus dialogic performances on the same tasks. The studies presented in this book try to fill this gap. In the process an intriguing question may be answered. As discussed in section 1.4.4 Robinson’s theory gathers interactive factors under the term ’task condition’. From a cognitive point of view, however, Tavakoli and Foster (2008) suggest that ‘a monologic task [. . . ] makes greater demands on attentional resources than an interactive task’ (Tavakoli and Foster 2008: 461). In other words, it may be that language production in interaction is cognitively simpler than language processing in a monologic situation, since the absence of an interlocutor in a monologue possibly generates an increase in cognitive demands by itself. Also psycholinguistic research into interaction among native speakers points into this direction. Pickering and Garrod (2004) state that in interaction, interlocutors tend to mirror each others speech on all linguistic levels: syntax, semantics, phonology, and pragmatics. Their Alignment Hypothesis argues that this copying of the other’s speech ‘greatly simplifies production and comprehension in dialogue’ (Pickering and Garrod 2004: 169). The authors explain the simplification of the cognitive processes during speech production via alignment by referring to Levelt’s (1989) model of language production. First, in a dialogue the speaking partner’s turn creates planning time. While listening to the speaker the hearer can conceptualize his own speech act. As a result, the hearer has more cognitive capacity for formulating during his own speaking turn because less attention for online planning and conceptualization is needed. Furthermore, research of Fiksdal (2000) has shown that interactive tasks push fluency because interlocutors tend to ‘help out’ as soon as the partner falls silent in order to keep a constant flow of interaction. Therefore, interactive tasks show fewer moments of silence than monologic tasks. In contrast, in a monologue all processes of language production – including the effortful conceptualization of a message – need to be performed at the same time. As this is a complex cognitive task that requires attention, speakers may process information in a mostly serial way which is likely to generate hesitations and pauses.17 Second, in addition to this supposed extra planning time in interaction, another simplification in dialogues may occur due to priming. Priming refers to the easier availability of items (e.g., words or syntactic structures) that have been pre-activated through related items or structures (Friederici, Steinhauer, and Frisch 1999, Meyer and Schvaneveldt 1971). For example, the vocabulary item ‘knife’ 17 Pickering and Garrod (2004) argue that also the amount of self-repair is higher in dialogues than in monologues. In interaction, speakers by default are in a listener’s position such that their attention is focused on the spoken output of the speaking partner but also of themselves. As they are likely to notice more errors in their own output, the authors consider self-repairs as ’a natural byproduct of dialogues’.

34


pre-activates the vocabulary item ‘fork’ because these words are semantically related. Similarly, ‘knife’ pre-activates the word ‘wife’ because they share phonological features. Priming may occur on semantic, syntactic, phonological, and morphological levels. In a dialogue, vocabulary items and syntactic structures used by a speaker pre-activate these words and forms in the speaking partner. When turns change, the current speaker needs less activation in order to incorporate the actual and related items and structures into his own turn. As priming affects all levels of linguistic production, message generation in a dialogue presumably needs less cognitive resources than in a monologue. After all, in a monologue one has to conceptualize, formulate, and articulate a message from scratch on one’s own and speakers cannot rely on primes of a partner. A third effect of alignment is termed ‘routinization’, that is, interlocutors establish and agree on specific phrases or words they may use and keep using within a conversation (Pickering and Garrod 2004). Put differently, it is likely that interactants copy each others lexical and syntactic forms and use them again and again during the same dialogue. Costa, Pickering, and Sorace (2008) applied the Alignment Hypothesis to L2-interaction. They state that L2-learner’s may not be affected by alignment to the same extent at the same linguistic levels as L1-speakers. Also Ejzenberg (2000) points out that priming processes in second language learners may only be possible when their level of proficiency in the L2 is high enough. For example, vocabulary items that are not known in the L2 cannot be primed.18 Even so, also L2-learners may benefit from extra planning time and primed language use during interaction. In sum, this alternative perspective on the factor ± monologic suggests that dialogues are cognitively less effortful than monologues because interaction frees attentional capacity. Dialogic tasks can give extra planning time during the speaking partner’s turn, and processes of alignment and priming may ease language production. As a result, interactive, dialogic tasks as opposed to non-interactive, monologic tasks, may be seen as cognitively simple tasks.19 Applying this view to the Cognition Hypothesis (Robinson 2005), the factor ± monologic then possibly could be related to the resource-dispersing factor ± planning time rather than to the interactive factor of task condition one–way / two–way flow of information. This alternative perspective would predict that dialogues push accuracy and fluency of L2-speakers (due to cognitive ease on a resource-dispersing variable) while speech performances may be linguistically less complex because of recycling of words and clauses via routinization. To recap, the Cognition Hypothesis predicts with respect to interactive factors of task condition that accuracy increases while linguistic complexity and fluency decrease in interactive tasks (see section 1.4.4 and Table 1.2). Research evaluating these two perspectives would be welcome. In light of Robin18 Similarly,

it may be that primed items of the L1 need to be suppressed and thus priming uses rather than frees attentional capacity. 19 This statement may only hold from a cognitive point of view. Taking a pragmatic perspective on interaction, a dialogue may put up many unknown conversational demands, which can complexify speech production.


35

son’s (2005) Cognition Hypothesis also combined effects of increased cognitive task complexity and interaction may ask for more empirical work.

1.6.3

Investigating task specific measures

Only a few studies so far do explore data with respect to the use of task specific measures in relation to the claims of the Cognition Hypothesis because Robinson raised this issue only recently (Cadierno ´ esz ´ 2008, in press, Robinson et al. 2009, Robinson and Gilabert 2007). These and Robinson 2009, Rev investigations consistently report that while the global CAF-measures did not discriminate between the task manipulations, task specific measures did find supporting results for the Cognition Hypothesis (see chapter 5). New work in this direction may be able to yield more conclusive results with respect to the use of task specific measures.

1.6.4

Investigating native speaker task-based performance

Surprisingly few studies tested L1-speakers in order to have a baseline for the evaluation of L2-learner’s ¨ task performance. To the best of my knowledge only Dornyei and Kormos (1998), Foster (2001), Foster and Tavakoli (2009) included native speakers in their analyses. Especially in light of the Cognition Hypothesis, which bases its predictions on cognitive and attentional processes during task-based performance, it is important, however, to interpret L2-production against a native speaker control group. Levelt (1989) presents the native language production system as a modular construct consisting of a conceptualizer, a formulator, and an articulator. He assumes that formulating and articulating a message in the L1 relies on automatic, incremental, parallel, and mostly unconscious and therefore fast processes whereas the conceptualization and monitoring of a message need attentional control. Oral production in a second language differs in at least three points from native language production ¨ (a.o. de Bot 1992, Costa et al. 2008, Dornyei and Kormos 1998, Poulisse 1997, Poulisse and Bongaerts 1994). First, L2-speakers’ production is guided by attention at all levels of speech production, that is, also lexical retrieval, morphosyntactic formulation, and articulation are non-automatic and therefore rely on serial processing – especially at lower levels of L2-proficiency. Second, as L2-learners possibly cannot express the originally intended message with their incomplete L2-system they may need to revise their propositional messages in the conceptualizer. Third, L2-learners need attention to prevent automatic L1-activation that may interfere with L2-production (cf. Kroll and Sunderman 2003, MacWhinney 2001, Segalowitz 2003). Consequently, in contrast to task performance in the L1, task-based L2-performance is a mostly conscious and therefore slow and effortful cognitive activity (Schmidt 2001). Taking into account this difference between native and non-native language production possibly enables us to understand taskbased L2-performance in a better way. For example, we may be able to come to more reliable inter-

36


pretations of measures of linguistic complexity, accuracy, and fluency when evaluating L2-performance against a reference of L1-speakers (e.g., Ellis 2009, Foster and Tavakoli 2009, Pallotti 2009, Skehan 2009). Therefore, it is important to have a native speaker baseline in task-based cognitive research into L2-production.

1.7

Concluding remarks

This chapter has given the theoretical basis of this book with a focus on the Cognition Hypothesis (Robinson 1995b, 2001a, b, 2003b, 2005). In addition, it reviewed global and specific measures of task performance. Furthermore, it reviewed earlier empirical work investigating the Cognition Hypothesis. The conclusion from this theoretical overview may be that in order to test the claims of the Cognition Hypothesis with respect to cognitively complex interactive tasks there is a need for studies that systematically evaluate effects of cognitive task complexity and interaction, both on their own as well as in combination. In the process, research may examine L2-learner performances by means of global and task specific measures of task performance in light of L1-speaker baseline data. The research in this book aims at providing empirical data that address these open issues. The next chapter introduces the empirical work of this book, which is presented in the chapters 3, 4, and 5.

Chapter

2

The studies in this book

The aim of this chapter is to give an overview of the research presented in this book as a whole. First, it frames the goal of this book and summarizes the theoretical claims under investigation. Second, it formulates the research questions and hypotheses that guide the present research. Finally, the overall design, population, and measures of the experimental investigations in chapters 3, 4, and 5 are presented. At this point it is important to ascertain that the different investigations mirror the growth of this work as a whole, that is, study 1 (presented in chapter 3) yielded methodological insights that were used to improve the investigation of study 2 (presented in chapters 4 and 5). Chapter 5 itself is a follow-up of chapter 4 as it evaluates the same data set but elaborates the analysis by means of global CAFmeasures of task performance (which was reviewed in chapter 4) by examining a task specific measure in chapter 5. These growing insights account for the fact that the predictions, for example, from chapters 3 and 4 show some differences. They reflect the chronological stages of the work presented in this book. Furthermore, the reader may be reminded that the empirical chapters are intended as individual scientific papers that have been or will be published outside the context of this book. Therefore, they show some overlap in their content, especially when reviewing the theoretical framework or when explaining the design of the experimental investigations.

2.1

Goal

The goal of the studies presented in this book is to investigate the claims of the Cognition Hypothesis by Robinson (1995b, 2001a, 2001b, 2003b, 2005) with respect to the effects of cognitive task complexity and interaction. Furthermore, this book explores whether there are any combined effects of these two factors on the oral task-based performance of L2-learners. In addition, it evaluates the use of task specific measures as a complement to global CAF-measures and the added value of a native speaker baseline.

38


2.2

The claims under investigation

As discussed in section 1.4.3 the fundamental claim of the Cognition Hypothesis (Robinson 2005) is that so-called resource-directing factors of cognitive task complexity (e.g., ± few elements) are able to focus the L2-learner’s attention towards task relevant linguistic aspects such that task performance becomes linguistically more complex than upon simple tasks. As L2-performers can rely on multiple attentional pools (Wickens 2002, 2007) accuracy does not suffer from the increased cognitive complexity. Rather, cognitively complex tasks have the potential to also promote accuracy because L2-learners attention is focused to form. Consequently, cognitively complex tasks may foster a parallel increase of linguistic complexity and accuracy at the cost of fluency. As section 1.3.4 revealed, the Limited Attentional Capacity Model predicts a different outcome of task manipulations by means of cognitive factors of task complexity (Skehan 1996, Skehan and Foster 2001). Crucially, Skehan and Foster expect trade-off effects between dimensions of task performance, in particular between linguistic complexity and accuracy, upon cognitively complex tasks. As explained in section 1.4.4 the Cognition Hypothesis does not make overt predictions with regard to effects of the factor ± monologic. However, following the claims concerning the interactive factor of task condition one–way / two–way flow of information (Robinson 2001a, 2003b, 2005) and the literature on interaction (e.g., Long 1990) it was assumed that an interactive task reduces linguistic complexity and fluency but pushes the accuracy of L2-performance. This follows from the fact that in a dialogue mutual understanding is crucial so that both L2-participants may focus their attention towards form. This increased attention results in a higher accuracy of speech production at the cost of fluency in dialogues. As speaking partners ask each other for clarification and give feedback, speech production in interaction presumably is linguistically less complex than monologic task performances, because turn-taking behavior prevents speakers from building elaborate syntactic constructions. An alternative perspective on the factor ± monologic would predict a slightly different effect of dialogic tasks (see section 1.6.2 and Costa et al. 2008, Pickering and Garrod 2004, Tavakoli and Foster 2008). These accounts assume that dialogues are cognitively less demanding than monologues and therefore may expect interactive task settings to push accuracy and fluency while linguistic complexity may decrease due to routinization in alignment. The intriguing claim by Robinson (2001b, 2003b, 2005) concerns a combination of the cognitive and interactive factors of task design. As cognitively complex tasks may need more clarification work than simple tasks do, cognitively complex interactive tasks may yield more interaction than cognitively simple interactive tasks. Accordingly, the linguistic complexity in cognitively complex interactive tasks may be decreased. In contrast, both factors – increased cognitive task complexity and the interactive task setting – will push accuracy and reduce fluency. In sum, Robinson’s theory predicts cognitively complex interactive tasks to yield L2-speech of a low linguistic complexity and fluency but a high accuracy.

2.2 The claims under investigation

39

Table 2.1: Predicted effects of task complexity and interaction of the present studies

TASK COMPLEXITY

INTERACTION


MEASURE L2-LEARNERS

L1-SPEAKERS

complex

dialogue

complex dialogue

↑

⇓

↓

accuracy

↑

⇑

⇑

fluency

↓

⇑

↑

ling. complexity

≈

↓

≈

accuracy

≈

≈

≈

fluency

≈

↑

≈

ling. complexity

a

Note. ↑ = increase; ↓ = decrease; ⇑ = large increase; ⇓ = large decrease; ≈ = no effect; L2 = second language; L1 = first language; ling. complexity = linguistic complexity; a As there are no different predictions for structural and lexical complexity, the two sub-dimensions are combined into the single measure ‘linguistic complexity’ in this table.

Table 2.1 gives a graphical summary of the predictions of the present book that are based on a combination of the different claims listed above.1 As Robinson’s predictions concerning effects of cognitive task complexity are clearly stated the studies in this book follow the Cognition Hypothesis: Cognitively complex tasks manipulated on a resourcedirecting factor results in higher linguistic complexity and accuracy, at the cost of fluency (see Table 1.2 and Robinson 2001b, 2003b, 2005). With respect to the interactive factor ± monologic the present studies combine the different perspectives put forward: Dialogues are expected to yield a higher accuracy but a lower linguistic complexity than monologues as both the Cognition Hypothesis and e.g., the Alignment Hypothesis support this prediction. Concerning the contrasting accounts with respect to fluency, the empirical investigations in this book follow the alternative perspective, that is, fluency is expected to be pushed by a dialogic task setting. The rationale behind this choice is first, that as mentioned in section 1.4.4 the Cognition Hypothesis makes no overt statements about the factor ± monologic. Second, the predictions of the alternative accounts advocate a processing oriented perspective on L2-performance, which seems to be appropriate considering the performative nature of the measure of fluency. Concerning combined effects of cognitive task complexity and interaction this book assumes that manipulating cognitive task complexity (e.g., the factor ± few elements) generates a smaller effect than manipulating interaction (e.g., the factor ± monologic). The rationale follows from the intuition that the effect of interaction on task-based L2-performance is quite strong. That is, whether someone acts on his 1 Note, that this table accordingly expects a slightly different outcome than the Cognition Hypothesis, in particular with respect to effects of interaction as compared to Table 1.2 on page 25.

40


own or whether two speakers interact has a large impact on L2-performance. In contrast, the difference between cognitively simple and complex tasks on the resource-directing factor (± few elements) may be of a small nature only.2 When combining the predictions with respect to cognitive factors of task complexity and interactive factors of task condition the present book therefore expects that the pushing or reducing effects of interaction (marked by double arrows ⇑ or ⇓) may successfully mitigate against counter influences of cognitive task complexity (expressed by a single arrows ↑ or ↓).3 Consequently, cognitively complex interactive tasks are expected to yield speech productions of a lower linguistic complexity, but a higher accuracy and fluency than cognitively simple interactive tasks, or any task in a monologic setting. Last but not least, although the Cognition Hypothesis presents itself as a theory on task performance by L2-learners, the table shows predictions for native-speaker performance too. Chapters 4 and 5 include L1-speaker data as a baseline. For L1-speakers, no effects of cognitive task complexity are expected. As explained in section 1.6.4 native speakers rely on mostly automatic speech processes (Levelt 1989) and therefore may not be affected by a higher cognitive task complexity as manipulated in the present studies. Effects of interaction though are expected to similarly influence L1-speakers and L2-learners. After all, effects of turn-taking and alignment may hold for both populations, though the impact may be smaller in natives than in non-natives (Costa et al. 2008, Pickering and Garrod 2004). See chapters 4 and 5 for a more elaborate discussion of the predicted effects on native speakers’ task performance.

2.3

Research questions and hypotheses

The following research questions and hypotheses guide the empirical investigations of the present work: Research Question 1 What is the effect of increased cognitive task complexity on L2 oral task performance? Hypothesis 1 Increased cognitive task complexity results in higher accuracy and higher linguistic complexity, but lower fluency of L2 oral task performance. Research Question 2 What is the effect of interaction on L2 oral task performance?

2 Earlier

work of Gilabert (2005) gives support to this assumption as in this study the resource-directing factor ± here-and-now showed smaller effects on task performance than the resource-dispersing factor ± planning time. 3 N.B. this symbolism by single and double arrows does only assume ‘larger’ effects of interaction but not that it doubles the impact of cognitive task complexity.

2.4 Design

41

Hypothesis 2 Interactive tasks raise the accuracy and fluency of L2 oral task performance while linguistic complexity decreases.

Research Question 3 Are there any combined effects of increased cognitive task complexity and interaction on L2 oral task performance?

Hypothesis 3 Increased cognitive task complexity enhances interaction and therefore further decreases the linguistic complexity of cognitively complex interactive task performances. They increase the accuracy and fluency of L2 oral task performance.

L2 oral task performance is evaluated by means of global measures of linguistic complexity, accuracy, and fluency in chapters 3 and 4. In addition, chapter 5 evaluates the speech performances by means of a task specific measure (the use of conjunctions). Moreover, although the data collection focuses on L2-learners’ task performance, chapters 4 and 5 include a group of native speakers as a baseline. These two elaborations generate two more questions and hypotheses:

Research Question 4 What is the effect of increased cognitive task complexity on the frequency and occurrence of conjunctions in L2 oral task performance?

Hypothesis 4 Increased cognitive task complexity leads to an increase in the frequency and occurrence of conjunctions in L2 oral task performance.

Research Question 5 What are the effects of increased cognitive task complexity and interaction on their own as well as in combination on L1 oral task performance in contrast to L2 oral task performance?

Hypothesis 5 Increased cognitive task complexity does not affect the highly automatic oral task performance of L1-speakers. Accordingly, no combined effect of cognitive task complexity and interaction is expected. In contrast, interaction shows similar but smaller effects on L1-speakers than on L2learners.

42


Table 2.2: Experimental design of the present studies INTERACTION (between participants) monologue

dialogue

simple

CONDITION 1 + few elements + monologic

CONDITION 3 + few elements – monologic

complex

CONDITION 2 – few elements + monologic

CONDITION 4 – few elements – monologic

TASK COMPLEXITY (within participant)

2.4

Design

Based on the Triadic Componential Framework (Robinson 2005) the present studies investigate effects of an increased cognitive task complexity by means of the factor ± few elements. Furthermore, they explore effects of the manipulation of the interactive factor ± monologic. Both factors are systematically examined on their own as well as in combination. In a 2×2 design cognitive task complexity is implemented within participants and interaction between participants. This generates the four different conditions depicted in Table 2.2: (1) cognitively simple monologic (+ few elements / + monologic ), (2) cognitively complex monologic (– few elements / + monologic) , (3) cognitively simple dialogic (+ few elements / – monologic), and (4) cognitively complex dialogic (– few elements / – monologic). Task design and setting are operationalized such that the empirical investigations singled out these two factors. Great effort was taken to control or counterbalance all other factors named in Robinson’s Triadic Componential Framework (see Figure 1.1). For example, visual and linguistic input material of the task instructions are kept as similar as possible. Pre-task and online planning time, gender, and language background are controlled over participants and tasks. As the investigations use large groups of participants inequalities due to individual differences on learner factors may be overcome by group means.

2.5

Participants

The data collection took place from May to October 2006 (study 1) and from September 2007 to January 2008 (study 2). The experimental group of participants of the present investigations are adult learners of Dutch as a second language who had their first contact with Dutch after puberty. Table 2.3 summarizes the background information for all the participants that contributed to the studies presented in this book. In the first study there were 44 and in the second study 64 L2-learners respectively. They were all of Turkish or Moroccan background. This was thought to simplify the data collection of two homogeneous

2.5 Participants

43

Table 2.3: Background information for all participants of the present studies LANGUAGE

STAY IN THE

AGE

GENDER

PROFICIENCY

NETHERLANDSa

ORIGIN

N

Mean (SD)

m

f

Mean (SD)

Mean (SD)

Tur

Mor

44

27.7 (6.4)

17

27

21.5 (9.3)b

3.8 (4.5)

15

29

L2-LEARNERS

64

27.6 (6.2)

29

35

53.8 (17.2)c

3.8 (4.2)

31

33

L1-SPEAKERS

44

20.6 (3.5)

9

35

96.3 (3.2)c

study 1 L2-LEARNERS study 2

Note. SD = Standard Deviation; m = male; f = female; a in years; Tur = Turkish; Mor = Moroccan; b participants performed on a cloze task where every eleventh word was eliminated with a total of 50 gaps; c in eight short texts participants had to choose among three possible words at a gap with a total of 100 gaps

learner groups as these form the two largest immigrant populations in the Netherlands (i.e., about 4% of the Dutch national population is constituted by Turks and Moroccans, with the former group slightly outnumbering the latter, CBS 2010).4 The participants were recruited from different language institutes in the Netherlands, mostly in Amsterdam but partially also in Utrecht and the surrounding cities. According to a background information questionnaire they had resided in the Netherlands on average since three and a half years (see Appendices A.3, B.2). As they were attending or had just finished higher education most of them were in their twenties. All were attending classes for students with a higher educational background. At the moment of testing they were about to take or just had taken the State Examination for Dutch as a second language. Accordingly, they were classified by their teachers to be at an intermediate level of proficiency (i.e., level B1/B2 of the Common European Framework of Reference for Languages (CEF), Witte and Mulder 2006). In both investigations their estimated level was assessed by a written proficiency task. In study 1 this was a cloze task where every eleventh word was eliminated with a total of 50 gaps (see Appendix A.1 and A.2). Study 2 used a word choice task with a total of 100 gaps (see Appendix B.1).5 This revealed that participants of the first study were at lower intermediate levels while participants of the second study were at a slightly higher intermediate level. The second data collection included a control group of 44 native speakers who performed the same tasks under the same conditions as the non-native participants.6 These L1-data established a baseline for the empirical work presented in chapters 4 and 5. The native speakers were mostly students at the 4 N.B. None of the investigations explicitly explores differences between these two populations. After all, cognitive approaches to task-based L2-performance would not assume Turkish learners of Dutch to act differently on cognitively complex or simple monologic and dialogic tasks than Moroccan L2-learners of Dutch. 5 I thank the Language Center of the University of Groningen that uses this task as a placement test for their language courses. 6 I thank Rachel Jobels for her help especially with respect to collecting the native speaker data.

44


University of Amsterdam and on average a bit younger than the non-natives (as established by means of a background information sheet, see Appendix B.2). They scored at ceiling on the language proficiency task. In general, more females participated in the studies. Within the L1-group the different numbers of males and females is the largest.

2.6

Measures

The data analyses of the present work includes global and specific measures of performance. The first two investigations (chapters 3 and 4) use similar global measures of linguistic complexity, accuracy, and fluency. In order to avoid redundancy the CAF-constructs are carefully chosen based on the literature (cf. section 1.5). The first study (chapter 3) evaluates 12 global measures of linguistic complexity, accuracy, and fluency. Based on these results the choice of measures was revised as correlational analyses revealed co-linearity of some of the measures. The second study (chapter 4) accordingly investigates 10 global CAF-measures. In order to complement these findings chapter 5 investigates the use of a task specific measure in the dataset of study 2.

Basic unit As a basic syntactic unit of reference the studies presented in this book use the Analysis-of-Speech (AS)-unit by Foster et al. (2000) with some exceptions when measures are related to the number of words or time. For example, in the analysis by means of a task specific measure raw scores are corrected for sample length by calculating a ratio to 100 words rather than to the AS-unit. This is done because the specific measure (i.e., the use of conjunctions) may be related to any syntactic unit (see chapter 5 for a more elaborate explanation with respect to this point).

Linguistic complexity The empirical investigations on global CAF-measures (chapters 3 and 4) take the basic distinction between structural and lexical complexity as a measure of linguistic complexity. The first investigation gauges syntactic complexity by means of two measures: the total number of clauses per AS-unit and by the Subordination Index, i.e., the ratio of subordinate clauses per total number of clauses. After reviewing the results and consulting the literature again (Norris and Ortega 2003, 2009a), co-linearity, and therefore redundancy, of these two measures was acknowledged. The second investigation therefore used only the Subordination Index (in a slightly different form though: the number of subordinate clauses per AS-unit). This measure gives insight into the ability of participants to use complex syntax. It is thought to be most indicative for speakers at an intermediate

2.6 Measures

45

stage of L2-proficiency (Norris and Ortega 2009a). In addition, the number of words per clause is calculated as the mean length of clause serves as an indicator for complexity within a phrase (Ferrari 2009). Especially at higher levels of L2-proficiency and in native speech structural complexity may be most often expressed by phrase-internal complexification rather than by the use of more subordination or clauses (Norris and Ortega 2009a). Following earlier work Guiraud’s Index is used as a global measure of lexical complexity for the empirical investigations in chapters 3 and 4 (Gilabert 2005, Guiraud 1954, Robinson 2001b). Guiraud’s Index is calculated by dividing the number of different types by the square root of the number of tokens. Taking the square root of the type-token ratio (TTR) is thought to correct for sample length and therefore Guiraud’s Index may be more appropriate than the TTR (Vermeer 2000).7 The first investigation uses a second measure of lexical complexity: the percentage of lexical content words related to the total number of words. This ratio is calculated in the tradition of earlier work (Gilabert 2005, Rahimpour 1997, Robinson 1995b). However, Vermeer (2000) does not consider the percentage of lexical words to be any more valuable than Guiraud’s Index. As redundancy due to colinearity of measures needs to be avoided (Norris and Ortega 2009a), the second empirical investigation abandons the percentage of lexical words in favor of Guiraud’s Index as the latter is widely used in research investigating the Cognition Hypothesis.

Accuracy The studies presented in this book use accuracy measures based on specific types of errors. This procedure was adopted in order to tap even slight differences of task performance and because the target participant population is at an intermediate level of language proficiency such that almost no errorfree units are expected, while the native speakers of the second investigation may produce speech that consists for almost 100% of error-free units. The present work chose error categories following earlier research by van Daele (2007), Gilabert (2005), and Robinson (2001b). The first study gives the total number of errors per AS-unit as a broad starting point. Furthermore, the number of lexical errors and the number of omissions (of articles, verbs, and subjects) are calculated. The second dataset is analyzed for lexical, morphosyntactic, and determiner errors. Both counts included a category of ‘other errors’, which turned out to be almost empty and therefore was excluded from statistical analyses. For both investigations any word choice errors (e.g., forms in another language than Dutch, wrong prepositions) are classified as lexical errors. The second study redefined and partially split the category of ‘omissions’ of the first investigation into two categories: morphosyntactic and determiner errors. The former addresses non-target-like word order, omissions of obligatory constituents and other syntactic 7 As the bias of sample length may not be completely eliminated by this procedure some researchers still question its reliability (Bulte´ 2007) and favor the use of D (Malvern and Richards 2002) or Lambda (as proposed by Meara cf. Skehan 2009).

46


problems as well as morphological failures on agreement and inflection. The latter counts any erroneous use or omissions of articles as well as mismatches on grammatical gender on determiners as ‘nontarget-like use of articles’ (Gilabert 2005, Rahimpour 1997, Robinson 1995b). This change from the first to the second study is made because in the first study the category of ‘omissions’ to a large extent consisted of determiner errors. As the second study expects a similar kind of error pattern (participants are at broadly the same level and tasks are similar), determiners are singled out as an own measure such that it cannot ‘pollute’ the other categories. Both studies furthermore calculate two measures with respect to self-repairs: the ratio of self-repairs in relation to the number of errors as well as the percentage of self-repairs related to the total number of words. Following Gilabert (2007a), who cites Kormos (1999), these measures are considered to be good reflections of the speakers’ ability to monitor the own speech. After classifying these measures in the first investigation as accuracy measures, the second study arranges them under what Tavakoli and Skehan (2005) call ‘repair fluency’. The explanation for this is given in the following paragraphs discussing fluency.

Fluency In order to analyze the speech production for measures of fluency the investigations presented in this book rely mostly on work by van Daele (2007), Gilabert (2005), Mehnert (1998), and Yuan and Ellis (2003) and it uses the division into the three sub-constructs presented by Tavakoli and Skehan (2005). The first sub-dimension termed pausing or breakdown fluency concerns the amount, location, and duration of silence. It presumably reflects the planning and conceptualization phase of speech production. Both investigations calculate breakdown fluency by means of the number of filled pauses (e.g., uh, uhm) per hundred words. As in natural speech the number of silent and filled pauses correlates with ´ each other using both would possibly lead to redundancy (Kormos and Denes 2004). Practical reasons favored the counting of filled pauses rather than unfilled pauses. The data are coded for two different measures of the second sub-dimension, i.e., speed. Both give an estimation of ‘how fast and dense the produced language is in terms of the time units’ (Tavakoli and Skehan 2005: 255). The ‘unpruned speechrate A’ is the ratio of all syllables produced per second or minute and sheds a light on how well a speaker can fill time with sound. The second speed measure, ‘pruned speechrate B’ is cleaned from reformulations, repairs and repetitions before relating the syllables to time. It accordingly is a measure of ‘meaningful’ speech per time unit. The third sub-dimension of fluency is named repair fluency by Tavakoli and Skehan (2005). It refers to the amount of repetitions of exact words and phrases, reformulations, false starts, corrections and partial repeats. It gives insight into the speakers’ ability to monitor their own speech (Freed 2000, Gilabert 2005, 2007a, Gilabert et al. 2009, Kormos 1999, 2000a, b). Monitoring, and thus repair fluency,

2.7 Concluding remarks

47

are associated with the processes in the conceptualizer during speech planning (Levelt 1989). Both investigations in the present work code for the number of self-repairs of errors and non-errors. The first investigation, however, associates self-repairs with accuracy rather than fluency following Gilabert (2005). Considering Tavakoli and Skehan’s (2005) construct of repair fluency, the second investigation revised this classification and gathers self-repairs under the construct of fluency.

The task specific measure In addition to an analysis by means of global CAF-measures (see chapters 3 and 4), chapter 5 of this book presents data based on a task specific measure. This analysis is performed following a suggestion of Robinson and colleagues to complement global CAF-measures with task-specific ones (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). By definition, every investigation exploring task specific measures has its own specific measure. The present work examines the frequency and occurrence of conjunctions. The rationale behind it is that task-based production is elicited by means of argumentative tasks manipulated on the factor ± few elements.8 That is, both the cognitively simple and complex task ask for the balancing of reasons when choosing between several possible combinations among the elements. It is expected that the higher number of elements in the complex task may induces a higher number of arguments for or against an option. This increase in cognitive complexity is hypothesized to be reflected in linguistic structures and forms that are related to the balancing of reasons, e.g., conjunctions as lexical markers of argumentation.9 See chapter 5 for a more elaborate explanation and a detailed description of the specific conjunctions under investigation.

2.7

Concluding remarks

This chapter has introduced the empirical work by formulating the research questions and hypotheses that guide the present investigations. The subsequent chapters 3, 4, and 5 will elaborate on the actual empirical studies. On the one hand these chapters partially repeat the theoretical content because they serve as independent scientific papers. On the other hand, they further develop the theory based on their own data. In the process they therefore adapt some of the predictions concerning effects of cognitive task complexity and interaction presented so far. Finally, chapter 6 combines the theory (chapter 1) with the evidence (chapters 3 to 5) as it presents the general discussion and conclusion of the present book and relates them to the hypotheses formulated in this chapter.

8 N.B.

As discussed in section 1.4.6 Kuiken and Vedder (2007b) argue that an increase of cognitive task complexity on the factor ± few elements almost automatically increases the reasoning demands of a task too. 9 There may be many other (e.g., lexical, syntactic, pragmatic) ways to linguistically mark a line of argumentation. For practical reasons, the work presented in this book limits itself to the investigation of conjunctions.

Chapter

3

Study 1: The influence of complexity in monologic versus dialogic tasks in Dutch L2

Abstract This study puts the Cognition Hypothesis (Robinson 2005) to the test with respect to its predictions of the effects of changes in task complexity (± few elements) and task condition (± monologic) on L2-performance. 44 learners of Dutch performed both a simple and a complex oral task in either a monologic or a dialogic condition. The performance of the L2-learners was analyzed with regard to linguistic complexity, accuracy, and fluency. As predicted by the Cognition Hypothesis, the complex task generated more accurate though less fluent speech. Linguistic complexity, however, was only marginally affected. Dialogic tasks triggered more accurate and fluent output though it was structurally less complex. The interaction of task complexity and task condition showed effects on measures of accuracy only. In the monologic but not in the dialogic condition task complexity did promote accuracy. As a consequence, our results only partially support the Cognition Hypothesis.

3.1

The Cognition Hypothesis

In the last decade task-based language teaching (TBLT) has become an important field in second language acquisition (SLA) research. In this approach a central role is assigned to tasks in L2-learning. Tasks have been studied from different perspectives, among which a cognitive, information-theoretic This chapter in adapted form was published earlier as Michel, M.C., F. Kuiken, and I. Vedder (2007) The influence of complexity in monologic versus dialogic tasks in Dutch L2. International Review of Applied Linguistics 45(3), 241–259. Retrievable from the De Gruyter’s platform www.reference-global.com.

50


approach is advocated by Skehan and Foster (e.g., 2001) and Robinson (e.g., 2001b, 2005, 2007b). This view investigates how (cognitive) task factors influence the performance of L2-learners. Robinson (2005) assumes that some particular factors of task demands direct the learner’s attention towards language form. Attention is crucial in L2-learning because ‘SLA is largely driven by what learners pay attention to’ (Schmidt 2001: 3). Robinson proposes a Triadic Componential Framework, also known as the Cognition Hypothesis, that assigns a crucial role to factors of task complexity and factors of task condition, which can be manipulated systematically in task design with beneficial effects on L2-performance. According to the Cognition Hypothesis, increases in task complexity along the so-called resourcedirecting dimensions lead to both more accurate and more complex L2-performance. Interactive tasks are also thought to direct the learner’s attention to language and thus promote more accurate speech. Robinson rejects the idea of trade-off effects in the linguistic output due to limitations of attentional capacity, as proposed by Skehan and Foster (2001). For a discussion of the predictions of the Limited Attentional Capacity Model in contrast to the Cognition Hypothesis see Kuiken and Vedder (2007b). The aim of the present study is to investigate whether there is empirical evidence for Robinson’s claims. The study, which is based on the Cognition Hypothesis as presented in Robinson (2005), focuses on the effects of increased task complexity in a monologic versus a dialogic task condition. In the following sections (3.1.1 to 3.1.4) we will briefly present the basic assumptions of the Cognition Hypothesis that are relevant for the study (see Robinson (2007b) for a discussion of the Cognition Hypothesis and the Triadic Componential Framework).

3.1.1

Task complexity: cognitive factors

One of the key constructs of the Cognition Hypothesis is cognitive task complexity, which refers to the amount of cognitive processing that is needed to perform a task. According to Schmidt’s Noticing Hypothesis (Schmidt 1990, 2001) cognitive task demands are strongly related to what is noticed and noticing is ‘the first step in language building’ (Schmidt 1990: 31). Within the dimension of cognitive task complexity the Triadic Componential Framework makes a distinction between resource-directing and resource-dispersing factors. Robinson’s claim is that when tasks are cognitively more demanding along resource-directing factors (± here and now, ± few elements, ± reasoning demands) L2-learning is promoted, since these tasks trigger linguistically more complex structures and a more varied lexis. For example, having to take into account numerous elements induces lexically more diverse and structurally more complex output be´ (1985), who cause more elements have to be distinguished and compared. This view is based on Givon states that structural and functional complexity are associated with each other. In addition, ‘uptake and incorporation of forms is more likely to be evident on more complex tasks, since these more effectively


51

direct learner attention to the targeted input’ (Robinson 2001a: 304) and as such SLA is promoted. To sum up, the Cognition Hypothesis predicts that cognitively complex tasks trigger both greater accuracy and greater linguistic complexity. In contrast, fluency suffers from increased task complexity, since complex tasks are thought to require more explicit and conscious language processing, affecting procedural dimensions like fluency (Robinson 2005).

3.1.2

Task condition: interactive factors

Interactive tasks, where students act in pairs, give the opportunity for negotiation for meaning, clarification requests, and comprehension checks. As ‘negotiation brings learners’ attention to L2-versions of their interlanguage utterances’ (Pica 1994: 514), earlier work relates this heightened (shared) attention for language form to more noticing and uptake (cf. Doughty 2001, Gass 2003 and Pica 1994). Robinson also attributes a central role to interactive factors in his Triadic Componential Framework as ‘interaction is an important context and opportunity for activating processes thought to contribute to SLA’ (Robinson 2007a: 14). Robinson (2005, 2007b) distinguishes two types of interactive factors: participation factors, making interactional demands (one-way/two-way flow of information, open/closed solution, and convergent/divergent solution) and participant factors, which pose interactional demands (e.g., same/different gender). Being inherent to the task itself, interactional demands are relevant for task design. They influence L2-performance irrespective of the individual learner characteristics a participant brings to the task. These interactional demands are nonetheless closely related to and consequently restricted by the target task in real-life situations. For example, talking to a friend on the phone (dialogic) in order to make a date to go to the movies is a two-way flow task, which asks for an open and hopefully convergent solution. The Cognition Hypothesis does not make clear predictions with regard to the effects of interaction2 on particular aspects of L2-performance. The relation between interaction, heightened attention and noticing, as hypothesized by Doughty (2001), Gass (2003), Schmidt (1990, 2001), and Pica (1994), however, suggests that interaction favors accuracy, while fluency is expected to decrease. Linguistic complexity will be lower, because the clarification requests and repetitions of the interlocutor’s speech will induce shorter and structurally simpler sentences as well as lexically less varied output.

2 In the original paper the term ‘interactivity’ was used. This was changed into ‘interaction’ in order to use a consistent terminology throughout all chapters of this book.

52


3.1.3

Cognitively complex interactive tasks

Robinson states that interaction and task complexity will generate a combined effect on the linguistic complexity of L2-performance. ‘Cognitively complex oral interactive tasks simply lead to greater quantities of interaction and modified repetitions’ (Robinson 2005: 11) as the cognitive load posed by the complex task requires even more clarification requests and comprehension checks. As a consequence, the Cognition Hypothesis claims that complex interactive tasks affect linguistic complexity of L2-performance negatively because they trigger structurally and lexically less complex speech. The beneficial influence of increased task complexity on accuracy, however, is not thought to be affected differently in dialogic tasks. Similarly to monologic tasks fluency is expected to decrease. Table 3.1 summarizes the predictions of the Cognition Hypothesis (Robinson 2001a) with regard to the influence of task complexity and interaction, both separately as well as with regard to their combined effect. Note that the Cognition Hypothesis does not predict any differences between simple monologic and simple dialogic tasks. However, that does contradict our own expectations. As outlined in section 3.1.2, we predicted higher accuracy, but lower fluency and complexity in dialogic tasks.

3.1.4

Previous research within the Cognition Hypothesis

Positive effects of increased task complexity as proposed by Robinson have been found in several studies in which the factor ± here and now was manipulated (Gilabert 2007b, Iwashihta et al. 2001, Rahimpour 1997, Robinson 1995b). In a series of experiments concerning L2-writing, Kuiken et al. (2005) and Kuiken and Vedder (2007a, b) operationalized the factor ± few elements. Their data partially confirmed the Cognition Hypothesis, as increased task complexity resulted in more accurate writing. With respect to complexity they report a trend for higher lexical variation in the more complex task while no significant effect was found on syntactic complexity. In an oral interactive task (Robinson 2001b) manipulated the factor ± few elements. The complex task prompted significantly more lexically varied speech than the simple task but neither structural complexity nor accuracy revealed any significant effects. Fluency decreased in the complex version. One Table 3.1: Predicted effects of task condition and task complexity (based on Robinson 2001a) TASK CONDITION

TASK COMPLEXITY

monologue

dialogue

ACCURACY

COMPLEXITY

FLUENCY

simple

–

–

+

complex

+

+

–

simple

–

–

+

complex

+

–

–

Note. + = increase; – = decrease

3.2 Research questions, method, and design

53

effect of task complexity on interaction was that participants needed significantly more comprehension checks and displayed a trend for more clarification requests in the complex task. Nuevo (2006) found contradictory evidence for the Cognition Hypothesis with respect to the influence of interaction. Her study analysed the amount of learning opportunities induced by the task in interactive simple versus complex tasks manipulated on the factor ± reasoning demands. Results reveal that the simple task led to more interaction and clarification requests while increased task complexity did not affect accuracy in a post test. To sum up, the empirical research up to now has not given a conclusive picture with respect to the claims of the Cognition Hypothesis. Especially the predicted effects of ± monologic task conditions in combination with increased task complexity have not yet been tested systematically.

3.2

Research questions, method, and design

In the present study we put the Cognition Hypothesis (Robinson 1995b, 2005, 2007b) to the test. Our aim is to collect data on oral tasks that manipulate two factors of the Triadic Componential Framework: cognitive task complexity, with respect to the factor ± few elements and task condition, concerning the factor ± monologic. Our research questions are: RQ1 What are the effects of increased cognitive task complexity, manipulated along the factor ± few elements on the oral performance of second language learners? RQ2 What are the effects of changes in task condition with respect to the factor ± monologic on the oral performance of second language learners? RQ3 Are there any combined effects of cognitive task complexity and a ± monologic task condition on the oral performance of second language learners? To analyze the L2-performance of the participants, measures of accuracy, complexity, and fluency were used. Our hypotheses H1 and H3 with respect to research questions (1) and (3) are based on the Cognition Hypothesis. The second hypothesis (H2), concerning research question 2, is based on our predictions pointed out in section 3.1.2. Our hypotheses are as follows: H1 Increased cognitive task complexity along the resource-directing factor ± few elements will have a beneficial effect on the performance of L2-learners in that their speech will be more accurate and linguistically more complex. Fluency will suffer from increased task complexity. H2 Changes in task condition along the factor ± monologic will affect the performance of L2-learners in so far that in dialogic tasks, the oral output will be syntactically and lexically less complex than in monologic tasks. Accuracy will be promoted in the interactive task condition but fluency will suffer.

54


H3 Combined effects of increased cognitive task complexity (± few elements) and changes in task condition (± monologic) will influence the performance of L2-learners in so far that in dialogic tasks linguistic complexity will be even more reduced by increased task complexity than in monologic tasks. However, similarly to complex monologic tasks, complex dialogic tasks will push learners to greater accuracy while fluency will decrease.

3.2.1

Method

Participants The participants of the study were 44 L2-learners of Dutch, 29 Moroccan and 15 Turkish, who had had their first contact with Dutch after puberty (mean age 27.7 years, SD 6.4). Moroccans and Turks form the two largest groups of immigrants in the Netherlands; 9% of the population of Amsterdam is constituted by Moroccans and 5% by Turks. The 27 females and 17 males were selected from four different language institutes in Amsterdam where they attended classes for students with a higher educational background. As they were about to take or just had taken the State Examination for Dutch as a second language, the participants were classified to be at an intermediate level of proficiency, i.e., level B1/B2 of the Common European Framework of Reference for Languages (cf. Witte and Mulder 2006). A cloze task where every eleventh word was eliminated revealed a mean score of 21 out of 50 (SD 9.3). See Appendix A.1 and A.2.

