Referent Similarity and Nominal Syntax in Task-Based Language Teaching PDF

This volume addresses an important gap in the literature on task design and second language use. Building on insights from over 50 years of research on the relationship between task demands and language use, it examines how referent similarity relates to developmentally-relevant variation in the use of nominal structures, comparative structures and abstract lexis among first and second language speakers of English. In addition to providing an empirical basis for future research on tasks, it shares both theoretical and practical information on task design, which will greatly benefit curriculum and material developers.

Autor Craig Lambert | Kantaro Ohmori | Satyajit Ratha | Aneeya Kumar Samantara

111 downloads 6K Views 2MB Size

Report

Download pdf

Recommend Stories

Empty story

Idea Transcript

Craig Lambert

Referent Similarity and Nominal Syntax in TaskBased Language Teaching

Referent Similarity and Nominal Syntax in Task-Based Language Teaching

Craig Lambert

Referent Similarity and Nominal Syntax in Task-Based Language Teaching

Craig Lambert Curtin University Perth, WA, Australia

ISBN 978-981-13-3088-9 ISBN 978-981-13-3089-6 (eBook) https://doi.org/10.1007/978-981-13-3089-6 Library of Congress Control Number: 2018965189 © Springer Nature Singapore Pte Ltd. 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Acknowledgments

I am grateful to Judit Kormos who provided input on all stages of this study and on earlier versions of the manuscript. Jenefer Philp, Dorit Ravid, and Andrea Revesz also read and commented on earlier drafts of the manuscript. I am also grateful to Ayumi Hara, who drew many of the task illustrations, and Sachiko Nakamura, who worked as my research assistant in coding the transcripts. Any remaining shortcomings are entirely my own.

v

Contents

1 Introduction�� 1 1.1 Tasks in Language Teaching and Research�� 1 1.2 Task-Based Language Teaching (TBLT)�� 3 1.3 Task Learning Versus Language Learning�� 4 1.4 Referent Similarity and TBLT�� 5 1.5 Overview of the Book �� 5 References�� 7 2 L2 Performance and Incidental SLA �� 11 2.1 Language Development from a Usage-Based Perspective �� 11 2.2 Frequency Effects �� 12 2.2.1 Cue Strength�� 13 2.2.2 Prototype Effects�� 14 2.2.3 Chunking�� 15 2.3 Syntactic Priming�� 16 2.4 Transferability�� 17 2.5 First and Second Language Acquisition �� 17 2.6 Summary �� 19 References�� 19 3 Referential Communication and L2 Production�� 21 3.1 Referential Communication Tasks�� 21 3.2 Communicative Demands �� 23 3.3 Contextual Constraints�� 24 3.4 Discourse Demands�� 26 3.5 Limitations of Tasks in Instructed SLA�� 27 3.6 Proficiency and L2 Performance�� 28 3.7 Summary �� 29 References�� 29

vii

viii

Contents

4 Tasks in L2 Syllabus Design�� 31 4.1 Attention �� 31 4.1.1 Detection �� 32 4.1.2 Awareness �� 32 4.1.3 The Output Hypothesis �� 33 4.1.4 Resource Pools�� 35 4.1.5 Selection (Saliency, Effort, Expectancy & Value)�� 37 4.2 Speech Production�� 38 4.3 Cognitive Processes in SLA�� 38 4.3.1 Analysis�� 39 4.3.2 Control�� 40 4.4 Cognitive Models of Task-Based L2 Instruction�� 40 4.4.1 The Limited Attention Capacity Hypothesis�� 41 4.4.2 The Cognition Hypothesis�� 42 4.5 Summary �� 44 References�� 44 5 Referent Similarity�� 47 5.1 Referent Identification�� 47 5.1.1 Frequency�� 48 5.1.2 Comparison �� 51 5.2 The Number of Elements�� 52 5.3 The Similarity Between Elements�� 55 5.4 The Present Study �� 57 5.5 Summary �� 58 References�� 59 6 Referent Similarity and L2 Production �� 61 6.1 Noun Phrase Complexity�� 61 6.2 Comparative Structures�� 63 6.3 Relative Clause Structures�� 65 6.4 Lexical Abstractness �� 67 6.5 Summary �� 69 References�� 69 7 Methods�� 71 7.1 Purpose of the Study �� 71 7.2 Design �� 72 7.3 Participants�� 73 7.3.1 Native Speakers of English �� 73 7.3.2 Non-native Speakers of English�� 73 7.4 Materials �� 80 7.5 Procedures�� 82 7.6 Analysis�� 83 7.6.1 Transcription �� 83 7.6.2 Noun Phrase Coding �� 83 7.6.3 Comparative Structures�� 84

Contents

ix

7.6.4 Relative Clause Types �� 85 7.6.5 Statistical Analyses �� 86 7.7 Summary �� 86 References�� 86 8 Results �� 89 8.1 Preliminary Data Screening�� 89 8.1.1 Descriptive Statistics for Similarity Effect on NP Complexity�� 95 8.1.2 Descriptive Statistics for Referent Effect on NP Complexity�� 95 8.1.3 Descriptive Statistics for Proficiency Effect on NP Complexity�� 96 8.2 Three-Way MANOVA on Noun Phrase Complexity�� 97 8.2.1 Referent Similarity Effects on Noun Phrase Complexity�� 97 8.2.2 Referent Effects on Noun Phrase Complexity�� 99 8.2.3 Proficiency Effects on Noun Phase Complexity �� 101 8.2.4 Interactions Between Factors on Noun Phrase Complexity�� 104 8.2.5 Summary of Noun Phrase Complexity Results�� 104 8.3 Two-Way ANOVA for Use of Comparative Structures�� 105 8.3.1 Referent Similarity Effects on the Use of Comparative Structures�� 106 8.3.2 Referent Effects on the Use of Comparative Structures�� 107 8.3.3 Proficiency Effects on the Use of Comparative Structures�� 108 8.3.4 Summary of Results for Comparative Structures �� 108 8.4 Non-parametric Tests for Use of Relative Clause Structures�� 109 8.4.1 Referent Similarity Effects on the Use of Relative Clauses�� 110 8.4.2 Referent Effects on the Use of Relative Clauses�� 110 8.4.3 Proficiency Effects on the Use of Relative Clauses�� 112 8.4.4 Summary of Findings for the Use of Relative Clauses�� 114 8.5 Summary �� 114 References�� 115 9 Referent Similarity and TBLT�� 117 9.1 Implications for Theory and Research on TBLT�� 117 9.1.1 Referent Similarity and the Use of English Nominal Structures�� 117 9.1.2 Referent Effects�� 119 9.1.3 Proficiency Effects�� 120 9.2 Implications for Task-Based Pedagogy and Materials Design �� 120 9.2.1 Designing Tasks Using Referent Similarity�� 120

x

Contents

9.2.2 Formatting Tasks for Different Instructional Objectives�� 124 9.2.3 Referential Similarity Tasks and Language Assessment�� 126 References�� 126 Appendices�� 129 Appendix 1: Materials for the Study�� 129 Appendix 2: Noun Phrase Coding Manual�� 139 Appendix 3: Full Descriptive Statistics for NP Complexity�� 145

List of Figures

Fig. 5.1 Model of hypothesized effects of increased referent similarity on production.................................................................................... 12 Fig. 9.1 Three picture sets for clothing description tasks based on dissimilar items............................................................................ 121 Fig. 9.2 Three picture sets for clothing description tasks based on concrete differences..................................................................... 122 Fig. 9.3 Three picture sets for clothing description tasks based on abstract differences...................................................................... 123 Fig. 9.4 Seven one-way tasks based on concrete differences in an integrated format...................................................................... 125 Fig. 9.5 Three two-way integrated format tasks at different levels of referent similarity......................................................................... 125

xi

List of Tables

Table 3.1 Two types of communication (based on Brown & Yule, 1983).................................................................................. 23 Table 3.2 Contextual influences on communication skill development (Yule, 1997)............................................................... 25 Table 3.3 Task type and functional language demands (Yule, 1997)............. 27 Table 6.1 Three types of comparison in English............................................ 64 Table 6.2 Developmental typology of relative clauses (Diessel & Tomasello, 2005).......................................................... 66 Table 6.3 Early and late emerging terms (Nippold, Hesketh, et al., 2005; Nippold, Ward-Lonergan, et al., 2005)................................. 68 Table 7.1 Table 7.2 Table 7.3 Table 7.4 Table 7.5