Procedure and tasks In a 2×2 design with ± few elements as a within-subject factor and ± monologic as a between-subject factor 22 participants performed in a monologic and 22 in a dialogic condition. They all did a simple (+ few elements) and a complex (– few elements) task. Table 3.2 presents a schematic overview of the task factors manipulated in the study at hand. Participants received a full-colour leaflet with two electronic devices (MP3 players or mobile phones) in the simple task and with six devices in the complex version. Table 3.2: Manipulated factors INTERACTION (between participants)

TASK COMPLEXITY (within participant)

monologue

dialogue

simple

+ few elements + monologic

+ few elements – monologic

complex

– few elements + monologic

– few elements – monologic

3.2 Research questions, method, and design

55

The gadgets differed from each other in seven relevant features (e.g., price, colour, capacity). See Appendix A.4. Two versions of the same leaflet were created concerning either MP3 players or mobile phones. In the two versions, the features of the mobile phones or MP3 players were either identical or replaced by analogous information. The order of presentation of the different versions was counterbalanced over participants. In the monologic setting, participants were told to leave a message on the answering machine of a friend who had asked for advice about the MP3 player or mobile phone he or she should buy. In the dialogic setting, participants discussed with each other on the phone about the type of MP3 player or telephone they would buy.

3.2.2

Production measures: accuracy, complexity, and fluency

Speech samples were transcribed using CLAN (MacWhinney 2000). The output was coded for measures of production in terms of accuracy, linguistic complexity, and fluency. The Analysis-of-Speech unit (AS-unit) by Foster et al. (2000) was chosen as the basic syntactic unit of analysis. The five measures of accuracy, four of complexity, and three of fluency respectively are listed in Table 3.3. With respect to accuracy, we employed one general performance measure, i.e., the total number of errors per AS-unit, and two specific measures, i.e., the number of lexical errors as well as the total number of omissions (of articles, verbs, and subjects), both in relation to the number of AS-units. Furthermore, two measures with respect to self-repairs were included: the ratio of self-repairs in reTable 3.3: Measures of accuracy, complexity, and fluency ACCURACY

COMPLEXITY

FLUENCY

total number of errors

total number of clauses

unpruned speechrate A

per AS-unit

per AS-unit

number of syllables/second

number of lexical errors

Subordination Index

pruned speechrate B

per AS-unit

number of syllables/second

number of omissions

percentage of

number of filled pauses

per AS-unit

lexical words

per 100 words

percentage of

Guiraud’s Index

self-repairs ratio of self-repairs to errors Note. AS = Analysis-of-Speech unit; % = percentage; Subordination Index = number of subor√ dinate clauses per total number of clauses; Guiraud’s Index = types/ tokens; speechrate A = syllables per minute in unpruned speech; speechrate B = syllables per minute in pruned speech

56


lation to the number of errors as well as the percentage of self-repairs related to the total number of words. These repair measures were chosen because repair behaviour is thought to reflect the speaker’s self-monitoring and therefore is an indication of learners’ attention to form (cf. Gilabert 2007a). Structural complexity was measured by means of the total number of clauses per AS-unit and by the Subordination Index: the ratio of subordinate clauses per total number of clauses. Lexical complexity was measured by Guiraud’s Index (Guiraud 1954) and the percentage of lexical words in relation to the total number of words. Guiraud’s Index, which is calculated by dividing the number of types by the square root of the number of tokens, is thought to be more appropriate than the type-token ratio (TTR), as it takes sample length into account (Vermeer 2000). The second measure, i.e., the percentage of lexical words, was used following earlier studies (Gilabert 2005, Rahimpour 1997, Robinson 1995b). Two measures of fluency were chosen on the basis of Mehnert (1998) and Yuan and Ellis (2003): speechrate A, i.e., the ratio of syllables per minute in unpruned speech (including reformulations, repetitions, and replacements), and speechrate B, the ratio of syllables per minute in pruned speech (without reformulations, repetitions, and replacements). Furthermore, the number of filled pauses (e.g., uhm) per hundred words was calculated as a measure of breakdown fluency (Skehan and Foster 2005).

3.3

Results

The statistical analysis consisted of a multivariate repeated measures analysis of variance (MANOVA) with task complexity (± complex) as within-subject factor and task condition (± monologic) as a betweensubjects factor. Three different MANOVAs were conducted on the five measures of accuracy, the four measures of complexity, and the three measures of fluency respectively. Table 3.4 gives the descriptive statistics of the means and standard deviations on the measures of accuracy, complexity, and fluency. The means of the total scores on the simple versus complex task indicate the direction of possible effects on L2-performance. The table shows that complex tasks generally yielded a higher accuracy, as measured by the number of errors, omissions and the ratio of repairs to errors, while the percentage of repairs went in the opposite direction in simple tasks, i.e., the percentage of repairs is lower. Structural complexity decreased, but lexical complexity increased in complex tasks. Fluency is higher in simple tasks with respect to both speechrates. Comparison of the monologic and dialogic condition suggests that dialogues yielded more accurate speech with regard to the number of errors, omissions, and the ratio of repairs to errors. The percentage of repairs, however, was higher in monologues. Monologues also produced a higher structural complexity. Lexical complexity increased in dialogues, but only with respect to Guiraud’s Index. Dialogic tasks yielded more fluent speech. In the following sections the results are presented of the three MANOVAs on measures of accuracy, complexity, and fluency in order to detect the statistical power of the observed differences.

3.3 Results

57

Table 3.4: Descriptives of all measures: Mean (SD) TASK COMPLEXITY simple

complex

monologue

dialogue

total

monologue

dialogue

total

2.03

0.83

1.43

1.50

0.85

1.17

(0.83)

(0.35)

(0.87)

(0.62)

(0.38)

(0.60)

0.49

0.15

0.32

0.31

0.17

0.24

(0.35)

(0.14)

(0.31)

(0.25)

(0.11)

(0.20)

0.64

0.25

0.45

0.47

0.27

0.37

(0.42)

(0.17)

(0.37)

(0.35)

(0.15)

(0.28)

3.62

3.74

3.18

3.76

3.93

3.34

(2.90)

(1.89)

(2.46)

(3.56)

(1.36)

(2.69)

17.14

31.01

24.08

21.20

24.11

22.65

(15.24)

(36.93)

(28.79)

(22.21)

(15.24)

(18.88)

1.52

1.26

1.39

1.42

1.27

1.34

(0.25)

(0.14)

(0.24)

(0.17)

(0.13)

(0.17)

0.14

0.08

0.11

0.11

0.07

0.09

(0.12)

(0.05)

(0.10)

(0.05)

(0.06)

(0.06)

54.51

52.61

53.56

55.12

56.09

55.61

(6.42)

(6.59)

(6.50)

(7.35)

(5.96)

(6.63)

5.88

6.19

6.04

6.12

6.27

6.19

(0.59)

(0.90)

(0.77)

(0.50)

(0.95)

(0.75)

134.96

170.60

152.78

132.20

158.07

145.14

(32.02)

(34.30)

(37.42)

(32.40)

(31.28)

(34.09)

103.93

139.78

121.86

104.38

131.51

117.95

(28.18)

(30.92)

(34.40)

(31.20)

(28.53)

(32.57)

24.30

16.38

20.34

25.34

15.82

20.58

(10.36)

(6.50)

(9.44)

(10.49)

(7.73)

(10.30)

ACCURACY errors per AS lexical errors per AS omissions per AS % repairs repairs per errors COMPLEXITY clauses per AS Subordination Index % lexical words Guiraud’s Index FLUENCY speechrate A (unpruned) speechrate B (pruned) filled pauses per 100 words

Note. SD = Standard Deviation; AS = Analysis-of-Speech unit; % = percentage; Subordination Index √ = number of subordinate clauses per total number of clauses; Guiraud’s Index = types/ tokens; speechrate A = syllables per minute in unpruned speech; speechrate B = syllables per minute in pruned speech

58


Table 3.5: Results of the repeated measures MANOVA on accuracy OVERALL EFFECTS ON MEASURES OF ACCURACY

df

F

p

TASK COMPLEXITY

5

5.78

0.001***

TASK CONDITION

5

7.62

0.001***

TASK COMPLEXITY X TASK CONDITION

5

2.28

0.07

errors per AS

1

6.63

0.01**

lexical errors per AS

1

3.25

0.08

omissions per AS

1

2.46

0.12

percentage of repairs

1

0.09

0.77

ratio of repairs to errors

1

0.08

0.76

errors per AS

1

41.72

0.001***


1

20.86

0.001***

omissions per AS

1

14.95

0.001***


1

2.53

0.12


1

2.49

0.12

errors per AS

1

7.63

0.01**


1

5.26

0.03*

omissions per AS

1

3.99

0.05*


1

0.00

0.97


1

1.23

0.27

EFFECTS ON DIFFERENT MEASURES OF ACCURACY TASK COMPLEXITY

TASK CONDITION

TASK COMPLEXITY X TASK CONDITION

Note. AS = Analysis-of-Speech unit; df = degrees of freedom; F = F-value of MANOVA; * = p < 0.05; ** = p < 0.01; *** = p < 0.001

3.3.1

Accuracy

Table 3.5 lists the statistics on the different measures of accuracy by means of the repeated measures MANOVA. With respect to accuracy there are significant main effects of task complexity (F(5,38) = 5.78, p < 0.001) and of task condition (F(5,38) = 7.62, p < 0.001), while the combined effect did not reach significance (task complexity X task condition: F(5,38) = 2.28, p = 0.07). Participants were significantly more accurate on complex tasks. Similarly dialogues yielded more accurate speech. Concerning the separate measures, a significant effect of increased task complexity is reflected only in the total number of errors per AS-unit (F(1,42) = 6.63, p < 0.01), while neither lexical errors or omissions, nor the repair behavior yielded significant results. The main effect of task condition can be

3.3 Results

59

detected in a robust effect on error countings. The interactive task generated significantly more accurate speech with regard to the total number of errors (F(1,42) = 41.72, p < 0.001), lexical errors (F(1,42) = 20.86, p < 0.001), and omissions (F(1,42) = 14.95, p < 0.001) per AS-unit. Again, repair behavior was not significantly affected. An interaction effect of task complexity and task condition was found with respect to three measures of accuracy. Increased task complexity did produce more accurate speech in the monologic task condition only concerning the total number of errors (F(1,42) = 7.63, p < 0.01), lexical errors (F(1,42) = 5.26, p < 0.05), and omissions per AS-unit (F(1,42) = 3.99, p = 0.05). Once again, the repair behavior of the participants was not significantly affected.

3.3.2

Complexity

In Table 3.6, the results of the repeated measures MANOVA on the different measures of complexity are listed. In the general analysis, linguistic complexity was not affected significantly by increased task complexity. Task condition however, did generate a significant main effect (F(4,39) = 10.12, p < 0.001). No combined effects of task complexity and task condition were found. Of the different measures of linguistic complexity, the percentage of lexical words was significantly affected by task complexity (F(1,42) = 4.47, p 0.33). For L1-speakers the number of lexical errors and determiner errors were significantly though moderately affected by the interaction factor (lexical: partial η2 = 0.256; determiners: partial η2 = 0.092). Both populations made fewer errors in dialogues than in monologues.

4.3.3

Effects on fluency

Table 4.10 gives the descriptives for the four fluency measures. From these raw numbers, comparing across the columns, no consistent difference between simple and complex task performances is visible in either group. Comparing monologue with dialogue, however, the figures suggest that L2-learners as well as L1-speakers are more fluent in dialogues than in monologues. This is reflected in both speed measures (higher numbers, i.e. faster in dialogues than in monologues) as well as in the pausing and repair behavior (lower numbers, i.e. fewer pauses and repairs in dialogues than in monologues). Table 4.11 summarizes the statistical outcome of the mixed GLM ANOVA on the fluency data. The p-values of this analysis indicate that there was no main effect for cognitive task complexity on any

84

Study 2a: Effects of task complexity and interaction on L2-performance

Table 4.10: Descriptives of the measures of fluency FLUENCY

L2-LEARNERS

L1-SPEAKERS

DESCRIPTIVES

TASK COMPLEXITY simple

Unpruned speechrate A

Pruned speechrate B

Filled pauses per AS-unit

Repairs per AS-unit

TASK COMPLEXITY

complex

simple

complex

INTERACTION

Mean

(SD)

Mean

(SD)

INTERACTION

Mean

(SD)

Mean

(SD)

mono (N=32)

2.22

(0.51)

2.10

(0.50)

mono (N=20)

3.41

(0.62)

3.43

(0.62)

dia (N=32)

2.43

(0.66)

2.59

(0.62)

dia (N=24)

3.87

(0.50)

3.62

(0.52)

total (N=64)

2.33

(0.59)

2.34

(0.61)

total (N=44)

3.66

(0.60)

3.53

(0.57)

mono (N=32)

1.75

(0.46)

1.65

(0.45)

mono (N=20)

3.14

(0.62)

3.17

(0.63)

dia (N=32)

2.01

(0.61)

2.15

(0.60)

dia (N=24)

3.58

(0.46)

3.34

(0.49)

total (N=64)

1.88

(0.55)

1.90

(0.58)

total (N=44)

3.38

(0.58)

3.26

(0.56)

mono (N=32)

1.63

(0.81)

1.58

(0.90)

mono (N=20)

0.78

(0.43)

0.85

(0.44)

dia (N=32)

0.89

(0.61)

0.78

(0.55)

dia (N=24)

0.28

(0.18)

0.30

(0.16)

total (N=64)

1.26

(0.80)

1.18

(0.84)

total (N=44)

0.51

(0.40)

0.55

(0.42)

mono (N=32)

0.28

(0.11)

0.31

(0.15)

mono (N=20)

0.12

(0.09)

0.11

(0.06)

dia (N=32)

0.18

(0.09)

0.17

(0.12)

dia (N=24)

0.11

(0.09)

0.11

(0.07)

total (N=64)

0.23

(0.11)

0.24

(0.15)

total (N=44)

0.11

(0.09)

0.11

(0.07)

Note. L2 = second language; L1 = first language; SD = standard deviation; mono = monologue; dia = dialogue; speechrate A = syllables per second in all speech; speechrate B = syllables per second in cleaned speech; AS-unit = Analysis-of-Speech unit

measure of fluency – neither in L2-learners nor in L1-speakers. However, once again, interaction did influence both groups. L2-learners were affected significantly on all measures, showing larger effect sizes on repair and pausing behavior (partial η2 > 0.25) than on the two speechrates (partial η2 > 0.11). Native speakers’ fluency was significantly affected by interaction with respect to the unpruned speechrate A (partial η2 = 0.092) along with a significant and large effect on pausing behavior (partial η2 = 0.482). Both, pruned speechrate B and repair fluency, were not affected significantly, but SRB showed a trend. The only combined effect for cognitive task complexity by interaction that turned out to be significant in the dataset appeared in the fluency measures of native speakers. Planned post-hoc comparisons evaluating this large effect on both speechrates showed that L1-speakers were faster in simple rather than complex dialogues while no effect for cognitive task complexity was visible in monologues (SRA: monologue simple–complex = -0.022, n.s.; dialogue simple–complex = 0.254, p < 0.01; SRB: monologue simple–complex = -0.027, n.s.; dialogue simple–complex = 0.238, p < 0.01). Non-native speakers showed a different trend on their speechrates: Increased cognitive task complexity tended to yield faster speech in dialogues, but slower speech in monologues. This comparison, however, was not statistically significant.

4.4 Discussion

85

Table 4.11: Statistics of the measures of fluency FLUENCY

L2-LEARNERS

STATISTICS

MEASURE

TASK COMPLEXITY

INTERACTION


L1-SPEAKERS

F

df, Error

SSQ

partial η2

p

MEASURE

F

df, Error

SSQ

partial η2

p

0.081

SRA

0.05

1,62

0.01

0.001

0.826

SRA

3.20

1,42

0.29

0.071

SRB

0.08

1,62

0.01

0.001

0.775

SRB

2.87

1,42

0.24

0.064

0.098

Pauses

1.14

1,62

0.18

0.018

0.290

Pauses

0.78

1,42

0.04

0.018

0.381

Repairs

0.22

1,62

0.00

0.003

0.644

Repairs

0.14

1,42

0.00

0.003

0.715

SRA

8.07

1,62

3.94

0.115

0.006**

SRA

4.24

1,42

2.28

0.092

0.046*

SRB

10.28

1,62

4.54

0.142

0.002**

SRB

3.90

1,42

2.03

0.085

0.055

Pauses

20.98

1,62

19.09

0.253

0.000***

Pauses

39.03

1,42

6.00

0.482

0.000***

Repairs

26.96

1,62

0.48

0.303

0.000***

Repairs

0.43

1,42

0.00

0.010

0.518

SRA

3.68

1,62

0.62

0.056

0.060

SRA

4.56

1,42

0.42

0.098

0.039*

SRB

3.50

1,62

0.46

0.053

0.066

SRB

4.50

1,42

0.38

0.097

0.040*

Pauses

0.15

1,62

0.03

0.002

0.696

Pauses

0.25

1,42

0.01

0.006

0.618

Repairs

0.68

1,62

0.01

0.011

0.412

Repairs

0.12

1,42

0.00

0.003

0.729

Note. L2 = second language; L1 = first language; F-value; df, Error = degrees of freedom and Error df; SSQ = Sum of Squares; partial η2 = effect size; SRA = Unpruned speechrate A in syllables per second; SRB = Pruned speechrate B in syllables per second; Pauses = Filled pauses per Analysis-of-Speech unit; Repairs = Repairs per Analysis-of-Speech unit; * = significant at p < 0.05; ** = significant at p < 0.01; *** = significant at p < 0.001

4.4

Discussion

This section discusses the results with respect to the Cognition Hypothesis concerning cognitive task complexity and interaction. Section 4.5 summarizes the findings by giving suggestions for future research and formulating the practical implications of the work at hand. In order to summarize the results of all measures, Table 4.2, which gave the predicted effects, is repeated here as Table 4.12 but based on the actual outcome of the study. Since the structural and lexical measures of linguistic complexity sometimes did reveal a different pattern, they are listed separately.

4.4.1

Effects of increased cognitive task complexity?

The Cognition Hypothesis, as formulated in the first and third hypothesis of the present study, predicts that an increase of cognitive task complexity along resource-directing factors results in higher accuracy, linguistic complexity, but lower fluency during L2-task performance. In interactive tasks, increased cognitive task complexity promotes interaction such that the syntactic and lexical complexity are decreased in complex interactive tasks while accuracy and fluency are enhanced. In the present study, cognitive task complexity manipulated on the factor ± few elements did only affect one measure of L2-performance. Higher scores on Guiraud’s Index were found in complex tasks when compared to simple ones. Native speakers, in addition, showed a combined effect for cognitive task complexity with interaction on fluency. In dialogues, but not in monologues, increased cognitive

86


Table 4.12: Manifested effects of cognitive task complexity and interaction

TASK COMPLEXITY

INTERACTION


MEASURE L2-LEARNERS

str. complexity

complex ≈

dialogue √ ⇓

complex dialogue ≈

√ lex. complexity

↑

↑

≈ √

accuracy

≈

↑

≈ √

fluency

≈

⇑ √

L1-SPEAKERS

≈ √

√

str. complexity

≈

⇓

≈

lex. complexity

↑

↓

≈

accuracy

≈

↑

≈

fluency

≈

√ ↑ ( )

√ ≈ ( )

√

√

√ √

√

Note. L1 = first language; L2 = second language; str. complexity = structural complexity; lex. complexity = lexical complexity; ↑ = increase; ↓ = decrease; ⇑ = large increase; ⇓ = large decrease; √ √ ≈ = no effect; = results reveal expected effect; ( ) = results partially reveal expected effect

task complexity slowed them down (as measured on the two speechrates). No other measures were affected by an increase in cognitive task complexity or by the combination of cognitive task complexity and interaction. The findings with respect to lexical complexity are in line with Robinson’s (2005) predictions, but no parallel increase in accuracy was found. A central claim of the Cognition Hypothesis, however is, that accuracy and complexity both are promoted by increased cognitive task complexity. Therefore, the present study gives little support for Robinson’s claims. Since there was neither an increase nor decrease in any accuracy measure, the data do not suggest the existence of trade-off effects between accuracy and linguistic complexity. Consequently, the present study does not support the Limited Attentional Capacity Model by Skehan (1998) either. This study is not in line with earlier work that manipulated the factor ± few elements. For example, Michel et al. (2007, chapter 3) and Robinson (2001b) did find significant increases of accuracy and lexical complexity in complex oral L2-performance, as did Kuiken and Vedder (2007b) for written L2tasks. Likewise, Gilabert (2007a) found an increase in repair behavior in complex narrative tasks. The question arises then, why the present study found such minor effects of cognitive task complexity on task performance. The subsequent paragraphs discuss four possible explanations that address (1) the difference in cognitive load between the simple and complex task, (2) different effects of cognitive task complexity in different task types, (3) general CAF-measures versus task specific measures, and (4)

4.4 Discussion

87

qualitative versus quantitative changes in linguistic performance. First, an explanation could be that perhaps the difference between the simple and the complex tasks was not large enough. The present study did single out the resource-directing factor ± few elements and it was manipulated based on findings from cognitive psychology. Halford et al. state that our working memory and reasoning limitations share a central capacity that ‘is limited to relations between four variables’ ((Halford et al. 2007: 240)). Accordingly, the simple task of the present study giving four elements/combinations should be within the human capacity limits whereas the complex task with six elements/nine combinations should be beyond it. On these theoretical grounds, the manipulation should put up a difference in cognitive load as proposed by Robinson’s Triadic Componential Framework and therefore critique addressing this point may be debatable. The work by Gilabert and colleagues (Gilabert 2007a, Gilabert et al. 2009) points towards a second possible explanation. In their studies, increased cognitive task complexity did affect task performance in an instruction giving task, but not in a decision making task. One may assume then that the factor ± few elements was overruled by the factor ± reasoning. Also the latter is a factor of cognitive task complexity itself (cf. the Triadic Componential Framework, Robinson 2005). For other task types, e.g., narrating a picture story (Robinson 2001b), giving instructions in a map task (Gilabert 2007a), or evaluating inanimate items (as e.g., the mobile phone tasks in Michel et al. 2007), increasing the cognitive task complexity on the factor ± few elements may differentiate L2-task performance. In a complex reasoning task, however, the manipulation of the factor by varying the number of elements may not substantially affect the attentional allocation. As a consequence, no difference in task performance would be visible. Results of the questionnaire on affective variables of the present study corroborate this explanation. Participants’ perceptions of task difficulty did not differ significantly between the simple and complex tasks.5 The possibility then of differential effects of cognitive task complexity across different task types may then provide the explanation. Robinson (2007b) mentions a third possibility. It may be that the global measures of linguistic complexity, accuracy, and fluency used in the present study are not sensitive enough to discern the differences in performance caused by increased cognitive task complexity. Although these measures have been used reliably in task-based research (e.g., Michel et al. 2007/chapter 3, Robinson 2001b, Skehan and Foster 2001) recent work corroborates the added value of using task specific measures. For ´ esz ´ (2008) evaluated L2-task performance by means example Cadierno and Robinson (2009) and Rev of reference to psychological and cognitive state terms, number of wh-clauses, or clausal conjoinings. Task complexity effects were more prominent on these task specific than on general measures. A more qualitative look at the data at hand therefore will focus on the use of (causal) connectives as a more specific measure for the argumentative tasks used in the present study (chapter 5). In the process, it 5 Mean (standard deviations) on a 5 point Likert scale (1 = difficult, 5 = easy) of L2-learners: monologues simple = 3.6 (0.6) and complex = 3.5 (0.7), dialogues simple 3.9 (0.6) and complex 3.9 (0.5) and L1-speakers: monologues simple = 3.1 (0.6) and complex = 3.0 (0.6), dialogues simple = 3.7 (0.7) and complex = 3.8 (0.5).

88


will evaluate this third possible explanation. All the same, data of the present study on general CAF-measures do not support the idea that increased cognitive task complexity manipulated on the factor ± few elements leads to more focused attention to language form by L2-learners. Hence, the fact that lexical complexity was enhanced in complex tasks can be accounted for by a rather simple idea. The complex tasks in this study consisted of more elements (= persons) that were each associated with more words (the six characteristics per person), cf. Appendix B.4. For example, the input in the simple task showed four different numbers (= the age of the depicted people) whereas the complex task mentioned six different numbers (= the age of six people). This pattern of input vocabulary holds for most characteristics mentioned per person, resulting in twelve more words in the input of the complex task than in the input of the simple task. Calculating Guiraud’s Index with twelve more types and tokens mirrors the gain from simple to complex tasks in the study at hand quite well.6 Native speakers show a similar gain in lexical complexity, which may explain the differences found between simple and complex tasks on the lexical complexity measure. A fourth and more plausible explanation for the effects of the factor ± few elements in the present study then is as follows. Rather than influencing the linguistic aspects of the task performances in terms of ‘different’ or ‘more elaborate’ language use, (i.e., a qualitative change), the manipulation of the single factor ± few elements results in a quantitative change only, i.e.,‘more of the same’.7 From a pedagogic perspective it may be interesting to have L2-learners practice a wider range of vocabulary in a more complex task by means of more elements. From a research-theoretical point of view, that seeks to understand the interplay between task characteristics and attentional allocation, the use of more words as a result of more words in the given input, however, is rather trivial. A last comment concerning Robinson’s Cognition Hypothesis addresses the combined effects of cognitive task complexity and interaction. While effects of cognitive task complexity were expected to be larger in L2-monologues than in L2-dialogues, in the current study, cognitive task complexity only mattered in L1-dialogues (on the fluency measures). Thus, as in Michel et al. (2007, chapter 3) the only significant combined effect contradicts the predictions of Robinson’s theory. To conclude, data of the present study give little support for claims of the Cognition Hypothesis with respect to the effects of cognitive task complexity on its own or in combination with interaction. Other than an increase in lexical complexity, which may be explained by the input given, the study at hand therefore suggests that increased cognitive task complexity manipulated on the single factor ± few elements does not affect L2-learners’ attentional allocation and task performance as proposed by the Cognition Hypothesis (Robinson 1995b, 2001b, Robinson and Gilabert 2007). 6 In

this study, L2-learners used in the simple task on average about 95 types and 232 tokens which equals a Guiraud of √ types/ tokens = 6.24. Hypothesizing that L2-learners use every input word about three times (12 more types, 36 more tokens) the value of Guiraud (= 6.54) comes close to the average Guiraud found for complex tasks. 7 I thank Rod Ellis for pointing this out.

4.4 Discussion

4.4.2

89

Interaction and L2-performance

The second hypothesis of the present study predicted that interactive tasks raise the accuracy and fluency of L2-task performance while syntactic and lexical complexity is decreased due to interactional turn taking and alignment processes. Results indeed show significant main effects for interaction on all measures. In dialogues L2-learners were more accurate, lexically more diverse and more fluent but structurally less complex than in monologues. L1-speakers display a similar pattern with the exception of lexical complexity. As hypothesized, Guiraud was lower in native dialogues than monologues. Accuracy measures show consistently that L2-learners made fewer errors in dialogues than in monologues. All effects were large, i.e., the number of errors were at least halved in the interactive setting, and especially with respect to lexical accuracy, dialogic performances generated more accurate speech than monologues. Since even the small amount of lexical and determiner errors in native speakers was significantly lowered in dialogues, the conclusion from the present work is that interaction has a beneficial effect on the accuracy of task performance of L2-learners and L1-speakers. Interaction also promotes the fluency of L2-learners with respect to speed, pausing as well as repair behavior. Effects are the largest for pausing and repair, where interactive tasks almost halve the scores. This positive effect of interaction is predicted by the Interaction Hypothesis (Long 1990), the Output Hypothesis (Swain 1985, Swain and Lapkin 1995), and Schmidt’s Noticing Hypothesis (Schmidt 1990). The joint focus on the language code as proposed by the Cognition Hypothesis (Robinson 2001b, 2005, 2007b) also predicts higher accuracy measures in dialogues. The parallel increase in fluency, however, allows another interpretation. Tavakoli and Foster (2008) assume that dialogues put less procedural pressure on L2-task performance than monologues. As speakers in dialogues can plan their performance during the interlocutor’s turn, they have more resources available for the actual speech production during their own turn. In a monologue, hesitations occur because the non-automatic speaking process in L2-speakers is interfered with by active onlineplanning. In dialogues, a joint gain in measures of accuracy and fluency (with respect to speed, pausing and repair) is possible because the formulator and articulator have more attentional resources available (Levelt 1989). The general assumption that dialogues are cognitively simpler than monologues is in line with the Alignment Hypothesis (Costa et al. 2008, Pickering and Garrod 2004) and fits into Skehan’s Limited Attentional Capacity Model, where increased planning time is expected to reduce trade-off effects between fluency and form, i.e., accuracy (Skehan 2001). The fact that for L1-speakers dialogues significantly reduced the amount of pauses and increased the unpruned speechrate A, but showed no effect on the pruned speechrate B or repair behavior, further corroborates this idea. Hence, pausing behavior may be most directly related to conceptualizing, a process that needs attentional resources in native speakers too. Spreechrate B is cleaned for repairs and reformulations such that it is not affected by online-planning, while the unpruned speechrate A

90


reflects this gain in processing resources. In addition, the task difficulty judgments of the participants revealed a greater perceived ease in monologues than in dialogues. As a whole, these results thus support the idea that increased planning time and alignment, as a natural byproduct of interaction, leave more attentional resources for speech production in dialogues. These cognitive aspects of the factor ± monologic thus may serve as an explanation for the effects of interaction on learner speech. The lower cognitive task complexity induced by the factor (– monologic) in interactive tasks has beneficial effects on form and meaning in the oral task performance of L2-learners and L1-speakers. Effects of interaction on linguistic complexity were large in both populations. L2-learners and native speakers used more complex clauses and more complex syntactic structures in monologues than in dialogues. Since these decreasing effects in native speakers are larger than in L2-learners, possibly a lower syntactic complexity in interactive tasks reflects native behavior. Due to turn taking, sentences become short and simple and fewer subordination occurs in dialogues. Accordingly, clarification work, comprehension checks and interruptions prevent speakers (whether in their L1 or L2) from using complex syntactic structures. Also Pallotti (2009) points out that lower linguistic complexity is sometimes more native-like than complex syntactic structures – which presumably is manifested by the data of the present study. The fluency measures in turn support the idea that interlocutors prevent each other from building complex clauses and syntactic structures. Dialogues generate fewer silences than monologues. Apparently, interactants start speaking as soon as the interlocutor hesitates. Thereby they reduce the structural complexity of the joint speaking performance because the speaking partner has no chance to perform complex syntactic operations. In contrast, the measure of lexical complexity (Guiraud’s Index) was higher in L2-dialogues than in L2-monologues. At first, the fact that L1-speakers show the reversed picture is puzzling. However, taking into account the parallel gains on (lexical) accuracy and fluency the following explanation may hold. Due to freed attentional resources in dialogues L2-learners benefit from each other on all linguistic levels, i.e., accuracy, fluency, AND lexical complexity. Pickering and Garrod (2004) assume that in dialogues interlocutors recycle each others speech to a large extent which they call alignment. One central process of alignment is routinization: interlocutors establish and agree on conversation specific phrases and words, routines, that they keep using during this conversation (Pickering and Garrod 2004). The second hypothesis of the present study therefore expected that linguistic complexity would decrease in dialogues, which indeed is manifested by the native speaker data. In L2-learners, alignment results in a different pattern. Non-natives do also copy each other’s words and phrases. However, the incomplete knowledge of the target language inhibits full application of routinization: In L2-learners, copying the words of the interlocutor leads to a more diverse vocabulary

4.5 Conclusion and directions for future research and practice

91

use, because they can use more different words than they would come up with on their own. In a monologue L2-learners only use their own limited lexicon. In a dialogue they profit from the input of the speaking partner by incorporating his or her lexical choices. As a result, the lexical complexity increases in joint L2-task performance.To recap, the present study shows that interactive tasks push the accuracy, lexical complexity, and fluency of L2-speakers while monologic tasks allow for the use of more complex syntactic structures. Different processing accounts (Costa et al. 2008, Levelt 1989, Pickering and Garrod 2004), perceived task difficulty judgments, as well as the comparison with L1-speaker’s performances indicate that monologues are cognitively more effortful than dialogues.

4.5

Conclusion and directions for future research and practice

To conclude, the study at hand gives only little support to the claims of the Cognition Hypothesis (Robinson 2005). Increased cognitive task complexity manipulated on the factor ± few elements resulted in a higher lexical complexity only while no combined effects of cognitive task complexity and interaction supporting the Cognition Hypothesis were found. The expectations about effects of interaction on its own, however, were largely confirmed. Interactive tasks did promote lexical complexity, accuracy, and fluency in L2-learners, but lowered the structural complexity. Based on the points discussed in section 4.4 the following three conclusions are formulated: First, this study manipulated different factors of task design (± few elements and ± monologic) on their own as well as in combination. This systematic approach to investigate cognitive and interactive variables of the Triadic Componential Framework in isolation, but also the focus on the interplay of the two task variables, allows powerful interpretations on attentional allocation during task performance. Hence, as increased cognitive task complexity on the single factor ± few elements as implemented in this study did only yield minor differences in the lexical complexity for L2- and L1-task performance, while interaction yielded large effects on all measures of task-based performance, it seems that this cognitive factor is not a task characteristic that substantially affects the allocation of attention. Rather than a qualitative change of linguistic behavior, the factor ± few elements presumably affects the speech of L2-learners in a quantitative way only, i.e., ‘more of the same language’. As the Cognition Hypothesis aims foremost at being a guideline for task sequencing and syllabus design for interlanguage development (Robinson and Gilabert 2007), whereas the present study investigated L2-task performance at one single moment in time, future research may focus on effects of cognitive task complexity on L2-production over time by continuing to investigate effects of different task characteristics on their own as well as in combination. Second, interactive tasks push L2-learners to greater accuracy, lexical complexity, and fluency while monologues give speakers the opportunity to build complex syntactic structures. For the practice of language teaching, this study reveals that both monologues and dialogues are valuable settings for

92


L2-production, as they both promote different aspects of oral L2-performance. Language teachers and testers, however, should be aware of these differences, because L2-learners are often evaluated in a monologic setting, especially in a testing environment, which apparently is likely to give an underestimated picture of the L2-learner’s competence with respect to lexical complexity, accuracy, and fluency. Third, the interpretation of learner speech in light of a native speaker baseline gives valuable insights into the different processes speakers are involved in when they perform oral tasks. From a cognitive perspective on L2-task performance, results of this comparison suggest that we possibly need to change our theoretical framework. Based on psycholinguistic models of language processing (Levelt 1989) and the Alignment Hypothesis (Costa et al. 2008, Pickering and Garrod 2004) the present study suggests that next to being an interactive variable the factor ± monologic has a cognitive impact on L2-task performance too.

Chapter

5

Study 2b: The use of conjunctions in cognitively simple versus complex L2-tasks

Abstract The present study explores the use of conjunctions in simple versus complex argumentative L2-tasks as a specific measure for the amount of reasoning. The Cognition Hypothesis (Robinson 2005) states that an increase of cognitive task complexity promotes L2-performance. This should especially become visible on task specific measures (Robinson and Gilabert 2007). This paper evaluates these claims as it investigates the task-based performance of 64 L2-learners on cognitively simple and complex oral argumentative reasoning tasks. The analysis focuses first on the overall frequency and occurrence of conjunctions. Next, five conjunctions are examined more specifically as these are considered to be particularly task relevant. Results are compared to the speech production of 44 native speakers who performed the same tasks under the same conditions. The discussion addresses implications for the cognitive approach to task-based L2-research. More specifically it discusses the claims of Robinson’s (2005) Cognition Hypothesis. Furthermore, it reviews the pros and cons of using global versus specific measures and highlights the benefits of having a native speaker baseline.

5.1

Cognitive task complexity and the Cognition Hypothesis

This chapter examines the use of conjunctions as a specific measure in simple and complex argumentative reasoning tasks. An earlier investigation of the same data using global measures of linguistic This chapter in adapted form is under review as Michel, M.C., Effects of task complexity on the use of conjunctions in L2-task performance.

94


complexity, accuracy and fluency found minor effects of cognitive task complexity (Michel, in press, chapter 4). This was in contrast to Robinson’s Cognition Hypothesis that claims that an increase of cognitive task complexity promotes the linguistic complexity and accuracy of an L2-learner’s task performance at the cost of fluency (Robinson 2005). More recently Robinson has proposed to evaluate L2-production by means of task specific measures as a complement to traditional global measures (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). The present work explores whether a task specific measure indeed reveals more support for Robinson’s predictions. In the last decade the issue of how changes in cognitive task complexity affect L2-task performance has lead to a series of publications investigating two prominent models: Skehan’s (1998) Limited Attentional Capacity Model and the Multiple Attentional Resources Model of Robinson (2005), that may be better known as the Cognition Hypothesis. The main contrast between the two models concerns the question whether in cognitively complex tasks human resource limitations result in trade-off effects between different dimensions of L2-performance, i.e., linguistic Complexity, Accuracy, and Fluency (in short CAF). Skehan, on the one hand, states that attention is limited in its capacity. Accordingly, in cognitively complex tasks linguistic complexity, accuracy, and fluency, are in competition with each other (Skehan 1998, Skehan and Foster 2005). When task demands exceed the available attentional resources, tradeoff effects emerge because some one of the three CAF-constructs is prioritized. Resource limitations thus prevent a parallel increase of e.g., linguistic complexity and accuracy in cognitively complex tasks. Robinson, on the other hand, claims in his Cognition Hypothesis that a parallel increase of accuracy and linguistic complexity may be possible because L2-learners can rely on multiple pools of attentional resources (Robinson 2005). He presents a Triadic Componential Framework of factors that influence task performance of L2-learners. The framework distinguishes (1) cognitive factors of task complexity from (2) interactive factors of task condition, and (3) learner factors of task difficulty. According to Robinson (2005) cognitive factors of task complexity are important for task design because they may be manipulated such that they guide the L2-learner’s attention towards the language form and push L2-production. His model distinguishes (a) resource-directing from (b) resourcedispersing cognitive variables. An increase of the latter disperses attention away from task relevant linguistic processes. This results in inefficient processing and a decrease in L2-performance. For example, removing planning time will hinder smooth language production because attention is needed for on-line planning. In contrast, an increase along resource-directing cognitive variables (e.g., ± reasoning-demands and ± few elements) focuses the L2-learner’s attention on language. For example, a task with many elements requires an L2-learner to use a more diverse lexicon and more comparative syntactic structures in order to distinguish all elements from each other. In contrast to a task with only a few elements the cognitively more complex task draws the attention towards relevant linguistic forms. During cognitively

5.2 Global versus specific measures

95

simple tasks L2-learners tend to stay in a ‘pragmatic mode’ where simple linguistic means suffice. In a cognitively complex task L2-learners operate in a ‘syntactic mode’ because the more complex functional ´ 1985, 1995). demands ask for complex linguistic realizations (Givon The critical claim of Robinson is that L2-learners do not have to cope with resource limitations because they can rely on different attentional pools during complex task performance (Wickens 2007). Consequently, an increased cognitive task complexity triggers a more elaborate language without having harmful effects on accuracy. The Cognition Hypothesis thus claims that increased cognitive task complexity along resource-directing factors pushes linguistic complexity and accuracy of L2-task performance in parallel (Robinson 2005). By now, a body of empirical work has investigated Robinson’s claims (Gilabert 2007a, Ishikawa 2007, ´ esz ´ 2009, Robinson 2001b, 2007a). By manipulating Kuiken and Vedder 2007b, Michel et al. 2007, Rev the resource-directing factors ± here-and-now, ± few elements, and ± reasoning-demands, most of these studies corroborated the Cognition Hypothesis as they revealed a higher accuracy or linguistic complexity due to an increased cognitive task complexity, without showing trade-off effects on the other measures. However, studies that present a parallel increase in linguistic complexity and accuracy are not very convincing. For example, Michel et al. (2007, chapter 3) attested accuracy and linguistic complexity to be both raised due to a higher cognitive task complexity. Even so, this effect manifested itself on only two out of nine measures. In Robinson (2001b) the support was a mere trend effect and data of Kuiken and Vedder (2007b) showed mixed results for different populations. In sum, these studies do not generate a clear picture. Robinson formulated a possible explanation: As most studies evaluated L2-performance by means of global CAF-measures it may be that the traditional measures of linguistic complexity, accuracy, and fluency are too global. Possibly, they are not sensitive enough to grasp the performance differences due to increased cognitive task complexity. Traditional measures of L2-production therefore should be ‘supplemented by [task relevant] specific measures of the accuracy and complexity of production’ (Robinson and Gilabert 2007: 166). The present study evaluates this suggestion as it investigates effects of increased cognitive task complexity on L2-performance by means of a task specific measure. In the process, it adds to the current discussion about global versus specific measures of task performance.

5.2

Global versus specific measures

The use of global measures of linguistic complexity, accuracy, and fluency has a long tradition in taskbased L2-research. The CAF-trichotomy is thought to adequately and comprehensively capture the multi-componential nature of L2-performance (Housen and Kuiken 2009). Furthermore, it allows the

96


comparison of different studies over various experimental conditions and tasks. Moreover, global CAFmeasures have been shown to successfully differentiate between L2-performances in different conditions (for example in the cited studies above). Even so, global CAF-measures have their limitations too. For example, global measures of lexical complexity like Guiraud’s Index would not show differences in task specific lexis if the simple and complex task both ask for uncommon words (Pallotti 2009). Similarly, global measures of structural complexity e.g., the number of clauses per unit, cannot reveal that more complex types of clauses are used (e.g., conditional versus relative clauses). Likewise, traditional accuracy measures (error-free units) do not distinguish between the seriousness of mistakes (e.g., spelling versus semantic errors). In addition, the CAF-trichotomy does not tell us how successful an L2-learner was at reaching the communicative goal of a task (see the special issue on CAF-measures of Applied Linguistics for more elaborate discussions, Housen and Kuiken 2009). Only a few studies published so far have evaluated the Cognition Hypothesis using task specific measures. Robinson (2007b) manipulated the amount of intentional reasoning in simple, medium, and complex dialogic story telling tasks. As well as global CAF-measures, the use of psychological state terms (e.g., believed, wanted), the use of complex syntactic structures (e.g., conjoined and infinitival phrases), and the amount of interaction (e.g., uptake, clarification requests) were examined. While global CAF-measures did not discriminate between task manipulations, task specific measures lent support to the Cognition Hypothesis. The complex task yielded more interaction and participants used more psychological state terms which in turn correlated significantly with the use of complex syntactic structures. Cadierno and Robinson (2009) and Robinson et al. (2009) report on non-prototypical uses of past tense, progressive morphology, and motion verbs in cognitively complex there-and-then tasks and found that indeed, increased cognitive task complexity did affect the L2-behavior on task specific but not on ´ esz ´ (in press) evaluated L2-learners’ task performance on simple versus global CAF-measures. Rev complex reasoning tasks. Results on global measures showed partial support for the Cognition Hypothesis as lexical complexity and accuracy increased in complex tasks whereas syntactic complexity decreased. Data on specific measures corroborate Robinson’s claims as the complex task yielded a higher amount of developmentally advanced conjoined clause types (i.e., more biclausal coordinated sentences and adverbial clauses than independent coordinated clauses) and generated more language related episodes. These few examinations suggest that by evaluating L2-production by means of specific measures we may be able to detect differences in task performance that do not become visible when using global CAF-measures. Furthermore, it may widen our understanding of how changes in cognitive task complexity affect L2-production - which is the aim of this study.