Background of native-speaking participants.................................. 74 Background of advanced non-native participants........................... 75 Background of intermediate non-native participants...................... 76 Comparative experience of the participants................................... 79 Proficiency and aptitude comparisons of the two non-native groups.................................................................... 79 Table 7.6 Structures excluded from the analysis of comparisons.................. 85 Table 7.7 Relative clause types and token frequencies in the database.......... 85 Table 8.1 Noun phrase descriptive statistics between conditions (N = 54)......................................................................... 92 Table 8.2 Noun phrase descriptive statistics between referents (N = 54)........................................................................... 93 Table 8.3 Noun phrase descriptive statistics between groups*...................... 94 Table 8.4 Pearson correlations for noun phrase variables (N = 54)............... 97 Table 8.5 Noun phrase means and confidence intervals for conditions (N = 54)......................................................................... 98 Table 8.6 Pairwise comparisons of condition effect on noun phrase production (N = 54)............................................................ 99

xiii

xiv

List of Tables

Table 8.7 Noun phrase means and confidence intervals between referents (N = 54)........................................................................... 100 Table 8.8 Pairwise comparisons of referent effect (N = 54).......................... 101 Table 8.9 Noun phrase means and confidence intervals for groups............... 102 Table 8.10 Pairwise comparisons of noun phrase production between groups............................................................................... 103 Table 8.11 ANOVAs for condition-referent interaction................................... 104 Table 8.12 Descriptive statistics for comparative structures............................ 106 Table 8.13 Means for condition effect on comparison (N = 54)...................... 107 Table 8.14 Central Tendency for Effect on Comparison across Referents......................................................................................... 108 Table 8.15 Post hoc Wilcoxon tests for referent effects on explicit comparative use.............................................................................. 108 Table 8.16 Means and confidence intervals for groups on comparative structures (N = 54).......................................................................... 109 Table 8.17 Pairwise comparisons of group on comparative structures (N = 54).......................................................................................... 109 Table 8.18 Friedman’s tests for condition effect on relative clause production............................................................................ 110 Table 8.19 Friedman’s tests for referent effect on relative clause production............................................................................ 111 Table 8.20 Post hoc Wilcoxon tests for referent effects on relative clause production............................................................................ 111 Table 8.21 Mean ranks for relative clause production between groups........... 113 Table 8.22 Results of post Hoc comparisons for groups on relative clauses........................................................................... 113 Table 8.23 Jonckheere’s test for trends in RC production across proficiency levels............................................................................ 114

Chapter 1

Introduction

This chapter positions the book within the field of task-based language teaching and learning. Tasks are defined in terms of interactive, affective and societal need. Two possibilities for using tasks within second language instruction are then outlined, and the need for research on task design factors and their relationship to language development is discussed. The chapter concludes with a summary of the chapters that follow.

1.1 Tasks in Language Teaching and Research Over the past three decades, tasks have taken on an increasingly more important role in second language (L2) research and pedagogy (e.g., Brown & Yule, 1983; Ellis, 2003, 2018; Long, 1985, 2000, 2015; Long & Crookes, 1993; Robinson, 2001, 2011a; Skehan, 1996, 2009, 2014; Yule, 1997). Tasks are important tools in allowing teachers and researchers to tap into and potentially build upon the range of personal and linguistic resources that learners bring to them. As Schmid, Verspoor, and MacWhinney (2011) note: …the full range of the linguistic repertoire can only truly be investigated on the basis of spoken or written data produced under relatively natural conditions – that is, data where all aspects of the linguistic production process (the selection of the vocabulary, the sentence frame, grammatical aspects such as tense, mood and voice, orthography or phonology and so on) are, as far as possible, fully under the control of the learner (Schmid et al., 2011: 39).

R. Ellis (2009) provides a useful interactive definition of tasks that distinguishes them from the other types of language learning exercises that teachers use in L2 classrooms. He argues that for a language learning activity to be a task rather than an exercise, there must be: (1) a primary focus on meaning, (2) a gap in information, opinion or inference that necessitates communication, (3) a communicative outcome beyond using language for its own sake, and crucially, as Schmid © Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_1

1

2

1 Introduction

et al. (2011) point out above, (4) learners must draw on the full range their own personal and linguistic resources in completing them. Tasks which meet these four criteria promote interactionally-authentic language use in the L2 classroom and may contribute to learners’ ability to process language without undue effort while their attention is focused on meaning. Ellis’s criteria thus work together to ensure the integrity of tasks for developing implicit (Lambert, 2018) or at least fully- automatized (Suzuki & DeKeyser, 2015) language knowledge as it relates to interactive competence in using a second language. The personal or affective dimension of tasks is also important in terms of interest and effort and the role that they play in the learning process (Dewey, 1913, 1938; Maehr, 1984). Many tasks used in L2 teaching and research clearly meet the four criteria proposed by Ellis, but they too often fail to engage learners at a level that necessitates pragmatically-authentic language use. Lambert and Zhang (2018), for example, demonstrate how tasks based on learner-generated content can result in the use of pragmatic devices which are virtually non-existent on comparable tasks based on the type of generic teacher-generated content typically used in L2 teaching and research. Such pragmatic devices included tag questioning to elicit agreement, uncompleted sentences to deliver veiled challenges, mitigated questioning, unarticulated confrontation, and softened tone to create pragmatic meanings such as tentativeness, self-protection, collaboration, cooperation and the maintenance of face (Sabet & Zhang, 2015). There is simply no need for this level of pragmatic meaning in connection with tasks based on generic content supplied by the instructor or materials writer. As Paradis (1994, 2004) points out, learners in L2 classrooms are too often motivated only by the fact that it is their turn to speak rather than by the types of physical, social or emotional needs for understanding that motivate authentic language use. In addition to addressing the interactive needs of learners, tasks must also generate personal investment in the learning process. If learners do not need to negotiate pragmatic meaning on tasks, they cannot be expected to acquire these skills incidentally during task-based L2 instruction. However, the aspects of tasks outlined by R. Ellis (2009) and Lambert (2004; 2017; Lambert, Philp, & Nakamura, 2017; Lambert & Zhang, 2018) focus only on the interactive and pragmatic authenticity of the language used on tasks. These may be primary considerations in educational contexts where learners’ needs for the target language are unclear, such as in the case of young learners (Shintani, 2016), or when learners’ future language needs are simply unappealing to them as a basis for instructional content (Lambert, 2010). However, situational authenticity will also be an important dimension of pedagogic tasks when learners have specific needs for the language (immigration, travel, business, study, etc.), and they are motivated to master the language to meet these needs. Long (2015) discusses procedures for identifying learners’ language needs and developing task-based instruction to address them. As we shall see in the chapters which follow, the lexis of pedagogic can moderate the effects of other factors in the design and sequencing of tasks. Situational authenticity might thus play an important role in the language that speakers acquire in conjunction with tasks. In addition, situational authenticity is important to the accountability of language curricula in addressing societal needs for L2 competence (e.g., normative, felt, expressed and comparative) (Bradshaw, 2013; Long, 2015).

1.2 Task-Based Language Teaching (TBLT)

3

1.2 Task-Based Language Teaching (TBLT) The use of tasks in L2 instruction is predicated on the notion that language learning is an incidental process that takes place in line with learners’ communicative needs while they are focused on achieving communicative outcomes. It is also a process that takes place in line with learners’ internal syllabuses, resources and experience. In terms of current instructional practices, however, tasks are probably most commonly used to support the instructional objectives of another approach to language instruction. The use of tasks in this way is referred to as a task-supported language teaching (TSLT). One example is the present, practice, produce (PPP) approach in which an instructor presents new language, provides students with exercises for structured practice of this new language, and then provides them with a further task- like exercises in order to give them a chance for more unstructured practice. In such an approach, however, the integrity of the tasks as learning tools is often compromised and tasks are reduced to contextualized language learning exercises. This is due to the fact that, when used in conjunction with teaching specific linguistic content, tasks frequently end up functioning as a means to practicing and automatizing specific forms (Dekeyser, 2007). In other words, a PPP approach can bias learners toward the use of pre-taught language forms, and learners can be prevented from drawing on the full range of their own resources. When this happens, learners a completing situated language exercises rather than tasks (R. Ellis, 2009; Lambert, 2018). In contrast, task-based language teaching (TBLT) preserves the integrity of tasks for L2 learning. Instruction is conceptualized in terms of tasks, and all of the curriculum components (needs, goals, objectives, syllabus, assessments, methods and materials) are organized around tasks. Long (2015) provides a detailed rationale for TBLT, arguing that the pedagogic tasks used as to organize instruction should be based on real-world tasks that learners need to complete and sequenced to gradually approximate the demands of these tasks based on nonlinguistic criteria such as the number of elements they involve, the degree of similarity between these elements, and the amount of shared context available to learners (Long, 2015, pp. 232–4). In addition to ensuring the integrity of tasks for incidental L2 acquisition, task-based language teaching (TBLT) has the advantage of ensuring the relevance of instruction in terms of the content and face validity of tasks. Robinson (2001) notes that a similar approaches to one advocated by Long are employed in other educational contexts in which instruction is focused on the acquisition of complex skills: … the approach to syllabus design proposed here is no different than that underlying educational decision making in other domains such as pilot training or mathematics education which approach the development of complex skills and problem solving abilities (landing an aircraft in a blizzard, doing calculus) though initially simple, then progressively more complex tasks. (Robinson 2001: 301)

Task learning and language learning inevitably overlap to some degree. However, there is not a one-to-one relationship between them. Tasks provide learners with the space to use language in their own ways and integrate task content and their own

4

1 Introduction

language resources, but “…learners do identify how to perform a task using the minimum necessary linguistic resources” (White & Robinson, 1995, p. 98). A frequent criticism of task-based language instruction is thus that learners run the risk of becoming fluent in the language at the expense of ongoing L2 development (Skehan, 1996).