5.2 Global versus specific measures

5.2.1

97

The use of conjunctions as a task specific measure

The choice of task specific measure by definition is given by its relevance for the task itself. It should be a structure that characterizes successful task performance and is induced by the manipulation of the task (Robinson and Gilabert 2007). The present study examined the use of conjunctions as a specific measure in argumentative reasoning tasks. The literature associates overt clause marking by conjunctions with argumentative reasoning. Newton and Kennedy state that ‘it is the reasoning or argumentation that requires conjunctions to mark the relationships between propositions’ (Newton and Kennedy 1996: 320). The authors came to this conclusion after a twofold exploration of L2-data. At first, they found in an L2-corpus that task-based performances on split information tasks yielded language that was mainly descriptive in nature and relied on the use of ‘and’ or no conjunctive marking at all. In contrast, tasks with shared information led to discussions of an issue and generated more reasoning. These tasks induced a greater use of conjunctive marking by ‘so’ and ‘if’. Results of a follow-up experiment corroborated the explorative data. A shared information task (+ reasoning) elicited more (subordinating) conjunctions than a split information task (- reasoning). Also Robinson (2005) argues that, while tasks that do not involve reasoning mainly use coordinate clauses linked by ‘and’, reasoning tasks promote the use of conjunctions like ‘but’ or adverbial clauses ´ esz ´ with ‘because’, ‘so’, and ‘if’. As discussed before, Rev (in press) examined clauses marked by these conjunctions in adult L2-production. She found that more complex argumentative tasks indeed promoted the use of more advanced clause types marked by causal and conditional conjunctions. Her choice of conjunctions was based on research on L1-acquisition (Diessel 2004). He proposes that children first mark clauses by the coordinative ‘and’ before they start using more advanced clause types marked by causal and conditional conjunctions like ‘because’, ‘so’, and ‘if’. Evers-Vermeul (2005) explains that the emergence of different conjunctions in L1-acquisition is associated with the child’s growing ability to deal with complex information. ‘And’ and ‘because’ are acquired earlier than ‘but’, ‘when’ or ‘if’ because the latter mark more complex coherence relations. As at younger ages children lack the cognitive capacity to build these complex structures they have not yet learnt the associated lexical markers. From a discourse processing perspective the use of conjunctions therefore may be seen as a window into the child’s reasoning abilities (Spooren and Sanders 2008). All these studies thus link reasoning directly or indirectly to the overt use of conjunctions. To the best of my knowledge there is no work on L2-data that explores the marking of reasoning by conjunctions. A few studies examine the order of acquisition and frequency of conjunctions in L2speech and -writing and reveal similar results as in the child data (Cornu and Delahaye 1987, Nemeth and Kormos 2001, Perrez 2004, Plomp 1997, Vedder 1998). However, none of them focused on the use of conjunctions as an indicator for reasoning.

98

5.3


The present study

The present study evaluates a new specific measure for the amount of reasoning. It explores the use of conjunctions in argumentative reasoning tasks in order to examine effects of increased cognitive task complexity on oral L2-performance. An earlier investigation (Michel, in press, chapter 4) of the same data by means of global CAF-measures challenged the Cognition Hypothesis (Robinson 2005). The present chapter investigates whether a task specific measure yields more support.

5.3.1

Research questions and hypotheses

This study addresses the following research questions and hypotheses:

Main Research Question What is the effect of increased cognitive task complexity on the use of conjunctions in L2 oral argumentative reasoning tasks?

Main Hypothesis Cognitively complex tasks are expected to increase the use of conjunctions in L2 oral task performance in argumentative reasoning tasks following the Cognition Hypothesis that predicts increased cognitive task complexity to draw the L2-learner’s attention towards task relevant linguistic structures.

A sub question arises because L1-speakers performing on the same tasks under the same conditions were tested as a baseline:

Sub Research Question What is the difference in the use of conjunctions between L2 and L1 oral task performances in cognitively simple versus complex oral argumentative reasoning tasks?

Sub Hypothesis This question is more explorative in nature but the comparison of L2- and L1-data may widen our understanding of task-based L2-production. As native speakers’ language production relies on mainly automatic cognitive processes (Levelt, 1989) their task performance are expected not to suffer from resource limitations in cognitively complex argumentative reasoning tasks. L1-speakers accordingly may not show differences in the use of conjunctions in simple versus complex oral argumentative reasoning tasks.

5.3 The present study

5.3.2

99

Method

In order to test the hypotheses, data of an earlier (Michel, in press, chapter 4) investigation among L2learners and L1-speakers were analyzed for the use of conjunctions. The data were collected in a 2 x 2 design where cognitive task complexity (simple versus complex) was the within-participant factor and interaction (monologic versus dialogic) was a between-participants factor (cf. section 4.2 for a detailed description of the method and design of this empirical study).1

Argumentative reasoning and the factor ± few elements Based on the Triadic Componential Framework (Robinson 2005) cognitive task complexity was manipulated on the resource-directing factor ± few elements. Kuiken and Vedder (2007b) suggest that an increase in the number of elements implies a higher amount of reasoning: As more items need to be argumentatively differentiated, more reasoning emerges in complex than in simple tasks. Robinson and Gilabert (2007) predict that increased cognitive task complexity focuses the L2learner’s attention towards task specific structures. In cognitively complex argumentative reasoning tasks more conjunctions are expected because they lexically mark reasoning. A manipulation on the factor ± few elements accordingly is predicted to affect the use of conjunctions in task-based argumentative reasoning.

Tasks Two different sets of tasks (simple and complex) were designed addressing different topics (dating and study). The tasks asked participants to decide which two out of four (simple) or six (complex) people would make the best love or studying couple. These people differed in characteristics like age, favorite music and hobby. Irrespective of the topic, the simple condition allowed for four and the complex condition for nine combinations. See Appendix B.4.

Participants 31 Turkish and 33 Moroccan L2-learners of Dutch participated in the study. They were at an intermediate level of Dutch as assessed by a written proficiency task. In eight short texts test-takers had to choose among three possible words at 100 gaps.2 44 native speakers of Dutch were included as a control group. They scored at ceiling on the proficiency task. All participants were attending or had finished higher levels of education. Table 5.1 summarizes their background information. 1 The

factor interaction (± monologic) is not under investigation in this study. However, as the data collection included a monologic and a dialogic setting, data will be analyzed and presented for both settings separately. 2 I thank the Language Center of the University of Groningen, which uses this task as a placement test for their language courses, for sharing their materials.

100


Table 5.1: Background information for all participants LANGUAGE

STAY IN THE

AGE

GENDER

PROFICIENCYa

NETHERLANDSb

ORIGIN

L2-LEARNERS

N

Mean (SD)

m

f

Mean (SD)

Mean (SD)

Tur

Mor

32 monologues

32

27.6 (6.0)

15

17

52.9 (15.6)

3.8 (4.3)

15

17

16 dialogues

32

27.6 (6.6)

14

18

54.4 (19.1)

3.9 (4.1)

16

16

total

64

27.6 (6.2)

29

35

53.8 (17.2)

3.8 (4.2)

31

33

L1-SPEAKERS

N

Mean (SD)

m

f

Mean (SD)

20 monologues

20

23.2 (3.7)

3

17

96.3 (3.5)

12 dialogues

24

18.4 (0.6)

6

18

96.3 (2.9)

total

44

20.6 (3.5)

9

35

96.3 (3.2)

Note. SD = Standard Deviation; m = male; f = female; a in eight short texts participants had to choose among three possible words at a gap with a total of 100 gaps; b in years; Tur = Turkish; Mor = Moroccan

Procedure All participants did a simple and a complex version of the experimental tasks. In between they performed on the Dutch proficiency task. The order of cognitive task complexity (simple versus complex) and task topic (dating versus study) was counterbalanced over participants. Half of the participants acted on their own (monologic), the other half performed in pairs (dialogic).3 For the proficiency task L2-learners had 30 minutes time, L1-speakers only 15 minutes. In order to control for time on task natives performed for another 15 minutes on a written Dutch dummy task. In the experimental tasks participants were asked to call a friend and explain their choice for the best dating or study couple. In the monologic setting, the friend was unable to answer the phone so they should leave a message of about three minutes on an answering machine. In the dialogic condition participants discussed with each other for about 6 minutes. All participants received two minutes of planning time. In order to assure that they use all the speaking time available and that they would consider all possible combinations in their reasoning process, they were encouraged not only to explain why a pair of people was best but also why others would not make a good couple.

Data analysis The speech samples of all 108 participants were transcribed and coded in CLAN (MacWhinney 2000) for global CAF-measures and the task specific measure, i.e., the use of conjunctions. The results on 3 In

dialogues interlocutors shared their mother tongue and gender.

5.3 The present study

101

the global measures are discussed elsewhere (Michel, in press, chapter 4).

Frequency and occurrence of conjunctions The analysis in this chapter investigates whether the predicted higher amount of reasoning in cognitively complex tasks is reflected in the use of conjunctions, i.e., the frequency and occurrence of conjunctions in L2-speech. These measures are defined as follows: Frequency is the number of conjunctions per 100 words in a task performance.4 Occurrence is the number of participants that use a conjunction at least once in a task performance. The frequency ratio gives an impression of how often participants make use of conjunctions in their speech acts at all. It therefore serves as an indicator for the amount of reasoning. However, the frequency measure does not differentiate whether participants use a different set of conjunctions in the two tasks. Cognitively complex argumentative reasoning tasks possibly elaborate the L2-language ´ esz ´ such that task performers start using more different conjunctions. Based on Rev (in press) the occurrence measure was introduced. A higher score reflects that more participants used a conjunction at least once, i.e. if the mean score is higher a larger set of conjunctions was used by more participants.

Conjunctions under investigation After a broad inspection of the data 28 conjunctions that were likely to be present in the speech performances were selected for a more detailed exploration (see Table 5.2).5 In a first step, the 28 conjunctions were automatically compared to the transcripts of the speech performances using CLAN (MacWhinney 2000).6 This revealed that eight conjunctions were not used at all. ‘Wanneer’ (English ‘when’) was removed from the counts as this is a homograph of the marker for wh-questions (see the third column in Table 5.2 for a list of the excluded conjunctions). Data of the resulting twenty conjunctions were sent to SPSS 16.1 for further analysis. The properties (normality, homogeneity of variance and sphericity) allowed a multivariate analysis of variance (MANOVA) with task complexity (± few elements) as a within-participant factor and interaction (± monologic) as a between-participants factor. L2- and L1-data were subject to separate analyses. The alpha level was set to p < .05 and effect sizes (partial η2 ) equal to or greater than .01, .06, and .14 were seen as small, moderate, and large respectively (Sink and Stroh 2006). 4 As subordinate conjunctions introduce subordinate clauses the number of conjunctions was related to the number of words rather than to the number of syntactical units. 5 For example, temporal conjunctions were excluded because the tasks did not ask for time reference. 6 ‘Als. . . dan’ and ‘om. . . te’ were counted if either of the two parts was present.

102


Table 5.2: Dutch conjunctions under investigation PRESENT

SPECIFICALLY

NOT PRESENT

AT LEAST ONCE

TASK RELEVANT

OR REMOVED

aangezien – (as)

(als). . . dan – ((if). . . then)

desalniettemin – (nonetheless)

alhoewel – (although)

daardoor – (therefore)

echter – (however)

bovendien – (moreover)

daarom – (therefore)

indien – (if)

doordat – (as)

omdat – (because)

nochthans – (still)

dus – (so)

want – (because)

ofschoon – (although)

en – (and)

tenzij – (except)

hoewel – (though)

totdat – (until)

maar – (but)

wanneer – (when)

of – (or) om . . . te – (in order to) terwijl – (while) toch – (even so) toen – (then) zodat – (so that) zowel – (as well)

Note. Dutch – (English translation)

Specifically task relevant conjunctions In a second step the analysis focused on five conjunctions that were expected to be particularly task relevant. As the tasks at hand required participants to balance reasons when arguing for or against a possible dating or study couple, the L2-learners’ attention may be drawn in particular towards conjunctions that mark causal and conditional relations. Typical talk on the tasks was of the schematic type (1) where conjunctions are typeset in italics: (1) If we combine person A with person B, then they do not match on property X. Therefore, I would combine B with C, because they match better on property X. As the evaluation and comparison of several options by means of causal and conditional reasoning was inherently part of a communicatively successful task completion, the four causal conjunctions ‘want’ and ‘omdat’ (both meaning ‘because’ in English) and ‘daarom’ and ‘daardoor’ (both be translated into English as ‘therefore’) and the conditional conjunction ‘als. . . dan’ (English ‘if. . . then’) were evaluated in

5.4 Results

103

Table 5.3: Absolute numbers for the frequency of conjunctions for L2-learners L2-LEARNERS

MONOLOGUE (N = 32)

DIALOGUE (N = 32)

FREQUENCY

simple

complex

total

simple

complex

total

als. . . dan/if. . . then

34

33

67

52

76

128

daarom/therefore

20

10

30

12

16

28

dus/so

89

115

204

80

119

199

en/and

475

489

964

443

576

1019

maar/but

94

127

221

140

195

335

of/or

39

55

94

47

74

121

omdat/because

38

29

67

20

24

44

om. . . te/in order to

36

30

66

16

26

42

toch/even so

2

5

7

12

26

38

want/because

37

45

82

38

35

73

tot. 10 low freq. conj.

4

7

11

3

11

9

L2: tot. all 20 conj.

868

945

1813

863

1173

2036


705

805

1510

744

774

1518

Note. L2 = second language; N = number of participants; tot. 10 low freq. conj. = sum over other 10 low frequent conjunctions; tot. all 20 conj. = absolute total of all 20 conjunctions; L1 = native speaker

more detail. These five conjunctions will be referred to as ‘specifically task relevant conjunctions’ (see the second column in Table 5.2). As the data on the specifically task relevant conjunctions were not normally distributed, they were subject to Wilcoxon Signed Ranks tests for the repeated measures of cognitive task complexity. Separate calculations were made for L2- and L1-monologues and dialogues. The alpha level was set to p < .05 and effect sizes (r) of .10 (small), .30 (moderate), and .50 (large) were acknowledged (Field 2005).

5.4

Results

The data for the L2-learners with respect to the frequency and occurrence are presented in Table 5.3 and 5.4, respectively. They list the absolute counts as well as the numbers (and percentages) of participants per conjunction in simple and complex monologues and dialogues for the ten most frequently used conjunctions and give the totals of ten hardly-used conjunctions and of all twenty conjunctions. For the comparison with the native baseline data the bottom row summarizes the totals of the L1-speakers. Although these raw scores are biased for sample length one can make some interesting observa-

104


Table 5.4: Absolute numbers for the occurrence of conjunctions for L2-learners L2-LEARNERS OCCURRENCE

MONOLOGUE (N = 32)

DIALOGUE (N = 32)

simple

complex

total

simple

complex

total

% (n)

% (n)

% (n)

% (n)

% (n)

% (n)

als. . . dan/if. . . then

59 (19)

53 (17)

81 (26)

63 (20)

72 (23)

84 (27)

daarom/therefore

28 (9)

22 (7)

38 (12)

28 (9)

22 (7)

41 (13)

dus/so

63 (20)

78 (25)

84 (27)

75 (24)

75 (24)

78 (25)

en/and

100 (32)

100 (32)

100 (32)

10 (32)

100 (32)

100 (32)

maar/but

69 (22)

94 (30)

97 (31)

91 (29)

97 (31)

100 (32)

of/or

59 (19)

56 (18)

81 (26)

59 (19)

78 (25)

84 (27)

omdat/because

59 (19)

38 (12)

66 (21)

31 (10)

38 (12)

50 (16)

om. . . te/in order to

53 (17)

50 (16)

75 (24)

31 (10)

47 (15)

59 (19)

toch/even so

6 (2)

13 (4)

19 (6)

31 (10)

31 (10)

41 (13)

want/because

44 (14)

50 (16)

56 (18)

56 (18)

47 (15)

63 (20)

6 (2)

13 (4)

16 (5)

9 (3)

16 (5)

25 (8)


13

12

14

13

13

16


15

15

15

16

15

17

tot. 10 low freq. conj.

Note. L2 = second language; N = number of participants; n = number of performances where a conjunction was used; % = percentage of performances; tot. 10 low freq. conj. = sum over other 10 low frequent conjunctions; tot. all 20 conj. = total number of different conjunctions used in all the performances; L1 = native speaker

tions: Generally complex tasks yield higher numbers than simple tasks. For L2-learners, dialogues also increase the number of conjunctions. They prefer a set of ten conjunctions with ‘en’ (‘and’), ‘maar’ (‘but’), and ‘dus’ (‘so’) at the top that they use very frequently (Note that natives prefer the same ten conjunctions, but use ‘als. . . dan’/’if. . . then’ at similar rates as ‘maar’/’but’). Contrary to expectations, ‘daardoor’ (‘therefore’) - one of the five conjunctions that was assumed to be specifically task relevant is not among these most frequent conjunctions in either population.

5.4.1

The use of conjunctions

Table 5.5 summarizes the means over all 20 conjunctions corrected for sample length for both populations.7 While L2-learners show a higher frequency, L1-speakers display a higher occurrence. With the exception of the frequency in monologues, increased cognitive task complexity in L2-speakers resulted 7 Average sample sizes of a single speakers performance in Analysis-of-Speech (AS) units (Foster et al. 2000) were for L2learners: monologue simple = 31, complex = 37; dialogue simple = 47, complex = 60; and for L1-speakers: monologue simple = 30, complex = 33; dialogue simple = 41, complex = 46.

5.4 Results

105

in higher scores. For L1-speakers complex tasks yielded higher scores in monologues but lower scores in dialogues. However, as summarized by Table 5.6, the results of the MANOVA performed on these data did not yield any significant multivariate effects in either population. Also broken down into univariate effects, the performances were neither affected significantly by cognitive task complexity on its own, nor in combination with interaction. Interestingly, the factor interaction – that is not under investigation here – reveals a significant main effect on L1-speakers’ frequency of conjunctions (F(1,42) = 27.707, p < 0.05, partial η2 = 0.116).

5.4.2

Focusing on specifically task relevant conjunctions

The second analysis focused on the five conjunctions that specifically mark causal and conditional reasoning: ‘want’/’omdat’ (‘because’), ‘daarom’/’daardoor’ (‘therefore’) and ‘als. . . dan’ (‘if. . . then’). The descriptives are given in Table 5.7 (L2-learners) and 5.8 (L1-speakers). For L2-learners the absolute numbers do not show an obvious pattern. The results of a Wilcoxon Signed Ranks test revealed one significant effect on the occurrence of ‘omdat’ (‘because’), that is higher in simple than in complex monologues (T = 2, z = -2.111, p < 0.05, r = -0.264). For L1-speakers the frequency measures of all specifically relevant conjunctions are lower in complex than in simple tasks. The statistical analyses, however, show only one significant result for the frequency of ‘daarom’ (‘therefore’) in dialogues, that decreases from simple to complex tasks (T = 1, z = -2.028, p < .05, r = -0.293).

106


Table 5.5: Descriptives on frequency and occurrence of all conjunctions DESCRIPTIVES

L2-LEARNERS

L1-SPEAKERS

monologue

dialogue

total

monologue

dialogue

total

(N =32)

(N =32)

(N = 64)

(N =20)

(N =24)

(N = 44)

FREQUENCY

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

simple

12.33 (3.38)

11.23 (2.69)

11.78 (3.08)

11.62 (2.43)

10.72 (1.78)

11.13 (2.12)

complex

11.57 (3.00)

11.82 (2.45)

11.70 (2.72)

11.74 (1.92)

10.38 (2.04)

11.00 (2.08)

total

11.85 (2.77)

11.51 (2.31)

11.68 (2.54)

11.68 (1.79)

10.63 (1.45)

11.11 (1.68)

OCCURRENCE

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

simple

5.47 (1.50)

5.75 (1.39)

5.61 (1.44)

7.70 (1.56)

7.54 (1.67)

7.61 (1.60)

complex

5.66 (1.66)

6.22 (1.60)

5.94 (1.64)

7.95 (1.23)

6.88 (1.87)

7.36 (1.69)

total

7.13 (1.34)

7.25 (1.63)

7.19 (1.48)

9.45 (1.36)

9.00 (1.91)

9.20 (1.68)

Note. L2-learner = second language learner; L1-speaker = native speaker; N = number of participants; frequency = mean number of conjunctions per 100 words; occurrence = mean number of conjunctions used at least once in a performance; Mn = Mean; SD = standard deviation

Table 5.6: Statistics on frequency and occurrence of all conjunctions EFFECTS MULTIVARIATE

L2-LEARNERS Pill. Tr. F

L1-SPEAKERS df

Err

p

p.η2

.235 .046 .017

0.346 2

41

.709

.017

61

.328 .036 .126

2.966 2

41

.063

.126

61

.135 .064 .036

1.135 2

41

.328

.036

p

p.η2

df

Err

p

TASK COMPLEXITY .046

1.482 2

61

INTERACTION

.036

1.135 2

TASK COMPLEXITY .064

2.073 2

p.η

2

Pill. Tr. F

× INTERACTION UNIVARIATE

meas.

F

df, Err MSQ

p

p.η2

meas.

F

df, Err MSQ

TASK COMPLEXITY freq.

0.057 1,62

0.221

.812 .001 freq.

0.083 1,42

0.277

.774

.002

occ.

2.909 1,62

3.445

.093 .045 occ.

0.691 1,42

0.947

.410

.016

freq.

0.446 1,62

5.763

.507 .007 freq.

5.517 1,42

27.707 .024* .116

occ.

1.596 1,62

5.695

.211 .025 occ.

2.149 1,42

8.297

.150

.049

TASK COMPLEXITY freq.

3.807 1,62

14.749 .056 .058 freq.

0.334 1,42

1.111

.566

.008

× INTERACTION

0.534 1,62

0.633

3.345 1,42

4.583

.074

.074

INTERACTION

occ.

.468 .009 occ.

Note. L2 = second language; L1 = native speaker; N = number of participants; meas. = measure; freq. = number of conjunctions per 100 words; occ. = number of conjunctions used at least once; task compl. = task complexity; Pill. Tr. = Pillai’s Trace; df, Err. = degrees of freedom, Error df.; MSQ = Mean Square Error; p.η2 = effect size; * = significant at p < 0.05.

5.4 Results

107

Table 5.7: Descriptives of specifically task relevant conjunctions for L2-learners DESCRIPTIVES

WANT

OMDAT

DAAROM

DAARDOOR

ALS. . . DAN

L2-LEARNERS

(because)

(because)

(therefore)

(therefore)

(if. . . then)

FREQUENCY

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

simple

0.54 (0.78)

0.55 (0.64)

0.30 (0.62)

0.00 (0.00)

0.42 (0.41)

complex

0.55 (0.85)

0.34 (0.53)

0.15 (0.39)

0.01 (0.07)

0.42 (0.54)

simple

0.52 (0.57)

0.26 (0.47)

0.17 (0.29)

0.02 (0.09)

1.75 (1.04)

complex

0.36 (0.51)

0.26 (0.52)

0.13 (0.28)

0.00 (0.00)

1.97 (0.99)

N (%)

N (%)

N (%)

N (%)

N (%)

simple

14 (44)

19 (59)

9 (28)

0 (0)

19 (59)

complex

16 (50)

12 (38)

7 (22)

1 (3)

17 (53)

simple

18 (56)

10 (31)

9 (28)

1 (3)

29 (91)

complex

15 (47)

12 (38)

7 (22)

0 (0)

31 (97)

monologue

dialogue

OCCURRENCE monologue

dialogue

Note. L2 = second language; frequency = number of conjunctions per 100 words; occurrence = absolute number N (and percentage %) of participants using a conjunction at least once; Mn = mean; SD = standard deviation

Table 5.8: Descriptives of specifically task relevant conjunctions for L1-speakers DESCRIPTIVES

WANT

OMDAT

DAAROM

DAARDOOR

ALS. . . DAN

L1-SPEAKERS

(because)

(because)

(therefore)

(therefore)

(if. . . then)

FREQUENCY

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

Mn (SD)

simple

0.35 (0.33)

0.59 (0.55)

0.00 (0.00)

0.12 (0.22)

1.26 (0.88)

complex

0.32 (0.40)

0.48 (0.52)

0.00 (0.00)

0.09 (0.16)

1.05 (0.56)

simple

0.52 (0.47)

0.55 (0.68)

0.13 (0.26)

0.04 (0.13)

1.50 (0.86)

complex

0.46 (0.51)

0.44 (0.60)

0.01 (0.05)

0.03 (0.09)

1.21 (0.84)

N (%)

N (%)

N (%)

N (%)

N (%)

simple

14 (70)

14 (70)

0 (0)

6 (30)

18 (90)

complex

10 (50)

13 (65)

0 (0)

5 (25)

19 (95)

simple

17 (71)

13 (54)

6 (25)

2 (8)

23 (96)

complex

16 (67)

13 (54)

2 (8)

3 (13)

21 (88)

monologue

dialogue

OCCURRENCE monologue

dialogue

Note. L1 = first language; frequency = number of conjunctions per 100 words; occurrence = absolute number N (and percentage %) of participants using a conjunction at least once; Mn = Mean; SD = standard deviation

108

5.5


Discussion

Robinson and colleagues argue that specific measures should be used as a complement to global CAF-measures in order to reveal differences due to increased cognitive task complexity (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). The work at hand has followed this suggestion and investigated the use of conjunctions in monologic and dialogic L2-performances on simple and complex argumentative reasoning tasks. Data are interpreted in light of an L1-speaker baseline.

5.5.1

The use of conjunctions in cognitively simple versus complex tasks

Concerning the main research question the results of the MANOVA on the frequency and occurrence of a large set of different conjunctions showed no significant main effect of cognitive task complexity, nor did the combination of cognitive task complexity by interaction turn out significant. The analysis on five specifically task relevant conjunctions revealed that ‘omdat’ (‘because’) was affected significantly by increased cognitive task complexity. This effect was found on the occurrence measure only and was in opposition to the hypothesized direction: Complex tasks yielded a lower score than simple tasks. Neither the occurrence of the other conjunctions nor the frequency of any conjunction was significantly influenced by cognitive task complexity. The data at hand therefore do not support the main hypothesis as no significant promoting effects of cognitive task complexity on the use of (specifically task relevant) conjunctions in L2 oral argumentative reasoning tasks were found. Concerning the sub question, whether native and non-native speakers behave similarly in the present study, there are some interesting contrasts. First, Table 5.5 reveals that L1-speakers generally show higher occurrences and L2-learners higher frequencies. A further contrast is the highly frequent use of ‘als. . . dan’ (‘if. . . then’) by L1-speakers which L2-learners do not mirror (see Table 5.7 and 5.8). Furthermore, interaction has an effect on natives but not on non-natives. In relation to the main research question, however, both populations are similarly influenced by cognitive task complexity. With the exception of lower scores for one specifically task relevant conjunction (‘omdat’ (‘because’) in L2-learners and ‘daarom’ (‘therefore’) in L1-speakers), increased cognitive task complexity did not affect the use of conjunctions in either population. Consequently, although the raw data point towards an increase in the amount of conjunctions from cognitively simple to complex tasks, the statistical analysis, where the task specific measure is corrected for sample length, does not yield confirmatory results. The overarching conclusion accordingly is that a cognitively more demanding task manipulated on the resource-directing factor ± few elements does not focus the L2-learner’s attention towards (specifically task relevant) conjunctions such that the frequency or occurrence of these conjunctions is substantially affected.

5.5 Discussion

109

The present study thus challenges Robinson’s claims about effects of cognitive task complexity on the use of task specific measures (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). As the only significant effect of a higher cognitive task complexity showed a decrease in performance (on the occurrence of the specifically relevant conjunction ‘omdat’ (‘because’) in L2monologues), one could attribute the findings to Skehan’s Limited Attentional Capacity Model (Skehan 1998, Skehan and Foster 2005). Yet, the data do not confirm Skehan’s approach either. Only one out of five specifically relevant conjunctions was significantly affected as predicted by Skehan and none of the comparisons in the overall analysis turned significant. The following paragraphs discuss other possible explanations.

5.5.2

The factor ± few elements

The present analysis by means of a task specific measure extends an examination of the same data by using global CAF-measures (Michel, in press, chapter 4). This earlier analysis found that lexical complexity was higher in cognitively complex tasks while accuracy and fluency were unaffected. Michel (in press, chapter 4) concludes that the data challenge Robinson’s Cognition Hypothesis, as a manipulation of the cognitive factor ± few elements had a mere quantitative effect. Rather than different language use the increase of the number of elements in the task input resulted in more but qualitatively the same linguistic behavior of L2-learners. As the present chapter investigates the same data this explanation may hold again. The analysis by means of the task specific measure, the frequency and occurrence of conjunctions, mirrors the work of Michel (in press, chapter 4). The absolute scores show that the complex tasks generated a higher number of conjunctions than the simple tasks (Table 5.3) but when speech samples are corrected for sample length this finding is non-significant (see Table 5.5 and 5.6). Like in Michel (in press, chapter 4), increased cognitive task complexity manipulated on the factor ± few elements yielded a mere quantitative change of L2-task performance. Consequently, also the data at hand challenge Robinson’s Cognition Hypothesis.8 ´ esz ´ (in press), who This raises the question why the findings of Kuiken and Vedder (2007b) and Rev investigated the factor ± few elements, support the Cognition Hypothesis. In contrast to the present chapter, these studies, however, manipulated the number of elements such that it explicitly involved an increase on the factor ± reasoning-demands too. For example, in Kuiken and Vedder (2007b) participants had to take into account more characteristics of the same number of elements when reasoning about a decision. As the authors argue, this induces ‘automatically’ a higher reasoning in the complex ´ esz ´ (in press) manipulated the number of elements as well as the factor ± reasoning-demands. task. Rev It may be that combined manipulations of the factors ± few elements and ± reasoning-demands have 8 The fact that occurrence (which is not dependent on speech length) is not affected towards a specific direction (simple versus complex) may corroborate this explanation.

110


the potential to affect cognitive processes during task-based L2-performance and qualitatively influence L2-performance. In contrast, the aim of the present study was to strictly operationalize the single factor ± few elements. It was manipulated in a straight-forward manner: In the simple task four people were each accompanied by six characteristics. In the complex task two people with six characteristics each were added. Probably, this strict manipulation did not entail an ‘automatic’ increase of reasoning demands. L2-production was only influenced in a quantitative way because the factor ± few elements did not affect the amount of reasoning. In contrast, in the earlier studies it presumably was the factor ± reasoningdemands that was responsible for the qualitative changes in task performance. From Michel (in press, chapter 4) and the present study it seems that mere manipulations of the number of elements do not guide the cognitive processes of L2-learners towards (task specific) linguistic means. As the data on specific measures mirror the global picture, the results of the present study and Michel (in press, chapter 4) show that focusing on a task specific measure does not revise the conclusions of the earlier analysis on global CAF-measures. In sum, they raise the question whether the factor ± few elements affects L2-learners’ attentional allocation during task-based performance as predicted by the Cognition Hypothesis (Robinson 2007b). For future work within the Triadic Componential Framework (Robinson 2005) a focus on the factor ± reasoning-demands possibly is more promising.

5.5.3

Good measures of task performance

As an alternative explanation one could doubt whether the present study used good measures for investigating the amount of argumentative reasoning. However, the exploration of the literature attests that conjunctions are markers of argumentative reasoning. Furthermore, L1-data confirm the L2-behavior, e.g., both populations show a preference for the same set of ten frequently used conjunctions and the exploration of the specifically task relevant conjunctions in L2-learners finds support in the L1-speech. The comparison with the native baseline reveals another interesting point. L2-learners generally used conjunctions more often than the L1-speakers (Table 5.5). It may be that natives used different linguistic means to express argumentative reasoning than overt lexical marking by conjunctions. Norris and Ortega (2009a) and Pallotti (2009) mention that with growing proficiency also L2-learners rely more on phrase-internal linguistic structures than on overt lexical marking. If the complex tasks made L2learners switch to a ‘syntactic mode’ of processing they probably started using other more complex linguistic features to meet the complex reasoning demands of the task. The present analysis on the frequency and occurrence of conjunctions, however, is not able to detect such a difference. This raises a more fundamental issue: As discussed in Housen and Kuiken (2009), quantitative measures of L2-performance are probably not as reliable as their wide-spread use in task-based research

5.6 Summary and conclusion

111

would suggest. Norris and Ortega (2003, 2009a) point out that when measuring L2-performance it is essential that we understand what we measure and why we do so. An extension by means of task specific measures or adding more and other global CAF-measures to the currently used ones, however, will not overcome this problem. Including more measures ‘instead of providing a wider picture, just makes it more blurred’ (Pallotti 2009: 599). Future research could take a different perspective in order to complement global CAF-measures. If we want to understand what L2-learners do when they perform a task it may be more promising to evaluate the communicative success of L2-performance (de Jong et al. 2007, Kuiken, Vedder, and Gilabert 2010). As shown by the present work (and suggested by e.g. Michel et al. 2007, Tavakoli and Foster 2008, and in Housen and Kuiken 2009) also the inclusion of and comparison with native speaker data presumably is a more fruitful approach than extending our toolkit with more but similar quantitative measures.

5.6

Summary and conclusion

This chapter presented an exploration of the use of conjunctions as a specific measure for cognitively simple versus complex L2-production on argumentative reasoning tasks. It elaborates on an earlier analysis of the same data by means of global CAF-measures (Michel, in press, chapter 4). Contrary to expectations, but just as in Michel (in press, chapter 4), only minor effects of the factor ± few elements could be attested. Consequently, it challenges the Cognition Hypothesis and claims of Robinson and Gilabert (2007) who argue that global CAF-measures should be complemented by task specific measures. The present study concludes: First, like in Michel (in press, chapter 4) it seems that a manipulation of the factor ± few elements results in a quantitative rather than a qualitative change in L2-task performance. When trying to guide attentional allocation towards language during task-based L2-production, other factors (e.g., ± reasoning-demands) may be more interesting. Second, when evaluating task-based L2-performances by means of quantitative measures – irrespective of whether these are global or task specific measures – there remains the risk of interpreting a higher score as ‘better’ language use. In order to better understand how manipulations of task variables affect L2-production, it may make more sense to complement the traditional tools by measures of communicative success or to compare L2- with L1-data.

Chapter

6

Summary of the findings, discussion, and implications

6.1

Introduction

The studies in this book investigated the oral task-based performance of L2-learners of Dutch. This chapter summarizes the findings of the empirical studies presented in chapters 3, 4, and 5 and relates them to the theoretical framework presented in chapter 1 and the hypotheses formulated in chapter 2. It starts by giving a summary of the theoretical background of the studies, in section 6.2. Next, section 6.3 gives an overview of the empirical investigations and summarizes the results. Section 6.4 discusses the outcomes in the light of Robinson’s Cognition Hypothesis (2005). It focuses on cognitive aspects of task complexity (simple versus complex) and interactive factors of task condition (monologic versus dialogic). In the process it elaborates on measures of task-based performance and highlights the benefits of including a native speaker baseline. Section 6.5 presents the theoretical implications for task-based research into L2-pedagogy and highlights new aspects of the factors under investigation, ± few elements and ± monologic, with respect to Robinson’s Triadic Componential Framework (Robinson 2005). Section 6.6 makes suggestions for the practice of language teaching and testing. Finally, section 6.7 closes with a summary of the conclusions of this book.

6.2

Theoretical basis

Chapter 1 explained in detail the theoretical framework of the studies presented here. Starting from a general description of the cognitive perspective on task-based L2-performance it focused on the theory under investigation: the Cognition Hypothesis by Robinson (2001a, b, 2003b, 2005). This theory includes a taxonomy of factors that may influence attentional allocation during task-based L2performance. The so called Triadic Componential Framework distinguishes cognitive factors of task complexity from interactive factors of task condition and from learner factors of task difficulty. The

114


following sections will give a summary of the theoretical claims under investigation that address the cognitive and interactive variables named in the Triadic Componential Framework.

6.2.1

The claims of the Cognition Hypothesis

The main goal of this book is to investigate effects of cognitive task complexity on the task-based performance of L2-learners. Following the Cognition Hypothesis there is a distinction between resourcedispersing and resource-directing factors of cognitive task complexity. Robinson (2005) argues that an increase on resource-dispersing variables (e.g., ± planning time, ± prior knowledge) diverts attentional allocation over various linguistic and non-linguistic task aspects. As a result, the linguistic output of L2-performers suffers due to the extra cognitive load of a complex task that focuses the attention on other task features than language. In contrast, the Cognition Hypothesis predicts that an increase on resource-directing cognitive variables, like ± few elements or ± reasoning demands, attracts attention and allocates it towards language. Increased cognitive task complexity by means of resource-directing factors initiates data-driven processes that focus the attention to task relevant linguistic aspects (Robinson 2003a). In terms of measures of linguistic complexity, accuracy, and fluency (in short global CAF-measures) Robinson predicts the following effects on L2-task performance. As L2-learners try to meet the extralinguistic cognitive demands of a complex task they will use more complex linguistic structures and a more varied lexis. Accordingly, the linguistic complexity increases in cognitively complex tasks. Crucially, L2-learners can meet these demands without losing control over attentional allocation because they can rely on several pools of cognitive resources (Wickens 1992, 2007). As a result of the focused attention, their speech performance is more accurate but less fluent. In short, cognitively complex tasks may push actual L2-performance such that linguistic complexity and accuracy increase in parallel at the cost of fluency. More recently, Robinson and colleagues have argued that effects of focused attention to form as a result of an increased cognitive task complexity should become visible in particular on those forms and structures that are specifically relevant for successful task performance (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). As they assume that global CAF-constructs may not be sensitive enough to the task manipulations proposed in Robinson’s framework they suggest complementing them by task specific measures. A further goal of this work is to examine effects of interaction on task-based L2-performance. In the Triadic Componential Framework Robinson (2005) presents different interactive factors of task condition. So called participation variables, like one–way / two–way flow of information, have predictive value for the amount and nature of interaction that is expected during task performance. A radical form of manipulating the flow of information compares monologic to dialogic tasks. In a

6.2 Theoretical basis

115

monologue, L2-learners act on their own such that the flow of information by default is one-way. After all, interaction is not possible in a monologic task. In contrast, a dialogue creates a platform for twoway interactions with frequent turn-taking due to comprehension checks, clarification requests, and negotiations for meaning and form (Ellis 2000, Long 1990). Even though the Cognition Hypothesis makes no overt statements about the factor ± monologic, as explained in section 1.4.4 Robinson’s assumptions about focused attention and interaction (for example based on Long’s 1990 Interaction Hypothesis) may predict the following outcome with respect to manipulations of this interactive factor. On the one hand, the linguistic complexity possibly decreases in dialogues because frequent interactional moves prevent speakers from constructing elaborate linguistic structures. On the other hand, as mutual understanding is crucial, an interactive task presumably focuses attention towards language such that it promotes accuracy at the cost of fluency. The third goal of the present studies is to explore the combined effects of cognitive task complexity and interaction. Robinson (2001a, b, 2005) claims that cognitively complex interactive tasks generate more interaction than cognitively simple interactive tasks because complex tasks presumably need more clarification work, which reduces linguistic complexity. In contrast, both the joint attention to language in a dialogic setting and the higher cognitive task complexity may push accuracy at the cost of fluency in cognitively complex interactive tasks.

6.2.2

Alternative accounts on cognitive task complexity and interaction

As discussed in section 1.3.4 the Limited Attentional Capacity Model predicts a different outcome upon cognitively complex tasks (Skehan 1996, Skehan and Foster 2001). Based on VanPatten (1990) Skehan and Foster argue that attentional resources are limited in capacity such that any increase in cognitive task complexity puts the different dimensions of L2-performance into competition with each other for the available resources. As L2-learners then prioritize either meaning or form, cognitively complex tasks generate trade-off effects. These may emerge in particular between linguistic complexity and accuracy. Section 1.6.2 brought forward an alternative perspective with respect to task-based performance in interactive tasks. Tavakoli and Foster (2008) suggest that monologues may be cognitively more complex than dialogues. They argue that in dialogic tasks the speaking turn of the interlocutor creates online planning time for the hearer. Also psycholinguistic research suggests that dialogic tasks put up a lower cognitive load than monologues because processes of alignment and priming ease language production in interaction (Costa et al. 2008, Meyer and Schvaneveldt 1971, Pickering and Garrod 2004). Alignment may have a twofold effect. As interactants mirror each others’ speech, this may have a decreasing effect on linguistic complexity in interactive tasks. In contrast, the lower cognitive load of a dialogic task will enhance fluency. Furthermore, in order to keep a constant flow of interaction interlocutors tend to ‘help out’ as soon as the partner falls silent, which again results in a higher overall fluency of dialogic tasks

116


(Fiksdal 2000). In sum, these alternative accounts with respect to the factor ± monologic predict that in dialogues linguistic complexity decreases while accuracy and fluency may be pushed by an interactive task setting.

6.2.3

The hypotheses under investigation

In order to test the claims of the Cognition Hypothesis and the alternative accounts, this book evaluates task-based L2-performance on cognitively simple versus complex tasks in either a monologic or a dialogic setting. It examines oral speech production by using both global CAF-measures and a task specific measure, that is the use of conjunctions (see chapter 5 for the rationale behind this choice). Furthermore, the present studies investigate the L2-learners’ task performances in light of L1-speaker data. The hypotheses that guide the empirical investigations in this book with respect to manipulations of resource-directing cognitive factors of task complexity are based on claims of the Cognition Hypothesis.1 The focus lies on manipulations of the factor ± few elements. Following Robinson (2005) Hypothesis 1 argued that increased cognitive task complexity results in a higher accuracy, a higher linguistic complexity but a lower fluency of L2 oral task performance. As the Cognition Hypothesis makes no clear statements about the factor ± monologic the present studies follow the predictions of e.g., Foster and Tavakoli (2009) and Costa et al. (2008) concerning effects of interaction. While cognitive ease may push speech performance with respect to accuracy and fluency, alignment and interactional moves may result in less complex linguistic structures and forms. Accordingly, Hypothesis 2 expected interactive tasks to raise the accuracy and fluency of L2 oral task performance while linguistic complexity may decrease. Following Robinson’s (2005) predictions Hypothesis 3 stated that increased cognitive task complexity enhances interaction and therefore further decreases the linguistic complexity of cognitively complex interactive task performance when compared to task-based performance on cognitively simple interactive tasks. As both factors may push accuracy, cognitively complex interactive tasks were predicted to further increase the accuracy of L2 oral task performance. With respect to fluency, effects of cognitive and interactive factors may compete with each other. As the factor ± monologic is expected to show larger effects that mitigate against the presumably smaller effects of the cognitive factor ± few elements, fluency is expected to increase in dialogues due to the cognitive ease and the tendency of speakers to keep the flow of speech constant. In line with more recent claims of Robinson and colleagues (e.g., Robinson and Gilabert 2007) Hypothesis 4 predicted that increased cognitive task complexity would lead to an increase in the frequency

1 See

section 2.3 on page 40 for the research questions concerning these hypotheses.

6.3 Empirically investigating effects of cognitive task complexity and interaction

117

and occurrence of conjunctions in L2 oral task performance.2 As the empirical studies include a native speaker baseline, the task-based L1-performance was the subject of investigation too. Hypothesis 5 predicted that the highly automatic oral task performance of L1-speakers would not be affected by an increase in cognitive task complexity. In contrast to the cognitive factors of task complexity, natives also were expected to display effects of changes in the interactive task setting though the influence may be smaller than in non-natives. So, Hypothesis 5 predicted that interaction shows similar but smaller effects on L1-speakers than on L2-learners. Hypothesis 5 expected no combined effects of cognitive task complexity by interaction for natives because cognitive task complexity was predicted to have no influence on L1-speakers.

6.3

Empirically investigating effects of cognitive task complexity and interaction

As introduced by chapter 2 chapters 3, 4, and 5 reported on the empirical studies into task-based L2-performance that form the experimental basis of the present book. Referring to the Cognition Hypothesis these studies focused on Robinson’s (2005) claims about effects of cognitive task complexity and interaction both on their own as well as in combination with each other. By means of a 2×2 design the experimental investigations manipulated cognitive task complexity as a within-participant factor and interaction as a between-participants factor. Based on the Triadic Componential Framework (Robinson 2005) they operationalized cognitive task complexity by means of the resource-directing factor ± few elements and interaction by means of the factor ± monologic. The simple tasks addressed only a few elements to be taken into account while the complex tasks concerned many elements. Half of the participants performed both tasks on their own (monologue), the other half acted in pairs (dialogue). Accordingly, all participants performed on a cognitively simple and a cognitively complex task – either both in a monologic or both in a dialogic setting.