1.3 Task Learning Versus Language Learning Criticism of TBLT frequently has to do with the distinction between task learning and language learning (Skehan, 1996; White & Robinson, 1995). Many L2 learners are not only interested in learning to complete tasks but in mastering the language system as it is used by proficient speakers in completing tasks. Many aspects of a language are not essential to arriving at successful task outcomes. In Long’s (2015; Long & Robinson, 1998) approach to task-based instruction, problems of language form are addressed incidentally as they arise in communication through the use of corrective techniques which preserve a focus on meaning (e.g., confirmation checks, clarification requests, and recasts). However, the sensitivity of learners across aptitude and motivational profiles to correction of erroneous language forms during meaning-focused communication has been questioned, and many researchers and practitioners are skeptical about the efficacy of incidental focus on form as a means of supporting acquisition (see Ellis & Shintani, 2014 for a recent review), arguing that more explicit means of instruction are necessary to support task-based L2 instruction (Ellis, 2018). As White argues above, the relationship between task learning and language learning is likely to be a complex one with tasks supporting language acquisition in some ways and working against it in others. Pedagogic tasks which are limited to simulating the demands of the tasks that learners may have to complete outside of the classroom may also result in a general slow-down in language learning when basic communicative competence has been acquired. A key factor in this slow-down is task frequency. In discussing the development of cue strength, or the ability of learners to access a specific form in relationship to a frequently encounter communicative function (see Chap. 2), MacWhinney (2001) notes that: Because most basic linguistic tasks [in L1 acquisition] are well above threshold frequency, the dimension of task frequency is seldom an important determinant of relative cue strength. However, in the case of a second language … task frequency could become a factor determining a general slow-down in acquisition (MacWhinney 2001: 71).

In other words, the demands of the tasks that L2 learners typically have to complete outside of the classroom do not always provide them with frequent enough exposure to the broader range of linguistic resources that give L1 speakers the headroom to complete these tasks with the precision and flexibility that is sometimes required. L2 learners typically do not have to complete highly demanding versions of tasks which occur only infrequently in real-world communication. However, L2

1.5 Overview of the Book

5

learners might benefit from be exposed to rarer, more demanding versions of tasks as part of the instructional process in order to push them to experiment with new ways of completing tasks, incorporating the full range of linguistic resources that proficient speakers of the language are able to incorporate (cf., Skehan, Xiaoyue, Qian, & Wang, 2012). The pedagogic tasks in typical L2 instructional materials often leave more proficient L2 learners unchallenged linguistically. Teachers may thus frequently be at a loss in providing them with challenging oral communication tasks in the classroom. In fact, there are typically learners enrolled in formal language instruction who are competent enough in the target language to complete the range tasks that they need to complete fluently but who still do not command the full range of linguistic resources that would be available to proficient speakers in completing these tasks.

1.4 Referent Similarity and TBLT Rather than turning prematurely to explicit instruction of the PPP type to address the gaps between task learning and language learning, this book investigates how tasks might be designed and sequenced in non-linguistic terms to narrow the gaps that exist between task learning and language learning when pedagogic tasks are limited to the level of communicative demand of real-world target tasks. Building on insights from over 50 years of research on the relationship between task demands and language use, this book examines how referent similarity relates to developmentally-related variation in the use of nominal structures, comparative structures and abstract lexis by L1 and L2 learners of English. In addition to providing an empirical basis for future research on the role of tasks in L2 development, the book provides practical information on the design of pedagogic task that will be of use to teachers and material developers.

1.5 Overview of the Book This chapter has defined tasks and discussed how the book will contribute to task- based language teaching. Chapter 2 discusses how task-based L2 production relates to acquisition. The focus of the chapter is on usage-based theories of learning. It is argued that the frequency with which contextualized language is used for specific communicative functions drives cue strength at all levels of L2 processing (e.g., Bybee, 2008; Ellis, 2002, 2008, Eskildsen, 2008; MacWhinney, 2001, 2008, Tomasello, 2000a, 2000b). It is also argued that language knowledge is represented as a complex set of weighted connections in the brain and that the mechanisms responsible for the development of this knowledge ultimately operate implicitly and result in the implicit probabilistic knowledge that forms the basis for performance (N. Ellis, 2002). Based on these theories of learning language through using

6

1 Introduction

language, it is argued that more extensive exposure to communicative task demands will result in the expansion and conventionalization of the form-to-function mappings which determine the range of resources available to speakers when they complete a specific type of task. Chapters 3 and 4 then discuss current perspectives on the role that tasks have been argued to play in L2 performance. Chapter 3 focuses on theories of how referential communication tasks relate to the mapping of L2 forms to the functional demands of different discourse genres such as description, instruction, narrative and opinion (e.g. Berman, 2008; Brown & Yule, 1983, Yule, 1997). Chapter 4 then discusses cognitive theories of task-based instruction which attempt to relate factors in the design of tasks to cognitive processes akin to those that Bialystok (1994) refers to as analysis and control (e.g., Robinson, 2001, 2011b) and that Levelt (1989, 1999; Kormos, 2006) postulates in his modular model of speech processing (Skehan, 2014). While the approaches to task-based instruction in both Chaps. 3 and 4 explain the learning process in cognitive terms, Chap. 3 focuses on the effects of tasks on aspects of the language system, whereas Chap. 4 focuses on the effects of tasks on learners’ allocation of attention during performance as manifest in general measures of the fluency, accuracy and complexity of their L2 production. The assumption underlying the models discussed in Chap. 4 is that balanced focus on these three aspects of L2 production will ultimately result in the balanced development of language proficiency over time. Chapters 5 and 6 then review research related to the specific variables in the study of referent similarity and English nominalization processes that will be reported in Chaps. 7 and 8. In Chap. 5, key studies of the effects of referent similarity on L1 production over the last 50 years are reviewed, and the results of these studies are discussed in terms of how increased similarity demands are likely to be met linguistically both in terms of general and specific measures of syntax and lexis. Following this, the construct of referent similarity is disambiguated from a related construct that have received considerable attention in the TBLT literature (the number of elements in tasks). Chapter 6 then summarizes L1 developmental research and identifies general and specific variables connected with nominalization processes that have been shown to emerge late in the language production of L1 speakers of English. Chapter 5 thus addresses the independent factors in the study to follow, whereas Chap. 6 addresses the dependent measures to be adopted. Chapters 7 and 8 present the methods and the findings of the study of the effects of referent similarity on nominalization processes in L1 and L2 speakers’ descriptive discourse. In Chap. 7, key research questions are identified and the specific hypotheses are formulated based on the research surveyed in Chaps. 5 and 6. The particulars of the participants, materials, procedures and analysis used in the study are then outlined in enough detail that the study might be replicated in whole or in part in future research on the role of referent similarity in TBLT. In Chap. 8, the results of the study are then summarized. The research questions are answered based on a combination of parametric and non-parametric statistical tests. Results supported the hypotheses relating to the effect of referent similarity on noun phrase complexity, the use of comparative structures, and the use of later-emerging relative

References

7

clauses. In addition, a moderating effect of the specific referent being described was identified, and the relationship between the dependent measures and proficiency of the participants in the study is examined. Finally, Chap. 9 discusses implications of the study for task-based L2 research and pedagogy. The findings regarding each of the research hypotheses are first discussed in detail, and key theoretical issues regarding the role of tasks in L2 production are identified. Practical task design principles are then outline and illustrated with examples that both teachers and researchers will find useful in applying the ideas discussed in the book to the design and sequencing of tasks for task-based L2 research and instruction.

References Berman, R. (2008). The psycholinguistics of developing text construction. Journal of Child Language, 35, 775–771. Bialystok, E. (1994). Analysis and control in the development of second language proficiency. Studies in Second Language Acquisition, 16, 157–168. Bradshaw, J. (2013). A taxonomy of social need. In C. Cookson, S. Sainsbury, & C. Glendinning (Eds.), John Bradshaw on social policy (pp. 1–11). York, UK: University of York. Brown, G., & Yule, G. (1983). Teaching the spoken language. Cambridge, UK: Cambridge University Press. Bybee, J. (2008). Usage-based grammar and second language acquisition. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 216– 236). New York: Routledge. DeKeyser, R. (2007). Skill acquisition theory. In B. Van Patten & J. Williams (Eds.), Theories in second language acquisition: An introduction (pp. 97–113). Mahwah, NJ: Routledge. Dewey, J. (1913). Interest and effort in education. Boston: Houghton Mifflin. Dewey, J. (1938). Experience and education. New York: Simon and Schuster. Ellis, N. (2002). Frequency effects in language processing: A review with implications for theories for implicit and explicit language acquisition. Studies in Second Language Acquisition, 24, 143–188. Ellis, N. (2008). Usage-based and form-focused language acquisition: The associative learning of constructions, learned attention and the limited L2 endstate. In P. Robinson & E. Nick (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 372–405). New York: Routledge. Ellis, R. (2003). Task-based language learning and teaching. Oxford, UK: Oxford University Press. Ellis, R. (2009). Task-based language teaching: Sorting out the misunderstandings. International Journal of Applied Linguistics, 19, 221–246. Ellis, R. (2018). Reflections on task-based language teaching. Bristol, UK: Multilingual Matters. Ellis, R., & Shintani, N. (2014). Exploring language pedagogy through second language acquisition research. New York: Routledge. Eskildsen, S. (2008). Constructing another language: Usage-based linguistics in second language acquisition. Applied Linguistics, 30, 335–357. Kormos, J. (2006). Speech production and second language acquisition. Mahwah, NJ: Routledge. Lambert, C. (2004). Reverse-engineering communication tasks. ELT Journal, 58, 18–27. Lambert, C. (2010). Task-based needs analysis: Putting principles into practice. Language Teaching Research, 14, 99–112.

8

1 Introduction

Lambert, C. (2017). Tasks, affect and second language performance. Language Teaching Research, 21, 657–664. Lambert, C. (2018). Tasks vs. exercises. In TESOL encyclopedia of English language teaching. Malden, MA: Wiley/Blackwell. Lambert, C., Philp, J., & Nakamura, S. (2017). Learner-generated content and engagement in L2 task performance. Language Teaching Research, 21, 665–680. Lambert, C., & Zhang, G. (2018). Engagement in the use of English and Chinese as second languages: The role of learner-generated content in instructional task design. Modern Language Journal. Levelt, W. (1989). Speaking from intention to articulation. Cambridge, MA: MIT Press. Levelt, W. (1999). Producing the language: A blueprint of the speaker. In C. Brown & P. Hagoort (Eds.), The neurocognition of language (pp. 83–122). New York: Oxford Press. Long, M. (1985). A role for instruction in second language acquisition: Task-based language teaching. In K. Hyltenstam & M. Pienemann (Eds.), Modelling and assessing second language acquisition (pp. 77–99). Clevedon, UK: Multilingual Matters. Long, M. (2000). Focus on form in task-based language teaching. In R. Lambert & E. Shohamy (Eds.), Language policy and pedagogy (pp. 181–196). Philadelphia: John Benjamins. Long, M. (2015). Second language acquisition and task-based language teaching. Oxford, UK: Wiley Blackwell. Long, M., & Crookes, G. (1993). Units of analysis in syllabus design: The case for the task. In G. Crookes & S. Gass (Eds.), Tasks in a pedagogical context (pp. 9–44). Cleveland, UK: Multilingual Matters. Long, M., & Robinson, P. (1998). Focus on form: Theory, research, and practice. In C. Doughty & J. Williams (Eds.), Focus on form in classroom second language acquisition (pp. 15–41). Cambridge, UK: Cambridge University Press. MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and second language instruction (pp. 69–90). Cambridge, UK: Cambridge University Press. MacWhinney, B. (2008). A unified model. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 341–371). New York: Routledge. Maehr, M. (1984). Meaning and motivation: Toward a theory of personal investment. In R. Ames & C. Ames (Eds.), Motivation in education: Student motivation (Vol. 1, pp. 115–144). San Diego, CA: Academic. Paradis, M. (1994). Neurolinguistic aspects of implicit and explicit memory: Implications for bilingualism and SLA. In N. Ellis (Ed.), Implicit and explicit learning of languages (pp. 393– 419). New York: Academic. Paradis, M. (2004). A neurolinguistic theory of bilingualism. Amsterdam: John Benjamins. Robinson, P. (2001). Task complexity, cognitive resources and second language syllabus design. In P. Robinson (Ed.), Cognition and second language instruction (pp. 287–318). Cambridge, UK: Cambridge University Press. Robinson, P. (2011a). Task-based language learning: A review of the issues. Language Learning, 61(Suppl. 1), 1–36. Robinson, P. (2011b). Second language task complexity, the cognition hypothesis, language learning and performance. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Sabet, P., & Zhang, G. (2015). Communicating through vague language: A comparative study of L1 and L2 speakers. Basingstoke, UK: Palgrave Macmillan. Schmid, M., Verspoor, M., & MacWhinney, B. (2011). Coding and extracting data. In M. Verspoor, K. de Bot, & W. Lowie (Eds.), A dynamic approach to second language development: Methods and techniques (pp. 39–54). Amsterdam: John Benjamins. Shintani, N. (2016). Input-based tasks in foreign language instruction for young learners. Amsterdam: John Benjamins.

References

9

Skehan, P. (1996). A framework for the implementation of task-based learning. Applied Linguistics, 17, 38–62. Skehan, P. (2009). Modeling second language performance: Integrating complexity, accuracy, fluency and lexis. Applied Linguistics, 30, 510–532. Skehan, P. (2014). Processing perspectives on task performance. Amsterdam: John Benjamins. Skehan, P., Xiaoyue, B., Qian, L., & Wang, Z. (2012). The task is not enough: Processing approaches to task-based performance. Language Teaching Research, 16, 170–187. Suzuki, Y., & DeKeyser, R. (2015). Comparing elicited imitation and word monitoring as measures of implicit knowledge. Language Learning, 65, 860–895. Tomasello, M. (2000a). First steps toward a usage-based theory of language acquisition. Cognitive Linguistics, 11, 61–82. Tomasello, M. (2000b). The item-based nature of children’s early syntactic development. Trends in Cognitive Sciences, 4, 156–163. White, R., & Robinson, P. (1995). Current approaches to syllabus design: A discussion with Ron White. Guidelines, 17, 93–101. Yule, G. (1997). Referential communication tasks. Mahwah, NJ: Lawrence Erlbaum.

Chapter 2

L2 Performance and Incidental SLA

This chapter provides a theoretical foundation for the connection between L2 production on tasks and second language acquisition (SLA). The chapter focuses primarily on functional category development from a usage-based perspective as this area of development is the most relevant to the effect of specific tasks factors on development, but the impact of meaning-focused language use on language representation has been discussed for nearly three decades (Bybee, 1985, 2006; Langacker, 1987, 2000, 2008).

2.1 Language Development from a Usage-Based Perspective Current usage-based perspectives on language development (e.g., Bybee, 2008; N. Ellis, 2002; Langacker, 2008; MacWhinney, 2001, 2008; Tomasello, 2000a, 2000b) view linguistic competence as an inventory of item-based constructions developed through context-based, functionally-driven imitation, analogy and abstraction. It is argued that in constructing complex utterances speakers integrate already mastered words, phrasal schemata, and utterances functionally in ways appropriate to the demands of the tasks that they have learned to complete and the social context in which they have learned to complete them. Bybee (2008) argues that language acquisition consists of the learning of constructions and the frequency-based abstraction of regularities connected with them. She defines constructions as stored pairings of form and function that range from units as small as the word to units as large as complex sentences. These constructions are conventionalized in the speech community and entrenched in the grammatical knowledge of the individual speaker. They serve well-defined linguistic functions and are independently represented in the speaker’s mind. Many carry meanings tied to specific lexical items, but others carry meanings that are independent of particular words.

© Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_2

11

12

2 L2 Performance and Incidental SLA

Tomasello (2000a, 2000b) argues that after an adequate number of such constructions are functionally mapped within the developing language system, abstract categories of language resources specifically suited to a learner’s functional needs are created. These categories represent language use at all levels and permit syntactic operations which integrate already mastered words, phrasal schemata, and abstract utterances functionally in ways appropriate to the local speech community. It is contextualized communicative functions that are the basis of these abstract categories. According to Tomasello (2000a, 2000b) people categorize language forms based on function. Constructions that serve the same or a similar communicative purpose are categorized as functionally equivalent. The process of language acquisition thus requires sufficient experience with meaning-focused language use and the ability to realize the functional equivalence of different expressions. Usage in this sense results in the ability to modify the things that one says based on abstract grammatical rules.

2.2 Frequency Effects It should be clear from the previous section that frequency plays a key role in the development of language representations in usage-based accounts of first and second language acquisition. N. Ellis (2002) provides a more detailed overview of the specific role that frequency plays in SLA. In a similar way to Tomasello, he argues that learning is the result of experience with language and its distributional properties. However, N. Ellis is explicit on his position that human categorization is a process of unconsciously tallying the frequencies of item types and tokens and determining probabilities based on their distributions. These probabilities then provide the implicit basis for action that language users draw on in assigning interpretations to input or choosing expressions during production. Thus, not only is the frequency-based knowledge that provides the basis for decision-making during L2 use implicit, but it is also implicitly acquired. However, this does not mean that explicit knowledge and conscious decision-making processes on the part of the learner are not also important in adult SLA. These processes will be addressed in Chap. 4. Elaborating his position, N. Ellis (2002) argues that the processing of frequency falls into the domain of what Posner and Snyder (1975) call automatic processes (i.e., without awareness, intention, or voluntary effort), and provides support for the position that tagging frequency, in the sense of unconsciously tallying item type and token frequencies and determining probabilities based on their distributions, has little impact on one’s ability to simultaneously attend to other aspects of the situation (e.g., Hasher & Chromiak, 1977: 173, referenced in N. Ellis, 2002). High- frequency items get higher activation than low-frequency items from the same evidence. For example, when one judges whether a specific word belongs to a given language or not, the new word is implicitly compared to examples stored in m emory.