6.3.1

Experimental studies

The first study investigated L2-learners’ task performance by means of global measures of linguistic complexity, accuracy, and fluency (see chapter 3). The second study focused on task-based L2performance using global CAF-measures in comparison with an L1-speaker baseline (see chapter 4). The third investigation elaborated study 2 by means of an analysis focusing on a task specific measure, the use of conjunctions (see chapter 5). The following paragraphs first explain each empirical study 2 As the investigations with respect to the task specific measure focused on effects of cognitive task complexity, no specifically task relevant structure based on the factor ± monologic was identified or tested in the present book. Therefore, Hypothesis 4 does not make any predictions about the effects of interaction on the use of conjunctions. Similarly, Hypothesis 4 did not include predictions about combined effects on the task specific measure.

118


in more detail and then summarize the results with respect to effects of cognitive task complexity and interaction on their own as well as in combination. Chapter 3 reported on study 1 where 44 L2-learners of Dutch acted as participants. Their task was to choose between two (simple) or six (complex) electronic devices (mobile phone or mp3-player). The devices differed from each other in seven features (e.g., price, capacity, color). Half of the L2-learners acted on their own, the other half performed in pairs. All of them performed the simple and the complex task (in a counterbalanced order). The oral data were transcribed, and coded for global measures of linguistic complexity, accuracy, and fluency. There were five measures of linguistic complexity, four accuracy measures, and three fluency counts. Three separate statistical analyses per construct (linguistic complexity, accuracy, and fluency) tested by means of a multivariate analysis of variance (MANOVA) for effects of cognitive task complexity (within-participant) and interaction (between-participants) in the L2-speaker data. Chapter 4 reported on study 2. It used a similar experimental design to study 1 but improved the reliability of its measures, analyses, and interpretations based on the findings of this first investigation. Apart from overcoming methodological problems and statistical limitations of study 1 it elaborated the participant groups. In the second study 64 L2-learners of Dutch acted on simple and complex tasks. Half of them performed both tasks in a monologic setting, the other half acted on both tasks in a dialogic setting. In order to induce more speech than in study 1, the tasks addressed a more challenging human topic (Ellis 2000). Participants were asked to identify the best pair out of four (simple) or nine (complex) couples. They could base their choice on six characteristics that the couples differed on (e.g., age, hobby, favorite music style). In order to have a better base to interpret the results attained by L2learners, 44 L1-speakers of Dutch acted as a baseline on the same tasks under the same conditions as the L2-learners. All data were examined by means of global CAF-measures. In both populations three different MANOVAs on each global CAF-construct separately tested for effects of cognitive task complexity and interaction. The analyses included three measures of linguistic complexity, three measures of accuracy, and four measures of fluency. Chapter 5 presented a more elaborate analysis of the L2- and L1-data of study 2 in order to meet Robinson’s suggestion to use task specific measures (e.g., Robinson and Gilabert 2007). The analysis focused on the use of conjunctions as markers of argumentation. As successful task performance depended on a convincing line of argumentation in order to defend one’s choice (the best couple), it was hypothesized that the cognitively more complex tasks including more elements asked for more arguments. Therefore, they were predicted to promote the use of linguistic means that mark argumentation, for example conjunctions. Separate MANOVAs for the learner and native speaker populations investigated effects of cognitive task complexity on the frequency and occurrence of conjunctions. Furthermore, Wilcoxon Signed Ranks tests examined the data for effects of the factor ± few elements on the frequency and occurrence of five specifically task relevant conjunctions.


6.3.2

119

Effects of cognitive task complexity

This section summarizes the findings with respect to effects of cognitive task complexity that was manipulated in the studies at hand by means of the factor ± few elements. Table 6.1 will give a graphical summary of these findings (see p. 123). The analyses by means of global measures (chapters 3 and 4) discriminate structural from lexical measures of linguistic complexity. Neither of the studies yielded an effect of increased cognitive task complexity on structural measures. In contrast, in both studies lexical complexity was raised in complex tasks. In study 1, the complex tasks generated a higher percentage of lexical words than the simple tasks while Guiraud’s Index was not significantly affected. In study 2, Guiraud’s Index showed a significant increase from cognitively simple to complex tasks. The investigations used different measures of accuracy. In study 1 the total number of errors per AS-unit was significantly lowered by an increase of cognitive task complexity. There were no effects on any of the other three error measures. Study 2 found that none of the three accuracy measures (number of lexical, semantic, or determiner errors) was influenced by cognitive task complexity. Fluency showed a significant effect in study 1 such that participants were slower (on the unpruned speechrate) in cognitively complex than simple tasks. None of the other measures of speed, pausing, and repair were affected. Study 2 displayed no effects of cognitive task complexity on any fluency measure. With respect to the task specific measure (chapter 5) there were hardly any significant results. Neither the frequency nor the occurrence of conjunctions was affected by cognitive task complexity. The only significant effect was found with respect to one specifically task relevant conjunction. The occurrence of ‘omdat’/‘because’ was significantly lowered from cognitively simple to complex tasks. Looking at native speaker performance by means of global measures of task performance in study 2, only Guiraud’s Index was significantly higher in cognitively complex L1-tasks (similar to the L2-data). There were no effects of cognitive task complexity on structural complexity, accuracy, fluency, or the use of conjunctions. Again, the only significant effect was found concerning one specifically task relevant conjunction: the frequency of ‘daarom’/‘therefore’ was significantly lowered by an increase in cognitive task complexity. To sum up, in study 1 cognitively complex tasks did yield higher scores with respect to lexical complexity and accuracy, while fluency displayed lower scores. These effects however were significant on three out of twelve sub-measures only. Accordingly, study 1 gives partial support for Hypothesis 1 that predicted a parallel increase of linguistic complexity and accuracy at the cost of fluency due to an increase of cognitive task complexity. In study 2, one out of ten sub-measures (Guiraud’s Index) shows an effect in the predicted direction. As there was no parallel increase on any other complexity or accuracy measure, nor a decrease on the

120


construct of fluency, the results of study 2 do not support Hypothesis 1. In the analysis established by the task specific measure, i.e., the use of conjunctions, there were no supporting effects at all. The only significant influence of cognitive task complexity contradicts Hypothesis 4 (a decrease of the occurrence of one out of five specifically task relevant conjunctions). Concerning native speaker performance, the analysis by means of global CAF-measures shows only one effect of increased cognitive task complexity in study 2 (on lexical complexity). All other CAFmeasures were not influenced. As a whole, these data support Hypothesis 5 that predicted no effects of cognitive task complexity in the native speaker population.

6.3.3

Effects of interaction

In this section the results are summarized with respect to the factor ± monologic, i.e., interaction, that was manipulated as a between-participants factor in the studies presented here. See again Table 6.1 (p. 123) for a graphical overview. In the investigations by means of global CAF-measures structural complexity decreased from monologic to dialogic task performances. Study 1 did not find any effects on lexical complexity, in study 2 dialogues displayed a higher lexical complexity than monologues. Interaction significantly affected L1-speakers’ structural and lexical complexity. Post-hoc pair-wise comparisons revealed that the size and direction of effects in natives and non-natives were different. L1-speakers show a larger decrease on both structural measures and unlike L2-learners also Guiraud’s Index decreases in interactive L1performances. Both studies consistently showed a large decrease in the number of errors in dialogic tasks when compared to monologic performances. This higher accuracy manifested itself on all different error types in both studies in the L2-learners. Unexpectedly, even in the native population dialogic task performances showed a significant gain in accuracy with respect to lexical and determiner errors but not concerning morphosyntactic errors. Interaction revealed a consistent effect in both studies with respect to fluency. Dialogues yielded faster speech, fewer pauses, and also fewer repairs (in study 2 only) than monologues. Mirroring L2learners, native speakers’ fluency increased from monologues to dialogues – with respect to unpruned speechrate A and in pausing behavior. The task specific measure (i.e., the use of conjunctions) was chosen based on its ‘specificity’ for the task with respect to cognitive task complexity. Therefore, the measure was not expected to react to a difference between monologic and dialogic settings. However, the analysis in chapter 5 included interaction as a between-participants factor such that these results are summarized here. As expected, none of the comparisons of monologic versus dialogic speech performances yielded a significant effect on the frequency or occurrence of conjunctions. In contrast to non-natives, native speaker’s frequency


121

of conjunctions was significantly affected by interaction. It was higher in monologues than dialogues. To sum up, most of the data confirm Hypothesis 2 that predicted a decrease of linguistic complexity but an increase of accuracy and fluency in interactive tasks. In study 1 both measures of structural complexity decreased and all error counts and fluency measures increased, while the measures of lexical complexity and repair were not affected. In study 2 lexical complexity showed an unexpected increase from monologues to dialogues. Data on other global CAF-measures confirm Hypothesis 2. As no effects on the use of conjunctions with respect to the factor interaction were expected the null-results for the L2-learners with respect to the task specific measure confirm the hypothesis. According to expectations, an interactive task condition did influence native speech production with respect to global CAF-measures. Even though not all sub-measures were affected the L1-data largely mirror L2-performance. Also the behavior with respect to the task specific measure is comparable in the two populations. There are two differences between L1-speakers and L2-learners. First, lexical complexity is affected in opposite directions in the two groups and second, the frequency of conjunctions does show an effect for L1-speakers but not for L2-learners. Still, in general the data are in line with Hypothesis 5 that predicted similar effects of interaction in natives and non-natives.

6.3.4

Combined effects of cognitive task complexity and interaction

The findings with respect to combined effects of cognitive task complexity and interaction, addressing Hypothesis 3 are given in this section. The graphical summary presents again Table 6.1 (p. 123). The present investigations hardly revealed any combined effects of cognitive task complexity by interaction – neither by means of global CAF-measures (chapters 3 and 4) nor when using the task specific measure (chapter 5).3 In study 1, there was only one significant combined effect of cognitive task complexity by interaction in the L2-learners’ task performance. Cognitively complex tasks in the monologic condition yielded significantly fewer errors and omissions while in the dialogic condition increased cognitive task complexity did not affect the accuracy measures. The only combined effect of cognitive task complexity by interaction that turned out to be significant in study 2 concerns the fluency of native speakers. As revealed by planned post-hoc comparisons, L1-speakers’ speechrates were higher on both measures in simple compared to complex dialogues. In the monologic condition no such effect of cognitive task complexity could be detected. As a whole, these data contradict Hypothesis 3 that expected combined effects of cognitive task complexity and interaction such that complex interactive tasks would further decrease linguistic complexity, while accuracy and fluency were expected to be pushed by both factors. In study 1, the results contradict the predictions because the pushing effects of increased cognitive task complexity found in the monologic task condition disappeared in the dialogic condition. Study 2 did not generate any sup3 Note that for the specifically task relevant conjunctions no such analysis was performed as interaction was not a main goal of the study described in chapter 5.

122


porting effects by means of global measures. Similarly, the analysis using the task specific measure did not confirm the hypothesis. In contrast to L2-learners, L1-speakers showed a combined effect of cognitive task complexity by interaction on fluency. This contradicts Hypothesis 5 that expected the two populations to show similar effects.

6.3.5

Summary of results

In sum, only four out of eleven predicted effects of cognitive task complexity on L2-speakers’ performance manifested themselves in the present studies. As this is based on partial support in study 1 – in the sense that effects were visible on a minority of the different sub-measures – the conclusion is that the empirical data presented in this book challenge Hypothesis 1: An increase of cognitive task complexity by means of the factor ± few elements did not result in a parallel increase of linguistic complexity and accuracy at the cost of fluency. Similarly, its related Hypothesis 4 that expected effects of cognitive task complexity to become even more visible on task specific measures is rejected by the data at hand. In contrast, the experimental investigations concerning effects of interaction support Hypothesis 2. An interactive task condition influenced all tested aspects of task-based L2-performance in the predicted direction with the exception of lexical complexity and repair behavior. Manipulating the factor ± monologic resulted in a lower structural complexity, but higher accuracy and fluency in dialogic than in monologic L2-performances. Unexpectedly, lexical complexity was increased by an interactive setting in study 2 (but not in study 1), while repair behavior was not influenced in either study. The data give no support for Hypothesis 3 that expected combined effects of cognitive task complexity by interaction. Cognitively complex interactive tasks did not further push the accuracy or fluency of L2-performance, nor did they further lower linguistic complexity. Only two results confirm the hypothesis, that is, as predicted there was no combined effect on the frequency nor the occurrence of conjunctions. Cognitive task complexity did not substantially affect native speaker task performance while effects of interaction in the L1-data mostly mirror L2-learners’ performance. Both findings confirm Hypothesis 5. Table 6.1 summarizes these results.


123

Table 6.1: Summary of results of the present studies L2-LEARNERS TASK COMPLEXITY INTERACTION STUDY 1 GLOBAL CAF

MEASURE str. complexitya lex. complexity accuracy fluency STUDY 2 MEASURE GLOBAL CAF str. complexity lex. complexity accuracy fluency TASK SPECIFIC frequency (conjunctions) occurrence spec. task relevant HYPOTHESISb

≈ √ ↑ ( ) √ ↑ ( ) √ ↓ ( )

√ ⇓ ≈ √ ↑ √ ⇑ ( )

≈ √ ↑ ≈ ≈ ≈ ≈ ↓

⇓ ↑ ⇑ ⇑ ≈ ≈ ≈

TASK COMPLEXITY × INTERACTION ≈ ≈ ≈ ≈

√

1&4 not confirmed

√ √ √ √ √

2 confirmed

≈ ≈ ≈ ≈ √ ≈ √ ≈ n.a. 3 not confirmed

L1-SPEAKERS TASK COMPLEXITY INTERACTION STUDY 2 GLOBAL CAF

MEASURE str. complexity lex. complexity accuracy fluency TASK SPECIFIC frequency (conjunctions) occurrence spec. task relevant HYPOTHESISb

√ ≈ ↑ √ ≈ √ ≈ √ ≈ √ ≈ √ ↓ ( ) 5 confirmed

⇓ ↓ ↑ ↑ ↓ ≈ ≈

√ √ √ ( ) √ √

5 confirmed

TASK COMPLEXITY × INTERACTION √ ≈ √ ≈ √ ≈ √ ≈ ( ) √ ≈ √ ≈ n.a. 5 confirmed

√ √ Note. L2 = second language; L1 = first language; = confirmed prediction; ( ) = partially confirmed prediction; Global CAF = global measure of linguistic complexity, accuracy, and fluency; str. complexity = structural complexity; lex. complexity = lexical complexity; task specific = use of conjunctions; spec. task relevant = specifically task relevant conjunctions; n.a. = not applicable; a The predictions did not make a difference between structural and lexical measures of linguistic complexity, however, the results sometimes revealed a different pattern. Therefore, they are listed separately in this table.; b A hypothesis is considered to be confirmed if there are more measures (partially) affected in the predicted direction than measures that are not affected or affected in the opposite direction.

124

6.4


Discussion

This section puts the results of the empirical chapters 3, 4, and 5 in relation to the theoretical framework of the cognitive approach to task-based research presented in chapter 1 and discusses their consequences for the hypotheses formulated in chapter 2.

6.4.1

Cognitive task complexity

The main goal of the present work was to investigate effects of cognitive task complexity on L2-learners’ oral task performance. In the process, the studies evaluated the Cognition Hypothesis (Robinson 2005) that predicted a parallel increase of linguistic complexity and accuracy at the cost of fluency in cognitively complex tasks. As summarized in Table 6.1, the results of the empirical investigations in this book that manipulated the factor ± few elements do not support Robinson’s claims. The only reliable change from simple to complex tasks appeared on the measure of lexical complexity: It was higher in cognitively complex tasks than in cognitively simple tasks. This finding is supported by both studies that employed global CAFmeasures (chapters 3 and 4) and on its own is in line with the Cognition Hypothesis. However, in order to confirm Robinson’s claims also accuracy should be enhanced in cognitively complex tasks (Skehan 2009). Although study 1 revealed a higher accuracy in complex tasks, this finding may be debatable because only one out of four sub-measures was affected significantly. In addition, study 2 does not show any effects on any accuracy measure. The present studies accordingly reject Hypothesis 1. Following the suggestion of Robinson and colleagues, chapter 5 examined the data for the frequency and occurrence of conjunctions as a task specific measure (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). Hypothesis 4 claimed that cognitively complex tasks would promote the use of conjunctions in the cognitively complex tasks. As the results revealed no effect of cognitive task complexity on this measure also Hypothesis 4 is rejected. As a whole, these investigations fail at finding confirmatory results for the Cognition Hypothesis and question Robinson’s predictions. However, before jumping to conclusions, this section highlights four possible explanations for the results at hand. They can be summarized as (1) a quantitative effect of the factor ± few elements, (2) a confound of the resource-directing factors ± few elements and ± reasoning demands, and doubts about (3) cognitive task complexity or (4) limited attentional capacity.

The factor ± few elements generates more speech The studies presented in this book aimed at singling out the factor ± few elements. Therefore, it manipulated the number of elements in a straightforward way. A simple task included a few elements (two elements in study 1; four combinations in study 2) and by adding some more elements a complex ver-

6.4 Discussion

125

sion of the same task concerned many elements (six elements in study 1; nine combinations in study 2). Robinson (2005) states that an increase on the factor ± few elements does change the attentional allocation during task-based L2-performance. He bases his predictions on the cognitive perspective on memory and attention during L2-task performance (see chapter 1). This claim finds support in cognitive psychology as Halford et al. (2007) argue that our cognitive capacity is limited to a maximum of four elements or combinations.4 Accordingly, the simple tasks in the present book addressed a number of elements that were within the assumed limitations whereas the complex tasks included an amount that should be beyond these capacity limits. The data suggest however, that increasing the number of elements did not have the expected effect on cognitive processes. Hardly any significant effects of the factor ± few elements were found. Possibly, ‘just adding more elements’ did not place ‘greater functional or conceptual communicative demands’ (Robinson and Gilabert 2007: 162) on the learner. If the complex task does not put up a greater cognitive load, L2-learners will not show traces of trying to meet the increased demands. That is, no parallel increase of linguistic complexity and accuracy, or a difference in the use of conjunctions, becomes visible in the data. When dealing with many elements in the complex tasks at hand an effective strategy could be to consider one element after the other. Such a linear approach to more elements most likely would not affect the cognitive load of a task. Only if all elements are taken into account at the same time may this produce a higher cognitive load. Meeting the increase in the number of elements by addressing one element after another possibly would manifest itself in longer speech productions. In other words, if talking about nine rather than four options to combine people into pairs, one may just speak more in order to talk about every possible combination. The longer speech samples then would be characterized by similar, recurring linguistic structures and forms rather than showing different linguistic means than when talking about a few elements. This explanation finds support in the data. A consistent finding of the empirical work at hand is that the complex tasks generated more speech than the simple tasks. Looking at raw scores of, for example, the number of subordinate clauses therefore suggests that cognitively complex tasks promote subordination and therefore complexify linguistic production. However, quantifying linguistic complexity, accuracy, and fluency or a task specific form by means of measures that correct for sample length, the differences between the simple and complex task conditions disappear. This explanation may also account for the fact that in both studies that employed global CAFmeasures, lexical complexity was significantly raised from simple to complex tasks. As discussed in chapter 4, the increased number of words in the input of the complex task fits to the manifested increase on Guiraud’s Index from simple to complex task performances. It seems that complex tasks did 4 N.B.

The seminal paper by Miller (1956) assumes a limitation of seven elements or chunks.

126


generate speech with a higher lexical complexity, but, to repeat the earlier explanation, participants in the complex tasks addressing many elements did not produce linguistically ‘different’ output. Apparently they just spoke ‘more’. In line with this assumption, also the data gained by the task specific measure may be explained. As discussed in chapter 5, the analysis by means of the specific measure did not revise any findings of the analysis examining global CAF-measures. Apparently, the tasks addressing more elements did not result in participants using substantially more or a different set of conjunctions. As the raw scores indicate higher numbers (see Tables 5.3 and 5.4), it may be that the longer speech samples of the complex tasks addressing more elements increased the absolute numbers of conjunctions. Correcting them for sample length, however, generates no significant differences any more (see Tables 5.5 and 5.6). It is more likely that the L2-learners relied on recurring linguistic forms, that is they used the same conjunctions more often. Repeating the conclusions of chapters 4 and 5, the studies as a whole therefore suggest that participants produced in the complex versions of the tasks at hand more speech, i.e., ‘more of the same’ language, rather than a linguistically ‘different’ output. Put differently, it seems that the factor ± few elements results in a mere quantitative change of L2-speech production.

A confound of the factor ± reasoning demands Interestingly, earlier work investigating the factor ± few elements did find corroborating results for the Cognition Hypothesis. For example, the work of Gilabert (2007a) and Gilabert et al. (2009) did yield supporting results by manipulating the number of elements in an instruction giving task. Paradoxically, in the same studies this effect was not found in a decision making task. As an explanation chapter 4 hypothesized that in the decision making task there was a ‘confound’ by means of the factor ± reasoning demands. After all, the Triadic Componential Framework includes reasoning as a resource-directing variable (Robinson 2005). Robinson and Gilabert (2007) distinguish even three different types of reasoning demands (causal, spatial, intentional) that focus attention towards different specific linguistic means. Robinson (2007b) found that complex intentional reasoning tasks triggered L2-learners to make greater use of e.g., cognitive state terms, which was the task specific measure in this study. The decision making tasks in Gilabert (2007a) and Gilabert et al. (2009) as well as the present investigations included reasoning tasks that were manipulated by means of the factor ± few elements. It may be, that in these studies all tasks were quite complex in terms of the factor ± reasoning demands. Possibly, if participants act on highly complex reasoning tasks, an additional change of the number of elements does not affect attentional allocation any more. In other words, possibly the factor ± few elements did not differentiate between two tasks that were both complex on the dimension of reasoning.

6.4 Discussion

127

Importantly, the Cognition Hypothesis argues that the factor ± reasoning differentiates between tasks that do and tasks that do not ask for reasoning (e.g, argumentative tasks versus description tasks). ´ esz ´ (in press), who found corroborating results for Robinson’s claims, systematically Earlier work by Rev ´ esz ´ combined increases of the reasoning demands and the number of elements. Rev changed the number of elements in her tasks such that it explicitly and deliberately included a parallel increase of the reasoning demands. In Kuiken et al. (2005) and Kuiken and Vedder (2007b) participants were asked to base a decision on either three (simple) or six (complex) criteria. All of these criteria had to be taken into account in parallel when performing the task. The authors argue that this manipulation of the number of elements almost automatically generated a higher amount of reasoning. Their data are partially in line with Robinson’s claims (see discussion chapters 4 and 5). As a whole, this suggests that the earlier support for the Cognition Hypothesis possibly stems from the factor ± reasoning demands, that was a result of the increase on the factor ± few elements, rather than from the latter factor on its own. Put differently, it may be that in earlier work the factor ± few elements in combination with the cognitive factor ± reasoning demands influenced cognitive task complexity such that it affected task-based L2-performance. The aim of the present studies, however, was to investigate the single factor ± few elements. Even though all the tasks were argumentative, that is the simple and the complex tasks were all + reasoning tasks, great effort was made to increase the number of elements without a ‘confound’ of the amount of reasoning demands. However, it may be that this very fact eliminated effects of increased cognitive task complexity by means of the factor ± few elements. After all, the data of the present studies challenge the Cognition Hypothesis (Robinson 2005) that predicts the resource-directing cognitive factors of task complexity to affect the cognitive processes of L2-learners during task-based performance. More specifically, they question the use of the factor ± few elements as a cognitive factor of task complexity.

Cognitive task complexity? This discussion raises another pending question: How exactly do we manipulate cognitive task complexity? Put differently, how do we know that a task manipulation, e.g., by means of the factor ± few elements, indeed induces a higher cognitive load than its simple counterpart? The operationalization of cognitive task complexity in this book is based on the Triadic Componential Framework (see section 1.4.6 and Robinson 2005). Robinson explains how changes on different task design factors of this taxonomy influence the cognitive complexity of L2-tasks and how this in turn influences task-based L2-performance. As such the framework functions as a research agenda for empirical investigations. But how do we know its assumptions are correct? As discussed in chapter 4 the scores on the affective variables questionnaire of the second study suggest that participants did not perceive a substantial difference in task difficulty between the task

128


addressing a few and the one addressing many elements (see section 4.4). It may be, that the manipulation of the number of elements did not establish the intended higher cognitive task complexity. Again, this supports the idea that the factor ± few elements, as it was manipulated in this book, may not have the potential to directly affect cognitive processes and attentional allocation during task-based L2-performance. Moreover, the studies in chapter 4 and 5 of this book included a native speaker baseline. As explained in section 1.6.4, native speakers can rely on mostly automatic processes of language production. Therefore, Hypothesis 5 expected L1-speakers to show hardly any effects of cognitive task complexity. This assumption concerned global CAF as well as task specific measures of performance. Results of the present studies confirmed Hypothesis 5 as they revealed that indeed, the L1-speakers showed only one effect of the factor ± few elements on global CAF-measures. Like L2-learners, their lexical complexity was higher in cognitively complex than simple tasks. Also regarding the task specific measure L1-speakers mirrored L2-learners’ behavior. There was one effect on one specifically task relevant conjunction in the unexpected direction, that is, it was higher in simple than complex tasks. As predicted, L1-speakers were hardly affected by this task manipulation, which we may explain by the automaticity of their speech processing. But also the L2-learners show hardly any influence of the number of elements in their task performances. The similarities between the two populations therefore point towards the earlier assumption that also for L2-learners the factor ± few elements as manipulated in this book possibly did not directly affect the cognitive load of a task. As it does affect task performance – it generates a quantitative change of task-based production – it may be seen as a factor of general task design that pushes the amount of speech in task-based performance (of L2-learners and L1-speakers) rather than as a resource-directing factor of cognitive task complexity.

Limited attentional capacity? A last possible explanation may come from the alternative account on task-based performance. As the studies do not confirm the Cognition Hypothesis, it may be that the Limited Attentional Capacity Model gives a better fit to the results (Skehan 1996, Skehan and Foster 2001). This model assumes trade-off effects, in particular between linguistic complexity and accuracy upon cognitively complex tasks. However, the present investigations do not point towards trade-off effects. For example, even though the effects may not be significant on all measures the descriptives of study 1 revealed a parallel increase of linguistic complexity and accuracy from cognitively simple to complex tasks (see Table 3.4). As discussed in chapter 4, also the scores in study 2 do not support Skehan’s model. That is, there was neither an increase nor a decrease on any accuracy measure while lexical complexity was higher in the cognitively complex task performances. Consequently, the present studies do not support Skehan and Foster either who predict trade-off effects.

6.4 Discussion

129

Concluding remarks As a whole, the studies at hand indicate that the factor ± few elements on its own (without affecting reasoning demands) may not have the potential to directly influence cognitive processes during taskbased performance. In contrast to the predictions of the Cognition Hypothesis (Robinson 2005), no effects on attentional allocation due to a higher number of elements could be traced. There were no consistent results in cognitively simple versus complex L2-tasks other than an increased lexical complexity. Similarly, no confirmatory effects on a task specific measure were found in the data. It follows that the data do not confirm Hypothesis 1 and Hypothesis 4. In light of the confirmed Hypothesis 5 that expected no effects of cognitive task complexity on L1-speaker data, the factor ± few elements may no longer be characterized as a cognitive factor of task complexity. Apparently, the factor ± few elements can influence the amount of speech produced upon a certain task input: A task concerning + few elements generates a small amount of speech. A task that addresses many elements (– few elements) induces more speech, asks for a wider lexicon and possibly creates more instances and thus learning opportunities for similar recurring linguistic structures. Section 6.5 will elaborate on the theoretical implications of this point.

6.4.2

Interaction

The second goal of the present book was to investigate effects of interaction on L2-learners’ oral task performance. Hypothesis 2 predicted interactive tasks to raise the accuracy and fluency of L2 oral task performance while linguistic complexity may decrease. These predictions were in line with Robinson’s (2005) assumptions about interactive factors of task condition in his Triadic Componential Framework. Accordingly, the factor ± monologic may be seen as a radical form of the factor one-way/two-way flow of information. Robinson (2001b) states that dialogues may reduce the linguistic complexity of task performance because of frequent interactional moves and interruptions. As interaction creates a need for mutual understanding and focuses both speakers’ attention to language, dialogic tasks may push accuracy at the cost of fluency. Alternative perspectives propose that dialogues enhance L2-speech performance because they put up a lower cognitive load than monologues based on more planning time and processes of alignment and priming (Costa et al. 2008, Pickering and Garrod 2004, Tavakoli and Foster 2008). However, linguistic complexity may decrease because of the copying and mirroring of words and clauses through alignment and priming. As a whole, these accounts expect dialogues to show a lower linguistic complexity but a higher accuracy and fluency than monologues. The summary of results (Table 6.1, p. 123) shows consistent effects of the factor ± monologic. The studies confirm Hypothesis 2 concerning effects of interaction. Also Hypothesis 5 that expected similar results in the native population was confirmed. In more detail the data may support a combination of

130


the predictions by Robinson and the alternative accounts with respect to the factor ± monologic.

The factor ± monologic: interactive and cognitive aspects As dialogues resulted in a lower structural complexity than monologues, the data suggest that Robinson’s predictions concerning linguistic complexity were correct. Apparently, turn taking behavior cuts off the production of complex syntactical structures and forms. However, whether this indeed is the reason for the lower structural complexity cannot be answered by the data at hand because only global measures were used to investigate effects of interaction and the analysis did not include countings of, e.g., comprehension checks and clarification requests. Even so, the fact that the average number of AS-units in general was higher in dialogues than monologues lends further support to this assumption. In addition, L1-speakers show a similar behavior as the L2-learners. As addressed in chapter 4, this fact points to the conclusion that a low structural complexity in dialogues possibly is a natural byproduct of interactive speech. In other words, to produce structurally less complex language in dialogues than monologues is what native speakers do. Therefore this may be seen as target language use. Robinson’s perspective and for example Long’s (1985, 1989) Interaction Hypothesis may also explain the higher accuracy in dialogues. The need for mutual understanding may focus both speakers’ attention to form. This joint attention to language predicts a reduction of the amount of errors in interactive settings – which is what we find in the data. However, in contrast to Robinson’s predictions, also lexical complexity and fluency increased from monologic to dialogic tasks in L2-learners. It may be that the alternative accounts on interaction possibly present a better fit to these results (e.g., Costa et al. 2008, Foster and Tavakoli 2009, Pickering and Garrod 2004, Tavakoli and Foster 2008). As discussed in chapter 4, assuming processes of alignment and priming can account for the increase of lexical complexity in the dialogic task condition (Costa et al. 2008). It may be that due to copying and mirroring of words L2-interactants could profit from each other’s lexical input in dialogues. In the monologic condition, participants had to rely on their own lexical knowledge while in the dialogic condition speakers presumably incorporated words of the interlocutor they may not have come up with on their own. Possibly, this kind of priming resulted in an elaborate lexicon. Again, this explanation gains support from the native speaker data. As assumed by Pickering and Garrod (2004) the lexical complexity in L1-speakers decreases in interactive tasks compared to monologues. The authors explain that as the L1-lexicon is very large by itself it may be that through routinization and recycling alignment of vocabulary items in L1-dialogues decreases rather than increases lexical complexity (see chapter 4 for a more detailed discussion of this point). Robinson’s perspective can account for the fewer errors in L2-dialogues by assuming joint focused attention to language. This view may be supported by seminal hypotheses on SLA that predict interaction to promote attention to form (e.g., the Interaction Hypothesis Long 1985, the Output Hypothesis

6.4 Discussion

131

Swain and Lapkin 2001 and the Noticing Hypothesis Schmidt 1990, see section 1.3.2). However, also native speakers showed a gain in accuracy (especially with respect to lexical choices). Therefore, the explanation by Costa et al. (2008), Pickering and Garrod (2004) and Tavakoli and Foster (2008) possibly finds more support. This view predicts a lower cognitive load in dialogues due to online planning time during the speaking partner’s turn. This possibly accounts for the gain in accuracy. Also the fact that dialogues generated more fluent speech than monologues (in both populations) is in line with the assumption of a lower cognitive complexity in dialogues. In sum, both accounts serve partially as an explanation for effects of the factor ± monologic in the data presented here. It seems that apart from being an interactive factor of task condition, the factor ± monologic has a cognitive dimension that possibly affects task complexity at the resource-dispersing dimension.

Concluding remarks Taken together, the studies in this book give a new perspective on the factor ± monologic. As defined in Robinson’s Triadic Componential Framework (Robinson 2005) it may serve as an interactive variable of task condition. However, the present studies highlight the cognitive aspects of this factor, which could be corroborated by the comparison of the L2-learner data with a native speaker population. In sum, the data presented here suggest that the factor ± monologic has a cognitive dimension that may be related to the resource-dispersing factor of cognitive task complexity ± planning time. Again, the theoretical implications of this discussion will be addressed in section 6.5.

6.4.3

Cognitive task complexity and interaction in combination

The third goal of the present book was to explore the combined effects of cognitive task complexity and interaction on L2 oral task performance. Hypothesis 3 followed Robinson’s claims by expecting cognitively complex tasks to even further decrease linguistic complexity but even further increase the accuracy of oral L2-task performance. As effects of interaction were expected to mitigate against the effects of cognitive task complexity, fluency was predicted to be high. Hypothesis 5 expected no joint effects of cognitive task complexity and interaction in native speakers because no main effect of the cognitive factor was expected. Like in L1-speakers, L2-learners showed hardly any combined effects of cognitive task complexity and interaction. Study 1 revealed that the promoting effect of cognitive task complexity on accuracy, that was found in monologic tasks, disappeared in the dialogic setting. Study 2 found no joint effects with the exception of native speaker fluency. As summarized in Table 6.1 the studies therefore do not support Robinson’s theory. In light of the discussion of the factor ± few elements this is not surprising. How could almost no effects (of the cognitive factor of task complexity) create combined effects with the

132


factor interaction? Based on the discussion about the cognitive aspects of the factor interaction it may be, though, that factors of cognitive task complexity can generate positive synergies with interaction. If dialogues indeed are cognitively simpler than monologues, an increase of cognitive task complexity by means of a resource-directing factor in an interactive setting possibly can benefit from the freed attentional capacity in the dialogue. In other words, if interaction eases the cognitive load during task performances and consequently makes more attentional resources available, it may be that more attention can be focused on task relevant linguistic forms by increasing the cognitive task complexity by means of a resourcedirecting factor.5 Future research may reveal whether such synergies exist. It seems, however, that the resource-directing dimension then should be manipulated on other factors than the number of elements as the work in this book does not support Robinson’s claims of combined effects of cognitive task complexity and interaction.

6.4.4

Global versus specific measures of task performance

In the process of investigating effects of cognitive task complexity and interaction the present work aimed at evaluating the use of global versus specific measures of performance. Robinson and colleagues postulated that global measures possibly fail at detecting slight differences due to task manipulations and therefore called for task specific measures (Cadierno and Robinson 2009, Robinson et al. 2009, Robinson and Gilabert 2007). The work at hand does not confirm the added value of investigating specific structures and forms as the findings by means of the task specific measure did not change the results obtained by the global CAF-constructs. This finding suggests, on the one hand, that focusing on task-specific measures may be more useful when one is interested in L2-development as they probably give a better insight in how ´ esz ´ in press). task performers develop over time (Norris and Ortega 2009a, Rev On the other hand, when interpreting the data a limitation of the present studies must be considered. The analysis in chapter 5 did evaluate only one task specific measure, that is, the use of conjunctions as a marker for argumentation in simple and complex tasks manipulated by means of the factor ± few elements. Even though the present studies intended to strictly manipulate only the number of elements, it is possible that there is a confound of the two factors ± few elements and ± reasoning. That is, both the simple and the complex tasks were argumentative tasks. As conjunctions are seen as lexical markers of argumentation, their use may be in particular sensitive to manipulations of the factor ± few elements in tasks that involve reasoning even if the number of elements was manipulated. Possibly, other task specific measures that can be related to the factor ± few elements may show the predicted 5 For example, it is worth noticing that in chapter 5, there is an apparent approximation of the L2-performances to the native speakers’ use of the specifically task relevant conjunction ‘als. . . dan’/‘if. . . then’ (see Tables 5.7 and 5.8 on page 107). As no statistics were calculated with respect to this point the data at hand cannot give a conclusive answer. Even so, a trend on the frequency of all conjunctions (cf. Table 5.6, p. 106) possibly supports this idea.

6.4 Discussion

133

pushing effect on L2-performance (e.g., the use of attributive adjectives or relative clauses as proposed by Robinson 2001a, 2005). Future work using specific measures in tasks manipulating the number of elements that do involve reasoning may focus on these measures accordingly. For the foremost goal of the task-based research presented here the global CAF-measures, however, may suffice in order to understand how manipulations of task characteristics influence task-based L2-performance. After all, global constructs are thought to catch the most important facets of L2-speech (Housen and Kuiken 2009). Moreover, CAF-measures are widely used in research and therefore allow comparisons over tasks, populations, and settings. For example, the comparison of the data of nonnative and native speakers that reveals as expected that L1-speakers outperform L2-learners. These observations highlight that global CAF-measures give valuable insights for example when evaluating the speech performances in monologic versus dialogic conditions. This fact may serve as a validation for the use of global CAF-measures, while the consistency in the data over studies and populations may confirm their reliability.6

6.4.5

The benefits of a native speaker baseline

Taken together, the present studies show how valuable it is to include a native speaker baseline in research into task-based L2-performance. At all times, the L2-performance could be interpreted in light of L1 task specific performances. This presumably has more confirmatory strength than evaluating L2-speech by means of an external standard (which is often based on prescriptive written norms). For example, the similarities between the two populations with respect to effects of the factor ± few elements led to the assumption that this factor may not directly affect the cognitive load of a task. As the non-native experimental group and the native control group displayed comparable gains in lexical complexity this interpretation was corroborated. Without the control group, such an interpretation possibly would be hard to defend. Also the interpretation of effects of the factor ± monologic gained from having the native speaker baseline. For example, the disparities between native and non-native speaking performances concerning lexical complexity in L2- and L1-interaction served as an explanation for the effect in L2-learners. To recap, an interactive task setting decreased the lexical complexity of native speakers but increased it in L2-learners’ performances. The comparison of the L2- and L1-data allowed the assumption that dialogues put up a lower cognitive load than monologues (see the discussion in chapter 4). Similarly, the interpretation of the interactional aspects, i.e., turn-taking, interruptions, clarification requests, were corroborated by the native speaker data. For example, the decrease of structural complexity due to the dialogic task setting was larger in natives than in non-natives, which led to the inter6 N.B. The examination by means of the task specific measure did not intend to test for effects of interaction therefore this factor cannot contribute to the discussion for or against the use of specific measures of task performance when evaluating monologues versus dialogues.

134


pretation that a decrease of structural complexity is a natural byproduct of interactive speech rather than a drop in performance in dialogues. In sum, the L1-baseline corroborated tentative explanations of the L2-data and as such helped understanding the L2-learner results.

6.5

Theoretical implications

This section discusses the theoretical implications by evaluating cognitive and interactive factors of task design in light of Robinson’s Cognition Hypothesis and formulates directions for future research based on the present studies.

6.5.1

Cognitive and interactive factors of task design

As discussed in section 6.4 the present book generated insights with respect to the main interventions, i.e., cognitive task complexity and interaction. With respect to Robinson’s Triadic Componential Framework and the Cognition Hypothesis the findings point to a change of the predictions related to the factor ± few elements and the factor ± monologic. After all, the data suggest that both factors possibly address other aspects than expressed in the original framework of Robinson (2005). As a summary of the discussion in sections 6.4.1 and 6.4.2, Table 6.2 (on page 135) displays the proposed changes. First, summarizing section 6.4.1, the present work suggests that the factor ± few elements may be characterized as task design factor that affects the amount of speech while attentional allocation may not be directly influenced by manipulating this factor. Adding more elements to a simple task possibly has the result that the – few elements task creates more speech than the + few elements task. Accordingly, manipulating the number of elements may create the single CAF-effect of yielding a higher lexical complexity. Based on the assumption that producing more speech can be related to more recurring but similar structures and forms, no other effects of an increase in the number of elements may be found upon manipulations of the factor ± few elements – as long as measures correct for sample length. Second, as a summary of section 6.4.2, the studies in this book propose that the factor ± monologic possibly has two sides. In terms of Robinson (2005) it is related to (a) interactive factors of task condition, e.g., the participation variable one–way / two–way flow of information. Manipulations of this factor accordingly result in frequent interactional moves, which induce a lower structural complexity while the joint attention to language presumably promotes lexical complexity and accuracy. In addition, the factor ± monologic may be related to (b) resource-dispersing factors of task complexity. The present studies suggest that an interactive task setting creates natural pauses for speakers during the turn of the interlocutor which increases the amount of planning time (Tavakoli and Foster 2008). In addition, processes

6.5 Theoretical implications

135

Table 6.2: Summarizing effects of the factors ± few elements and ± monologic Factors

± FEW ELEMENTS

± MONOLOGIC

CAF

≈ structural complexity

− structural complexity

effects

+ lexical complexity

+ lexical complexity

≈ accuracy

+ accuracy

z

≈ fluency }|

{

z

+ fluency }|

{

Possible

• more speech

• more interactional moves

• time for online planning

expla-

• more lexical items

• interruptions & clarifications

during interlocutors’ turn

nations

• similar, recurring

• negotiation of meaning

• priming and alignment

and form

on all linguistic levels

• joint attention to language

• freed attentional capacity

In terms of

TASK CONDITION

TASK COMPLEXITY

Robinson

interactive factors

cognitive factors

(2005)

(a) participation

(b) resource-dispersing

linguistic structures • hardly affecting attention

of priming and alignment may decrease the cognitive load in dialogic tasks (Costa et al. 2008, Pickering and Garrod 2004). In general, dialogues therefore may free attentional capacity such that it pushes L2-performance.7 As a whole, the studies presented here suggest that these two aspects of the factor ± monologic together affect speaking performance. Possibly, routinization and copying, that are associated with alignment, increase lexical complexity of task-based L2-production while the turn-taking behavior reduces structural complexity. Accuracy and fluency are fostered in dialogic performances. Finally, in light of the discussion in section 1.4.6 it is questionable whether these additions to Robinson’s Triadic Componential Framework make the model more feasible. In contrast, the results may rather give support to the critique by Kuiken and Vedder (2007b) and Ellis (2009). That is, the existing proposals of the framework (e.g., Robinson 2005, 2007b) create a large variety of different manipulations and research designs, so that it may be difficult to find consistent results. The results of the present studies therefore possibly support the call for more precise explanations how exactly researchers could operationalize and weight the different factors named in the Triadic Componential Framework. Even so, future research, which uses the Triadic Componential Framework as a research agenda, can benefit from the considerations concerning the factors ± few elements and ± monologic as it was found in the present studies. 7 N.B. In this analysis pragmatic aspects of the factor ± dialogic are not taken into account. As addressed in chapter 1, it may be that the pragmatic conventions needed in a dialogue increase the cognitive load.