2.2 Frequency Effects

13

The closer it matches their characteristics, the more likely it is to be judged wordlike. N. Ellis argues that SLA at all levels is based on frequency and probability knowledge of this sort. N. Ellis’s (2002) argument is that sensations are not simple facts, but that they are supplemented by the ideas associated with them in the individual mind. When a sensation is associated with more than one such concept, unconscious processes weigh the odds and the most probable alternative is perceived. Accurate and fluent L2 use thus requires an appropriately weighted range of associations. However, this will not be a product of raw frequency. On the one hand, progress will occur at different rates at different points in development due to a power law of practice which predicts that progress on any skill will be more strongly affected by frequency or practice in the early stages of acquisition than in the later stages of acquisition. On the other hand, any complex system such as a second language also requires that the mediating effects of other elements in the system must be taken into account in predicting progress on each of these elements individually. Nevertheless, N. Ellis argues that progress in the system as a whole will be the direct result of frequency in that users will choose the most probable utterance in a given situation based on the frequency of previous utterance representations and also that they will perceive the most probable syntactic and semantic analyses in a given situation on the basis of their previous experiences. However, N. Ellis’s basic argument can be further elaborated in terms of the specific subsystems or mechanisms by which learning progresses in response to frequency. Examples of these mechanisms relate to cue strength, prototype effects and chunking.

2.2.1 Cue Strength MacWhinney (2001) argues that cue validity plays a crucial role in successful language performance and determines the basis for possible action in a given context. According to MacWhinney, cue validity is determined by four dimensions of a cue’s distribution in the language that a learner experiences: (1) the simple frequency with which the linguistic operation is performed (e.g., determining the agent of a verb, determining the antecedent in anaphoric reference, etc.), (2) the frequency with which a given cue is contrastive in determining the meaning of the utterances that it is associated with (e.g., preverbal positioning is crucial for determining the meaning of a sentence, but subject-verb agreement is often redundant due to the presence of the subject noun or pronoun), (3) the reliability with which the cue is connected with the correct functional choice, and (4) the reliability of a cue in relation to other cues. The important point about cue validity as it relates to the connection between task-based L2 production and SLA concerns Factor 1 (the frequency with which a given linguistic operation is performed) because Factors 2–4 refer to fixed features of the target language code. As MacWhinney notes, “Because most basic linguistic tasks are well above threshold frequency, the dimension of task frequency is seldom an important determinant of relative cue strength (in L1 speakers). However, in the

14

2 L2 Performance and Incidental SLA

case of a second language … task frequency could become a factor determining a general slow-down in acquisition (2001: 74).” In other words, the typical functional demands of communication in an L2 may not necessitate the use of many of the forms that L1 use supports. As will be argued in Chaps. 3, 4 and 5, task-based ISLA might address this issue. MacWhinney (2008) further elaborates the role of cue strength in language learning. He argues that long-term linguistic knowledge is organized into a series of self-organizing maps (SOMs). Within this model, learning involves three phases: (1) all units receive activation from the input and each unit computes its current activation, (2) units compete through local inhibitions of their neighbors, and the best matching unit emerges as the winner, and (3) weights of the responding units are adjusted to increase the precision of activation in future trials so that neighboring units become increasingly responsive to similar input patterns. Similar inputs thus come to activate neighboring units in the map giving SOMs their self-organizing quality. During the first stages of learning, there is great plasticity because particular areas are only weakly committed to particular parts of speech. Connections are only weakly organized and lightly entrenched. Over time, however, specific areas gain strength and begin attracting like items. As the lexicon becomes larger, for example, the network becomes more entrenched, and the precision of recall increases. In MacWhinney’s model, learning is thus grounded on self-organization within associative maps. However, the processes of buffering, chunking, and resonance are argued to modulate the effect of frequency in learning in these maps. Buffering works to provide short-term storage of material to allow the processor to compare competing forms and to extract consistent patterns as inputs to learning. Chunking facilitates the fluent integration of information between maps. Resonance works to consolidate representations within maps. Thus, the selection of forms during production will be governed by the strength with which cues have been established based on previous use.

2.2.2 Prototype Effects Bybee (2008) further elaborates the process of category formation and extension in SLA, arguing that type and token frequency play complementary roles in the establishment and extension of linguistic prototypes. According to Bybee, token frequency refers to how often particular items are experienced and has a conserving effect on language. In other words, the repetition associated with token frequency strengthens memory representations for linguistic forms and makes them more accessible. For example, frequently encountered irregular past tense verbs in English will be conserved in speech and function as prototypes for the production of other forms, whereas infrequently encountered irregular forms will tend to either regularize or conform to prototypical paradigms. Bybee also notes that in the case of extremely common tokens, rote learning can take place and lead to an autonomy effect whereby the expression ceases to be associated with the units that comprise it (e.g., ‘gonna’ for ‘going to’ in English). In such cases, these items will presumably

2.2 Frequency Effects

15

become of limited value as prototypes. Bybee also argues that token frequency will have a reducing effect on language. She argues that in all human activities, frequent repetition of neuro-motor sequences leads to greater overlap and reduction of the component gestures. High frequency production of specific tokens can thus be expected to result in more efficient as well as more fluent articulation. As articulation becomes more efficient, output will be more affected by assimilation. In contrast, type frequency refers to how many variations a certain pattern or construction supports. Bybee argues that when a construction is encountered with different items occupying a specific position, parsing of the construction is facilitated, and the learner is better able to apply the pattern productively to forms not encountered before. In other words, the productivity of a given construction, or the learners’ ability to use it creatively, is a function of its type frequency. However, the complementary effects of token and type frequency also have important implications for the formation and extension of categories. Bybee (2008) argues that L1 and L2 learning patterns are both traceable to the nature of categorization, a general property of which is that one high-frequency item functions as a prototype around which others organize. In other words, after high-frequency exposure, a prototypical structure is built up and items used in new contexts are evaluated by analogy with these stored instances of the prototype. For example, many adjectives only occur with one noun. After a certain number of token occurrences, such collocations can be argued to be conventionalized and stored in memory. The tokens that are more frequent have stronger representations in memory and serve as the analogical basis for forming novel instances of the constructions. These novel instances are also represented in memory, although not as robustly, and may have an effect on the category, causing it to extend in new directions. However, as mentioned above, a very high-frequency pattern might be learned and stored autonomously from the more general pattern. Thus, when prototypes have been established through token frequency, the variations made available through type frequency function to extend categories. Lower-frequency forms are analyzed and learned dependent on higher-frequency forms. The frequency distribution of forms in terms of token and type relationships (i.e., prototypes and variations on them) will thus be crucial to L2 development.

2.2.3 Chunking Procedural knowledge is knowledge of how to do something that is acquired through the repetition of commonly occurring sequences of separate actions. With practice these separate actions are repackaged into a single action. Once automatized in this way, it is difficult to unpack them again. The process by which sequences of actions that occur together repeatedly are combined into a single action is called chunking or automatization (see DeKeyser, 2001, for an overview). Bybee (2008) argues that chunking occurs naturally and unconsciously with practice, and that the degree of fusion between actions is based on the frequency with which they occur together without exception. According to Bybee, construc-

16

2 L2 Performance and Incidental SLA

tions in a language are chunks or neuro-motor routines with movable parts. Chunks larger than standard analytic units (i.e., words, morphemes) are the units of storage and processing and higher-level systematic organization is possible through ongoing experience. The importance of automatic access in task-based L2 production is summarized by Pawley and Snyder (quoted in N. Ellis, 2002): In the store of familiar collocations there are expressions for a wide range of familiar concepts and speech acts, and the speaker is able to retrieve these as wholes or as automatic chains from the long-term memory; by doing this he minimizes the amount of clause- internal encoding work to be done and frees himself to attend to other tasks in talk-exchange, including the planning of larger units of discourse (Pawley & Snyder, 1983: 192).