136

6.5.2


Manipulating cognitive task complexity

Considering the explanations for the results of the data with respect to the cognitive factor ± few elements, several suggestions for future research can be formulated. Two of them are named more explicitly concerning (1) the confound of the factor ± few elements and ± reasoning demands and (2) the need for a better definition and investigation of cognitive task complexity. As discussed, it may be that the earlier research confirming Robinson’s claims with respect to the factor ± few elements suffered from a confound of the associated factor ± reasoning demands. Kuiken and Vedder (2007b) argue that every increase of the number of elements automatically induces a larger amount of reasoning. The studies at hand tried to eliminate this confound. Accordingly, their findings seem to indicate that the mere addition of some elements may not affect reasoning demands per se. Rather, they suggest that the factor ± few elements can be seen as a factor of task design that influences the amount of speech. Future research could systematically investigate which of the two analyses receives more support by empirical data. Importantly, the present work suggests that with respect to factors of cognitive task complexity there is a need for an objective way to empirically determine the cognitive load of a task. Based on the Cognition Hypothesis there is a body of studies that have established an empirical base. However, Robinson (2003a) defines cognitive task complexity as the amount of cognitive processing that is needed in order to perform successfully on a task. As such it is dependent on task inherent features and characteristics that increase or decrease the mental effort needed, which in turn affects task performance as established by for example CAF-measures. Norris and Ortega (2009b) point out the circularity in this definition: How can we objectively determine the inherent cognitive load of a task, the independent variable, if it is defined by successful task performance, the dependent variable? Therefore, future research may aim at having some external means that confirm the theoretically assumed cognitive load even if the foremost goal is to test its effects. Norris and Ortega (2009a) name several possibilities to do so, e.g., gauge perceived task difficulty or use imaging techniques to register brain activity. Another solution would be to establish beforehand the amount of cognitive processing that is needed for a certain task. Considering other fields of research (computer science, cognitive psychology) possibly gives fruitful suggestions in the search for how to determine the intrinsic cognitive load of a task.8

6.5.3

Manipulating interaction

With respect to manipulations of the factor ± monologic there are at least three major implications of the present work that are worth mentioning. Also some limitations of these studies may serve as directions 8 For example Cognitive Load Theory (Kirschner, Kester, and Corbalan 2010, Sweller 1988) or Computational Complexity Theory (Goldreich 2008, Papadimitriou 1994).


137

to future research. First, as this is one of the few investigations that have systematically manipulated the factor ± monologic, there is a need for more work looking at the difference between task-based performance by one person on his or her own in contrast to task-based interactions of pairs or even groups of L2-learners on the same tasks under the same conditions. In the process, more attention could be given to the difference between learner-learner-interactions and native-learner-interactions. Moreover, it may be a limitation that, when coding dialogues, the studies presented here looked at a single speaker’s individual performances. A future examination may choose to see how the CAFmeasures look when the dialogic speech performances are analyzed based on the joint dialogic performance. Second, the studies presented here adopted a quantitative perspective on data processing. As a result, speech performances were evaluated by means of quantitative measures, e.g., global measures of linguistic complexity, accuracy, and fluency. Even though some global measures (e.g., accuracy) may have a ‘communicative’ aspect, it would make good sense to include judgments on communicative success (like e.g., Ferrari 2009, de Jong et al. 2007, Kuiken et al. 2010, Pallotti 2009). In addition, it would be interesting to have a more qualitative look on interactions to find out whether the variety in type of interaction differed between the simple and complex performances. Especially from a communicative account like the task-based approach, studies should be more interested in whether L2-speakers managed to achieve the communicative goal of a task. This is a limitation of the work at hand. Third, when focusing on the factor ± monologic, the present book suggests to take into account the cognitive aspects of interaction (see also Tavakoli and Foster 2008). The Triadic Componential Framework Robinson (2005) assumes interactive factors to be factors of task condition. The present studies, however, indicate that this factor seems to affect cognitive task complexity too. It would be interesting to take a more detailed look into interactive task-based performance by addressing its effects on the cognitive load of a task, for example, by investigating the amount of online planning time. Also when considering effects of alignment and priming, the present work calls for other measures (and possibly different experimental techniques) that would catch the cognitive aspects of interaction. Possibly, the existing tools (e.g., counting LREs) may not be sufficient. With regard to differential effects on non-natives and natives it would be in particular valuable to find out whether, and if so how exactly, alignment and priming ease L2 dialogic speech. Systematic investigation of these issues sounds promising also in light of earlier work addressing alignment (Costa et al. 2008, Pickering and Garrod 2004). Recently, CAF-constructs have been interpreted in terms of Levelt’s (1989) blueprint of the speaker that focuses on the underlying processes during speech production (Ellis 2009, Skehan 2009). Skehan (2009), for example, relates processes of message generation in the conceptualizer to measures of lexical and structural complexity while accuracy and fluency (of lexical retrieval) are accounted for by

138


the formulator. This perspective can partially account for the results of the present studies. The (visual) input of the argumentative tasks at hand required the use of some infrequent lexical items. Possibly, the lexical retrieval of these items swallowed most of the attentional capacity of the formulator. As a result, accuracy suffered – especially in the complex version where more lexical items were compulsory for successful communicative task performance. It may be that this effect mitigated against accuracyraising effects of a higher cognitive task complexity. Especially in light of a native speaker baseline, this interpretation gains support. In contrast to L2-learners, L1-speakers have no problems meeting the functional demands of the conceptualized message with their native language competence. As they need hardly any attention for lexical retrieval, structural and lexical complexity accordingly show the same tentative directions – which is what we see in native speaker’s performances comparing e.g., monologues and dialogues.

6.5.4

Manipulating cognitive task complexity by interaction

Concerning Robinson’s claims with respect to a combined effect of cognitive task complexity and interaction there is definitely more work to be done. It may be worth researching whether indeed an interactive task setting frees attentional capacity such that increased cognitive task complexity in complex interactive tasks has more attention available that in turn may be focused towards task relevant aspects (as was hypothesized in section 6.4.3). Again, such an investigation would benefit from measures that capture the cognitive nature of interaction.

6.5.5

Measuring task-based L2-performance

Based on the findings but also limitations of the present studies there may be some suggestions for future research concerning measures of task performance. The following paragraphs address each global construct (linguistic complexity, accuracy, and fluency) separately before highlighting some general points regarding measuring L2-performance. First, concerning linguistic complexity, the comparison of native and non-native speech with respect to structural complexity reveals that a higher score does not always imply a more native-like behavior (see discussion in chapters 4 and 5). This corroborates Pallotti (2009) who postulates that researchers should be careful when interpreting ‘higher’ scores on a CAF-measure as ‘better’. Second, with regard to accuracy the present studies reveal how the interpretation of the accuracy measures gained from the comparison of the L2- with the L1-data. For example, the data showed that native speakers make mistakes if errors are defined as a violation of a (written) standard form. Therefore, future L2-error coding may preferably use adequate oral task-based L1-performance as a reference point.


139

Third, with respect to fluency the division into the sub-measures of speed, silence, and breakdown fluency as proposed by Tavakoli and Skehan (2005) seems to address the intuitive dimensions of this construct. Even so, de Jong and Wempe (2009) have introduced an automatic syllable counter into the program PRAAT (Boersma and Weenink 2006). This allows a quick and easy determination of the phonation time ratio, a measure that subsumes the fluency measures of speed and silence. The present work has shown that sometimes it may be difficult to distinguish between different measures because they are intertwined with each other. For example, breakdown fluency is related to accuracy as repair is triggered by an error. A non-error repair presumably is induced by problems in lexical retrieval, which in turn is related to pausing behavior. As a whole, study 2 of this book improved the reliability of its measures, analyses, and interpretations as it overcame some methodological and statistical limitations of study 1. The comparison of the two studies therefore reminds us to be cautious, for example, with interpretations based on one out of many measures (see also Pallotti 2009). Similarly, studies may gain in statistical power from calculating statistics on measures that are corrected for sample length, from testing measures for collinearity and redundancy and giving effect sizes (as suggested by Norris and Ortega 2003, 2009a). Luckily, computer applications allow less time-consuming ways such that we can use more reliable performance measures. For example, CLAN (MacWhinney 2000) has the possibility to measure linguistic complexity by means of the lexical measure D (Malvern and Richards 1997) rather than the type-token-ratio. In sum, the present investigations show a need for new measures that take into account the cognitive aspects of language in use. Even so, adding new measures may not be the solution either. As addressed by Pallotti (2009), ideally researchers should aim at generating a small but reliable set of measures that are used by most studies. After all, ‘what we measure in our individual primary studies will need to be understood in relation to what is being measured in other studies of CAF in other learning contexts’ (Norris and Ortega 2009a: 556). One step into this direction may be to always include a native speaker baseline when investigating task-based L2-performance. In general, the present studies show how valuable it is to collect native speaker data. As discussed in section 6.4.5 the L1-data many times served as a corroboration of tentative assumptions based on the L2-findings. In addition, comparing L2-performance with L1-performance on the same tasks under the same conditions probably creates a more reliable base for interpretations. In terms of Daller et al. (2003) who distinguish internal and external measures of lexical complexity (cf. section 1.5.2), having native speaker baseline data may produce a ‘text-internal’ (or at least a task specific) reference for any kind of measure. Therefore, native speaker data give important insights for the interpretation of L2-performance measures. Finally, a limitation of the studies in this book is that they measure L2-performance at a certain moment in time. Accordingly, it is difficult to say anything about L2-development. An aim of task-based research, however, is to give implications for L2-pedagogy. Especially the theory under investigation,

140


the Cognition Hypothesis, presents itself as a guideline for the sequencing of L2-tasks in order to push L2-development (Robinson and Gilabert 2007). Pallotti (2009) warns against making interpretations about development as long as no longitudinal data are collected. Yet, apart from Ferrari (2009) up to now not much work has been based on investigations over time. As these few studies furthermore focus on the evaluation of CAF-constructs, there is a need for studies that investigate developmental data – in particular evaluating Robinson’s predictions about sequencing effects on interlanguage development (Robinson 2010).

6.6

Practical implications

Even though the studies in this book do not share the pedagogical aim of the Cognition Hypothesis its results may contribute to L2-pedagogy. This section will highlight some practical implications for the L2-classroom.

6.6.1

Challenging tasks may include many elements

The studies presented here fail at giving confirmatory results to Robinson’s (2001b, 2005) Cognition Hypothesis. Accordingly, a task including many elements did not push L2-performance with respect to accuracy and linguistic complexity in parallel at the cost of fluency. Only a gain in lexical complexity could be attested. Turning this conclusion around, there is an interesting implication for the practice of task-based language teaching. After all, the higher number of elements in complex tasks did not harm L2-performance either. That is, also in the task addressing many elements the measures of structural complexity, accuracy, and fluency were stable while lexical complexity increased. This means that L2-learners can work on tasks concerning many elements without showing trade-off effects on other measures, by means of, for example, using simpler syntactic structures, making more errors, or suffering from fluency problems. A suggestion for language teaching then is that it is good to have tasks with many different elements. At least, the studies at hand show that while L2-learners use similar structures, make similar errors, and keep their pace of speaking, they are able to elaborate their lexical complexity in a task with many elements. In addition, a task addressing many elements made participants speak more. Similarly, the tasks in study 2 generated more speech than the set of tasks in study 1. Probably the fact that in study 2 the task topic concerned human beings rather than inanimate electronical devices was the reason for this finding (Ellis 2000). Another explanation may be that the tasks in study 2 were more challenging because participants were asked to argue for a combination of people. Probably, combining-tasks are more interesting than giving advice for one single item. In sum, the studies at hand therefore suggest to favor tasks in the classroom that challenge L2-

6.6 Practical implications

141

learners, e.g., because they address many different items that need to be paired or combined. In the end, more speech also means that L2-learners have more instances of recurring linguistic forms or structures they use. Accordingly, challenging tasks that induce more speech are predicted to create more chances for L2-learners to practice and eventually learn task relevant language.

6.6.2

Learning and testing by monologic and dialogic tasks

With respect to the factor interaction, the studies in this book point towards a pushing effect of interactive tasks. In dialogues, L2-learners used a wider range of lexical items, they made fewer mistakes, and were more fluent, while monologues let task performers produce more structurally complex language. This suggests that in the L2-classroom mostly dialogues should be used for actual speaking tasks. After all, even if they are still learning, students apparently can perform at higher levels of their L2 when they act in pairs. In contrast, monologic settings probably serve as a platform for creating complex syntactic structures. Also for the practice of grading L2-learners’ task-based performance this finding is important. Most oral language testing is based on individual assessment sessions. The studies in this book, however, suggest that L2-learners possibly display more of their competences when they act in pairs. Ideally, oral performance testing therefore should include a monologic and a dialogic part. It would give second language learners the chance to show their structural competence during monologues and their knowledge with respect to lexical complexity, accuracy, and fluency in dialogues. In addition, a dialogic setting would give insights into the pragmatic abilities of L2-learners.

6.6.3

Sequencing tasks

Combining the findings with respect to the factor ± few elements and the factor ± monologic, the results of the studies at hand may be used for the sequencing of tasks. ‘The fundamental pedagogic claim of the Cognition Hypothesis is that pedagogic tasks should be designed, and then sequenced for learners on the basis of increases in their cognitive complexity’ (Robinson and Gilabert 2007: 162). Robinson (2007a, 2010) claims that the L2-learner will gradually approximate target language use in real life by performing a sequence from cognitively simple to complex tasks that are manipulated on resourcedirecting and resource-dispersing variables. Table 6.3 on page 142 presents a possible sequence of tasks that is likely to gradually approach language use outside the classroom. It is based on the results of the present studies that point to some concrete manipulations how to raise cognitive task complexity by means of resource-directing and resource-dispersing factors. First, one may start with a simple task, including only a few elements such that L2-learners can get familiar to a task. In a monologic setting with plenty of pre-task planning time this possibly makes L2-

142


Table 6.3: Example of task sequencing (following Robinson 2010) DIMENSIONS OF

SIMPLE

COMPLEX

TASK DESIGN

step 1

step 2

step 3

step 4

step 5 (≈ real life)

± few elements

+

-

-

-

-

± planning time

+

+

-

-

-

± no reasoning demands

+

+

+

-

-

± monologic

+

+

+

+

-

structural

lexical

fluency

accuracy &

linguistic complexity,

complexity

complexity

ling. comp.

accuracy & fluency

CAF-EFFECT

learners produce language that, on the one hand, is characterized by a low lexical complexity, accuracy, and fluency. On the other hand, the unlimited time gives the opportunity to build complex linguistic structures. The second step, would confront L2-learners with an increased number of elements such that the L2learners produce more speech and use a wider range of lexical items with similar syntactic structures. In a third round, pre-task planning time is removed, L2-learners will have to produce the earlier rehearsed complex structures under online time pressure, which enhances fluency. This increases task complexity by means of a resource-dispersing factor which may affect the automatization of earlier internalized linguistic means Robinson and Gilabert (2007). The fourth manipulation raises the reasoning demands of the task (e.g., from describing photographs of people to a task where participants need to give arguments for a pair of people). This increase on a resource-directing factor requires the use of structurally more complex features. The focused attention to language results in a parallel increase of the accuracy. Finally, task performers act in pairs. The results of the present studies suggest that a dialogic task performance creates online planning time while alignment and priming lead to a more elaborate lexis that is accompanied by greater accuracy and fluency. In the end, L2-learners are able to produce, even upon a complex task that includes many elements, a large amount of speech that is elaborate concerning lexical complexity, accuracy, and fluency due to the interactive setting. Because of the earlier training phases from step 1 to 4 on their own, the speech performance is also structurally complex. Sequencing tasks according to the proposed order gradually brings L2-learners towards real life task performance. Consequently, frequent task cycles of this kind present an L2-learning strategy that possibly pushes L2-development.9 The table shows, what task characteristic may promote what dimension of task-based L2-performance. That is: planning time may enhance all CAF-dimensions; a higher number of elements apparently increases lexical complexity; increasing the reasoning demands 9 By requiring task performers to finally present (as a monologue) their findings to the whole class, these claims concerning task sequencing possibly serve as a theoretical justification for the think-pair-share-method of teaching (Lyman 1987).

6.7 Summary

143

possibly pushes lexical and structural complexity; monologues promote the use of complex linguistic structures while dialogues push accuracy, lexical complexity, and fluency. Finally, in the classroom it may not be practicable to consider all these steps because teachers may not have their students perform the same task more than once. Therefore educators could chose to include or leave a manipulation and use different tasks with the same underlying structure, like the dating- and study-tasks used in study 2 (see chapters 4 and 5 and appendix B.4).

6.7

Summary

The data of the present book challenge Robinson’s Cognition Hypothesis. The factor ± few elements, as manipulated here, did not result in a parallel increase of linguistic complexity and accuracy at the cost of fluency, nor were there any effects on the task-specific measure. In addition, there were no supporting combined effects of cognitive task complexity by interaction. In light of the mere raise of lexical complexity, the discussion argues that by only increasing the number of elements a task may not affect attentional allocation during task performance as predicted by the Cognition Hypothesis. Rather than a different language use, the manipulation resulted in ‘more of the same’ language, i.e., a quantitative change. The factor ± few elements accordingly may not have the potential to affect the cognitive load of a task and direct attentional resources towards task relevant linguistic means. In contrast, considering the consistent effects of interaction, this work suggests that the factor ± monologic, apart from being an interactive factor, has a cognitive impact too. Manipulated by means of the interactive factor of task condition, dialogues reduce the structural complexity of task performance presumably because they induce frequent interactional moves and interruptions. Taking a cognitive perspective on interaction, the factor ± monologic may be related to resource-dispersing factors of task complexity, like ± planning time. A lower cognitive task complexity induced by a dialogic task therefore may result in a pushing effect with respect to lexical complexity, accuracy, and fluency. The use of a task-specific measure did not change the interpretation of the data based on global CAF-constructs. In contrast, the availability of a native speaker baseline helped greatly to interpret the data of the present studies. In sum, the perspective adopted in this book highlighting cognitive and interactive aspects of taskbased production, seemed to be a productive way to evaluate the oral performance of L2-learners.

Appendix

A

Appendices to study 1

146

A.1


Proficiency task

The proficiency task in study one was a cloze task where every 11th word was deleted. Participants received 30 minutes time to fill in a total of 50 gaps. The original text was taken from a Dutch woman’s magazine (Libelle, 2005, nr. 52, p.37).

gatentekst

deelnemersnummer ........

Beste deelnemer, Hieronder vind je een tekst waarin op een aantal plaatsen gaten zijn gevallen. Het is de bedoeling dat jij uit het zinsverband probeert af te leiden welk woord op de puntjes zou kunnen staan. Je moet telkens precies ÉÉN WOORD invullen. De woorden zijn soms kort en soms lang. De tekst gaat over de verandering van het klimaat op de aarde. Het onderstaande interview met Rob van Dorland, meteoroloog aan het KNMI, gaat over de gevolgen hiervan voor Nederland. Vul in het lijstje rechts het juiste woord in. Je hebt maximaal 30 minuten de tijd. Succes ermee! Ons klimaat wordt warmer en natter New Orleans, Pakistan, Mexico, Katrina, Rita en Wilma: natuurrampen lijken het afgelopen jaar aan de orde van de dag. Regelmatig worden we met rampscenario’s om de oren geslagen. 1 dat er over vijf jaar vijftig miljoen mensen

De Verenigde Naties dakloos zullen

2 door milieurampen. Kunnen we in de toekomst

meer natuurgeweld verwachten

3 wat betekent dat eigenlijk voor

Nederland? We vragen het aan Rob van Dorland

1................................... 2................................... 3...................................

4 zich bij het

4...................................

5 we inderdaad meer te maken gehad met

5...................................

KNMI bezighoudt met klimaat-veranderingen. Het afgelopen jaar natuurrampen dan in

6 jaren. Deels kunnen wij mensen daar niets

6...................................

aan doen. Aardbevingen

7 tsunami’s zijn bijvoorbeeld een gril van

7...................................

de natuur. Voor orkanen

8 dat iets ingewikkelder. Of het er meer of

8...................................

minder worden, waarschijnlijk wel dat is de

9 we niet, maar ze worden in de toekomst 10. Dat komt doordat de aarde warmer wordt en

10.................................

11 het gevolg van menselijk handelen. De verwachting is dat

11.................................

12 de komende eeuw tussen de twee en zes graden stijgt.

13 boosdoeners zijn vooral de gassen (CO2) die vrijkomen door het verkeer

9...................................

14 de industrie. Deze gassen komen in de atmosfeer

terecht waardoor

15 warmer wordt. Warmere lucht bevat meer

vocht en daardoor zal

16 wereldwijd meer neerslag vallen. Hierdoor

ontstaan er meer overstromingen. Dat hebben, komt ook door de keuzes helft van de wereldbevolking

17 daar zo’n last van

18 we zelf maken. Zo woont de

19 een kustgebied. Dat heeft

economisch gezien misschien allerlei voordelen, maar

20

12................................. 13................................. 14................................. 15................................. 16................................. 17................................. 18................................. 19................................. 20.................................

maakt ons wel kwetsbaarder.

z.o.z.

→

A.1 Proficiency task

147

gatentekst

deelnemersnummer ........

Voor orkanen hoeven we in Nederland te maken met wateroverlast ons land gedeeltelijk al onder meters worden als we niets

we nu bang

24 zeespiegel ligt. Deze eeuw zal het

24................................

25 dat kunnen in de toekomst

25................................

26. Nu zijn het nog onder water gelopen

26................................

28 flink ophogen. Dat betekent overigens niet dat

29 zijn voor een vloedgolf, zoals in New Orleans. Onze

dijken zijn

30 stuk beter dan in New Orleans. Bovendien stijgt het

zeewater hier

22................................ 23................................

27 misschien niet meer in Nederland blijven

kelders - dan kunnen

21................................

23 dat is behoorlijk zorgwekkend omdat

zeewater 10 tot 90 centimeter stijgen,

wonen tenzij we de

21 bang te zijn. Die komen in

22 we krijgen waarschijnlijk wel steeds meer

ons gebied niet voor.

31 langzaam en we hebben nog tijd genoeg om er iets

32 te doen.

27................................ 28................................ 29................................ 30................................ 31................................ 32................................

Maar we merken er wel al iets van het ook bij ons vooral

33 de temperatuur stijgt. Zo zal

34 de winter meer gaan regenen. De kans op

33................................ 34................................

een koude

35 met een Elfstedentocht wordt steeds kleiner. In de

35................................

zomer valt

36 juist minder neerslag, maar als het regent, zijn het

36................................

steeds

37 stortbuien. Dat hebben we vorig jaar in juli en augustus

38 meemaken.

37................................ 38................................

We kunnen de opwarming niet meer stoppen, maar er zorgen dat de opwarming vrij laag blijft. Hiervoor moeten

39 voor 40 heel

39................................ 40................................

goede afspraken maken over het verminderen van schadelijke gassen. 41 werken veel landen daar niet genoeg aan mee. Eigenlijk is

42 een geluk bij een ongeluk dat olie schaars is en

43 duur. Hierdoor gaan grootverbruikers zoals Amerika zoeken naar alternatieven die

41................................ 42................................ 43................................

44 schadelijk zijn voor ons klimaat. Er kan in de

44................................

toekomst

45 alles gebeuren. Het warmer worden van de aarde is

45................................

trouwens

46 alleen maar negatief. Sommige streken waar het nu

46................................

extreem koud

47, worden er juist een stuk aangenamer door. Dat

geldt bijvoorbeeld graden erbij het niet

48 delen van Canada en Rusland. Met een paar

49 daar misschien landbouw mogelijk. En voor ons zou

50 zijn voor het toerisme als het zomers wat warmer wordt

in Nederland. Maar dat is natuurlijk geen reden om maar niets te doen. We moeten juist nu maatregelen nemen, zodat we ook in de toekomst prettig kunnen leven. Alleen als we vroeg beginnen, kunnen we het ergste voorkomen. Bron: Libelle nr. 52 2005, p. 37

Bedankt voor het invullen van de tekst.

47................................ 48................................ 49................................ 50................................

148

A.2


Score sheet for the proficiency task

This table lists for every gap in the cloze task the words that were in the original text (column 1) and that were accepted as alternative answers (column 2). Column 3 lists examples of answers that were given, Antwoorden & goede alternatieven: Cloze task study 1

which were counted as a mistake. origineel

alternatieven

fouten

1verwachten

voorspellen, denken, schatten, stellen

gaan ervan uit, zijn

2 zijn

worden

3 en 4 die 5 hebben 6 andere

namelijk vroegere, vorige, voorgaande, voorliggende, afgelopen, eerdere

7 en

zoals

8 ligt

is, zit

9 weten

weet

10 heftiger

sterker, heviger, groter, krachtiger

frequenter, meer

11 grotendeels

wel, dus, weer, ook, evident, indirect, dan,

ook

natuurlijk, gedeeltelijk 12

warmte

aarde

grote, voornaamste

grootste

temperatuur 13 De 14 en 15 het

van deze, lucht

16 er 17 we

de aarde, hij zijn

wij, mensen

18 die

is voor

19 in

langs

aan

20 het

dat

zij

21 niet

minder

gelukkig

22 Maar

wel

23 en 24 de 25 maar

en

26 doen

veranderen, ondernemen

27 we

mensen, wij

28 dijken

boel

29 moeten

hoeven

doe

te hoeven

A.2 Score sheet for the proficiency task 30 een

149 zeker

31 heel

maar, erg, echt, nogal, vrij, nog, slechts

32 aan

tegen

33 dat

omdat, als, wanneer

34 in

tijdens

werkbaars

35 winter 36 er

daarentegen, dan

dan, het

37 vaker

korte, meer, hevigere, grote, forse, zware

38 kunnen

moeten, mogen

ook, al

39 wel

slechts, nu

bestaat

40 we

wij, overheden

snel

41 Helaas

nog, momenteel

dus, jammer

42 het 43 dus

ook, zo, vrij, heel, erg, daardoor

44 minder

niet, on-

45 van 46 niet

zijn nooit

47 is 48 voor

grote

49 is

wordt

50 slecht

nadelig, erg, ongunstig

moeilijk

geslacht

man □

vrouw □

................ jaar ........... maanden

.......... weken ........uur per week

8. Hoe lang volg je Nederlandse les ?

taal ....................... taal ....................... taal ....................... taal ....................... taal .......................

gevorderd □ gevorderd □ gevorderd □ gevorderd □ gevorderd □

7. Hoe lang ben je in Nederland?

6. Welke opleiding heb je in welke taal afgerond? Lager onderwijs ja □ nee □ Middelbaar onderwijs (Havo/VWO) ja □ nee □ Middelbare beroepsopleiding (MBO) ja □ nee □ Hogere beroepsopleiding (HBO) ja □ nee □ Universiteit ja □ nee □

5. Welke andere talen ken je op welk niveau? Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □

4. Mijn moedertaal of moedertalen zijn (meerdere antwoorden mogelijk) □ Marokkaans Arabisch □ Marokkaans Berber □ Turks □ Koerdisch □ Frans □ anders, namelijk.....................................

3. In welke landen heb je hoe lang gewoond? ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ......................................................................

2. Mijn geboorteland is ...................................................................................

1. Leeftijd ............. jaar

Bedankt voor het invullen van de vragenlijst

Hieronder heb je ruimte voor opmerkingen. ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................

10. Maak je van onderstaande media gebruik in het Nederlands? Televisie nee □ soms □ ja □ Radio nee □ soms □ ja □ Krant/Magazine nee □ soms □ ja □ Boeken nee □ soms □ ja □ Internet nee □ soms □ ja □

9. Welke talen spreek je regelmatig waar en met wie? (meerdere talen mogelijk)  Op je werk ...............................................................................................  Op school ................................................................................................  Met je ouders ..........................................................................................  Met je broers en zussen .........................................................................  Met je partner ..........................................................................................  Met je kinderen .......................................................................................  Met je vrienden .......................................................................................  In andere situaties, namelijk ................................................................... spreek je .................................................................................................

vul deze vragenlijst alsjeblieft volledig in. Je antwoorden zullen anoniem blijven en alleen gebruikt worden in het kader van dit onderzoek.

Beste Deelnemer,

deelnemersnummer ..................

A.3

Vragenlijst voor de deelnemer 1

150 Appendices to study 1

Background information sheet

Every participant filled in the following background information sheet (original Dutch version only). Apart

from age and sex the form asked detailed questions about the language background, e.g., their mother

tongue(s), the language use of their mother tongue in contrast to their use of Dutch or other languages.

A.4 Tasks and instructions

A.4

151

Tasks and instructions

The two texts give examples of the instruction for a monologic and a dialogic performance. The figures on the following pages show the visual input belonging to the different tasks.

Example of instruction for a monologue De beste mp3-speler (original Dutch version) Een vriend of vriendin van je heeft jou om advies gevraagd. Hij of zij wil een mp3-speler kopen. Hiervoor wil hij of zij ongeveer 200 euro uitgeven. Deze folder laat de mogelijkheden zien. Maak nu een keuze en bel hem of haar op. Maar helaas, je krijgt alleen het antwoordapparaat. Je moet dus een bericht inspreken. Spreek een bericht in waarin je vertelt welk apparaat je zou kiezen. Leg goed uit waarom jij juist dat apparaat zou kopen en niet een ander. Doe dat met behulp van de genoemde kenmerken. Geef aan in welke situatie dit belangrijk is. Benadruk waarom dit volgens jou de beste keuze is. Overtuig de ander van jouw keuze.

The best mp3-player (English translation) A friend of you has asked you for advice. He or she would like to buy an mp3-player. He or she is willing to spend about 200 Euro. The leaflet shows the possibilities. Make a choice and give him or her a call. Unfortunately, the voice mail is answering your call. So you will leave a message and tell what electronic device you would choose. Explain why you would prefer this one above others. Use for the argumentation the characteristics given on the leaflet. Indicate in what circumstances these properties would be important. Argue why this is the best device to buy according to you. Try to convince your friend.

Example of instruction for a dialogue De beste mobiele telefoon (original Dutch version) Jullie willen een mobiele telefoon kopen. Hiervoor willen jullie ongeveer 300 euro uitgeven. Deze folder laat de mogelijkheden zien. Maak nu een keuze en bel elkaar vervolgens op. Vertel aan elkaar welk apparaat je zou kiezen en bespreek het uitgebreid met elkaar. Geef goede argumenten waarom je juist dat apparaat zou kopen en niet een ander. Doe dat met behulp van de genoemde kenmerken. Geef aan in welke situatie dit ´ apparaat uitkomen, dus overtuig de ander van jouw keuze. belangrijk is. Jullie moeten uiteindelijk op e´ en

152


The best mobile phone (English translation) You would like to buy a mobile phone. You would be willing to spend about 300 Euro. The leaflet shows the possibilities. Make a choice and then give each other a call. Tell each other what electronic device you would choose and discuss this with each other. Explain why you would prefer this one above others. Use for the argumentation the characteristics given on the leaflet. Indicate in what circumstances these properties would be important. Argue why this is the best device to buy according to you. You should come to a single answer so try to convince each other.

A.4 Tasks and instructions

153

Mp3-player tasks Simple mp3-player task

naam

Apple iPod

naam

Philips

prijs

249 €

prijs

157 €

kleur

wit

kleur

zwart

grootte (H x B x D)

10,4 x 6,1 x 1,6 cm

grootte (H x B x D)

8,6 x 5,4 x 1,6 cm

gewicht

167 gram

gewicht

87 gram

scherm

2 inch

scherm

1,5 inch

capaciteit

20 GB

capaciteit

4 GB

max. speeltijd

15 uur

max. speeltijd

17 uur

Complex mp3-player task naam

Creative Zen

naam

I River

prijs

212 €

prijs

195 €

kleur

oranje of blauw

kleur

rood

grootte (H x B x D)

8,3 x 5,1 x 1,7 cm

grootte (H x B x D)

9,5 x 5,5 x 1,5 cm

gewicht

115 gram

gewicht

96,2 gram

scherm

1,5 inch

scherm

1,5 inch

capaciteit

8 GB

capaciteit

5 GB

max. speeltijd

15 uur

max. speeltijd

12 uur

naam

Apple iPod

naam

Philips

prijs

249 €

prijs

157 €

kleur

wit

kleur

zwart

grootte (H x B x D)

10,4 x 6,1 x 1,6 cm

grootte (H x B x D)

8,6 x 5,4 x 1,6 cm

gewicht

167 gram

gewicht

87 gram

scherm

2 inch

scherm

1,5 inch

capaciteit

20 GB

capaciteit

4 GB

max. speeltijd

15 uur

max. speeltijd

17 uur

naam

Packard Bell

naam

Kenwood

prijs

196 €

prijs

230 €

kleur

zilver of zwart

kleur

wit of zwart

grootte (H x B x D)

8,8 x 5 x 1,5 cm

grootte (H x B x D)

10,4 x 6,1 x 1,7 cm

gewicht

95 gram

gewicht

140 gram

scherm

1,8 inch

scherm

2,2 inch

capaciteit

6 GB

capaciteit

20 GB

max. speeltijd

10 uur

max. speeltijd

16 uur

154


Mobile phone tasks Simple mobile phone task

naam

I-Mate

naam

Siemens

prijs

349 €

prijs

257 €

kleur

wit

kleur

grijs

grootte (H x B x D)

10,4 x 6,1 x 1,6 cm

grootte (H x B x D)

8,6 x 5,4 x 1,6 cm

gewicht

167 gram

gewicht

87 gram

scherm

2 inch

scherm

1,5 inch

capaciteit

64 MB

capaciteit

12 MB

max. stand-by tijd

300 uur

max. stand-by tijd

340 uur

Complex mobile phone task naam

Qtek

naam

Nokia

prijs

312 €

prijs

295 €

kleur

grijs of blauw

kleur

blauw

grootte (H x B x D)

8,3 x 5,1 x 1,7 cm

grootte (H x B x D)

9,5 x 5,5 x 1,5 cm

gewicht

115 gram

gewicht

96,2 gram

scherm

1,5 inch

scherm

1,5 inch

capaciteit

24 MB

capaciteit

15 MB

max. stand-by tijd

300 uur

max. stand-by tijd

240 uur

naam

I-Mate

naam

Siemens

prijs

349 €

prijs

257 €

kleur

wit

kleur

grijs

grootte (H x B x D)

10,4 x 6,1 x 1,6 cm

grootte (H x B x D)

8,6 x 5,4 x 1,6 cm

gewicht

167 gram

gewicht

87 gram

scherm

2 inch

scherm

1,5 inch

capaciteit

64 MB

capaciteit

12 MB

max. stand-by tijd

300 uur

max. stand-by tijd

340 uur

naam

T-Mobile

naam

Sony Ericsson

prijs

296 €

prijs

330 €

kleur

zilver of zwart

kleur

grjis of zwart

grootte (H x B x D)

8,8 x 5 x 1,5 cm

grootte (H x B x D)

10,4 x 6,1 x 1,7 cm

gewicht

95 gram

gewicht

140 gram

scherm

1,8 inch

scherm

2,2 inch

capaciteit

18 MB

capaciteit

64 MB

max. stand-by tijd

200 uur

max. stand-by tijd

320 uur

Examples of transcripts

@Begin Hilal: spreekt met Hilal. eh jouw vriend Hilal. ik vind vor voor u Siemens. de de prijs is eh tweehonderdzevennegentig. het kleur is grijs. gewicht is eh helemaal lekker. is zeventa eh tachtig gram. ´ en vijf. eh sss scherm is e´ en capacite´ is minder. is niet eh niet zo goed. maar maximum standby de tijd is voor. mmh is eh. heeft voordeel. is driehonderdveertig. maar voor mij is goed. is eh de de ander. Mate Maat Mate. de prijs is beetje omhoog. is drie driehonderdnegenenveertig. de kleur is wit. maar er zijn er ook de anders kleures. tah het verschil kleur is eh zwart of grijs of veel. gewicht is honderzevenenzestig. capacite´ is heel goed.

Example of a male Moroccan L2-learner (name changed).

Monologue simple mobile phone

A.5

is vier vieren eh zestig m MB. maximum staat by de tijd is driehonderd uur. is veel voor voordeel. maar je je helpts jou. ke ja keus voor jou. onderzoeker: waarom is de capaciteit belangrijk? ´ Hilal: capacite? voor alles voor. eh. la capacite´ is belangrijk in voor voor eh. ´ als jij hebt geen capacite. je hebt niks. capacite´ is eh. daarom is duurder. ´ voor capacite. ´ meer capcite. eh. meer eh meer geleer? capacite´ eh MB eh capacite´ eh MB. voor eh. voor mp3 trois. voor alles voor z ff. voor eh voor . ik denk zo ja. @End

A.5 Examples of transcripts 155

@Begin Sibel: eh met Sibel. eh hi Leyla. eh eh jij jij wilde eh een eh I-pod eh I-pod gekoch eh ko kopen. en ik ik heb een onderzoek gedaan voor eh voor jou. en ik heb een eh folders gevonden. eh er zijn twee soorten I-pod. maar eh maar eh. ik vond eh eentje heel goed. eh deze eh vo deze is Philips. eh het prijs eh honderdvijfen eh honderdzevenenvijtig euro. maar eh eh. capaciteit is veel veel eh lager dan de andere. ehm mmh eh. vier me eh vier eh G GB capaciteit. dus misschien eh wilt eh wilt niet wilt je niet deze eh I-pod. maar eh di dit prijs is beter voor jou. mmh en eh eh. hij gewicht eh he hij eh is eh ve zevenentachtig gram. ´ komma vijf inch. en scherm eh e´ en maar heel heel eh heel eh goed eh heel goe goed is deze I-pod. maar eh speeltijd is veel langer dan de andere. @End

Example of a female Turkish L2-learner (name changed).

Monologue complex mp3-players


@Begin ¨ Huseyin: goeie morgen ik heb gezien een eh ik heb gezien een eh. hoe heet het die die dingetje? Bilal: mp3. ¨ Huseyin: mp3. die heet eh honderd eh honderdvij eh honderdzevenenvijftig de prijs. Bilal: ja. ¨ Huseyin: en eh ik wil graag hum deze apparaat eh hebben. Hoe kan ik die bestellen? ja eh ik vind eh die witte mp3 een beetje duurder dan eh de zwart mp3. Bilal: Pilips. ¨ Huseyin: ja. Bilal: maar eh dat is ook een goeie merk. eigenlijk, Pilips. ¨ Huseyin: ehum. Bilal: en het is een kleine mp3. maar eh de capaciteit is weinig. eh die witte heef twintig GB. maar eh de zwarte heeft vier GB. en eh mts de speeltijd is ook een beetje ja. meerder dan eh de witte mp3. wat denk je? ¨ Huseyin: ja, ik begrijp deze goed. de witte is wel eh meer kwaliteit.

Example of two male Turkish L2-learners (names changed).

Dialogue simple mp3-players

ja. maar de prijs is een groot verschilt. ja. kijk, de zwarte ook heeft eh is eh be eh de helft van de prijs. ja. en eh de geheugen staat eh de capaciteit van de vier GB. maar de maximaal speeltijd is zeventien uur. Bilal: uur ja. ¨ Huseyin: ik denk eh wel aan de een kant wel is kleiner, dan is de witte. maar volgens mij is eh de zwarte ook gew goed de prijs. Bilal: ja, denk je. de zwart. ¨ Huseyin: en de prijs is niet zo duur. maar alleen wel er is is verschilt eh eh. zwarte verschillen dan witte, maar. ja, kijk maar, ongeveer een klein beetje verschilt bij elkaar. ´ wat wil je kopen? Bilal: oke, wil je zwart of wit? ¨ Huseyin: ik wil gew liever de zwarte. Bilal: anders ik ga ook eh de zwart kopen. ja. goedkoper dan eh de andere. en een goeie merk. dat was het. @End

Bilal: ¨ Huseyin: Bilal: ¨ Huseyin: Bilal: ¨ Huseyin:


Jamila: Khalida: Jamila: Khalida: Jamila:

Jamila: Khalida:

Jamila: Khalida:

Jamila: Khalida:

Khalida: Jamila: Khalida:

@Begin Jamila: Khalida: Jamila: Khalida: Jamila:

eh goeden avond met Jamila. hoi Jamila hoe is het? hoi Khalida, hoe is het met jou? goed. goed. en met jou met jou? ja. ja ook goed. ja, wat doe je vandaag? niks, eh rustig blijven. thuis televisie kijken. heb jij misschien een mobiele telefoon gekozen. ja dat is waar. ik heb ik heb een eh folder van eh een bedrijf gekregen. en eh ik heb hem een beetje doorgelezen. en eh ik vind Nokia t is de best de beste. ja mobiel telefoon ja. waarom vind jij dat? ja eh, door het prijs. ik vind de prijs is is een redelijk prijs. en ook de kleur. eh blauw. ik hou weet je ik eh ik hou van eh de blauw kleur. mja. en ook eh haar cap zij capaciteit is eh het is eh ook eh redelijk. eh vijftig het is eh mega, het is goed. vijftien bedoel je. ja vijftien sorry vijftien. maar eh eigenlijk ik zie dat eh de kwaliteit is niet zo groot. sorry. de capaciteit is niet zo groot toch?

Example of two female Moroccan L2-learners (names changed).

Dialogue complex mobile phones

Khalida: maar voor mij eh, voor mij ik ben alleen student eh. ik heb niet eh en ook ook ik ben allochtoon, ik heb eh nog niet zo naam van. eh vriend en vriendin, daarom eh het is handig voor mij, en ook. Jamila: ja ik kies eh ja. weet ik niet maar eh. Khalida: en jij wat denk jij? Jamila: mmh, ik vind eh de Qtek, Qtek. Khalida: weet jij moet jij kiezen altijd de beste merk. het is een sterk merk en eh een universal merk. ik vind eh de Nokia het is de beste. Jamila: maar ik vind Qtek eigenlijk een een goede merk want eh de capaciteit is te hoog. vierentwintig eh megabite. en eh ook er zijn twee kleuren. je kunt kiezen tussen grijs of bla blauw. en de prijs is niet eh ja er is niet een groot verschil tussen de Nokia en eh de Qtek, toch? Khalida: sorry, welk, welke Jamila: de eerste, bedoel. Khalida: Qtek? Jamila: ja. Khalida: hmm. Jamila: en ook hij is niet eh te zwart. Khalida: maar eh hij kost eh driehonderd-eh-twaalf. Jamila: ja dus er is niet zo’n groot verschil. Khalida: en eh ik moet ook reiskosten toevoegen, dat is ook kost ge geld. ik weet het niet eh. Jamila: mja, maar ik vind echt, dat de eerste de best is. en eh ook eh hij eh hij kan eh. Khalida: misschien hij is ook eh niet te het is zwaar.


Jamila: Khalida:

Khalida:

Khalida: Jamila: Khalida: Jamila:

Jamila: Khalida: Jamila:


Khalida: Jamila:

Jamila:

Khalida:

Jamila:

honderd-eh-vijftig gram het is eh vijftien gram, t is eh weet je, heel zwaar voor mij. vijftien. ja ja, soms ik pak een kleine tasje. kan ik niet. maar dat is een mobiel telefoon. dat is niet een eh een gro te groot, toch? ´ ah oke. ja? ben je niet overtuigd heel veel. nee, nee, eigenlijk ik wil. nee, je blijft altijd in jouw mening. Qtek sorry, maar sorry, Jamila maar Qtek, t is niet eh een bekend merk. het is. dat is niet een kwestie van bekend of niet bekend. ik heb het niet gehoord gehoord, dat is de eerst keer Qe Qtek. heb je nooit ge ja maar dat is een advertentie dus eh andere merken mogen komen en eh. ik kan niet eh het risiko nemen met een eh een nieuw merk. maar waarom niet iets anders proberen. ik weet Nokia wat is het Nokia. waarom niet iets anders proberen, waarom niet eh altijd de dezelfde merk kiezen. maar ik ik probeer met te weinig geld. niet eh zo duur. ja maar eh, hoe ja hoe groot is het verschil ik zie. kan niet. tweehonderd. Jamila:

Khalida:

Jamila:

Khalida:

Jamila:


Jamila: Khalida: Jamila: Khalida: Jamila: Khalida: Jamila:

Jamila: Khalida: Jamila: Khalida: Jamila: Khalida:

ja zeventien zeventien euro dat is niet te veel. nee, staat ook het is geld altijd, voor mij is geld. ja, maar je kan dan beter kiezen voor Siemens. Siemens? ja hij is alleen voor eh tweeehduizendehzevenvijftig euro. ik heb een Siemens eh gekocht. en ik heb een Siemens eh geprobeerd maar het lukt niet. ja. echt waar het is niet go zo goed merk. oke´ dus je hebt al ervaring met eh Siemens. ja. dus je kiest liever voor Nokia. No beste ik vind het No Nokia. dus eh, misschien kunnen wij morgen samen gaan, om die telefoon eh te kopen? ja ja graag. ja wanneer eh, wanneer heb jij ja wij gaan, ik zal, ik zal ik zal eh eh zaterdagmiddag, dat is goed voor jou? mmh ja, om twee uur, zo? kunnen wij samen gaan. is goed, oh is goed. we gaan ook lekker eten. ´ oke. ja, waarom niet. ´ oke. oke´ hmm, dag, doei. dag doei, doedoei. @End


160


Appendix

B


162

B.1


Proficiency task

The language proficiency task, that was used in study 2, is developed by the language center of the Rijksuniversiteit Groningen (Talencentrum RUG). The language center uses this kind of task as a placement test. The test below shows an example (Dutch version only). The original may be collected at the language center of the RUG. As shown in the example below the proficiency task consisted of short texts taken from newspapers. At some places in the text (marked by bold face) participants had to choose the correct wording among three possible options. In total total there were eight short texts with a total of 100 ’gaps’ where participants needed to make a choice. L2-learners received 30 minutes to perform the task, L1-speakers 15 minutes.

Naam ......................................... Datum ........................................ Deze test bestaat uit acht korte teksten. Op sommige plaatsen in deze teksten staan drie woorden achter elkaar dik gedrukt. Je moet dan kiezen, want slechts één van deze drie woorden is goed. Je moet een cirkel om het goede woord zetten. Dus zo: Je moet het goede woord wegstrepen / omcirkelen / opschrijven . Je moet altijd één woord omcirkelen. Geen antwoord geven wordt fout gerekend. Twee of drie antwoorden geven wordt ook fout gerekend. Voorbeeld 1. AMSTERDAM Een 59-jarige chauffeur van een vuilniswagen is vrijdagmiddag ernstig gewond geraakt/geworden/gevonden toen hij in het Gelderse Almen door de afrastering van een brug reed en in het water terechtkwam. Een 38-jarige collega die in/naast/met hem zat raakte lichtgewond. Dat meldt de politie. De vrachtwagen raakte rond 14.30 uur door onbekende richting/wijziging/oorzaak van de Vordenseweg en stortte enkele meters naar beneden van de Spitholterbrug. Beiden konden door de voorruit, die kapot was geslagen door de klap op het water, uit de vuilniswagen komen. Zij werden door een passerende motorrijder aan/onder/bij de kant geholpen. etc. In total eight short texts with a total of 100 gaps to choose between three options each. (L2-learners) 15 minutes (L1-speakers) to accomplish the task. The correctParticipants answers hat in 30 theminutes example would /be: Correct answers in example:

1 geraakt 2 naast 3 oorzaak 4 aan

1. geraakt 2. naast 3. oorzaak 4. aan

1 2 3 4

geslacht

man □

vrouw □

.............. jaar ......... maanden



9. Hoe veel uur per week heb je gemiddeld les? ................. uur

8. Hoe lang volg je al Nederlandse les?.............. jaar..........maanden

7. Hoe lang ben je in Nederland?

6. Welke opleiding heb je in welke taal afgerond? basisonderwijs ja □ nee □ Middelbaar onderwijs (Havo/VWO) ja □ nee □ Middelbare beroepsopleiding (MBO) ja □ nee □ Hogere beroepsopleiding (HBO) ja □ nee □ Universiteit ja □ nee □

5. Welke andere talen ken je en welk niveau geef je jezelf Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □

4. Mijn moedertaal of moedertalen zijn (meerdere antwoorden mogelijk) □ Marokkaans Arabisch □ Marokkaans Berber □ Turks □ Koerdisch □ Frans □ anders, namelijk.....................................

3. In welke landen heb je hoe lang gewoond? ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ......................................................................




Hieronder heb je ruimte voor opmerkingen. ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................


10. Welke talen spreek je regelmatig? Vul in waar en met wie je welke taal gebruikt (meerdere talen mogelijk).  Op je werk ...............................................................................................  Op school ................................................................................................  Met je ouders ..........................................................................................  Met je broers en zussen .........................................................................  Met je partner ..........................................................................................  Met je kinderen .......................................................................................  Met je vrienden .......................................................................................  In andere situaties, namelijk ................................................................... spreek je .................................................................................................


Beste Deelnemer,


B.2

L2

B.2 Background information sheet for L2-learners and L1-speakers 163

Background information sheet for L2-learners and L1-speakers

All participants filled in a sheet about their background. Apart from age and sex the form asked detailed

questions about their language background, e.g., their mother tongue(s) and the use of their mother

tongue in contrast to Dutch (L2-learners) or other language use apart from Dutch (L1-speakers). The

next pages shows the forms for L2-learners and L1-speakers respectively (original Dutch versions).


geslacht

man □

vrouw □

6. Welke opleiding heb je in welke taal afgerond? Basisonderwijs ja □ nee □ Middelbaar onderwijs (Havo/VWO) ja □ nee □ Middelbare beroepsopleiding (MBO) ja □ nee □ Hogere beroepsopleiding (HBO) ja □ nee □ Universiteit ja □ nee □



5. Welke andere talen ken je? Welk niveau geef je jezelf? Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □ Taal ....................................beginner □ gemiddeld □

4. Mijn moedertaal of moedertalen zijn (meerdere antwoorden mogelijk) □ Nederlands en ........................................................... en ...........................................................

3. In welke landen heb je gewoond en hoe lang? ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ...................................................................... ................................jaar in ......................................................................




Hieronder heb je ruimte voor opmerkingen. ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................ ........................................................................................................................


7. Spreek je regelmatig een andere taal dan het Nederlands? Zo ja, vul in welke taal je waar en met wie gebruikt (meerdere talen mogelijk).  Op je werk ...............................................................................................  Op school ................................................................................................  Met je ouders ..........................................................................................  Met je broers en zussen .........................................................................  Met je partner ..........................................................................................  Met je kinderen .......................................................................................  Met je vrienden .......................................................................................  In andere situaties, namelijk ................................................................... spreek je .................................................................................................


Beste Deelnemer,

L1


B.3 Perceived task difficulty assessment sheet

B.3

165

Perceived task difficulty assessment sheet

Immediately after performing the experimental tasks (simple or complex dating or study task) every participant filled in the following sheet asking how they perceived the difficulty of the task (adapted from Gilabert 2005). It asked participants to judge the task on a 5 point Likert scale with respect to five affective variables: 1. ease, 2. frustration, 3. task accomplishment, 4. interest, and 5. whether they would like to do more of these kinds of tasks. This page shows the form (original Dutch version only).

datum__________

taak____

deelnemersnr._____

Vragenlijst over de taak die je net uitgevoerd hebt Geef de titel van de taak Over hoe veel mensen ging de taak Hier onder volgen een paar korte uitspraken over de opdracht die je net uitgevoerd hebt. Op een schaal van 1 (helemaal niet mee eens) tot 5 (helemaal mee eens) moet je aangeven in hoeverre je het met de uitspraak eens bent. Denk er niet te lang over na maar geef een spontaan antwoord. Ik vond deze taak makkelijk. helemaal niet mee eens

1

neutraal

2

3

mee eens

4

5

Ik raakte gefrustreerd door deze taak. helemaal niet mee eens

1

neutraal

2

3

mee eens

4

5

Ik heb deze taak goed uitgevoerd. helemaal niet mee eens

1

neutraal

2

3

mee eens

4

5

Deze taak was interessant. helemaal niet mee eens

1

neutraal

2

3

mee eens

4

5

Ik wil meer zulke opdrachten maken. helemaal niet mee eens

1

neutraal

2

3

mee eens

4

5

DANK JE WEL study 2

Marije Michel, UvA

herfst 2007

166

B.4


Tasks and instructions

Examples of the instructions for a monologic and a dialogic performance and the visual stimuli of the different tasks.1

Example of instruction for a monologue (dating) Met de liefde in zee Op televisie is er een nieuwe datingshow: Met de liefde in zee. Vier (simpel) / Zes (complex) deelnemers krijgen kort de tijd om elkaar te leren kennen. Daarna moeten ze iemand kiezen die ze leuk vinden. Alleen als twee mensen voor elkaar kiezen, mogen ze samen een week op vakantie in Spanje. Ook jij kunt een prijs winnen. Als je de winnaars goed voorspelt, krijg je 500,-. Samen met een vriend(in) heb je besloten om mee te doen. Over 2 minuten moet je hem/haar opbellen om jouw voorstel door te geven. Hij/zij zal de telefoon niet kunnen opnemen dus je moet een bericht op de voicemail inspreken. Je hebt net de beschrijvingen van de kandidaten gelezen. Bekijk ze nog eens goed en maak een keuze welke twee het beste bij elkaar passen en dus een grote kans hebben om te winnen. Als je je vriend(in) belt, krijg je 3 minuten de tijd om zo uitgebreid mogelijk uit te leggen wie jij zou kiezen. Zorg ervoor dat je goede argumenten hebt en leg precies uit waarom jouw voorstel de beste keuze is. ´ stel Leg ook uit waarom andere koppels niet zullen winnen. Let wel, jullie moeten uiteindelijk voor e´ en kiezen dat jullie doorgeven aan ‘Met de liefde in zee’ dus neem rustig de tijd en overtuig je vriend(in) in je verhaal.

Love is in the air (English translation) There is a new dating show on tv. Four (simple) / Six (complex) contestants receive a couple of minutes to get to know each other. Afterwards they have to choose someone they like. Only if a man and a woman choose each other do they win a sailing trip to Spain. The audience at home can also win a prize: if you predict the winning couple you will receive 500 euro. Together with a friend you decided to take part in the viewer competition. In two minutes from now you will call him/her to give your opinion. As your friend will not be at home, you have to leave a message on her/his voice mail. You have just read the descriptions of the candidates. Look at them again and make a decision about which two out of them (a man and a woman) make a good couple and therefore are likely to win the show. When you call your friend you have 3 minutes time to explain in detail who you would choose. Make sure you have good reasons and are able to explain why your choice is the best. Also explain 1 The depicted people agreed on using their photograph for this research project. The listed names and characteristics are fictitious. Any resemblance to their real identity is purely coincidental.

B.4 Tasks and instructions

167

why other couples are less likely to win. Note, you and your friend will have to agree on one couple in the end. Thus, take your time and have a convincing story for your friend.

Example of instruction for a dialogue (study) Naar Antwerpen (original Dutch version) Samen met een vriend(in) organiseer je een uitwisseling voor studenten Nederlands als tweede taal uit Nederland en Belgi. Elke Belgische student krijgt iemand uit Nederland met wie zij samen via e-mail, chat en telefoon Nederlands oefent en opdrachten maakt. Volgend weekend gaan jullie met zn allen naar Antwerpen, om elkaar voor de eerste keer te ontmoeten. Er zijn nu nog twee plekken vrij, voor ´ student uit Nederland en e´ en ´ student uit Belgi, maar er hebben zich vier (simpel) / zes (complex) e´ en mensen aangemeld. Je hebt met je vriend(in) afgesproken om hem/ haar over 2 minuten op te bellen om jouw voorstel door te geven. Jullie moeten aan de telefoon met elkaar bespreken wie er mee mag naar Antwerpen. Je hebt net de beschrijvingen van de kandidaten gelezen. Bekijk ze nog eens goed en maak een keuze welke twee het beste bij elkaar passen en dus een grote kans hebben om goed samen te werken. Als je je vriend(in) belt, krijgen jullie 6 minuten de tijd om zo uitgebreid mogelijk aan elkaar uit te leggen wie je zou kiezen. Zorg ervoor dat je goede argumenten hebt en leg precies uit waarom jouw voorstel de beste keuze is. Leg ook uit waarom andere koppels niet goed zijn. Let wel: Jullie moeten uitein´ stel kiezen dat jullie mee nemen naar Antwerpen dus neem rustig de tijd en overtuig je delijk voor e´ en vriend(in) in je verhaal.

Travelling to Antwerp (English translation) Together with a friend you are organizing a study exchange program. Students of Dutch as a second language in the Netherlands and Belgium will be paired into studying couples. Together they will practice the Dutch language by using e-mail, chat and telephone. Next weekend all the participants of the program will get to know each other in Antwerp. There are two more places left but you received applications of the four (simple) / six (complex) girls you see above. Together with your friend you need to decide who will be accepted for the program. In two minutes from now you will call your friend and discuss on the phone which pair is your favorite. You have just read the descriptions of the candidates. Look at them again and make a decision about which two of them (one from the Netherlands and one from Belgium) would make a good couple and therefore are likely to work happily together. When you call your friend you have 3 minutes time to explain in detail who you would choose. Make sure you have good reasons and are able to explain why your choice is the best. Also explain why other couples are less likely to win. Note, you and your friend

168


will have to agree on one pair in the end. So, take your time and have a convincing story for your friend. There is only one more couple joining you to Antwerp.

B.4 Tasks and instructions

169

Dating tasks Simple dating task

Patrik leeftijd: houdt van: muziek: sport: leest: roken:

Stefan 24 jaar koken Metal regelmatig de krant soms

leeftijd: houdt van: muziek: sport: leest: roken:

Merel leeftijd: houdt van: muziek: sport: leest: roken:

22 jaar boeken lezen Hip Hop vaak van alles ja

Susan 23 jaar films kijken Reggae nooit veel boeken nee


21 jaar reizen R&B vaak weinig soms

Complex dating task

Patrik leeftijd: houdt van: muziek: sport: leest: roken:

24 jaar koken Metal regelmatig de krant soms

Merel leeftijd: houdt van: muziek: sport: leest: roken:

Stefan

Martin leeftijd: houdt van: muziek: sport: leest: roken:

26 jaar uitgaan Electro soms boeken nee

Janet 23 jaar films kijken Reggae nooit veel boeken nee



22 jaar boeken lezen Hip Hop vaak van alles ja

Susan 25 jaar winkelen Pop soms tijdschriften ja


21 jaar reizen R&B vaak weinig soms

170


Study tasks Simple study task studenten uit België Petra leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

Sofie 25 jaar Polen pedagogiek regelmatig 8 maanden misschien

leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen

23 jaar Frankrijk farmacie vaak 12 maanden ja

studenten uit Nederland Dzifa leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

Marta 24 jaar Ghana Frans nooit 5 maanden nee

leeftijd: 22 jaar komt uit: Duitsland studie: geschiedenis leest: vaak NL les sinds: 14 maanden staatsexamen: misschien

Complex study task studenten uit België Petra leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

25 jaar Polen pedagogiek regelmatig 8 maanden misschien

Nicky

Sofie

leeftijd: 27 jaar komt uit: Australië studie: Spaans leest: soms NL les sinds: 10 maanden staatsexamen: nee

leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

23 jaar Frankrijk farmacie vaak 12 maanden ja

studenten uit Nederland

Weili

Dzifa leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

24 jaar Ghana Frans nooit 5 maanden nee

leeftijd: komt uit: studie: leest: NL les sinds: staatsexamen:

Marta 26 jaar China dierenarts soms 7 maanden ja

leeftijd: 22 jaar komt uit: Duitsland studie: geschiedenis leest: vaak NL les sinds: 14 maanden staatsexamen: misschien

Examples of transcripts

@Begin Farida: hallo. met Farida. eh volgens mij Marta en eh Marta en Sofie eh kunnen eh eh passen eh passen wel bij ek bij elkaar. ze kunnen eh echt eh s goed eh samenwerken. want eh ze hebben ongeveer dezelfd leeftijd. Marta verb tweeentwintig jaar. Sofie drieentwintig jaar. eh zij wonen eh allebei eh ongveer n eh n jaar in eh Nederlan. eh Marta in Nederland. ¨ en s en ehm Sofie in Belgie. dus eh ze gaan ongveer dezelfd niveau in de Nederlandse taal eh eh hebben. en eh ze lezen ook eh allebei vaak. dus eh ze hebben ongveer eh de zelfde in eh interesse. over eh Petra. ik denk, dat Petra bij eh Dzi Dzifa w past. want eh ze hebben ook ongeveer dezelfde leeftij. en ze wonen eentje te weinig eh, dan eh Marta en Sofie. D Dzifa woont word eh vijf maanden. en Petra negen maanden.

Example of a female Moroccan L2-learner (name changed).

Monologue simple study

B.5

dus zij hebben ongveer ook eh eh misschien ongveer dezelfde Nederlands niveau. dus ze kunnen eh samen helpen. en samen studeren. t was het. onderzoeker: hm hm. wie van de twee vind je het beste? Farida: van de twee vind ik eh Sofie en Marta. want eh ze lezen al ze lezen allebei vaak. en ze wonen word allebei eh veertien maanden en twaalf maanden. dus de zit ongveer genoeg, om eh het Nederlands nog te kunnen eh beheersen. onderzoeker: hm hm. en als je kijkt naar de landen waar ze vandaan komen? Farida: eh Duitsland en Frankrijk. dat is eh eh ook eh d d. ik denk, dat de tweede taal in Duitsland is toch het Frans? onderzoeker: hm hm. Farida: dus ze kunnen ook eh met een andere taal eh communiceren. dat was het. @End

B.5 Examples of transcripts 171

@Begin Chris: he´ Mirjam. met Chris. ik zou je nog even bellen over ehm die datingshow met de liefde in zee. wa waarmee we die vijfhonderd euro toch eh echt moeten gaan winnen. ik ehm heb d inmiddels de de cv van cv’s van de deelnemers voor me liggen. twee jongens. twee meisjes. en die heb je ook op tv eh, denk ik, wel gezien ehm. nu inderdaad aan ons de taak, om eh zo slim mogelijk mensen bij elkaar te zoeken. en als ik kijk, dan ja dan denk ik toch in eerst instantie Stefan en Merel. en mts waarom? nou ja. ze houden allebei van boeken lezen ehm. die jongen is wel iets sportiever. m misschien wel aanzienlijk sportiever dan het meisje. maar misschien, ehm dat dat op zich geen breekpunt is. ik weet alleen niet. zij rookt niet. en hij rookt wel. dus moglijk, dat dat nog eh nog lastig kan zijn. maar op zich qua muzieksoort. ik moe je eerlijk zeggen. ik wee nie eens, of hiphop en reggae wel erg veel uit elkaar liggen. maar goed. eh d ze hebben in ieder geval, denk ik, wel wat wat raakvlakken, om om om t wel eh wel ov over te hebben. het alternatief, laat zeggen Stefan met Susan.

Example of a female Dutch L1-speaker (name changed).

Monologue simple dating

ja zou kunnen. maar dat daar zit dan wel inderdaad de de overeenkomst op t gebied van sport. ook allebei wel eh wat wat roken. muzieksoort. R en B en hiphop. misschien, dat dat dichter bij elkaar ligt. als ik kijk naar die naar die Patrik. de de tweede jonge. dan zou die dan misschien met met Susan kunnen ehm. waarom? ja ehm. beiden wat sportief. k zie, dat t enige wat daar wat m wat wat wat mist, is de de de muziek eh keuze, die nogal erg ver uit elkaar ligt volgens mij. ´ houdt van koken. de e´ en en de ander van reizen. nou dat kan ik me voorstellen, dat eh. nou ja het ligt eraan, wat voor soort keuken hij van houdt. maar dat je dan misschien wel wat ge¨ınteresseerder bent in verschillende culturen. nou ja je hoort het. ik ben er nog nie helemaal uit. maar ehm ik hoor graag, eh wat jij ervan denkt. en misschien zie jij weer hele andere dingen. maar we moeten tuurlijk wel een waterdicht verhaal hebben. want die vijfhonderd euro die eh k. die gaat van ons. @End


@Begin ´ Deniz: oke. ik ehm ik kies voor Sofie en Marta. Deniz: ehm meestal, omdat die hee die hebben bijna de hetzelfde niveau van Neders de Nederse taal. and die allebei wa Sofie wil de staatsexamen ehm nemen doen. maar Marta misschien gaat et doen. ehm ik denk, dat het staatsexamen willen doen, is een motivatie geeft motivatie, voor de taal beter te leren. and ook die hebben de ehm hetzelfde niveau. dus het gaat goed met hun. ´ is niet zo geworden, dan die andere. dus de e´ en en ook denk ik, dat het is erg goed, als je van ehm studeer hele v ehm andere ehm vakken. geschiedenis en farmacie. dus i die kunnen ehm interessante dingen aan ekaar ehm vertellen. and die zijn ook bijna hetzelfde leeftijd. dus die gaat goed met elkaar. en ook ehm Frankrijk en Duitsland. misschien die waren in ehm. ja Sofie was in Duitsdand. and ook Marta was in Frankrijk. du die kunnen ook praten over de ehm hun reizen naar de andere land. ehm ik dach, dat Weili en Sofie kunnen ook doen. maar omdat ja. ´ komt uit China. de e´ en en die andere uit Frankrijk. hele verschillende landen. maar and die beiden willen de staatsexamen doen. maar ja hun Nederse taalniveau is niet gelijk. and ik denk, dat ja diernarts en farmacie zijn beetje ja. ehm is niet zo anders dan geschiedenis en farmacie. dus ja het wordt niet zo interessant dan geschiednis en

Example of a male Turkish L2-learner (name changed).

Monologue complex study

farmacie mensen. and ja de m die. ik denk, dat, als ze willen de staatexamen doen, dat is deh ehm hoofdreden. daarom het moet in mij opinie Sofie en Marta. of Sofie en Weili. en ik ga voor Sofie en Marta. maar niet voor die anderen. oh. ehm ja. misschien die Nicky en zifa Dzifa Dzifa kunnen ook meedoen. maar ja die beiden willen niet staatsexamen doen. dus er is geen motief. ehm die beiden s studeer studeren sp ehm taal talen. Spaans en Frans. die is goed, voor leren Neders. maar ja. als ze v geen doel heef, dan is ja. dat, denk ik, gaat niet goed. ehm ja Petra. Petra kon het ook doen. hm pedagogie. misschien ja met Weili. dat kan ook. dat kan ook ja. ehm taalniveau zijn gelijk. and het staatsexamen misschien een ja. en de studie zijn zo verschillend. ja. maar ik toch ga voor de Sofie en Marta, omdat die hebben veel te praten. ik denk meer, dan de andere koppels. @End


@Begin Jamal: goeie dag. hoe is t met jou? ik hoop, dat het goed gaat. ehm ik heb goed gekeken naar de kandidatentlijst. eh volgens mij eh. mmh ja. ik eh het echt moeilijk, om eh een keus te maken. maar eh. het is echt eh. ik ga meer aan de kant van eh meneer Martin. de misschien is echt eh meer kans eh meerder kans ook. ik zit te twijfel gewoon over eh voor als Janet of Merel. je kan bij Merel passen. ze samen gewoon een paar dinges al gemeen. leeftijd ook is gewoon eh is een goeie leeftijd tussen ze t allemaal. van ma eh bijvoorbeeld Martin en eh Merel zijn van eh de ui. Martin houdt van uitgaan. en eh Ma Merel van films films kijken. dus ze kunnen gewoon samen vaak naar eh de bioscoopt gaan. ook van het soort muziek. is gewoon ja. is geen groot verschil. ze roken allebei niet. ze houden allemaal gewoon eh van boeken. ze hebben echt best eh veel dinges algemeen. ehm Janet vind ik een beetje. is ehm een ander soort van muziek. ze sport ook we wel soms. precies als hem.

Example of a male Moroccan L2-learner (name changed).

Monologue complex dating

maar ze Janet rookt. Martin niet. dus eh ik ga meer we tussen Martin en Merel. die vind ik echt eh. ze kunnen gewoon eh een goeie paar eh maken. van maken, volgens mij. ja dat was t. ik denk t. onderzoeker: kun je nog iets zeggen over de mensen, die je nog niet genoemd hebt? Jamal: ja voor de rest van eh van de kandidaten. ze is bijna gewoon eh allemaal verschillende. dus een beetje moeilijk, iemand van die houdt van koken. de ander van eh reizen. is een beetje echt moeilijk, om eh zo bij elkaar te komen. de andere, die houdt van sport. de ander niet. misschien de leeftijd gewoon is goed zo. voor ze allmaal ze kun allmaal. er is geen probleem van de leeftijd. ja volgs mij, ik ga gewoon deze. meer aan de kant van Martin en Merel. die vind ik echt de best paar. t kan gewoon echt eh. kunnen echt eh een goeie paar van maken. ja dat was dat is allem alles, was kan ik zeggen. ik hoop, de rest van van u wat te horen. ik hoop eh, fijne dag nog voor jou. dag. @End


@Begin Abel: hallo met Abel. Bart: he´ met Bart. ehm effetjes, eventjes over eh de keuze, die wij nog moeten maken over ons eh uitwisselingsproject. Abel: ja. Bart: had jij al een eh ideetje van de vier aanmeldingen, die eh jij het beste bij elkaar vindt passen? Abel: ja ik vond het wel onw heel lastig eigenlijk. Bart: ja het is lastig. Abel: eh misschien is het leuk, om eh Dzifa en Sofie aan mekaar elkaar te koppelen. Abel: want eh Sofie komt natuurlijk uit eh Frankrijk. Bart: ja. Abel: en eh Dzifa, die eh studeert Frans. dus Dzifa zou daar wel een hele hoop van kunnen leren, eh denk ik zo. Bart: dat is zeker waar ja. aan de andere kant dacht ik, dat eh Dzifa is wel de enige, die eh nooit leest. Abel: nee dat eh getuigt niet echt van een eh bijster goeie inzet. Bart: nee precies en eh ja ze wil ook geen staatsexamen doen, dus en ze ja ze kan ook niet zo goed Nederlands. en het gaat toch een beetje, om Nederlands ook te leren, dus ik de ja ik denk, dat zij eh ze wil dan wel hiervoor aangemeld, maar ik weet niet echt, of zij eh is heel enthousiast een partner is eh in dit project. ik zit toch meer te denken aan Sofie en Marta, want die eh ja toch allebei Europees ook eh. allebei best lang Nederlandse les al. zullen allebei waarschijnlijk wel staatsexamen doen. eh nou of even oud. na nou dat zijn ze allemaal wel, dus maakt niet uit.

Example of two male Dutch L1-speakers (names changed).

Dialogue simple study

Abel: Bart:

Bart:

Bart: Abel:

Bart: Abel:

Abel:

Bart: Abel: Bart:

Abel: Bart: Abel: Bart: Abel: Bart: Abel: Bart: Abel: Bart: Abel:

daar zit ik meer aan te denken. ja dat is op zich wel ook wel een een goed idee natuurlijk. en Petra, die eh ja moet je daar nou van zeggen. ja ik zou niet weten, wat je daarover wil zeggen. zou ook nog wel met Marta kunnen eventueel. dat zou ook nog kunnen. die Marta kan dus wel met allebei. ja over Marta zijn we het wel eens. die Dzifa eh eventueel eventueel met eh Sofie. ja. maar ja. hm ja Sofie en Marta hebben wel allebei de langste Nederlandse les. ja. dus die zouden wel in staat moeten zijn, om het eh. enne en. die Dezifa Dzifa en Sofie, die gaan natuurlijk Frans met elkaar praten dan. en dat is niet toch niet helemaal de bedoeling van eh. nee want het is natuurlijk de bedoeling, dat ze Nederlands leren natuurlijk. daarom. dus eh ja ik denk dat Marta en Sofie wel het meest eh Nederlands kunnen. ja denk ik ook ja. het zou wel het makkelijkste moeten zijn, eh wat betreft communicatie. Petra ja, die studeert pedagogiek eh ja. dat ze wel allebei meer een beetje de Sofie is wel echt beta met der haar medicijnenstudie. ja inderdaad. maar eh dat maakt niet zo heel veel uit, denk ik.


Abel: Bart: Abel:

Abel: Bart:

Bart:

Bart: Abel:

Bart: Abel:

Abel: Bart: Abel: Bart: Abel:

nee. denk niet echt, dat dat een eh. het gaat toch meer om, dat ze Nederlands leren, neem ik aan. dat gaat er wel, ja denk ik ook. en niet, dat ze wat eh leren van elkaars ehm, hoe heet het, van elkaars studie, want daar heb je toch niet bijster veel aan. nee precies en eh. denk wel, dat ze ook nog het meest gemotiveerd zijn, Marta en Sofie. ja denk het ook. want Dzifa, die wil sowieso al geen staatsexamen doen, dus ik zou niet weten, waarom ze hier zit, nee toch? nee en, hoe heet het, eh ja. ja ik zou ook eh eerder Sofie aan Marta eh koppelen dan eh dan Petra, want ik denk toch, dat Sofie eh ja denk toch, dat die wat betere klik hebben met elkaar. alhoewel Duitsland Duitsland en Polen, ja dat is natuurlijk ook wel het eh. ja volgens mij liggen die mekaar niet zo. dat is wel gevaarlijk, weet je. nee precies. nee dat is eh. dat hoeft dan natuurlijk niet altijd zo te zijn, maar eh. wel een beetje haat en nijd tussen die landen natuurlijk. Bart: Abel: Bart:

Abel:

Abel: Bart: Abel: Bart: Abel: Bart: Abel: Bart:

Abel: Bart:

Bart: Abel: Bart:

ja precies ja. de oorlog en al die grappen natuurlijk. ja maar ja en Petra, die eh ja nee ik zou toch voor Sofie gaan, eh denk ik. ja dat lijkt me ook het eh. als ik ook es naar zo een eh naar ze kijk, dan zullen die wel goed met elkaar overweg kunnen, denk ik eh. ja dat denk ik op zich ook wel, die liggen mekaar elkaar wel. ja. houden allebei van lezen. ja precies eh. ik denk, dat ze eh. ik denk, dat die wel snel klik met elkaar kunnen vinden eh. kunnen beter communiceren. in Antwerpen, ja. misschien ze veel zelfde interesses misschien eh nou die eh ja. ja ik denk, dat we die het beste kunnen meenemen naar Antwerpen. die hebben het wel leuk samen. lijkt me wel eh lijkt me wel gezellig. toch wel he` @End


@Begin Canan: hallo met Canan. Dunya hallo Canan, dit is, je spreekt met Dunya. Canan: ja hoi. Dunya: hoi, hoe gaat het met jou? Canan: ja heel goed, en met jou? Dunya: ja ik ook, eh eh luister, ik wilde jou wat vragen. Canan: ja? Dunya: kijk je ook wel eens naar de televisieprogramma? Canan: mja. Dunya: die over de liefde gaat. Canan: eh ja. Dunya: oke´ en ik wilde jou even v eh vragen, als jij met mij ook eh even iemand kiezen, en misschien kunnen we ook iets winnen. Canan: oh oh heel goed, ik heb, ja ik heb dat niet gedacht. Dunya: ja? ´ oke. Canan: en j jij bent slim. Dunya: en wat da wat dacht je over de mensen? Canan: ehm het is moeilijk. ehm ik denk, dat ehm ehm Stefan en eh Merel eh passen eh bij elkaar. Dunya: ja? Canan: denk ik, maar het is moeilijk. die ene eh de Stefan rookt. Dunya: ja. Canan: maar Merel rookt niet. Dunya: ja ik vind het ook lastig, als je met iemand eh samen woont, die rookt, als je niet rookt. Canan: ja. Dunya: ik dacht, Stefan en Suze passen bij elkaar eigenlijk. Canan: ja maar t is ook eh vervelend misschien ja, de Stefan leest niet nou tjonge een weinig heel weinig.

Example of two female Turkish L2-learners (names changed).

Dialogue simple dating

Dunya: ja. Canan: maar eh S Stefan leest van alles. ja Stefan is. ´ Dunya: ja ja maar eh weet je, wat ik dacht, eh ene van de e´ en van de mensen luistert naar hiphop. Canan: ja. Dunya: en de andere naar R en B. Canan: hm hm. Dunya: die ook een beetje passen ook be eh een beetje bij elkaar. Canan: mmh. Dunya: maar patrik bijvoorbeeld luistert naar metaal. Canan: metal. ja. Dunya: en die t het is echt lastig. Canan: ja dus. Dunya: ja met de andere. Canan: wat is de andere? Dunya: ja. Canan: bedoel jij Merel? Dunya: wat denk jij over Merel? Canan: eh Merel ja, hm misschien Merel en Stefan, denk ik. ik weet het niet waarom so. Dunya: ja i vind. Canan: maar eh z ze maken eh ah nee, ja even kijken. Dunya: ja maar Stefan doet aan sport vaak doet aan sport. ` Canan: he. ja. Dunya: en eh Merel doet het niet. Canan: ja dat klopt, ik ik zie dat nu. Dunya: ja. Canan: ja je hebt gelijk. Stefan en eh Susan passen bij elkaar beter dan Merel. Dunya: ja hm ja beter dan Merel.


Dunya: Canan: Dunya: Canan: Dunya: Canan: Dunya: Canan: Dunya: Canan:

Canan:

Dunya: Canan: Dunya:

Dunya: Canan: Dunya: Canan:

Dunya: Canan:

Canan:

Canan: Dunya: Canan: Dunya: Canan: Dunya: Canan: Dunya:

ja ja maar. en ja eh ja misschien ook Patrik en Susan, dacht ik. mts mmh eh. metaal en R en B vind ik niet echt ja. ja ja. ehm. ze kunnen niet in een k concert gaan samen. nee niet samen. en eh. dus denk ik ja, je hebt gelijk, Stefan en Susan passen bij elkaar heel goed. ja lijkt mij ook de best eigenlijk. maar ja Merel mooier dan Susan. wat moet hij doen Stefan? maar ja Susan Susan vind ik best jong eigenlijk en. hm hm. die is vers, weet je. ja v oh oh. wat betekent? ik ben zevenentwintig, wat bedoel je. eh mooie hm. ehm. ik weet het niet eigenlijk. m mijn keuze is Susan en Stefan, maar ik weet niet, wat je denkt ervan. ja i ik dacht eerder Merel en Stefan, maar i ik denk nu ook als jou. ja. ja. Merel en Stefan. Stefan en Susan passen bij elkaar. En Patrik. oke´ en Patrik. ja. nou. arme Patrik. nee Patrik leest eh de krant. hm hm. Dunya: die betekent, dat dat hij iets dagelijk leest, zeg maar. Canan: mmh ja ja. niet regelmatig. Dunya: en then Sus Susan eh leest nie eh leest bijna niet n nooit. Canan: leest ook weinig. nooit, maar het is ook. Dunya: ja dat dat. Canan: niet ehm. Dunya: en ze beiden roken soms, niet altijd. Canan: sorry, hm hm. Dunya: en eh. Canan: soms. Dunya: sport eh aan sport doet Patrik regelmatig. en eh Susan. Canan: ja Susan doet het ook vaak. Dunya: ja. Canan: ja dat is ook eh. ´ Dunya: dat is ook eigenlijk oke. Canan: eh ja. Dunya: ehm ja ik weet niet eigenlijk. ik ik kan ik kan niet kiezen. Canan: ik ook eh eh. Dunya: eh en Stefan houdt van boeken lezen, maar Susan houdt van reizen. Canan: ja. Dunya: misschien Susan wil meer actie dan dan Stefan natuurlijk. Canan: ja ja dat denk ik, eh, kijk Merel houdt van thuis blijven blijven of ergens anders. Dunya: ja films kijken. ` boek lezen. Canan: ja Stefan ook eh zoiets he, Dunya: ja. Canan: ja eh hij l l leeft eh eh meestal thuis, denk ik. Dunya: ja ja maar vo. Canan: daarom zei ik, dat eh Stefan en Merel bij elkaar passen. Dunya: nee weet je wat, volgens mij is het eh grootste probleem, dat dat Stefan rookt en Merel niet.


Canan: mmh denk ik waarom. ja misschien eh Merel eh accepteert dat. Dunya: ja maar dat weten we niet. eh ja ja weet ik weet ik niet eigenlijk. k kan ik kan niet kiezen. Canan: ik zal dat nooit zo accepteren, maar ja. Dunya: maar ja ja ik vind dit ook lastig. daarom z zei ik Stefan en Susan. Canan: ehm ehm ja denk ik ook, denk ik ook, maar eh dat is ook eh een beetje vervelend, denk ik, de boek lezen. Dunya: de. Canan: de ene eh houdt van boek le lezen. Dunya: ja ja andre ja ja ja. Canan: de ene houdt van reizen. Dunya: dat klopt. maar eh weet je wat, eh Stefan en Susan zijn ehm misschien het enige, die bij elkaar passe. Canan: eh hm hm ja ja. Dunya: die andere ja die dit is de de grootste mogelijkheid, zeg maar. Canan: dat denk ik ook denk ik ook. Dunya: ja. Canan: ja. ja ehm ja, dat denk ik ook. iemand kan niet eh kijk Stefan sport heel vaak. Dunya: ja. Canan: ja eh Merel niet, dit is een hele grote verschil eh. Dunya: ja. ja nooit. Canan: denk ik. ´ Dunya: ja ja oke. Canan: eh zij kunnen samen eh kijk, e als Susan wil ergens eh reizen. Dunya: hm hm. Canan: na eh Stefan kan een boek lezen ja eh. Dunya: ja da da ja ja dat klopt eigenlijk ja. en ja Stefan leest van alles, maar ja Susan leest niet. da dat vind ik een beetje misschien een probleem maar. Canan: maar dat is eh minder een probleem, dan roken. Dunya: ja ja Canan daarom. Canan: daarom eh. t is redelijk goed. Dunya: ja ja dat vind ik ook eigenlijk. Canan: ja. Dunya: en als Susan ook af en toe rookt. Canan: ja. Dunya: is het niet een grote ja. Canan: samen es. ja samen soms de eh roken. ´ Dunya: oke´ oke. Canan: ze ja eh denk ik ook. Dunya: zullen we dan eh zullen we een smsje sturen aan de programma. Canan: oh hm hm. Dunya: dat we die Susan en Stefan een koppel gemaakt. Canan: ja het s goed. ´ Dunya: oke. Canan: maar ben je zeker? Dunya: ja eigenlijk eigenlijk ik vind het best. en ja Stefan is ook knap. Canan: oh? Dunya: ik eh ja ik vind Stefan niet, weet je wat, eh lelijk of zo. Canan: mh mh. Dunya: en ze misschien ze vinde elkaar leuk. Canan: ja ja dat denk ik ook. Dunya: mh mh. Canan: en eh de leeftijd ook goed. Dunya: ja de leeftijd ook goed. Canan: goed. mmh heel goed. ´ Dunya: oke. ´ Canan: ja oke. Dunya: dan zullen we het doen. ´ Canan: oke. Dunya: oke´ doeg. Canan: oke´ bye doe doei. @End


@Begin Emre: eh met Emre. Serdar: met Serdar. Emre: met Emre. Serdar: hoi Emre, met Serdar. Emre: eh hoi Serdar. Serdar: ik wil over eh de reis naar Antwerpen eh bespreken. Emre: hm hm. Serdar: ik heb een twee eh eh sudenten gekozen. ´ Emre: oke. Serdar: en volgens mij eh Sofie en Weili eh is het best kunnen met elkaar omgaan. Emre: en? waarom denk jij zo? Serdar: en eh ze zijn eh allebei het staatsexamen gehaald. Emre: ja. ´ jaar. Serdar: en eh ze woon in Nederland bijna een e´ en en eh de studierichting ook hetzelfde. farmacie en dierenarts. Emre: hm hm. Serdar: eh dichtbij. de leeftijd is ook eh niet zo groot eh. ze hebben ook eh geen eh leeftijdverschil grote leefleeftijdverschil. da denk ik zo. en jij? Emre: ja ik vind het ehm Sofie en Weili is eh ja een goeie match. en eh en eh volgens mij ehm moeten we samen samen studeren of samen m eh omgaan. en ehm en wat vind je over eh de andere studenten dan? ik heb nog eh twee studenten, die eh eh gematcht kunnen worden. Serdar: welke student bedoel je? Emre: eh ik heb eh nog eh Marta. eh Dzifa, ik heb nog n twee studenten.

Example of two male Turkish L2-learners (names changed).

Dialogue complex study

Serdar: Marta? Emre: Marta en eh de Dzifa. ´ Serdar: oke. Emre: en eh heb jij nog studenten? Serdar: even k, even kijken. Polen. en eh eh Petra uit Polen. eh Marta uit Duitsland. Emre: Marta komt uit Duitsland ja. eh Dzifa komt uit eh Ghana Ghana. Serdar: ja. Emre: en eh. Serdar: Dzifa. Emre: nee Dzifa ja. Serdar: en? Emre: en eh ja n eh en eh Marta studeert eh geschiedenis. en ze komt uit Duisland. en eh ze ehm ze is veertien maanden in ehm Nederland. Emre: en eh soms eh ja he ze leest heel vaak. Serdar: hm hm. Emre: en eh het staatsexamen. ja misschien eh eh ka kan ze het halen. of eh eh dat weet ik niet. en eh Dzifa is eh leest nooit. en eh ze kom dan uit Ghana. en eh ze studeert eh Frans. maar eh en eh ze heeft eh ze eh ze heeft eh het staats eh examen niet gehaald. dus eh m i ik vind eh eh ik vind eh met eh eh met ja eh ze kunnen matchen met Petra. eh de Dzifa en met Petra. en eh niet omdat eh hun eh eh hun eh richting dezelfde is. maar wel eh nou.


Petra Petra kan eh het staasexamen gaan halen. en eh dat is ook belangrijk. en eh ze nou ze leest or Petra lees regelmatig. dus ik vind eh ja, als ze samen matchen eh, misschien eh kunnen ze ja ehm kunnen ze goeie voorbeelden zijn eh met elkaar. Serdar: maar een goeie idee. maar volgens mij wij moeten eh stududenten, die het staatsexamen gehaald hebben, gekozen. dat is het best ve eh, volgens mij. Emre: we moeten niet anderen al al a alle studenten kiezen? Serdar: volgens mij deze spreken goed Nederlands. het staatsexamen gehaald. ze kunnen goed eh met elkaar omgaan. ´ Emre: oke. Serdar: en eh. nou ik vind zo. ´ Emre: mmh oke. Serdar: als wij ehm twee studenten moeten kiezen, eh de dan eh moeten wij die gene kiezen, die eh goe eh goed Nederlands spreken, of en eh die eh het staatexamen halen. dus eh ja. Emre: volgens mij is het ook ja is het belangrijk. Serdar: ja vind je het leuk? Emre: dat vind ik ook ja. Serdar: ja? Emre: maar ik dacht eh dat wij alle studenten m mede moeten matchen. daarom dacht ik van ja. ja ik vind het een goed idee. ´ Serdar: oke. afgesproken. eh Sofie en eh Weili. ´ oke? ´ Emre: oh oke. Serdar: die twee studenten komen met ons naar Antwerpen. ´ oke? Emre: ja. Serdar: afgesproken. Emre: afgesproken. ´ tot devolgende keer. Serdar: oke. Emre: tot devolgende keer. Serdar: dag. Emre: dag. Onderzoeker: ik wil nog heel graag, dat jullie nog heel even iets over Nicky zeggen. daar hebben jullie nog helemaal niks over gezegd. Serdar: Nicky? Onderzoeker: ja. ¨ Serdar: Nicky uit eh Australie. zevenentwintig jaar. studie Spaans. mmh Nicky heeft eh de staatsexaam niet gehaald. Emre: ja. Serdar: en ehm zij spreek Spaans, leest soms. Emre: ik vind de leeftijd ook belangrijk. Serdar: ja. Emre: dus eh zevenen. Serdar: volgens mij nee nee, het is niet zo belangrijk. maar. Emre: daarom ka ka kan zij niet meedoen. omdat ze een beetje ouder is. en eh ja ze leert eh. ja ja weet ik niet. Serdar: nee. kijk de Weili eh woont pas eh zeven maanden in Nederland. zij woont in eh zij woont tien maanden in Nederland. hoewel eh Weili pas sinds kort wo eh in Nederland woont, eh heeft hij zij eh het staatsexamen. Emre: gehaald. Serdar: maar Nicky niet. Emre: ja. Serdar: dat is b volgens mij een grote verschil. wij moeten eh zo kijken zo denken. ¨ Emre: ze komt eh uit Australie. en eh ze spreken daar eh eh Duits.