According to Sinclair (1991: 110), a formula is a semi-pre-constructed phrase that constitutes a single choice even though it might appear to be analyzable into segments. Due to processing time limitations on L2 production on tasks, speakers have very little time to plan what they will say and are likely to rely on utterances constructed as intonation units which are predictable in terms of lexical concordance (N. Ellis, 2002: 156). Furthermore, the greater the working memory demands of the processing task, the greater will be the need to rely on formulas. Both production and comprehension are thus mediated by learners’ representation of L2 chunks, and, according to N. Ellis, these representations develop through experience during both L2 comprehension and L2 production. L2 production on difficult tasks might thus be characterized as a piecing together of available units to effectively meet the communicative demands of the tasks a learner faces, and increased frequency of more advanced or elaborate phrases in line with these demands could be seen as essential to ongoing language development. Thus, it can be argued that learning is a result of meaning-focused language use of the type provided by L2 task performance. Learning is frequency driven.  It is a process of unconsciously tallying the frequencies of item types and tokens and determining probabilities based on their distributions. This process is supported by the mechanisms associated with developing cue strength, prototypicality and chunking.

2.3 Syntactic Priming However, in addition to increasing the frequency of certain items directly through the functional demands for specificity and elaboration that they place on learners, certain L2 tasks may also have priming effects on certain form to function mappings which results in the increased use of these combinations during subsequent production when contexts arise to which they are suited. N. Ellis (2002), for example, notes that the use of lexical items, thematic roles, or word sequences prime themselves in memory over stretches of discourse approximately ten sentences in length or intervals of up to approximately 20 min. In other words, language users tend to prefer reusing utterances for a given function which have been recently primed in memory rather than constructing novel utterances. Thus, the probability of using a specific form or utterance to achieve a given function increases based on how recently the utterance has been used within the discourse context (e.g., Pickering & Ferreira, 2005).

2.5 First and Second Language Acquisition

17

From a slightly broader perspective, De Bot and Larsen-Freeman (2011), based on the arguments of Larsen-Freeman (1997, 2010), claim that: …there is basically no difference between using the language and language change: Every time an element of a language is used, its status in the system is changed and the whole system is transformed, even if only to increase the probability of the element’s being selected the next time (De Bot & Larsen-Freeman, 2011: 16).

It follows that functionally demanding tasks which reduce the time between opportunities to use a more precise form to function combination (e.g., a sequence of very precise descriptions) may facilitate use of these combinations and their incidental acquisition. As MacWhinney (2001) points out (Sect. 2.2.1 above), this will be particularly important in the case of forms which L2 learners would otherwise use only intermittently, if at all, in accomplishing a task. Priming effects may thus play a crucial role in L2 learning outcomes on tasks such as those used in the present study. When tasks require learners to formulate descriptions at a level that they rarely encounter outside of the classroom, TBLT can push learners to aquire forms incidentally on tasks that they would otherwise not acquire.

2.4 Transferability The opportunity for task-based L2 output in itself may also have positive effects on the stability and transferability of implicit learning that occurs in conjunction with the performance of tasks. Erlam, Loewen, and Philp (2009) provide evidence that gains in implicit knowledge connected with L2 production are more stable over time and result in the ability to generalize to examples not explicitly practiced than implicit knowledge gained from processing L2 input alone. In their study, input processing instruction and instruction incorporating L2 output activities resulted in equal gains in implicit or explicit knowledge on immediate post-tests. However, on delayed post-tests of implicit knowledge, the output group outperformed the input group and showed more transfer of implicit knowledge to unpracticed parallel items. These finding suggest that meaningful L2 production may result in more stable gains in implicit learning and improve the transferability of this knowledge to new contexts. In other words, some benefits of L2 production may not become manifest in performance immediately, but involve long-term effects on the developing language system.

2.5 First and Second Language Acquisition N. Ellis (2002) argues that the basic learning sequence from the formula, through the slot-and-frame pattern, to the full constructions, will be the same in both L1 and L2 learning. Furthermore, Bybee (2008) argues that the analogical mechanism used to generalize language knowledge, the categorization mechanisms that are used for both linguistic and non-linguistic learning, and the chunking mechanisms needed to

18

2 L2 Performance and Incidental SLA

gain fluency in a language will occur naturally with practice whether one is learning a first language as a child or an additional language as an adult. Finally, as usage is argued to be the driving force behind cue strength and consequently emergence (MacWhinney, 2001, 2008), the emergence of language forms in the acquisition of an L1 and an L2 should be highly comparable provided that the frequency distributions in the language that the learner comprehends and produces are similar to those of their L1 peers. This makes exposure a key factor in the comparability of L1 and L2 speakers. As both Bybee (2008) and MacWhinney (2001, 2008) point out, it is difficult for L2 learners in a pedagogic or foreign language context to get the exposure characteristic of a L1 context. In any natural frequency distribution, high-frequency items might be argued to be present in sufficient quantities for acquisition. Semantic primes relating to fundamental referents and event types that are basic to the human experience of the world (e.g., cause, motion, state, possession, location), for example, will always be frequent. However, exposure to lower-frequency constructions, collocations and lexical items is very likely to be lacking in many contexts of L2 use, and the full range of forms that can serve a given task-essential function may not be represented. However, differences in exposure can potentially be addressed through well- planned instruction. Other factors related to conceptual development that distinguish L1 and L2 learning cannot be addressed in this way. The first major factor is the entrenchment of an L1 system. As MacWhinney (2008) notes, the brain of the L2 learner is already committed to the task of processing the first language, and Bybee (2008), for example, claims that some adults may simply lack the “neural plasticity” needed to set up the motor routines required for native-like proficiency in an additional language. According to MacWhinney (2008), these difficulties do not arise due to maturation, but rather due to entrenchment of the L1 system. Network mappings become progressively more complex over the lifespan, and it thus becomes progressively more difficult to re-organize fundamental factors. The second major factor distinguishing L1 and L2 learners is motivation. Adults learning an additional language already have a full understanding of the world and human society, and their conceptual development provides them with a set of standards by which to judge events with some degree of objectivity. As Bybee (2008) points out, the L1 learner acquires language as part of acquiring their identity in the culture in which he or she is raised. The L2 learner, on the other hand, has already matured as a person and is often simply trying to communicate. Children learning their L1 assimilate not only the language of those around them, but also the gestures, postures, facial expressions and other behavioral paraphernalia of their culture in minute detail. There are identity issues involved in the learning process that are far beyond the scope of effective L2 use, and L2 learners may not find it desirable to imitate the behavior of members of the target culture at this level. In short, the essential developmental processes, the learning mechanisms, and the input are potential the same for both L1 and L2 learners. However, learning will be moderated by differences related to factors in conceptual development. The most important of these factors might be argued to be L1 entrenchment and learner motivation.

References

19

2.6 Summary This chapter has attempted to connect task-based L2 production to incidental SLA. In line with usage-based accounts of SLA, it is has been argued that tasks provide learners with valid opportunities to map language forms to specific communicative functions. It has been argued that this requires syntax and lexis to be fully integrated with pragmatic task outcomes within a given discourse context. If task design factors can be shown to increase the functional need for precision, and consequently the use of developmentally-advanced and later-emerging language forms, tasks might be argued to play an important role in the restructuring of the language system. Furthermore, tasks might provide L2 learners with motivated opportunities to use language at a level to which they might otherwise have limited exposure. Finally, tasks may prime emergent forms within a given communicative context and result in increased stability and transfer of implicit learning gains. The next chapter will attempt to relate tasks to L2 use inside and outside of the classroom.

References Bybee, J. (1985). Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins. Bybee, J. (2006). From usage to grammar: The mind’s response to repetition. Language, 82, 711–733. Bybee, J. (2008). Usage-based grammar and second language acquisition. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 216– 236). New York: Routledge. De Bot, K., & Larsen-Freeman, D. (2011). Researching second language development from a dynamic systems perspective. In M. Verspoor, K. de Bot, & W. Lowie (Eds.), A dynamic approach to second language development: Methods and techniques (pp. 5–23). Amsterdam: John Benjamins. DeKeyser, R. (2001). Automaticity and automatization. In P. Robinson (Ed.), Cognition and second language instruction (pp. 125–151). Cambridge, UK: Cambridge University Press. Ellis, N. (2002). Frequency effects in language processing: A review with implications for theories for implicit and explicit language acquisition. Studies in Second Language Acquisition, 24, 143–188. Erlam, R., Loewen, S., & Philp, J. (2009). The roles of output-based and input-based instruction in the acquisition of L2 implicit and explicit knowledge. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, & H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing and teaching (pp. 241–261). Bristol, UK: Multilingual Matters. Hasher, L., & Chromiak, W. (1977). The processing of frequency information: An automatic mechanism? Journal of Verbal Learning and Verbal Behavior, 16, 173–184. Langacker, R. (1987). Foundations of cognitive grammar: Theoretical prerequisites. Stanford: Stanford University Press. Langacker, R. (2000). A dynamic usage-based model. In M. Barlow & S. Kemmer (Eds.), Usage- based models of language (pp. 1–63). Stanford: CSLI. Langacker, R. (2008). Cognitive grammar as a basis for language instruction. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 27–38). New York: Routledge.

20

2 L2 Performance and Incidental SLA

Larsen-Freeman, D. (1997). Chaos/complexity science and second language acquisition. Applied Linguistics, 26, 141–165. Larsen-Freeman, D. (2010). The dynamic co-adaptation of cognitive and social views: A complexity theory perspective. In R. Batstone (Ed.), Sociocognitive perspectives on language use and language learning. Oxford, UK: Oxford University Press. MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and second language instruction (pp. 69–90). Cambridge, UK: Cambridge University Press. MacWhinney, B. (2008). A unified model. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 341–371). New York: Routledge. Pawley, A., & Snyder, F. (1983). Two puzzles for linguistic theory: Native-like selection and native- like fluency. In J. Richards & R. Schmidt (Eds.), Language and communication (pp. 191–225). London: Longman. Pickering, M., & Ferreira, V. (2005). Structural priming: A critical review. Psychological Bulletin, 134, 427–459. Posner, M., & Snyder, C. (1975). Attention and cognitive control. In G. Bower & J. Spence (Eds.), Psychology and learning motivation (Vol. 3, pp. 55–85). San Diego, CA: Academic. Sinclair, J. (1991). Corpus, concordance, collocation. Oxford, UK: Oxford University Press. Tomasello, M. (2000a). First steps toward a usage-based theory of language acquisition. Cognitive Linguistics, 11, 61–82. Tomasello, M. (2000b). The item-based nature of children’s early syntactic development. Trends in Cognitive Sciences, 4, 156–163.

Chapter 3

Referential Communication and L2 Production

The present chapter discusses some of the roles that referential communication tasks have been argued to play in L2 performance and acquisition. Referential communication tasks are first defined in the way that it will be used throughout the book. A distinction is then made between two types of communication: (1) interactional communication, which is listener focused, and (2) transactional communication, which is message focused. This distinction is then elaborated into a discussion of specific constraints on task performance which optimize communication regardless of whether learners are using their L1 as children or using an additional language as adults. Following this, an approach to classifying and sequencing referential communication tasks in L2 instruction is outlined. In this approach, the functional demands of different discourse genres (i.e., description, instruction, narration and exposition) are argued to be related to conceptually motivated aspects of the developing language system. A discussion of the limitations of tasks in pushing some redundant aspects of language is then provided together with some possible ways of addressing this problem. The chapter closes with a discussion of the potential mediating effects of proficiency in task performance.

3.1 Referential Communication Tasks Beginning with Kraus and Weinheimer (1964, 1966, 1967), referential communication tasks have been widely used over the past 50 years in studies of L1 speech production. These tasks require learners to identify a target referent within a context of other referents which are more or less similar to the target referent. Yule (1997) defines referents as the range of objects, actions, attributes, states and relationships that must be identified and kept distinct by the learner in order to complete a task. As referential demand increases, he argues that L2 learners must go beyond their conventional or routine lexical responses to tasks and create original responses that require the integration of thought, language, and context during performance. © Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_3

21

22

3 Referential Communication and L2 Production

In terms of language processing, Yule (1997) argues that this requires two groups of abilities. The first group relates to the identification of the referent as distinct from non-referents and involves three sub-skills: (1) perceptual ability to discern the range of attributes of a given referent, (2) comparative ability to identify the differences and similarities between the referent and the non-referent, and (3) linguistic ability to encode these attributes and features into the target language. The second group of abilities relates to perspective-taking and involves four sub-skills: (1) the ability to recognize the need to consider the interlocutor’s perspective, (2) the ability to make accurate inferences about the interlocutor’s perspective, (3) the ability to make use of these inferences in developing the message, and (4) the ability to monitor performance and use feedback to edit and improve subsequent messages. In other words, acts of reference must occur within the context of authentic, meaning-focused language use in order to provide a valid perspective on performance variation as it relates to the L2 learning process. Furthermore, Yule (1997) argues that demanding referential communication tasks will frequently require reference to features for which no conventional label exists, and learners will be forced to develop a shared means of identification on a post hoc basis. The experience of having to refer to entities for which a conventional label is unknown, or of having to form a perspective not previously taken on a particular entity, leads learners to manipulate available linguistic resources to accomplish an immediate referential goal. The result of this process is the categorization of referents on an effective but temporary basis using an agreed-on means of communication without ever needing to consider special terminology from the dictionary. Yule provides an example of a native speaker of English asking for a part to fix his window screens and referring to it metaphorically as a ‘wing’ which the proprietor of the shop not only understands, but subsequently adopts for lack of a better term. In short, Yule (1997) argues that the adult acquisition of L2 referential skills is not discrete and linear. It requires individualized development through the manipulation of available linguistic resources to accomplish immediate referential goals with a range of interlocutors. Learners perform tasks in their own ways by taking control over the choice of linguistic forms and the discourse structure, focusing primarily on meaning rather than on using specific linguistic structures, and, importantly in Yule’s argument, taking responsibility for the outcome. As will be seen in Chap. 4, Swain (1985, 1995) also argues that the decision on the part of the learner to voluntarily allocate attention to a specific self-set learning goal is an important step in the learning process. For the time being, let it suffice to say that, if accessing language for the achievement of a non-linguistic outcome is instrumental in facilitating the process of SLA, referential communication tasks might play a central role in both L2 learning theories and L2 pedagogy.

3.2 Communicative Demands

23

3.2 Communicative Demands The notion of L2 referential communication relates closely to the distinction made by Brown and Yule (1983) between interactional and transactional uses of language. These distinctions can also be argued to relate directly to ongoing lexical, morpho- syntactic and pragmatic development in communication. Table 3.1 summarizes some general tendencies distinguishing these two types of communication. According to Brown and Yule (1983), interactional communication is important for maintaining social relations, but does little to promote the ongoing syntactic and lexical development of the language system. Interactional communication relies on the immediate physical context as well as shared assumptions, background knowledge and experience and thus allows for a considerable amount of implicit understanding in message formulation. Expressions such as “you know”, “it’s like” or “and so on” are typical of interactive discourse in English and express the desire for implicit understanding on the part of the listener. The focus of attention in interactional communication is typically on the other person’s feelings and involvement in the conversation rather than on the message itself. Turn-taking is thus generally balanced, and agreement is sought above all else with ambiguity or abrupt topic switches being acceptable means of avoiding any potential conflicts. While skills connected with interactional communication are important in using a language effectively, ongoing lexical and syntactic development might better be served by the demands of transactional communication tasks. In transactional communication, the goal is to communicate a specific message. To this end, it may be acceptable for an information giver to take extended turns or to exclude the information receiver altogether. In other words, the roles of speaker and listener are clearly defined in transactional communication, the interest is in achieving a pre-established goal, and it is unacceptable to switch topics until that goal is reached. Clarification requests, comprehension checks and even direct contradictions are all acceptable provided they work toward the shared goal of an explicit mutual understanding of the required information.

Table 3.1 Two types of communication (based on Brown & Yule, 1983) Transactional Institutional (work/school) Message-oriented Transfer of information New information Seeks clarification Longer turns Independent turns Defined roles (Info giver-receiver) Monologues Objective perspective

Interactional Social (home/friends) Listener-oriented Maintenance of social relations Shared knowledge Seeks agreement Shorter turns Co-dependent ‘scaffolding’ of discourse Balanced contributions Dialogues Subjective perspective

24

3 Referential Communication and L2 Production

Finally, in interactional discourse, the perspective is usually subjective, personal and centralized in the first and second person. In transactional discourse, however, people are expected to decentralize their perspective and speak objectively to a general audience. This means that less can be assumed in terms of shared experience and immediate context. The ideas, relationships, and relevant contextual factors must all be made explicit to the degree that the speaker is an effective communicator. In other words, transactional discourse requires the speaker to produce a message which is largely, if not entirely, independent of the listener’s input. The standards governing these two types of communication have important implications for what will be considered successful L2 performance in any given situation. The ability to distinguish when each type of communication is appropriate and to perform accordingly is an important aspect of L2 proficiency. When in social situations, interactional norms of communication tend to determine what is appropriate and constitutes successful performance. On the other hand, whenever important issues need to be addressed, transactional norms will determine what is appropriate and constitutes successful performance. Recognizing and being able to cope with these two types of discourse is a very important part of communicative competence in a language. For example, if a colleague, upon meeting in the elevator at work, says, “Nice day isn’t it?” it would probably be inappropriate to take this as an invitation to establish explicit criteria for evaluating the weather. However, if a colleague on a high stakes test development project, while looking over prototypical items for a section of the exam you are working on together, points to a specific item and says, “Nice question, isn’t it?” it would probably be equally inappropriate to agree casually and inquire about her children or a recent vacation. For the purpose of the present argument, however, transactional communicative demands are more important in pushing the development of L2 learners’ syntax and lexis than the demands of interactional communication.