Australie¨ of Engels? Serdar: Engels. Emre: Engels. Emre: en eh ze wilt eh Spaans of. Serdar: met wie praat jij? Emre: eh over Nicky. Serdar: tussen haakjes. Emre: eh over Nicky. oh met jou natuurlijk. ´ Serdar: oke. ¨ Emre: eh ze komt uit eh Australie. dus ehm spreekt ze Engels. en eh haar stu eh haar studie is Spaans. maar ehm voor het staatsexamen moest ehm moet moet ze ne eh Nede ook Nederlands spreken. dus eh misschien is het een beetje moeilijk voor haar. mmh ja. dus ja Sofie vind ik ehm Sofie en eh Weil Weili vind ik het beste. ´ Serdar: oke. Emre: ze spreken eh eh ze ze hebben allem allebei ne het staatsexamen gehaald. en eh a a ook eh hun eh richting is ehm zo. Serdar: dichtbij? ja. Emre: ja dichtbij. en leeftijd ook eh niet eh niet zo groot. het verschil is niet zo groot. Serdar: hm hm ja. ´ oke. afgesproken. Emre: afgesproken. ` Serdar: dat was t he? Emre: ja. ´ Serdar: oke. dag. Emre: dag. @End


@Begin Yasmin: eh ehm hallo met Yasmin. Malika: he´ Yasmin. met Malika. alles goed? Yasmin: ja goed. prima. met jou? Malika: ja ook. fine. ´ Yasmin: oke. ehm welke koppel vind jij, dat eh best past bij l eh bij elkaar? die past eh het best bij elkaar? Malika: ja. Yasmin: eh die gaat eh winnen? Malika: ik twijfel nog. maar eh ik heb een eh indruk over ehm ja Martin en ehm Susan. Yasmin: hm hm. en waarom? Malika: ja. ze passen bij elkaar in de leeftijd. hij is een eh ouder dan haar. en ook ze houden ehm ze elkaar van eh a. dus z zij eh zij houdt van eh reizen. Yasmin: hm hm. Malika: en hij eh houdt van uitgaan. dus is een beetje vlakbij. toch? Yasmin: maar eh. Malika: en jij? wat denk je? Yasmin: ja eh. van Martin?

Example of two female Moroccan L2-speakers (names changed).

Dialogue complex dating

Yasmin: Malika: Yasmin: Malika:

Malika: Yasmin: Malika:

Malika: Yasmin: Malika: Yasmin:

Yasmin: Malika: Yasmin:

Malika:

Yasmin:

Malika: Yasmin: Malika: Yasmin: Malika:

hm hm. hij leest boeken. ja. en eh Susan weinig. weinig toch maar toch, le eh sowieso leest zij toch? en eh wat is jouw eh. en ook Martin eh sport soms. en Susan vaak. nee. Susan rookt soms. en Martin rookt niet. toch? ja. wat is jouw idee? soms nee. eh ik vind Susan en Stefan. Stefan? ja. Susan en Stefan. ja. want eh de leeftijden is goed. hm hm. ja. maar hun ehm, zeg maar, hobbies zijn heel eh ver. S Stefan houdt van boeken lezen. dus hij houdt alleen van. maar eh zij wilt alleen naar eh de bibliotheek of thuis blijven, om eh boeken te lezen. hm hm. en zij houdt veel van reizen. reizen. ja is toch groot een groot verschil tussen eh hun. en Patrik is ehm houdt van koken.


Yasmin: van koken. dus ook eh. Malika: ja. en dan? wat vind je van de koppel? Yasmin: films kijken. dat is ook thuis blijven. en koken. dat is goed. Malika: wat? Yasmin: Patrik en Merel. Malika: ja. Yasmin: zij kan films kijken w. Malika: Merel? Yasmin: sport nooit. Malika: bedoel je Merel toch? Yasmin: ja. de leeftijden goed. maar zij l zij veel eh boeken lezen. ik weet het niet. Malika: ja volgens mij vind ik Martin en Susan. of eh ja. Yasmin: je kon ook Martin of eh waarom niet eh Martin nie eh en Janet? Malika: Martin en Janet? Yasmin: ja ze eh ze ook eh van winkelen eh houden? da ook van eh uitgaan. ze ze sport soms. zij ook. Malika: ja. Yasmin: zij houdt van eh tijdschrijf eh tijdschriften lezen. Malika: hm hm. Yasmin: en hij van boeken. maar ja zij rookt. en hij niet. ´ oke. Malika: ja weet je wat? Martin, die houdt niet van boeken veel boeken lezen. Yasmin: Malika:

.

Yasmin: Malika:

Yasmin: Malika:

Yasmin: Malika:

hij leest ja. maar toch hij houdt. de principle ve van hem is uitgaan. dus als eh Susan wilt naar eh ergens gaan, hij zegt niet eh nee, je wil niet, je wil thuis blijven, je wil koken, of ik wil eh boeken lezen, of rustig zitten of zo. Martin houdt ook van uitgaan. dus ook reizen of zo. en Susan leest ze leest eh niets nooit of zo. ze leest weinig. s sowieso leest, toch? en roken. ´ oke. rookt soms. niet vaak. niet eh ja altijd. en Martin nee. dus kan Martin eh invloed. weet je? om eh Susan. maar zij is zo zo verslaafd van eh roken. nee kan kan eh. volgens mij kan. dat roken is toch niet zo goed voor eh de gezondheid. ik vind zo. maar eh moet je goed nadenken. welke koppel past goed met elkaar. soms nee. ik vind Merel. alleen, ze wilt altijd films kijken. en ma Patrik ook eh. dus kan toch niet. de Martin in de keuken. en Merel alleen films kijken. ja so vind ik. dus Martin en Susan? ja.


Yasmin: Malika:

Malika:

Yasmin: Malika: Yasmin:

Yasmin: Malika:

Malika: Yasmin: Malika:

Malika: Yasmin:

Yasmin: Malika: Yasmin: Malika: Yasmin:

als je wil. wat vind je? ja? we gaan Martin en Susan eh kiezen? eh voor maar ik eh vind Martin en Merel. wat? Martin en Merel. Merel? ja. zij lezen boeken. ja. zij roken niet. zij houdt van uitgaan. zij houdt van films kijken. zij kunnen met elkaar uitgaan, om eh naar eh bis eh de bioscoop. volgens mij. films kijken. ja maar volgens mij ja. ik ben het niet helemaal eens met jou. en waarom niet? ja. ja zo vind ik dat Susan en Martin een goeie goe koppel kunnen. ik vind Martin wel goeie kop goed. wat? ja. ik vind eh Martin wel het beste koppel. of niet? ja moet je ehm goed nadenken. ik vind so moet je ook. dit is toch moeilijk om te doen. ja. dat is mijn eh idee. sowieso dat is een eh goeie kans voor Susan. want zij houdt veel van reizen. en ook Martin van uitgaan. dus zelfde bedoeling. uitgaan en reizen. ik vind eh dat de leeftijd is goed tussen hun. ja. wat is jouw antwoord dan? Yasmin: het is moeilijk. Malika: ja. Yasmin: echt moeilijk. Malika: wat? Yasmin: t ik weet het niet. ik vind Martin en Merel. en jij vindt Martin en Susan. Malika: beter dan Martin en Susan? Yasmin: ja. Malika: Martin toch sowieso is de de man? Yasmin: de beste of eh. ja? Malika: ja. of ja Patrik met Merel of. maar ja wat vind je dan? ik twijfel. ¨ voor zijn? onderzoeker: kunnen jullie er met zijn tweeen Malika: ja wel. ik vind Martin en Susan. Yasmin: ik ben begon bij Janet eh te kiezen. Martin moet eigenlijk een neutrale oplossing. Malika: Martin se en Susan. Yasmin: ja jij Martin enSusan. ik Martin en Merel. @End


English summary

This thesis investigates the effects of cognitive task complexity and interaction on oral performance of Turkish and Moroccan second language (L2) learners of Dutch within the task-based framework.

Chapter1: Theoretical framework Chapter 1 presents the theoretical framework of the studies at hand. It starts with a general discussion of the task-based approach to L2-research and focuses on its cognitive strand. There follows an in-dept discussion of the Cognition Hypothesis by Robinson (2001a, b, 2003b, 2005) as this book investigates some of Robinson’s claims. It first explains the notions of cognitive capacity and attentional allocation during task performance. Afterwards, this chapter highlights two factors of task design that, according to Robinson, affect attentional allocation during task-based L2-performance: cognitive factors of task complexity and interactive factors of task condition (cf. the Triadic Componential Framework on p. 23). The Cognition Hypothesis claims that so-called ’resource-directing’ factors of task complexity focus the attention of L2-learners on the language form (Robinson 2001a, b, 2003b, 2005). Accordingly, a complex task which, for example, asks to give a description of many elements, will need more different lexical items and complex syntactic structures than a task with only a few elements, where simpler structures may suffice. In a simple task, the sentences ‘the left one’ may be enough while a complex task with many elements may need descriptions of the form: ‘the one in the corner next to the other one, which is round’. Robinson therefore predicts that a cognitively complex task will push L2-learners to use linguistically complex structures and a wide vocabulary. As the learners’ attention is focused on language, also accuracy may benefit such that L2-learners will make fewer mistakes. It may be though, that these complex processes make task performance less fluent than upon cognitively simple tasks. Also interactive tasks may focus the learner’s attention to form. In interaction understanding is crucial. Therefore, L2-learners will be specifically focused on the linguistic code, which results in a higher accuracy. Due to turn-taking and interactional moves there will be fewer opportunities to build complex syntactic structures or search for specific lexical items. Accordingly, linguistic complexity will be lower in interactive tasks than in monologic settings, where L2-learners act on their own. Again, the heightened attention to form will decrease fluency in interactive tasks.

188

English summary

An intriguing claim of the Cognition Hypothesis concerns combined effects of cognitive task complexity and interaction. What happens, if L2-learners act in pairs on cognitively complex tasks? Robinson predicts that both factors (cognitive task complexity and interaction) push task performance so that the accuracy may increase while fluency decreases when task performance is compared to simple interactive tasks or any task in a monologic setting. Robinson expects complex tasks to result in more interactional moves than simple tasks. As a result linguistic complexity will be low in complex interactive tasks. These are the claims that are under investigation in the present book. Furthermore, this chapter discusses the measures of task performance that are widely used in task-based research. The traditional measures of production, that is linguistic complexity, accuracy and fluency (in short CAF-measures), are under discussion recently (e.g., Housen and Kuiken 2009). Especially, in light of the current discussion about the added value of task-specific measures (Cadierno and Robinson 2009, Robinson and Gilabert 2007) the present studies tries to give some further insights. The first chapter finishes with a review of earlier work on effects of cognitive task complexity and interaction on task-based performance of L2-learners. The focus lies on open issues that form the basis of the current research. For example, it highlights the fact that up to now hardly any research has been undertaken into task-based performance of native (L1) speakers.

Chapter 2: The studies in this book The second chapter explains in detail the empirical work that is presented in this book. This chapter formulates the research questions and hypotheses (see table 2.1 on p. 39 and the explanations on p. 40). It focuses on the actual variables that are manipulated in this book: the resource-directing cognitive factor of task complexity ‘± few elements’ and the interactive factor of task condition ‘one-way versus two-way flow of information’ that is transformed into the factor ‘± monologic’ in the work at hand. The hypotheses follow Robinson’s Cognition Hypothesis as explained above but they make slightly different predictions concerning the effects of interaction, and consequently also about the combined effect of task complexity and interaction. The present studies take an alternative view on interaction as they integrate insights form psycholinguistic research, that is the Alignment Hypothesis (Costa et al. 2008, Pickering and Garrod 2004) and task-based research by Tavakoli and Foster (2008). Based on this work this thesis proposes that dialogic tasks may be cognitively less complex than monologic tasks. Assuming dialogues to be less complex than monologues would mean that, on the one hand, one may expect the L2-learners’ accuracy and fluency to increase in interaction. On the other hand, alignment and routinization may make interactive tasks structurally and lexically less complex. Moreover, it is assumed that the effects of interaction are larger than those of increased cognitive task complexity so that in complex interactive tasks fluency will be increased (in contrast to Robinson’s predictions).

189 Contributing to the discussion about task-specific measures of performance, this thesis predicts that a specific measure, i.e., the use of conjunctions in cognitively simple versus complex tasks, will be affected by task complexity such that a complex task will increase the frequency and occurrence of conjunctions. With respect to L1-speakers, no effects of task complexity are expected – L1-production is highly automatic. However, interaction may similarly affect task performance by natives and second language learners. The remainder of this chapter gives a detailed description of the method and design of the empirical studies presented in chapters 3, 4, and 5.2 Finally, this chapter explains how the studies grew from one to the other and why there may be some changes and extensions (for example with respect to the measures) concerning the design of the different empirical chapters.

Chapter 3: The influence of complexity in monologic versus dialogic tasks in Dutch L2 (Study 1) Chapter 3 gives the details of a first empirical investigation into effects of cognitive task complexity and interaction on L2-learners’ speech performance. 46 highly educated L2-learners of Dutch participated in the study. They all were of Turkish and Moroccan origin and at an intermediate level of their L2 Dutch. They acted on cognitively simple and complex tasks where they were asked to give an advice about an electronic device. Following the Cognition Hypothesis task complexity was manipulated as the resource-directing factor ± few elements. That is, the simple task showed two devices and the complex task six devices respectively (see Appendix A.4). Half of the participants acted on their own in the monologic condition the other half acted in pairs in the dialogic condition. The 92 monologic and dialogic speech samples were transcribed and coded for 12 measures of linguistic complexity, accuracy, and fluency (see Table 3.3, p. 55). An analysis of variance (ANOVA) tested for effects of task complexity and interaction, both on their own and in combination. The results showed, as predicted, a higher linguistic complexity (with respect to one lexical measure), a higher accuracy (with respect to the total number of errors), and a lower fluency (as measured by syllables per minute). All effects were visible on one measure for each CAF-construct (in total three out of twelve measures). All other measures did not yield significant results. Interaction showed a statistically significant effect for both syntactic measures but none of the lexical measures of linguistic complexity (monologues were syntactically more complex than dialogues). Dialogues compared to monologues yielded more accurate (by means of three accuracy measures) and fluent speech (by means of all fluency measures, that is monologues generated slower speech with more pauses than dialogues). 2 N.B. As the empirical chapters 3, 4, and 5 have been or will be published as individual papers outside the context of this book there is only limited space in the chapters itself to report all background information. This fact also caused some overlap in their content, especially when reviewing the theoretical framework or when explaining the design of the experimental investigations.

190

English summary

There was one significant combined effect of cognitive task complexity by interaction. That is, the increasing effect of a complex task in monologues disappeared in the dialogic condition. The discussion explains that these data are partially in line with the claims of the Cognition Hypothesis. However, the confirmation may be not so strong as it was visible on three out of twelve measures only. Moreover, no combined effects were found that confirm Robinson’s claims. In contrast, interaction shows the expected increasing effects on accuracy and decreasing effects on (syntactic) complexity. There was an unexpected pushing effect on fluency. As this is the first empirical investigation that systematically looked at (combined) effects of task complexity and interaction, this chapter calls for new experimental studies that may generate a more conclusive picture with respect to these two factors. The conclusion furthermore asks for task-based investigations that include baseline data of native speakers.

Chapter 4: Effects of task complexity and interaction on L2-performance (Study 2a) The fourth chapter describes a second empirical study that investigates effects of task complexity and interaction with 64 Turkish and Moroccan intermediate L2-learners of Dutch and 44 native speakers of Dutch. All participants were highly educated. As the tasks of the first study sometimes failed at generating enough output, two new task were designed (the ‘study’ and the ‘dating’ task, see Appendix B.4). Participants were asked to find the best combination of two persons (a study pair in the study task and a love couple in the dating task). The difference in task complexity was increased as the simple task gave the opportunity to build four combinations whereas the complex task asked for a decision among nine possible pairs. Again, all participants acted on a simple and a complex task. One half performed both task on their own as a monologue, the other half acted in pairs in the dialogue. The 216 speech samples were transcribed and coded for measures of linguistic complexity, accuracy, and fluency. As some of the measures of study 1 suffered from methodological problems (e.g., co-linearity and redundancy), some new measures were taken. There were two measures of syntactic and one of lexical complexity respectively. Accuracy was accounted for by morphosyntactic, lexical, and determiner errors. The data were coded for speed, pausing, and breakdown fluency. The statistical analysis included separate ANOVAs for each CAF-construct and for each language group (L2 versus L1). Results reveal in both language groups one single significant effect of task complexity. That is, complex task yielded a higher lexical complexity. No other measures were significantly affected. L2-learners did not show any combined effects either. For the native speakers task complexity had no effect in monologues, but in dialogues a more complex task slowed them down. Effects of interaction were large in both language groups (for example, see Table 4.7 on p. 81).

191 In dialogues, L2-learners are syntactically less complex but lexically more complex. They are more accurate and more fluent with respect to all accuracy and fluency measures. L1-speakers showed similar effects but their effect sizes were smaller and not all measures were affected. Interestingly, interactive tasks made them more accurate too and (in contrast to L2-learners) they were lexically less complex in dialogues. This study did not yield confirming effects for Robinson’s Cognition Hypothesis (see the summarizing Table 4.12 on p. 86). Only one out of ten measures was affected by cognitive task complexity. However, this increase of lexical complexity may be accounted for by the extra amount of speech or by the task input. As L1-speakers show a similar effect, the conclusion may be that a complex task, manipulated by means of the factor ± few elements does not qualitatively influence task performance. It seems that there is a quantitative effect only: complex tasks yield more speech and, therefore, show a larger lexical complexity. Interaction shows a completely different picture. Dialogues improve the accuracy, lexical complexity, and fluency of L2-learners. Only syntactic complexity may be decreased by a dialogic task condition. As L1-speakers show similar effects, this seems to be a robust finding. The fact that L1-speakers show a large decrease of syntactic complexity and a decrease of lexical complexity too, may allow the conclusion that a low linguistic complexity be a ‘natural byproduct’ of interaction. The discussion suggests that these findings may be explained by the Alignment Hypothesis (Costa et al. 2008, Pickering and Garrod 2004) and the work of Tavakoli and Foster (2008) that proposes dialogues to be cognitively simpler than monologues such that they give L2-learners more attentional capacity for the language code. As a result, L2-learners make fewer errors in dialogues than in monologues. Furthermore, dialogues may give two learners the opportunity to benefit from each other by copying vocabulary items and help each other out at pauses in order to increase fluency. These are important insights, in particular for language testing. It seems that L2-learners show more of their competences when they act in pairs than when they work on their own. Monologues, however, give the opportunity to build complex syntactic structures. Finally, this study highlights the added value of including a native speaker baseline in task-based research. The interpretation of the L2-data gained a lot from the comparison with L1-data.

Chapter 5: The use of conjunctions in cognitively simple versus complex L2tasks (Study 2b) Chapter 5 reviews a more elaborate analysis of the data generated by study 2. In a reaction to critiques of the Cognition Hypothesis, Robinson claimed that the current global measures of linguistic complexity, accuracy, and fluency, may not be sensitive enough for differences in cognitive task complexity. Robinson and Gilabert (2007), therefore, call for task specific measures. Measures that are

192

English summary

specifically investigating the structures induced by a task may show the predicted effects of cognitive task complexity. In order to test this claim, this study investigated the data of the L1-speakers and L2-learners of the dating and study task of study 2. This time the focus was on a task specific measure. As these are argumentative tasks, you may expect complex tasks to induce a higher amount of reasoning. Reasoning may be marked by lexical elements like conjunctions. This investigation, therefore, tested the data for the frequency (the number of conjunctions per 100 words) and occurrence (the number of conjunctions that was used at least once per performance) of a total of 28 conjunctions. Five specifically task relevant conjunctions, that were expected to be in particular evoked by the argumentative tasks at hand, were investigated in more detail: Dutch ‘want’, ‘omdat’ (both meaning English ‘because’), ‘daarom’, ‘daardoor’ (both translate into English ‘therefore’), and ‘als. . . dan’ (English ‘if. . . then’). Again, the analysis looked at the frequency and occurrence of these specifically task relevant conjunctions. The statistical analysis focused on effects of task complexity because conjunctions are not considered to be specific with respect to the factor interaction. Results of multivariate ANOVAs showed no effects of task complexity – neither for the frequency nor the occurrence of conjunctions. A Wilcoxon Signed Ranks test that investigated the five specifically task relevant conjunctions did not yield any significant results that would support the hypothesis. In both language groups there was one exception only. Complex tasks generated a significantly lower use of one specifically task relevant conjunction than simple tasks with respect to ‘omdat’ for L2-learners and ‘daarom’ for L1-speakers. The discussion critically reviews the measures used in the current study and discusses the pros and cons of global CAF-measures versus task specific measures. This chapter gives no support to Robinson’s claims that specific measures have an added value above global CAF-measures. The global and task specific measures point towards the same direction. Apparently, cognitively complex tasks that are manipulated by means of the resource-directing factor ± few elements may not affect attentional allocation during task-based performance so that L2-learners (or L1-speakers) are pushed towards a qualitatively different language use. This conclusion may hold for both, global CAF-measures (as reported in chapter 4) and for task-specific measures of task performance. This study, therefore, suggests to, rather than using specific measures, focus on measures of communicative adequacy (as in de Jong et al. 2007, Kuiken et al. 2010). Another more favorable approach may be to interpret L2-performance in light of L1-speakers’ baseline data – as has been done in the studies at hand.

Chapter 6: Summary, discussion, and practical implications The final chapter summarizes the empirical work as presented in chapters 3, 4, and 5. It discusses the data in light of the theoretical framework given in chapter 1 and puts it in relation to the hypotheses

193 formulated in chapter 2. Table 6.1 on p. 123 gives an overview of the results of the empirical work. In sum, there are three measures out of twelve in study 1, one measure out of ten in study 2a, and none of the task specific measures of study 2b that give support to the Cognition Hypothesis. This holds with respect to effects of cognitive task complexity on its own as well as for task complexity and interaction in combination. Therefore, the present work does not support Robinson’s theory. Together, this leads to the conclusion that the factor ± few elements, as manipulated in the present dissertation, hardly affects attentional allocation in L2-learners (and L1-speakers) during task-based performance. The discussion highlights several possible explanations for these results. First, it may be that the only robust effect found (an increase in lexical complexity) is based on the longer speech samples or the input material of the complex tasks. Second, it may be that earlier work that found confirming results for the Cognition Hypothesis suffers of a confound of the two factors ± few elements and ± reasoning demands. Third, there seems to be a circularity when trying to define task complexity (Norris and Ortega 2009a). Fourth, the data do not support Skehan’s model of Limited Attentional Capacity either (Skehan 1996). The present studies largely support the hypothesis concerning effects of interaction, that is a manipulation of the factor ± monologic. In both study 1 and study 2a, interactive tasks generate a higher accuracy and fluency, but a lower syntactic complexity. In study 2a, also lexical complexity decreases from monologic to dialogic tasks. With the exception of lexical complexity, L1-speakers mirror this behavior. Accordingly, monologic tasks give the opportunity to build complex syntactic structures, while dialogues give L2-learners the chance to profit from each other. In the end, dialogues may let them show more of their competencies concerning lexical complexity, accuracy, and fluency. The discussion reviews possible explanations. It may be that the factor ± monologic affects attentional allocation of L2-learners. The Alignment Hypothesis (Costa et al. 2008, Pickering and Garrod 2004) states that dialogues are cognitively simpler as they allow priming and alignment so that speakers need less cognitive capacity (i.e., attention) for the underlying speech production processes. Tavakoli and Foster (2008) predict that in dialogues interactants may benefit from planning time during the speaking partner’s turn so that again, dialogues generate more attention for the language form. Finally, it seems that the factor ± monologic apart from the interactive aspects that are predicted by the Cognition Hypothesis (Robinson 2005) and the Interaction Hypothesis (Long 1989) also cognitive aspects of task performance may be affected by this factor. See Table 6.2 on p. 135 for a summary of this discussion. Concerning the measures of task performance, the discussion highlights the value of global CAFmeasures. In the present studies these global constructs yield the same information as task specific measures. Furthermore, the added value of L1-baseline data is discussed. Finally, this chapter names some practical implications of the research at hand. Focusing on the two

194

English summary

factors ± few elements and ± monologic there are at least two important insights. First, even though a higher cognitive complexity by means of a higher number of elements did not result in the predicted effects on accuracy, they did not harm task performance either. As a complex task may be more challenging than a simple task, this research gives the advice to use tasks with many elements that may ask for reasoning. The present studies show that L2-learners produce more speech on complex tasks, which increases the opportunity for uptake and intake of task relevant information. Second, it seems that during monologic tasks L2-learners generate more complex syntactic structures, while all other linguistic aspects (lexical complexity, accuracy, and fluency) may be pushed during dialogic tasks. It is important to be aware of these differences. Especially in a testing context, where L2-learners mostly act on their own, this research may indicate that a dialogic testing setting possibly would be more favorable. It seems that L2-learners are able to show more of their competences when they act in pairs. For the L2-classroom the studies at hand suggest that tasks can be systematically manipulated by the factors ± few elements and ± monologic – generating cycles of simple to complex and monologic to dialogic tasks. Different aspects of task performance may benefit from the various settings. Concerning dialogic tasks it is important to consider the interactive and the cognitive sides of this manipulation.

Conclusion To sum up, the studies in this dissertation give no support for the Cognition Hypothesis (Robinson 2005) with respect to the factor ± few elements. The task-based performances of L2-learners and L1-speakers showed hardly any effects of a more focused attention to form induced by a higher number of elements and a related higher cognitive complexity. With respect to the factor ± monologic the present studies have shown that interactive tasks lead L2-learners to increase their lexical complexity, accuracy, and fluency, while monologic tasks yield more complex syntactic structures. It may be that this is based on more attention to form in interactive tasks. Importantly, the factor ± monologic influences interactive and cognitive aspects of task-based performance in a second language.


Dit proefschrift onderzoekt de effecten van cognitieve taakcomplexiteit en interactie op de mondelinge prestaties van Turkse en Marokkaanse leerders van het Nederlands als tweede taal (NT2) binnen het kader van de taakgerichte aanpak.

Hoofdstuk 1: Theoretisch kader Hoofdstuk 1 geeft het theoretisch kader van dit proefschrift. Het begint met een algemene uitleg over de taakgerichte aanpak van onderzoek naar het leren van een tweede taal (T2) maar richt zich in het bijzonder op de cognitieve tak van taakgericht onderzoek. Het hoofdstuk mondt uit in een uitgebreide bespreking van de Cognitiehypothese van Robinson (2001a, b, 2003b, 2005) aangezien dit proefschrift een aantal specifieke stellingen van deze theorie onderzoekt. Eerst worden de noties cognitieve capaciteit en aandachtsverdeling tijdens taakprestaties besproken. Daarna gaat het hoofdstuk dieper in op twee factoren die volgens Robinson van invloed zijn op de verdeling van aandacht tijdens taakgerichte T2-productie: cognitieve factoren van taakcomplexiteit en interactieve factoren van taakconditie (vgl. ook het Triadic Componential Framework op p. 23). De Cognitiehypothese voorspelt dat zogenaamde ’resource-directing’ factoren van taakcomplexiteit de aandacht van een T2-leerder op de talige vorm van een taakprestatie richten (Robinson 2001a, b, 2003b, 2005). Een complexe taak, waarin je veel elementen moet beschrijven, vraagt bijvoorbeeld om specifieke woorden en veel verschillende structuren om alle elementen van elkaar te kunnen onder´ en scheiden. In een taak met bijvoorbeeld alleen twee elementen zou je kunnen volstaan met ‘de e´ en de ander’. In een complexe taak met bijvoorbeeld zes elementen moet je vaker zinnen gebruiken als ´ ‘die e´ ene links boven naast die ander onder in de hoek’. Robinson voorspelt dat in een cognitief complexe taak leerders ook lingu¨ıstisch complexere structuren en een uitgebreidere woordenschat moeten gebruiken. Hierdoor richten ze hun aandacht op de vorm zodat ze ook minder fouten maken. Dit alles gaat wel ten koste van de vloeiendheid, d.w.z. ze worden langzamer dan in cognitief simpele taken. Ook interactieve taken richten de aandacht van leerders op de vorm. Immers is onderling begrip in interactieve taken cruciaal. Hierdoor letten T2-leerders meer op de talige vorm en gaan minder fouten maken. Door onderbrekingen, vragen om opheldering en ander beurtwisselingsgedrag wordt de

196


lingu¨ıstische complexiteit in interactieve taken naar verwachting lager dan als mensen in hun eentje te werk gaan. Robinson voorspelt dat de verhoogde aandacht voor de vorm ten koste van de vloeiendheid werkt, zodat deze daalt in interactieve taken. Een interessante voorspelling maakt de Cognitiehypothese met betrekking tot een combinatie van de twee factoren, cognitieve taakcomplexiteit en interactie. Wat gebeurt er als twee T2-leerders cognitief complexe taken in interactie vervullen? Robinson voorspelt dat T2-leerders minder fouten maken dan in simpele interactieve taken of elke vorm van monologische taak. Immers dragen allebei de factoren bij tot een verhoogde accuratesse – ten koste van de vloeiendheid die daalt in complexe interactieve taken. Met betrekking tot lingu¨ıstische complexiteit verwacht Robinson dat complexe taken meer vragen opwerpen dan hun simpele tegenhangers. Dit resulteert in meer interactie, zodat de lingu¨ıstische complexiteit daalt. Deze voorspellingen van Robinson worden in het onderhavige proefschrift onderzocht. Dit hoofdstuk wijdt een sectie aan de maten waarmee de taalproductie van tweedetaalleerders gescored worden. De traditionele maten van productie lingu¨ıstische complexiteit, accuratesse en vloeiendheid/fluency (in het kort CAF-maten) staan de afgelopen jaren ter discussie (zie bijvoorbeeld Housen en Kuiken 2009). Vooral met het oog op de evtentuele meerwaarde van zogenaamde taakspecifieke maten (Robinson en Gilabert 2007, Cadierno en Robinson 2009) levert het onderhavige proefschrift een bijdrage aan deze discussie . Het eerste hoofdstuk besluit met een bespreking van eerder onderzoek naar effecten van cognitieve taakcomplexiteit en interactie op de talige prestaties van T2-leerders. Hierbij ligt de focus op open vragen die de aanleiding gaven voor het onderzoek in dit proefschrift. Bijvoorbeeld wordt er op ingegaan dat er tot noch toe nauwlijks onderzocht is hoe de talige prestatie in de moedertaal (T1) be¨ınvloed wordt door cognitieve taakcomplexiteit en interactie.

Hoofdstuk 2: Het onderhavig onderzoek in dit boek Het tweede hoofdstuk geeft een gedetailleerde beschrijving van het empirische werk dat in dit proefschrift gepresenteerd wordt. Hier komen ook de onderzoeksvragen en hypotheses aan de orde (zie de tabel op p. 39 en de uitleg op p. 40). De focus ligt op de twee daadwerkelijk te manipuleren variabelen: de resource-directing cognitieve factor van taakcomplexiteit ‘± weinig elementen’ en de interactieve factor van taakconditie ‘one-way versus two-way flow of information’ die in dit proefschrift omgevormd wordt tot de factor ‘± monoloog’. De hypotheses volgen Robinsons Cognitiehypothese zoals hier boven beschreven maar wijken in een paar cruciale punten af aangezien zij uitgaan van een alternatieve visie op het effect van de interactieve factor ‘± monoloog’. Ook een eventueel gecombineerd effect van cognitieve taakcomplexiteit en interactie wijkt af von Robinsons visie. Baserend op psycholingu¨ıstisch onderzoek in het kader

197 van de Alignmenthypothese (Costa et al. 2008, Pickering en Garrod 2004) een taakgericht onderzoek van Tavakoli and Foster (2008) stelt het onderhavige proefschrift dat dialogische taken cognitief simpeler zouden kunnen zijn dan monologische taken. Een gevolg van deze aanname zou zijn dat in interactie de talige productie van T2-leerders op accuratesse en vloeiendheid vooruit gaat, terwijl alignment en talige routines tussen twee sprekers in dialoog ervoor zorgen dat de lingu¨ıstische complexiteit daalt. Aangezien er bovendien aangenomen wordt dat het effect van interactie sterker is dan dat van cognitieve taakcomplexiteit, zal ook in complexe interactieve taken de vloeiendheid (anders dan door Robinson voorspelt) verhoogd worden ten opzichte van simpele interactieve taken of elke vrom van monologische taak. Dit proefschrift voorspelt m.b.t. een taakspecifieke maat voor argumentatieve taken dat een complexe taak een hogere frequentie en voorkomen van connectieven zal oproepen in het talige gedrag van T2-leerders. Terwijl taakcomplexiteit verwacht wordt nauwelijks van invloed te zijn op het talige gedrag van T1-sprekers zal interactie vergelijkbare effecten vertonen als in T2-leerders. Het vervolg van hoofdstuk 2 beschrijft nauwkeurig de opzet en de uitvoering van het empirische werk in de daaropvolgende hoofdstukken.1 Ten slotte wordt in dit hoofdstuk uiteengezet hoe de verschillende studies bij elkaar aansluiten en waarom welke aanpassingen (bijvoorbeeld m.b.t. de maten) in het design van de chronologisch op elkaar volgende empirische onderzoeken gemaakt zijn.

Hoofdstuk 3: Het effect van cognitieve taakcomplexiteit in monologische versus dialogische taken in het Nederlands als T2 (studie 1) Hoofdstuk 3 beschrijft een eerste studie naar de effecten van taakcomplexiteit en interactie op het talige gedrag van T2-leerders. 46 hoog opgeleide NT2-leerders van Turkse en Marokkaanse komaf met een gemiddelde taalvaardigheid in het Nederlands presteerden op cognitief simpele en complexe taaltaken. Zij kregen de opdracht om een vriend te adviseren over de keuze voor een electronisch apparaat, bijvoorbeeld de beste mobiele telefoon. In navoling van de Cognitiehypothese was taakcomplexiteit gemanipuleerd m.b.t. de resource-directing factor ± weinig elementen, d.w.z. de simpele taak gaf twee apparaten ter keuze, de complexe taak zes (zie Appendix A.4). De helft van de deelnemers deed de taak in zijn of haar eentje (monoloog) de andere helft werkte in duo’s (dialoog). In dit 2×2 design werden cognitieve taakcomplexiteit en interactie gecounterbalanced aangeboden. De 92 monologische en dialogische taalprestaties van hoge en lage complexiteit zijn getranscribeerd en gecodeerd op 12 globale maten van lingu¨ıstische complexiteit, accuratesse en vloeiendheid (zie tabel p. 55). Een variantieanalyse controlleerde voor effecten van allebei de factoren, taakcomplexitiet en interactie, apart en in combinatie. Hieruit bleek dat complexe taken zoals voorspeld een hogere 1 Aangezien de empirische hoofdstukken 3, 4 en 5 als aparte artikelen of boekhoofdstukken elders gepubliceerd worden of zijn, is er in de betreffende hoofdstukken zelf relatief weinig plek om alle achtergrondinformatie te bespreken. Een ander gevolg van deze prepublicaties is dat de er enige overlap en redundantie is m.b.t. de uitleg van de Cognitiehypothese aangezien deze voor alle drie de empirische studies de basis vormt.

198


lingu¨ıstische complexiteit (op een lexicale maat), een hogere accuratesse (op het totaal aantal fouten) en een lagere vloeiendheid (in syllabes per minuut) scoorden. Deze effecten werden echter alleen voor ´ maat per CAF-construct gevonden (in totaal dus op 3 van de 12 maten). Alle andere maten telkens e´ en waren niet significant door taakcomplexiteit be¨ınvloed. Interactie had een statistisch relevant effect op allebei de syntactische maar geen van de lexicale maten van lingu¨ıstische complexiteit (monologen waren syntactisch complexer dan dialogen), op drie maten van accuratesse (die was hoger in dialogen dan monologen) en op alle drie de vloeiendheidsmaten (monologen veroorzaakten langzamere spraak met meer pauzes dan dialogen). Alleen accuratesse liet een significant gecombineerd effect van taakcomplexiteit en interactie zien: het gevonden verhogende effect van taakcomplexiteit op accuratesse in monologische taken verdween in dialogische taken. De discussie legt uit dat de data m.b.t. syntactische complexiteit, acuratesse en vloeiendheid ten dele in lijn zijn met de verwachtingen van de Cognitiehypothese al is de bevestiging (op drie maten van de twaal) niet erg sterk. Bovendien worden er geen bevestigende gecombineerde effecten gevonden. Interactie vertoonde daarentegen de voorspelde bevorderende effecten op accuratesse en had verlagende effecten op (syntactische) complexiteit. Er was een onverwachts bevorderend effect op vloeiendheid. Aangezien dit het eerste onderzoek is dat systematisch kijkt naar (gecombineerde) effecten van taakcomplexiteit en interactie pleit dit hoofdstuk voor nieuw empirisch werk dat probeert een duidelijker beeld te krijgen met betrekking tot deze twee factoren - elk apart en in combinatie. Bovendien stelt het voor om in toekomstig taakgericht onderzoek ook baseline data van moedertaalsprekers te verzamelen om zo de data van T2-leerders beter te kunnen interpreteren.

Hoofdstuk 4: Effecten van taakcomplexiteit en interactie op prestaties in de T2 (studie 2a) Het vierde hoofdstuk doet verslag van een tweede empirisch onderzoek naar effecten van taakcomplexiteit en interactie op de talige prestaties van T2-leerders. Deze keer werkten in totaal 64 Turkse en Marokkaanse NT2-leerders mee die wederom een gemiddeld taalniveau in het Nederlands bereikt hadden. Bovendien werkten 44 moedertaalsprekers van het Nederlands mee. Alle deelnemers waren hoog opgeleid. Aangezien de taken van de eerste studie soms maar weinig spraak opleverden, werd en tweetal nieuwe taken ontworpen (de ‘studie’ en de ‘dating’ taak, zie Appendix B.4) die meer spraak zouden moeten oproepen. Deelnemers werden gevraagd om de beste combinatie van twee mensen te kiezen (een studiekoppel in de studietaak en een liefdesstel in de datingtaak). Het verschil in complexiteit was verhoogd (met betrekking tot de taken in de eerste studie) door in de simpele taak om een afweging te vragen tussen vier mogelijke combinaties terwijl de complexe taak negen mogelijke stellen presen-

199 teerde. Wederom was het een 2×2-design, d.w.z. alle deelnemers deden een simpele en een complexe taak in een counterbalancede volgorde. De ene helft deed allebei de taken in hun eentje in een mono¨ in de dialogische conditie. logische conditie en de andere helft werkte telkens met zijn tweeen Alle 216 spraaksamples werden getranscribeerd en gecodeerd op maten van lingu¨ıstische complexiteit, accuratesse en vloeiendheid. Naar aanleiding van de eerste studie zijn wat onzuiverheden ´ lexicale maat van complexm.b.t. de maten aangepast zodat er deze keer twee syntactische en e´ en iteit gebruikt werden, er voor accuratesse gecodeerd werd met het oog op lexicale, morfosyntactische ´ van pauzegedrag en twee van en lidwoordfouten en voor vloeiendheid twee maten van snelheid, e´ en verbetering onderzocht werden. De statistische bewerking omvatte aparte variantieanalyses voor elk CAF-construct apart en voor elke taalgroep (T2 versus T1) apart met cognitieve taakcomplexiteit en interactie als factoren. De re´ enkel statistisch significant effect op de lexicale maat van sulaten geven in allebei de taalgroepen e´ en lingu¨ıstische complexiteit (die was hoger in complexe taken). Geen enkele andere maat liet verder een effect zien van taakcomplexiteit. Voor T2-sprekers zijn er ook geen gecombineerde significante effecten van taakcomplexiteit en interactie. T1-sprekers laten echter een gecombineerd effect zien: in monologen heeft taakcomplexiteit geen invloed op de vloeiendheid van T1-sprekers maar in dialogen zijn ze in de complexe taken langzamer dan in simpele taken. De effecten van interactie zijn in allebei de taalgroepen veelvuldig en sterk (zoals te zien is aan effectgroottes, bijvoorbeeld tabel 4.7 op p. 81). T2-leerders zijn in dialogen syntactisch minder complex maar lexicaal complexer, accurater op alle drie de maten, en vloeiender op alle maten. T1-sprekers laten een vergelijkbaar beeld zien al zijn zij niet op alle maten even sterk be¨ınvloed. Interessant is dat in dialogische taken ook hun accuratesse vooruit gaat terwijl de lexicale maat van complexiteit daalt (in tegenstelling tot T2-leerders). Deze studie levert als totaal nauwelijks bevestiging voor Robinsons Cognitiehypothese (zie ook de samenvattende tabel 4.12 op p. 86 die laat zien welke voorspelde effecten door de data bevestigd worden). Het enkele effect van taakcomplexiteit (op lexicale complexiteit) kan eventueel verklaard worden door de extra hoeveelheid spraak die de complexe taken opleverden en door het grotere aantal woorden in de input. Aangezien ook moedertaalsprekers dit effect laten zien, is de conclusie dat een complexe taak gemanipuleerd op de factor ± weinig elementen geen kwalitatief andere taalprestatie teweeg brengt. Het lijkt erop dat we alleen een kwantitatief effect vinden: complexe taken leveren meer spraak op en laten dus een grotere lexicale complexiteit zien. Anders zit het met het effect van interactie. Dialogische taken verhogen de accuratesse, lexicale complexiteit en vloeiendheid van T2-sprekers. Alleen de syntactische complexiteit neemt af. Aangezien ook T1-leerders deze invloeden laten zien, lijkt dit een robuust gegeven te zijn. Het feit dat moedertaalsprekers een lagere syntactische en lexicale complexiteit vertonen, duidt erop dat een lagere lingu¨ıstische complexiteit een ’natuurlijk’ gevolg van interactie is. De discussie vermoedt dat in lijn met

200


de Alignmenthypothese (Costa et al. 2008, Pickering and Garrod 2004) en Tavakoli and Foster (2008) dialogen cognitief simpeler zijn dan monologen waardoor T2-sprekers meer tijd een aandacht hebben voor hun talige productie. Hierdoor maken ze minder fouten. Bovendien kunnen ze in dialoog van elkaar profiteren doordat ze elkaars woordenschat overnemen en vloeiender worden. Dit zijn belangrijke inzichten voor onderwijs en vooral toetsing van T2-leerders. Kennelijk kunnen leerders in samenwerking meer van hun vermogens laten zien dan als ze alleen te werk gaan. Een monologische setting geeft wel de mogelijkheid tot het formuleren van complexe syntactische zinnen. Ten slotte laat deze studie zien hoe waardevol het is om in taakgericht onderzoek baselinedata van moedertaalsprekers te verzamelen. De interpretatie van de T2-data tegen het licht van de T1-gegevens resulteerde in interessante inzichten.

Hoofdstuk 5: Het gebruik van connectieven in cognitief simpele en complexe T2-taken (studie 2b) Hoofdstuk 5 doet verslag van een vervolgonderzoek van de data van studie 2. In een reactie op kritiek aan de Cognitiehypothese stelt Robinson dat de maten van lingu¨ıstische complexiteit, accuratesse en vloeiendheid wellicht niet gevoelig genoeg zijn om de verschillen te laten zien die een verhoogde taakcomplexiteit veroorzaakt. Onder meer Robinson and Gilabert (2007) pleiten daarom voor het gebruik van taakspecifieke maten. Maten die gerelateerd zijn aan de specifieke eisen van een taak zelf zouden de voorspelde effecten wellicht wel laten zien. Om aan deze oproep te voldoen zijn de data van de T1-sprekers en T2-leerders van de dating- en studietaak in studie 2 opnieuw gecodeerd. Deze keer is er gekeken naar een specifieke maat voor de taken die gebruikt werden. Aangezien het om argumentatieve taken ging, zou je kunnen verwachten dat een complexe taak meer argumentatie oproept. Argumentatie kan gemarkeerd worden door lexicale elementen zoals connectieven. Op grond van deze redenering is gekozen om de frequentie (het aantal connectieven per 100 woorden) en het voorkomen (het aantal deelnemers dat een connectief ten minste ´ keer gebruikt) van 28 verschillende connectieven te toetsen. Een vijftal connectieven werd in het e´ en bijzonder als relevant voor de voorliggende taken beschouwd omdat zij causale en conditionele relaties markeren (want, omdat, daarom, daardoor, als. . . dan). De statistische analyse onderzocht alleen effecten van taakcomplexiteit omdat connectieven taakspecifiek zijn voor deze factor maar niet voor een verschil in interactie. De resultaten van multivariate variantieanalyses laten geen effecten van taakcomplexiteit zien voor de frequentie en het voorkomen van alle 28 connectieven. Ook de Wilcoxon Signed Ranks tests voor de vijf specifiek taakrelevante connectieven liet geen significanties zien die de hypothese bevestigen. Er waren slechts twee significante effecten: het voorkomen van ‘omdat’ in de T2-data en de frequentie van ‘daarom’ in de T1-data waren hoger in simpele dan complexe taken.

201 Afsluitend stelt dit hoofdstuk de gebruikte maten ter discussie en bespreekt de voors en tegens van globale CAF-maten tegenover taakspecifieke maten. Deze studie levert geen onderbouwing voor Robinsons claim dat taakspecifieke maten een beter beeld schetsen dan globale CAF-maten. In dit onderzoek wijzen de resultaten van globale en taakspecifieke maten allemaal dezelfde kant op. Kennelijk richten cognitief complexe taken, die gemanipuleerd zijn op de factor ± weinig elementen, de aandacht tijdens de taakprestatie niet op taal zodat T2-leerders (of T1-sprekers) een ander taalgebruik vertonen. Dit geldt voor algemene maten (vgl. hoofdstuk 4) en voor het gebruik van connectieven. Dit onderzoek pleit er daarom voor om in plaats van meer een specifieke maten te gaan gebruiken eerder te proberen om communicatieve adequaatheid te laten beoordelen (zoals bijvoorbeeld in de Jong et al. 2007, Kuiken et al. 2010). Een andere optie is om de globale maten te interpreteren in het licht van een T1-data zoals gedaan is in het voorliggende proefschrift.

Hoofdstuk 6: Samenvatting, discussie, en praktische implicaties Het besluitende hoofdstuk geeft een samenvatting van de drie empirische studies zoals besproken in de hoofdstukken 3, 4 en 5 en bespreekt de data in het licht van het theoretisch kader zoals geschetst in hoofdstuk 1 en in relatie tot de hypotheses zoals geformuleerd in hoofdstuk 2. Tabel 6.1 op p. 123 geeft een overzicht van de resultaten van de empirische studies. Samenvattend ´ maat van de tien, en in studie 2b geven in studie 1 drie van de twaalf maten, in studie 2a alleen e´ en geen enkele van de taakspecifieke maten bevestiging voor de Cognitiehypothese. Dit geldt zowel voor het effect van cognitieve taakcomplexiteit apart als ook voor taakcomplexiteit en interactie in combinatie. De voorliggende studies vormen daarom geen onderbouwing voor Robinsons theorie. Dit samen leidt tot de conclusie dat de factor ± weinig elementen, zoals hij gemanipuleerd is in het voorliggende proefschrift, nauwelijks een effect heeft op de aandacht van T2-leerders (en T1-sprekers) tijdens taakgerichte productie. De discussie bespreekt verschillende mogelijke redenen voor de resultaten. Ten eerste valt het enige robuste effect dat gevonden werd (een verhoogde lexicale complexiteit) wellicht terug te brengen op de langere spreeksamples en de input in complexe taken. Ten tweede zou het kunnen dat in eerder onderzoek, dat wel bevestinging voor de Cognitiehypothese gevonden heeft, dit komt door de factor ± redeneren en niet de factor ± weinige elementen. Ten derde stelt de discussie een mogelijke circulariteit bij het bepalen van taakcomplexiteit vast (Norris en Ortega 2009a). Ten vierde wordt uitgesloten dat de data een bevestiging zouden kunnen leveren voor een model dat uit gaat van beperkte capaciteit van aandacht (Skehan 1996). Het voorliggende onderzoek geeft ruime onderbouwing voor de hypotheses met betrekking tot effecten van interactie, d.w.z. manipulaties van de factor ± monoloog. Zowel in studie 1 als in studie 2a heeft interactie een verhogend effect op acuratesse en vloeiendheid en een verlagend effect op syntactische complexiteit. In studie 2a gaat ook de lexicale complexiteit omhoog in dialogen. Dit plaatje wordt

202


met uitzondering van de lexicale complexiteit door T1-sprekers bevestigd. Aldus geven monologen de mogelijkheid tot het bouwen van complexe syntactische structuren terwijl dialogen T2-leerders in staat stellen om veelvuldig van elkaar te profiteren en meer van hun competenties laten zien m.b.t. lexicale complexiteit, accuratesse en vloeiendheid. De discussie brengt een mogelijke onderliggende reden ter sprake. Het zou kunnen zijn dat de factor ± monoloog juist wel een invloed heeft op de aandacht van T2-leerders. De Alignmenthypothese (Costa et al. 2008, Pickering en Garrod 2004) vermoedt dat dialogen cognitief simpeler zijn omdat priming en alignmnet ertoe leiden dat er minder cognitieve capaciteit en dus aandacht voor de onderliggende taalproductieprocessen nodig is. (Tavakoli en Foster 2008) voorspellen bovendien dat er in een dialoog tijdens de spreekbeurt van de partner planningtijd is om de eigen beurt voor te bereiden. Ook dit leidt ertoe dat er in dialogen meer cognitieve capaciteit en dus aandacht is voor taal. Tot besluit, lijkt het erop dat naast de interactieve aspecten die door bijvoorbeeld de Cognitiehypothese (Robinson 2005) en Interactiehypothese (Long 1989) voorspeld worden de factor ± monoloog ook cognitieve aspecten van taakprestaties be¨ınvloedt. Zie tabel 6.2 op p. 135 voor een samenvatting van deze discussie. Met betrekking tot de gebruikte maten noemt de discussie de waarde van globale CAF-maten – die in dit onderzoek dezelfde informatie opleveren als de taakspecifieke maat – en volgt er een pleidooi voor een controlegroep van moedertaalsprekers die veel nuttige informatie brengt als het om de interpretatie van T2-data gaat. Tot besluit gaat dit hoofdstuk kort in op praktische implicaties van dit onderzoek. Met de focus op de twee factoren ± weinig elementen en ± monoloog springen er twee belangrijke punten uit. Ten eerste, ook al leverde een verhoogde taakcomplexiteit met behulp van meer elementen niet de verwachte effecten op accuratesse, er waren ook geen nadelige effecten te bekennen. Aangezien een complexe taak met meer elementen naar verwachting uitdagender is dan een simpele taak, is het advies voor de praktijk om vooral taken te gebruiken met meer elementen die om redeneerprocessen vragen. Uit dit onderzoek blijkt dat T2-leerders dan meer spraak produceren wat de mogelijkheid tot uptake en intake van informatie verhoogt die nodig is voor de taakprestatie. Ten tweede blijkt, dat in monologische taken T2-leerders complexere zinsstructuren gebruiken terwijl alle andere gemeten aspecten van taal (lexicale complexiteit, accuratesse en vloeiendheid) juist toenemen in dialogen. Voor de praktijk is het belangrijk om dit verschil in het achterhoofd te houden. Met name als het gaat om toetsen, waar T2-leerders meestal in hun eentje aan de slag moeten, is het een idee om erover na te denken of het toetsten van twee leerders tegelijkertijd wellicht niet meer recht doet aan hun competenties, aangezien ze bij het samenwerken meer van hun kunnen lieten zien. Voor de lespraktijk in de T2-klas is er het idee om taken cyclisch aan te bieden en daarbij de factoren ± weinig elementen en ± monoloog systematisch te manipuleren. Zo kunnen verschillende aspecten van taakprestatie gestuurd gestimuleerd worden. Hierbij is het belangrijk om m.b.t. interactie met zowel de interactieve als de cognitieve aspecten van deze factor rekening te houden.

203

Conclusie In de laatste alinea’s wordt nog eens stil gestaan bij de conclusie dat de studies in het onderhavige proefschrift geen onderbouwing leveren voor de voorspellingen van de Cognitiehypothese (Robinson 2005) met betrekking tot de factor ± weinig elementen. In de taakprestaties van T2-leerders en T1sprekers zijn geen effecten van verhoogde aandacht voor de taal te vinden als gevolg van een hoger aantal elementen en de daaraan gerelateerde voorspelde hogere cognitieve taakcomplexiteit. Met betrekking tot de factor ± monoloog hebben de voorliggende studies echter laten zien dat interactieve taken T2-leerders ertoe aanzetten om hun lexicale complexiteit, accuratesse en vloeiendheid te verhogen, terwijl monologen het gebruik van complexe syntactische zinsstructuren bevorderen. Het zou kunnen dat dit een gevolg van aandachtsverdeling is tijdens interactieve taken. In ieder geval lijkt het erop dat de factor ± monoloog zowel interactieve als cognitieve aspecten van taakprestaties van tweede taalleerders en moedertaalsprekers be¨ınvloedt.

Bibliography

Baddeley, A. D. (2003). Working memory and language: An overview. Journal of Communication Disorders 36(3), 189–208. Baddeley, A. D. and G. Hitch (1974). Working memory. In G. Bower (Ed.), Recent Advances in Learning and Motivation, Volume 8, pp. 47–90. New York: Academic Press. Boersma, P. and D. Weenink (1992-2006). Praat: Doing phonetics by computer. Computer Program. Retrieved from http://www.fon.hum.uva.nl/praat/ in March 2006. de Bot, K. (1992). A bilingual production model: Levelt’s ’speaking’ model adapted. Applied Linguistics 13(1), 1–24. ´ B. (2007). Measure for measure: Why type/token ratio based measures are not valid to assess Bulte, lexical complexity/richness as a dimension of language proficiency. In S. van Daele, A. Housen, F. Kuiken, M. Pierrard, and I. Vedder (Eds.), Complexity, Accuracy, and Fluency in Second Language Use, Learning, and Teaching, pp. 27–35. Brussel: Koninklijke Vlaamse Academie van Belgie¨ voor Wetenschappen en Kunsten (KVAB). van den Branden, K. (2006). Task-Based Language Education: From Theory to Practice. New York: Cambridge University Press. van den Branden, K., M. Bygate, and J. Norris (2009). Task-Based Language Teaching: A Reader. Amsterdam / Philadelphia: John Benjamins Publishing Company. Bygate, M., P. Skehan, and M. Swain (2001). Researching Pedagogic Tasks: Second Language Learning, Teaching, and Testing. Harlow: Pearson Education. Cadierno, T. and P. Robinson (2009). Language typology, task complexity and the acquisition of L2 lexicalization patterns for reference to motion. Annual Review of Cognitive Linguistics 7(1), 245–276. Candlin, C. (1987). Toward task-based learning. In C. Candlin and D. Murphy (Eds.), Language and Learning Tasks. Englewood Cliffs, NJ: Prentice Hall. CBS (2010). Bevolking: Kerncijfers naar diverse kenmerken. Last consultation of www.cbs.nl in September 2010. Chambers, F. (1997). What do we mean by fluency? System 25(4), 535–544. Cornu, A. and M. Delahaye (1987). Variability in interlanguage reconsidered: LSP vs. non-LSP IL talk. English for specific purposes 6(2), 145–151. Costa, A., M. J. Pickering, and A. Sorace (2008). Alignment in second language dialogue. Language and Cognitive Processes 23(4), 528–556. Cowan, N. (1988). Evolving conceptions of memory storage, selective attention, and their mutual constraints within the human information-processing system. Psychological Bulletin 104(2), 163–191. Cowan, N. (1993). Activation, attention, and short-term memory. Memory and Cognition 21(2), 162– 167. Cowan, N. (2000). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Language and Cognitive Processes 24(1), 87–185. Crookes, G. (1986). Task Classification: A Cross-Disciplinary Review. Honolulu, University of Hawai’i: Center for Second Language Classroom Research, Social Science Research Institute.

206

Bibliography

Crookes, G. (1990). The utterance, and other basic units for second language discourse analysis. Applied Linguistics 11(2), 183–199. van Daele, S. (2007). Lingu¨ıstische vlotheid, accuraatheid en complexiteit in de verwerving en de ¨ verwerking van een tweede taal. Unpublished Ph.D. Thesis, Vrije Universiteit Brussel, Belgie. Daller, H., R. van Hout, and J. Treffers-Daller (2003). Lexical richness in the spontaneous speech of bilinguals. Applied Linguistics 24(4), 197–222. Diessel, H. (2004). The Acquisition of Complex Sentences. New York: Cambridge University Press. ¨ Dornyei, Z. and J. Kormos (1998). Problem-solving mechanisms in L2 communication. A psycholinguistic perspective. Studies in Second Language Learning 20(3), 349–385. Doughty, C. J. (2001). Cognitive underpinnings of focus on form. In P. Robinson (Ed.), Cognition and Second Language Instruction, pp. 206–257. New York: Cambridge University Press. Doughty, C. J. and T. Pica (1986). Information-gap tasks: Do they facilitate second language acquisition? TESOL Quarterly 20(2), 305–326. Eckerth, J. and S. Siekmann (2008). Task-Based Language Learning, and Teaching: Theoretical, Methodological, and Pedagogical Perspectives. Frankfurt: Peter Lang. Ejzenberg, R. (2000). The juggling act of oral fluency: A psycho-sociolinguistic metaphor. In H. Riggenbach (Ed.), Perspectives on Fluency, Chapter 15, pp. 287–313. Ann Arbor, MI: University of Michigan Press. Ellis, R. (2000). Task-based research and language pedagogy. Language Teaching Research 4(3), 193–220. Ellis, R. (2003). Task-Based Language Learning and Teaching. Oxford: Oxford University Press. Ellis, R. (2005). Planning and task-based performance. In R. Ellis (Ed.), Planning and Task-Based Performance in a Second Language, Volume 11, pp. 3–34. Amsterdam / Philadelphia: John Benjamins Publishing Company. Ellis, R. (2009). The differential effects of three types of task planning on fluency, complexity, and accuracy in L2 oral production. Applied Linguistics 30(4), 474–509. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Ellis, R. and X. He (1999). The roles of modified input and output in the incidental acquisition of word meanings. Studies in Second Language Acquisition 21(2), 285–301. Ellis, R. and F. Yuan (2005). The effects of careful within-task planning on oral and written task performance. In R. Ellis (Ed.), Planning and Task-Based Performance in a Second Language, Volume 11. Amsterdam / Philadelphia: John Benjamins Publishing Company. Evers-Vermeul, J. (2005). The Development of Dutch Connectives: Change and acquisition as windows on form-function relations. Utrecht: LOT-Dissertation Series. Ph.D. Thesis, Universiteit Utrecht, the Netherlands. Ferrari, S. (2009). Valutare le competenze orali in italiano L2: Variazione longitudinale e situazionale in apprendenti a livello avanzato. Unpublished Ph.D. Thesis, Universita degli studi di Verona, Italia. Field, A. (2005). Discovering Statistics Using SPSS (2nd ed.). London / Thousand Oaks / New Delhi: Sage Publications. Fiksdal, S. (2000). Fluency as a function of time and rapport. In H. Riggenbach (Ed.), Perspectives on Fluency, Chapter 8, pp. 128–140. Ann Arbor, MI: University of Michigan Press. Foster, P. (2000). Attending to message and medium: The effects of planning time on the task-based language performance of native and non-native speakers. Unpublished Ph.D. Thesis, King’s College London, UK. Foster, P. (2001). Rules and routines: A consideration of their role in the task-based language production of native an non-native speakers. In M. Bygate, P. Skehan, and M. Swain (Eds.), Researching Pedagogic Tasks: Second Language Learning, Teaching, and Testing, pp. 75–93. Harlow: Longman. Foster, P. and P. Skehan (1996). The influence of planning on performance in task-based learning. Studies in Second Language Acquisition 18 (3), 299–324. Foster, P. and P. Skehan (1999). The effect of source of planning and focus on planning on task-based performance. Language Teaching Research 3(3), 185–215.

Bibliography

207

Foster, P. and P. Tavakoli (2009). Native speakers and task performance: Comparing effects on complexity, fluency, and lexical diversity. Language Learning 59(4), 866–896. Foster, P., A. Tonkyn, and G. Wigglesworth (2000). Measuring spoken language: A unit for all reasons. Applied Linguistics 21(3), 354–375. Freed, B. F. (2000). Is fluency like beauty in the eyes (and ears) of the beholder? In H. Riggenbach (Ed.), Perspectives on Fluency, Chapter 13, pp. 243–265. Ann Arbor, MI: University of Michigan Press. Friederici, A. D., K. Steinhauer, and S. Frisch (1999). Lexical integration: Sequential effects of syntactic and semantic information. Memory and Cognition 27(3), 438–453. Gass, S. M. (2003). Input and interaction. In C. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, pp. 224–255. Malden: Blackwell Publishing Ltd. Gass, S. M. and E. Varonis (1994). Input, interaction, and second language production. Studies in Second Language Acquisition 16(3), 283–302. Gass, S. M. and A. Mackey (2006). Input, interaction, and output. AILA Review 19, 3–17. Gass, S. M., A. Mackey, and T. Pica (1998). The role of input and interaction in second language acquisition: Introduction to the special issue. Modern Language Journal 82(3), 299–307. Gilabert, R. (2005). Task complexity and L2 narrative oral production. Unpublished Ph.D. Thesis, Universitat de Barcelona, Spain. Gilabert, R. (2007a). Effects of manipulating task complexity on self-repairs during L2 oral production. International Review of Applied Linguistics 45(3), 215–240. Gilabert, R. (2007b). The simultaneous manipulation of task complexity along planning time and ± Here-and-Now: Effects on L2 oral production. In M. del Pilar Garc´ıa Mayo (Ed.), Investigating Tasks in Formal Language Learning, pp. 136–156. Clevedon: Multilingual Matters. Gilabert, R., J. Baron, and A. Llanes (2009). Manipulating cognitive complexity across task types and its impact on learners’ interaction during oral performance. International Review of Applied Linguistics 47(3–4), 367–395. ´ T. (1985). Function, structure, and language acquisition. In D. I. Slobin (Ed.), The Crosslinguistic Givon, Study of Language Acquisition, pp. 1008–1025. Hillsdale, NJ: Erlbaum. ´ T. (1995). Functionalism and Grammar. Amsterdam / Philadelphia: John Benjamins Publishing Givon, Company. Goldreich, O. (2008). Computational Complexity: A Conceptual Perspective. New York: Cambridge University Press. ` Guiraud, P. (1954). Les caracteres statistiques du vocabulaire. Paris: Presses universitaires de France. Halford, G. S., N. Cowan, and G. Andrews (2007). Separating cognitive capacity from knowledge: A new hypothesis. Trends in Cognitive Sciences 11(6), 236–242. Housen, A. and F. Kuiken (2009). Complexity, accuracy, and fluency in second language acquisition. Applied Linguistics 30(4), 461–473. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Ishikawa, T. (2007). The effect of manipulating task complexity along the ± Here-and-Now-dimension on L2 written narrative discourse. In M. del Pilar Garc´ıa Mayo (Ed.), Investigating Tasks in Formal Language Learning, pp. 136–156. Clevedon: Multilingual Matters. Iwashihta, N., T. MacNamarra, and C. Elder (2001). Can we predict task difficulty in an oral proficiency test? Exploring the potential of an information-processing approach to task design. Language Learning 51(3), 401–436. Jacobs, D. (2006). Scanner recorder 1.9. Computer Program. Retrieved from http://www.davee.com/ scanrec/ in January 2006. de Jong, N. H., M. P. Steinel, A. Florijn, R. Schoonen, and J. H. Hulstijn (2007). The effect of task complexity on fluency and functional adequacy of speaking performance? In S. van Daele, A. Housen, F. Kuiken, M. Pierrard, and I. Vedder (Eds.), Complexity, Accuracy, and Fluency in Second Language Use, Learning, and Teaching, pp. 53–64. Brussel: Koninklijke Vlaamse Academie van Belgie¨ voor Wetenschappen en Kunsten (KVAB). de Jong, N. H. and T. Wempe (2009). Praat script to detect syllable nuclei and measure speech rate

208

Bibliography

automatically. Behavior Research Methods 41(2), 385 –390. Kim, Y. (2009). The effects of task complexity on learner-learner interaction. System 37(2), 254–268. Kirschner, F., L. Kester, and G. Corbalan (2011). Cognitive load theory and multimedia learning, task characteristics and learning engagement: The current state of the art. Computers in Human Behavior 27(1), 1–4. Kormos, J. (1999). Monitoring and self-repair in L2. Language Learning 49(2), 303–342. Kormos, J. (2000a). The role of attention in monitoring second language speech production. Language Learning 50(2), 343–384. Kormos, J. (2000b). The timing of self-repairs in second language speech production. Studies in Second Language Learning 22(2), 145–167. ´ Kormos, J. and M. Denes (2004). Exploring measures and perceptions of fluency in the speech of second language learners. System 32(2), 145–164. Krashen, S. D. (1985). The Input Hypothesis: Issues and Implications. Harlow: Longman. Kroll, J. F. and G. Sunderman (2003). Cognitive processes in second language learners and bilinguals: The development of lexical and conceptual representations. In C. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, Chapter 5, pp. 104–129. Malden: Blackwell Publishing Ltd. Kuiken, F., M. Mos, and I. Vedder (2005). Cognitive task complexity and second language writing performance. In S. Foster-Cohen, M. del Pilar Garc´ıa Mayo, and J. Cenoz (Eds.), Eurosla Yearbook, Volume 5, pp. 195–222. Amsterdam / Philadelphia: John Benjamins Publishing Company. Kuiken, F. and I. Vedder (2002). The effect of interaction in acquiring the grammar of a second language. International Journal of Educational Research 37(3–4), 343–358. Kuiken, F. and I. Vedder (2004a). De relatie tussen cognitieve taakcomplexiteit en lingu¨ıstische performance bij het schrijven in T1 en T2. Toegepaste Taalwetenschap in Artikelen 72, 23–32. Kuiken, F. and I. Vedder (2004b). Il bed & breakfast piu bello d’Italia! Cognitieve taakcomplexiteit en tekstkwaliteit in Italiaans T2. Incontri 19(1), 31–39. Kuiken, F. and I. Vedder (2007a). Cognitive task complexity and linguistic performance in French L2 writing. In M. del Pilar Garc´ıa Mayo (Ed.), Investigating Tasks in Formal Language Learning, pp. 117–135. Clevedon: Multilingual Matters. Kuiken, F. and I. Vedder (2007b). Task complexity and measures of linguistic performance in L2 writing. International Review of Applied Linguistics 45(3), 261–284. Kuiken, F. and I. Vedder (2007c). Task complexity, task characteristics, and measures of linguistic performance. In S. van Daele, A. Housen, F. Kuiken, M. Pierrard, and I. Vedder (Eds.), Complexity, Accuracy, and Fluency in Second Language Use, Learning, and Teaching, pp. 113–125. Brussel: Koninklijke Vlaamse Academie van Belgie¨ voor Wetenschappen en Kunsten (KVAB). Kuiken, F. and I. Vedder (2008). Cognitive task complexity and written output in Italian and French as a foreign language. Journal of Second Language Writing 17(1), 48–60. Kuiken, F., I. Vedder, and R. Gilabert (2010). Communicative adequacy and linguistic complexity in L2 writing. In I. Bartning, M. Martin, and I. Vedder (Eds.), Communicative Proficiency and Linguistic Development. Intersections between SLA and Language Testing Research, Volume 1 of Eurosla Monographs Series, pp. 81–99. Lantolf, J. P. (2000). Sociocultural Theory and Second Language Learning. Oxford: Oxford University Press. Laufer, B. and P. Nation (1995). Vocabulary size and use: Lexical richness in L2 written production. Applied Linguistics 16(3), 307–322. Lennon, P. (1991). Error: Some problems of definition, identification and distinction. Applied Linguistics 12(2), 180–195. Levelt, W. J. (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Long, M. H. (1983). Native speaker/non-native speaker conversation and the negotiation of comprehensible input. Applied Linguistics 4(2), 126–141. Long, M. H. (1985). A role for instruction in second language acquisition: Task-based language teach-

Bibliography

209

ing. In K. Hyltenstam and M. Pienemann (Eds.), Modeling and Assessing Second Language Acquisition, pp. 77–99. Clevedon: Multilingual Matters. Long, M. H. (1989). Task, group, and task-group interaction. University of Hawai’i Working Papers in English as a Second Language 8(2), 1–26. Long, M. H. (1990). Task, group, and task-group interactions. In A. Sarinee (Ed.), Language Teaching Methodology for the Nineties: Anthology Series, Volume 24. Long, M. H. and P. Robinson (1998). Focus on form: Theory, research, and practice. In C. J. Doughty and J. Williams (Eds.), Focus on Form in Classroom Second Language Acquisition, pp. 15–41. New York: Cambridge University Press. Loschky, L. and R. Bley-Vroman (1993). Grammar and task-based methodology. In S. M. Gass and G. Crookes (Eds.), Tasks and Language Learning: Integrating Theory and Practice, Volume 93. Clevedon: Multilingual Matters. Lyman, F. (1987). Think-pair-share: An expanding teaching technique. MAACIE, Cooperative News 1(1), 1–2. Mackey, A. (1999). Input, interaction and second language development. Studies in Second Language Acquisition 21(4), 557–587. MacWhinney, B. (2000). The childes project: Tools for analyzing talk. Computer Program. Retrieved from http://childes.psy.cmu.edu/ in March 2006. MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and Second Language Learning, pp. 69–90. New York: Cambridge University Press. Malvern, D. and B. Richards (1997). A new measure of lexical diversity. In A. Ryan and A. Wray (Eds.), Evolving Models of Language, pp. 58–71. Clevedon: Multilingual Matters. Malvern, D. and B. Richards (2002). Investigating accommodation in language proficiency interviews using a new measure of lexical diversity. Language Testing 19(1), 85–104. Mehnert, U. (1998). The effects of different lengths of time for planning on second language performance. Studies in Second Language Acquisition 20(1), 52–83. Meyer, D. M. and R. Schvaneveldt (1971). Facilitation in recognizing pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology 90, 227–234. Michel, M. C. (2009). Hoe simpel is complex? - over de invloed van taakcomplexiteit en interactie op de ´ spraak van tweede taalleerders. In Conferentiebundel ANeLA-congres, pp. 236–245. Michel, M. C., F. Kuiken, and I. Vedder (2007). The influence of complexity in monologic versus dialogic tasks in Dutch L2. International Review of Applied Linguistics 45(3), 241–259. Miller, G. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63(2), 81–97. Nassaji, H. (2007). Elicitation and reformulation and their relationship with learner repair. Language Learning 57(4), 511–548. Navon, D. (1989). The importance of being visible: On the role of attention in a mind viewed as an anarchic intelligence system. European Journal of Cognitive Psychology 1, 191–238. Navon, D. and D. Gopher (1979). On the economy of the human-processing system. Psychological Review 86(3), 214–255. Nemeth, N. and J. Kormos (2001). Pragmatic aspects of task-performance: The case of argumentation. Language Teaching Research 5(3), 213–240. Neumann, O. (1996). Theories of Attention, Volume 3. San Diego: Academic Press. Newton, J. and G. Kennedy (1996). Effects of communication tasks on the grammatical relations marked by second language learners. System 24(3), 309–322. Niwa, Y. (2000). Reasoning demands of L2 tasks and L2 narrative production: Effects of individual differences in working memory, intelligence, and aptitude. Unpublished Masters Thesis, Aoyama Gakuin University. Norris, J. and L. Ortega (2003). Defining and measuring SLA. In C. D. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, pp. 717–761. Malden: Blackwell Publishing Ltd. Norris, J. and L. Ortega (2009a). Measurement for understanding: An organic approach to investiga-

210

Bibliography

ting complexity, accuracy, and fluency in SLA. Applied Linguistics 30(4), 555–578. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Norris, J. and L. Ortega (2009b). Thoughts on task complexity and measurement. Personal communication. Nuevo, A.-M. (2006). Task complexity and interaction. Unpublished Ph.D. Thesis, Georgetown University, Washington D.C., USA. Nunan, D. (1989). Designing Tasks for the Communicative Classroom. New York: Cambridge University Press. Ortega, L. (1999). Planning and focus on form in L2 oral performance. Studies in Second Language Acquisition 21(4), 109–148. Ortega, L. (2003). Syntactic complexity measures and their relationship to L2 proficiency: A research synthesis of college-level L2 writing. Applied Linguistics 24(4), 492–518. Pallotti, G. (2009). CAF: Defining, refining and differentiating constructs. Applied Linguistics 30(4), 590–601. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Papadimitriou, C. H. (1994). Computational Complexity. Addison Wesley Longman. Perrez, J. (2004). Het gebruik van connectieven door franstalige leerders van het Nederlands: Een eerste verkenning. Toegepaste Taalwetenschap in Artikelen 71(1), 81–92. Pica, T. (1994). Research on negotiation: What does it reveal about second language learning conditions, processes, and outcomes? Language Learning 44, 493–527. Pickering, M. J. and S. Garrod (2004). Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences 27(2), 169–226. Plomp, F. (1997). Een empirisch onderzoek naar het effect van lexicale structuurmarkeringen op tekstbegrip in de eerste en in de tweede taal. Toegepaste Taalwetenschap in Artikelen 56(1), 47–61. Polio, C. G. (1997). Measures of linguistic accuracy in second language writing research. Language Learning 47(1), 101–143. Poulisse, N. (1997). Language production in bilinguals. In A. M. de Groot and J. F. Kroll (Eds.), Tutorials in Bilingualism: Psycholinguistic Perspectives, pp. 201–224. Mahwah, NJ: Lawrence Erlbaum Associates. Poulisse, N. and T. Bongaerts (1994). First language use in second language production. Applied Linguistics 15(1), 36–57. Prahbu, N. (1987). Second Language Pedagogy. Oxford: Oxford University Press. Rahimpour, M. (1997). Task condition, task complexity, and variation in L2 discourse. Unpublished Ph.D. Thesis, University of Queensland, Australia. Repovs, G. and A. Baddeley (2006). The multi-component model of working memory: Explorations in experimental cognitive psychology. Neuroscience 139(1), 5–21. ´ esz, ´ Rev A. (2007). Focus on form in task-based language teaching: Recasts, task complexity, and L2 learning. Unpublished Ph.D. Thesis, Teachers College, Columbia University, New York, USA. ´ esz, ´ Rev A. (2008). Task complexity, focus on form-meaning connections, and individual differences: A classroom-based study. Paper presented at the AILA-conference, Essen, Germany. ´ esz, ´ Rev A. (2009). Task complexity, focus on form, and second language development. Studies in Second Language Acquisition 31(3), 437–470. ´ esz, ´ Rev A. (in press). Task complexity, focus on L2 constructions, and individual differences: A classroom-based study. The Modern Language Journal 95(4). Riggenbach, H. (1991). Toward an understanding of fluency: A microanalysis of nonnative speaker conversations. Discourse Processes 14(4), 423–441. Riggenbach, H. (2000). Perspectives on Fluency. Ann Arbor, MI: University of Michigan Press. Robinson, P. (1995a). Review article: Attention, memory, and the ’Noticing’ Hypothesis. Language Learning 45(2), 283–331. Robinson, P. (1995b). Task complexity and second language narrative discourse. Language Learning 45(1), 99–145.

Bibliography

211

Robinson, P. (2001a). Task complexity, cognitive resources, and syllabus design: A triadic framework for examining task influences on SLA. In P. Robinson (Ed.), Cognition and Second Language Instruction, pp. 287–318. New York: Cambridge University Press. Robinson, P. (2001b). Task complexity, task difficulty and task production: Exploring interactions in a componential framework. Applied Linguistics 22(1), 27–57. Robinson, P. (2003a). Attention and memory during SLA. In C. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, pp. 631–678. Malden: Blackwell Publishing Ltd. Robinson, P. (2003b). The Cognition Hypothesis, task design, and adult task-based language learning. Second Language Studies 21(1), 45–105. Robinson, P. (2005). Cognitive complexity and task sequencing: Studies in a componential framework for second language task design. International Review of Applied Linguistics 43(1), 1–32. Robinson, P. (2007a). Criteria for grading and sequencing pedagogic tasks. In M. del Pilar Garc´ıa Mayo (Ed.), Investigating Tasks in Formal Language Learning, pp. 7–27. Clevedon: Multilingual Matters. Robinson, P. (2007b). Task complexity, theory of mind, and intentional reasoning: Effects on L2 speech production, interaction, uptake, and perceptions of task difficulty. International Review of Applied Linguistics 45(3), 193–213. Robinson, P. (2010). Situating and distributing cognition across task demands: The SSARC model of ¨ and L. Sicola (Eds.), Cognitive Processing and Second Lanpedagogic task sequencing. In M. Putz guage Acquisition: Inside the Learner’s Mind. Amsterdam / Philadelphia: John Benjamins Publishing Company. Robinson, P., T. Cadierno, and Y. Shirai (2009). Mind, time, and motion: Measuring the effects of the conceptual demands of tasks on second language speech production. Applied Linguistics 30(4), 533–554. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Robinson, P. and R. Gilabert (2007). Task complexity, the Cognition Hypothesis and second language learning and performance. International Review of Applied Linguistics 45(3), 161–176. Samuda, V. and M. Bygate (2008). Tasks in Second Language Learning. Basingstoke: Palgrave Macmillan. Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics 11(2), 17–46. Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and Second Language Learning, Chapter 1, pp. 3–32. New York: Cambridge University Press. Schoonen, R. (2005). Generalizability of writing scores: An application of structural equation modeling. Language Testing 22(1), 1–30. Segalowitz, N. (2003). Automaticity and second languages. In C. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, Chapter 13, pp. 382–408. Malden: Blackwell Publishing Ltd. Shehadeh, A. (2004). Modified output during task-based pair interaction and group interaction. Journal of Applied Linguistics 1(3), 351–385. Sink, C. H. and H. R. Stroh (2006). Practical significance: The use of effect sizes in school counseling research. Retrieved from http://www.schoolcounselor.org/files/9-5-401%20Sink.pdf in May 2009. Skehan, P. (1996). Second-language acquisition research and task-based instruction. In J. Willis and D. Willis (Eds.), Challenge and Change in Language Teaching, pp. 17–30. Oxford: Heinemann. Skehan, P. (1998). A Cognitive Approach to Language Learning. Oxford: Oxford University Press. Skehan, P. (2001). Tasks and language performance assessment. In M. Bygate, P. Skehan, and M. Swain (Eds.), Researching Pedagogic Tasks: Second Language Learning, Teaching, and Testing, Chapter 8, pp. 167–185. Harlow: Longman. Skehan, P. (2003). Task-based instruction. Language Teaching 36(1), 1–14. Skehan, P. (2009). Modeling second language performance: Integrating complexity, accuracy, fluency, and lexis. Applied Linguistics 30(4), 510–532. Special Issue: Complexity, accuracy, fluency (CAF) in second language acquisition. Skehan, P. and P. Foster (1997). Task type and task processing conditions as influences on foreign

212

Bibliography

language performance. Language Teaching Research 1(3), 185–211. Skehan, P. and P. Foster (1999). The influence of task structure and processing conditions on narrative retellings. Language Learning 49(1), 93–120. Skehan, P. and P. Foster (2001). Cognition and tasks. In P. Robinson (Ed.), Cognition and Second Language Learning, pp. 183–205. New York: Cambridge University Press. Skehan, P. and P. Foster (2005). Strategic and on-line planning: The influence of surprise information and task time on second language performance. In R. Ellis (Ed.), Planning and Task Performance in a Second Language, Volume 11, pp. 193–218. Amsterdam / Philadelphia: John Benjamins Publishing Company. Skehan, P. and P. Foster (2007). Complexity, accuracy, fluency, and lexis in task-based performance: A meta-analysis of the Ealing Research. In S. van Daele, A. Housen, F. Kuiken, M. Pierrard, and I. Vedder (Eds.), Complexity, Accuracy, and Fluency in Second Language Use, Learning, and Teaching, pp. 207–226. Brussel: Koninklijke Vlaamse Academie van Belgie¨ voor Wetenschappen en Kunsten (KVAB). Spooren, W. and T. Sanders (2008). The acquisition order of coherence relations: On cognitive complexity in discourse. Journal of Pragmatics 40(12), 2003–2026. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology 18, 643–662. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output in its development. In S. M. Gass and C. Madden (Eds.), Input and Second Language Acquisition, pp. 64–81. Rowley, MA: Newbury House. Swain, M. (1998). Focus on form through conscious reflection. In C. J. Doughty and J. Williams (Eds.), Focus on Form in Classroom SLA, pp. 64–81. New York: Cambridge University Press. Swain, M. and S. Lapkin (1995). Problems in output and the cognitive processes they generate: A step towards second language learning. Applied Linguistics 16(3), 371–391. Swain, M. and S. Lapkin (2000). Task-based second language learning: The uses of the first language. Language Teaching Research 4(3), 251–274. Swain, M. and S. Lapkin (2001). Focus on form through collaborative dialogue: Exploring task effects. In M. Bygate, P. Skehan, and M. Swain (Eds.), Researching Pedagogic Tasks: Second Language Learning, Teaching, and Testing, pp. 99–119. Harlow: Longman. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science 12(2), 257–285. Talmy, L. (2008). Aspects of attention in language. In P. Robinson and N. Ellis (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition, pp. 27–38. New York / London: Routledge. Tavakoli, P. and P. Foster (2008). Task design and L2 performance. Language Learning 58(2), 429–473. Tavakoli, P. and P. Skehan (2005). Strategic planning, task structure, and performance testing. In R. Ellis (Ed.), Planning and Task Performance in a Second Language, Volume 11, pp. 239–273. Amsterdam / Philadelphia: John Benjamins Publishing Company. Ullman, M. (2004). Contributions of memory circuits to language: The declarative/procedural model. Cognition 92(1), 231–270. VanPatten, B. (1990). Attention to form and content in the input: An experiment in consciousness. Studies in Second Language Acquisition 12(3), 287–301. VanPatten, B. (1996). Input Processing and Grammar Instruction in Second Language Acquisition: Theory and Research. Norwood, NJ: Ablex Publishing Corporation. Vedder, I. (1998). Dunque l’Italia ha perso il suo fascino? Io penso di no. Argumenterend schrijven in een tweede taal. Een analyse van argumentatieve teksten van Nederlandse studenten Italiaans. LOT-Dissertation Series. Ph.D. Thesis, Universiteit van Amsterdam, the Netherlands. Vermeer, A. (2000). Coming to grips with lexical richness in spontaneous speech data. Language Testing 17(1), 65–83. Watson-Gegeo, K. A. and S. Nielsen (2003). Language socialization in SLA. In C. Doughty and M. H. Long (Eds.), The Handbook of Second Language Acquisition, pp. 155–177. Malden: Black-

Bibliography

213

well Publishing Ltd. Wickens, C. D. (1989). Attention and skilled performance. In D. Holding (Ed.), Human Skills, pp. 71–105. New York: John Wiley. Wickens, C. D. (1991). Processing resources and attention. In D. L. Damos (Ed.), Multiple Task Performances, pp. 3–34. London / Washington: Taylor and Francis. Wickens, C. D. (1992). Engineering Psychology and Human Performance. New York: Harper Collins. Wickens, C. D. (2002). Multiple resources and performance prediction. Theoretical Issues in Ergonomic Science 3(3), 159–177. Wickens, C. D. (2007). Attention to the second language. International Review of Applied Linguistics 45(3), 177–191. Witte, F. and F. Mulder (2006). Staatsexamen NT2 op de Arbeidsmarkt. Informatiebulletin over het diploma Nederlands als tweede taal in bedrijf en beroepsopleidingen [State Examination Dutch as a Second Language (NT2) on the Job Market. Information Bulletin about the Certificate NT2 in Business and Vocational Education]. Retrieved from the Expertise Centrum NT2 website: http://www.expertisecentrumnt2.nl/staat/downloads/NT2%20nr%2018%20-%2009-2006.pdf in April 2007. Wolfe-Quintero, K., S. Inagaki, and H. Kim (1998). Second Language Development in Writing: Measures of Fluency, Accuracy, and Complexity. Honolulu, University of Hawai’i: Second Language Teaching and Curriculum Center. Yuan, F. and R. Ellis (2003). The effects of pre-task planning and on-line planning on fluency, complexity, and accuracy in L2 monologic oral production. Applied Linguistics 24(1), 1–27.

Curriculum vitae Marije Cornelie Michel was born in 1976 in Lachen in the German part of Switzerland where she was raised by Dutch parents. Growing up bilingually made her decide to study Dutch and German language and literature at Utrecht University in the Netherlands. Specializing in (second) language acquisition and multilingualism she wrote her MA-thesis at the Nijmegen Institute for Cognition and Information (NICI) about trilingual Dutch-German-English word recognition. After teaching Dutch as a second language at the Free University of Berlin she studied cognitive neurolinguistics at Potsdam University in Germany within the Graduate School ‘Economy and Complexity in Language’. She returned to the Netherlands as a PhD candidate at the Amsterdam Center for Language and Communication (ACLC) of the University of Amsterdam, which resulted in the present dissertation. Marije worked as a (guest) lecturer and/or research associate at the University of Cologne, the ¨ Technical University of Dortmund, and the Eberhard Karls University of Tubingen in Germany and the Hogeschool of Amsterdam in the Netherlands. Currently, she is a research associate at the University of Mannheim. Her research interests concern the multilingual society with a focus on cognitive and interactive aspects of second language acquisition, learning, and teaching.

Publications • Michel, M.C. (under review) The cognitive impact of task-based interaction in Dutch as an L2. • Michel, M.C. (under review) Effects of task complexity on the use of conjunctions in L2 task performance. • Michel, M.C. (in press) Effects of task complexity and interaction on L2-performance. In: P. Robinson (ed.) Second Language Task Complexity: Researching the Cognition Hypothesis of Language Learning and Performance. John Benjamins Publishing Company, Amsterdam/Philadelphia. • Michel, M.C. (2009) Hoe simpel is complex? - Over de invloed van taakcomplexiteit en interactie ´ op de spraak van tweede taalleerders, Conferentiebundel van het 6e ANeLA-congres 2009, p. 236–245. • Michel, M.C., F. Kuiken, & I. Vedder (2007) Effects of task complexity and task condition on Dutch L2. International Review of Applied Linguistics (IRAL), 45(3), p. 241–259. ¨ • Lemhofer, K., A.F. Dijkstra, & M.C. Michel (2004) Three languages, one ECHO. Cognate effects in trilingual word recognition. Language and Cognitive Processes, 19 (5), p. 585–611.

Task-based performance in a second language

Task-based performance in a second language

Suggest Documents

Speaking in a Second Language

LEARNING A SECOND LANGUAGE

developments in validity research in second language performance ...

Second Language Acquisition and Second Language Learning ...

Oral-performance Language Tasks for CSL Beginners in Second Life

oral-performance language tasks for csl beginners in second life

Relative Performance of English Second Language Students in ...

RESEARCH METHODS IN SECOND LANGUAGE ACQUISITION: A

Communicating in a second language: Investigating

RESEARCH METHODS IN SECOND LANGUAGE ACQUISITION: A

Reading in a Second Language - Heinemann

Variation in Second Language Spanish Textbooks: A

Rethinking Academic Language in Second Language Instruction

Language Technologies in Speech-Enabled Second Language ...

Language Technologies in Speech-Enabled Second Language ...

Oral Language Development in Second Language Acquisition

Rethinking Academic Language in Second Language Instruction

ENGLISH AS A SECOND LANGUAGE

Swedish as a second language

Language play in a second language: Social media ... - CyberLeninka

Native language interference in learning a second language ...

Language in the English as a Second Language and

Native language interference in learning a second language - CiteSeerX

Seminar in Second Language Writing