3.3 Contextual Constraints According to Yule (1997), it is the ongoing need for more explicit, de-contextualized communication characteristic of transactional or referential language use that pushes on-going language development regardless of whether learners are acquiring their first language as children or additional languages as adults. Furthermore, he argues that these referential communication skills can be improved with training. If young children are (1) required to reformulate ambiguous messages until they are unambiguous, (2) required to demonstrate their awareness of ambiguity, and (3) given verbal feedback regarding message effectiveness (e.g., “but there are four like that”), he argues that their ability to both identify and produce unambiguous messages will increase (1997: 20–22). However, the ability to de-centralize messages and to see things from others’ perspectives is also dependent on cognitive development. Yule (1997) argues that between the ages of 6 and 7 children develop from a latent sense of ambiguity to an explicit understanding of ambiguity. Although children 4–5 years old have a sense

3.3 Contextual Constraints

25

of variability between intended and conveyed meaning as evinced by longer reaction times and the significant effect of warnings and feedback, children 7–8 years old have this knowledge fully available and are able to manifest it explicitly in their language use. Beginning adult learners of additional languages will, of course, differ in this respect from beginning first language learners due to their level of cognitive development. Nevertheless, the need for explicit, de-contextualized communication can ultimately be expected to push L2 performance in either case. It is also interesting to note that the three conditions specified by Yule (1997) as crucial factors in referential training as well as the developmental shift between latent and explicit understanding of ambiguity are concomitant with the onset of formal schooling in most countries. The conditions imposed on communicative effectiveness in schools may, in fact, serve an important function in the L1 language development of children as well as provide some insight into how the constraints and conditions governing the performance of L2 learning tasks may be optimized to push L2 development in a classroom context. Table 3.2 contrasts some key factors that tend to characterize language use in formal educational settings as opposed to language use at home. The formal educational context requires children to embrace complex, unfamiliar topics in oral and written texts. These texts frequently consist of extended sequences of ideas with relationships between them made explicit lexically and morpho-syntactically. Formal education also pushes learners to demonstrate their understanding of new ideas clearly and explicitly. Taken together, these increased communicative demands require learners to pay attention, to some extent, to the ways that language functions while listening, reading, speaking and writing rather than focusing only on the meaning of what is said. It also requires them to attempt to be objective, to some degree, and take independent responsibility for the clarity of their language. Finally, the presence of linguistic devices by which learners’ make their speech more explicit has been shown to be a key element in L2 development crosslinguistically (Perdue, 1993). Based on longitudinal studies of adult L2 learners from several L1 backgrounds learning several different languages, Klein and Perdue Table 3.2 Contextual influences on communication skill development (Yule, 1997) School Extended instructional sequences Demonstration of understanding demanded Decentralized/objective perspective Conscious attention to language form Unfamiliar topics High information processing load More information representation Novel discourse structures Explicitness valued De-contextualized discourse Independent responsibility for clarity

Home Short interactional sequences Implicit understanding accepted Personal/subjective perspective Meaning-focused communication Familiar topics Low information processing load Less information representation Predictable discourse structures Intuitiveness valued Shared discourse context Co-dependent responsibility for clarity

26

3 Referential Communication and L2 Production

(1993), for example, conclude that a shift from an initial pragmatic organization of words (e.g., the juxtaposition of content words) to the use of explicit markers of grammatical relations (e.g., prepositions, verbal particles) is a defining characteristic of communicative L2 development across learners and languages. Although some of the learners observed continued to progress to a third stage of development marked by phrasal constraints such as finiteness and person/number agreement markings, development at this level was not universal and showed more individual variation than development from the first to the second stage. The development documented by Klein and Perdue provides insight into what Givon (1985) refers to broadly as a transition from an earlier and simpler pragmatic mode of production to a later and more complex syntactic mode of production, which, he claims, will be characterized by the use of more morphology, more syntactic subordination, and a higher noun to verb ratio.

3.4 Discourse Demands Although empirical research on the relationship between the discourse demands of tasks and the language forms used to complete them is currently lacking, Yule (1997) provides some interesting initial insights in this direction. He identifies three task types (descriptive, instructional, and narrative) which he argues will place progressively higher discourse demands on speakers. Yule also makes a distinction between these three types of tasks and opinion or expository tasks, but does not elaborate on the linguistic demands of this fourth task type in relation to the others. Yule’s (1997) claims that description tasks are the linguistically least demanding of these task types. Description tasks require learners to identify a particular referent within a larger set of referents. Description tasks thus require minimally only a head noun or a simple noun phrase (e.g., an adjective-noun or noun-prepositional phrase combination), and they frequently allow learners to cooperate and scaffold their discourse. They are thus particularly well-suited for use at the beginning level because in their simplest form they potentially allow for successful task completion with little more than a handful of relevant lexical items. As it is hoped the present study will demonstrate, however, description tasks can be designed to require the use of more developmentally advanced language. Although the demands of description tasks per se may be largely limited to variation in the nominal system, as will be argued in Chaps. 5, 6 and 7, it may nevertheless be possible to push this aspect of the developing language system through the use of tasks requiring differing levels of precision in description. One example might be the manipulation of the similarity between the referents that must be identified. Instruction tasks (e.g., giving directions based on a map, explaining how to install or assemble a product, explaining how to cook a specific dish) are more linguistically demanding than description tasks in that they require learners not only to identify specific referents but to do so within a sequence of actions. These tasks

3.5 Limitations of Tasks in Instructed SLA

27

Table 3.3 Task type and functional language demands (Yule, 1997) Sentence-level processing Longer, independent turns Variable thematic roles Variable verbal structures

Description No No No No

Instruction Yes Yes No No

Narrative Yes Yes Yes Yes

entail clear information giver and receiver roles which require speakers to take longer, independent turns and produce complete ideas. They thus result in sentence- level processing and the sequential linking of events in the discourse (e.g., first, next, after that, finally). However, they have the dual advantage of being limited to a fixed verbal structure in the impartive mode which consisting of simple verb forms (i.e., first do X, then do Y, finally do Z) with little or no need to keep track of adjustments for grammatical agreement, past tense forms (both regular and irregular) or to mark specific tense-aspect relationships. In this sense, instruction tasks provide a good intermediate step between description and narration tasks. Finally, narration tasks (e.g., recounting a traffic accident or relating a problem that occurred at work) will be more demanding that both description and instruction tasks in that they require learners to distinguish multiple referents and follow them through sequences of events requiring both regular and irregular past tense verb forms all marked for subject-verb agreement, and possibly requiring tense-aspect relationships to be clarified. More advanced narrative tasks might be designed to push the development of linguistic devices to foreground and background information in the narrative (Skehan, 2009), motivate events by specifying causes, intentions, or mental states (Lambert & Robinson, 2014; Robinson, 2011), and distinguish between one’s own and other perspectives on the relevant events linguistically (Robinson, 2011). Narrative tasks might thus lead to opinion or expository tasks (e.g., motivating and supporting certain perspectives and drawing appropriate inferences basis of this reasoning) which Yule (1997) mentions only briefly without much speculation on their linguistic demands (see Berman, 2008). Table 3.3 provides a summary of Yule’s (1997) claims regarding the discourse required by tasks of differing discourse demands. Yule’s model thus provides an initial example of how the discourse demands of tasks might be related to conceptually motived aspects of the developing L2 system. Another approach to this problem will be discussed in Chap. 4.

3.5 Limitations of Tasks in Instructed SLA Although tasks may be useful in promoting many aspects of SLA, they also have limitations. The ability to communicate information clearly and unambiguously provides the primary criterion of successful L2 performance in most contexts of

28

3 Referential Communication and L2 Production

natural language use, and tasks have the potential to push learners to reach these goals. However, due the inherent redundancy of language, coupled with the rich availability of contextual clues in most contexts, some linguistic forms that are natural or useful in the completion of a task will not be required in order to simply complete tasks (Lochsky & Bley-Vroman, 1993; Nobuyoshi & Ellis, 1993). While some tasks may require very precise descriptions, for example, there will never be a one-to-one relationship between such task demands and specific language forms. The means of dealing with these communicatively redundant features of language within a task-based approach to L2 instruction has been the subject of much discussion in the literature (see, for example, Doughty & Williams, 1998; Ellis, Basturkmen, & Loewen, 2002; Long & Robinson, 1998). In relation to task design, two primary suggestions for dealing with these features of the language are: (1) formatting input-based tasks to highlight problematic forms in aural and written texts and increase their salience (input enhancement); and (2) elaborating these texts in order to increase the frequency of these features (input flooding). In addition to such input-based focus on form, however, there is also general agreement on the necessity of well-timed methodological interventions during the performance of output-based tasks in the L2 classroom. Minimally, these interventions will involve techniques such as clarification requests, confirmation checks, and the recasting of utterances containing errors in communicatively redundant aspects of the language system. Ellis et al. (2002) provide a practical survey of these methodological interventions and their potential implications for ISLA.

3.6 Proficiency and L2 Performance Task design cannot be expected to push learners to produce forms that they do not know. In order to isolate the language demands of tasks empirically, it is important that the speakers in question are capable of using the range of forms involved. Thus, in addition to establishing that dependent variables in task-based L2 research are motivated conceptually by the demands of the task, motivated theoretically by the developmental literature on language emergence, and motivated empirically by previous research on the effects of comparable tasks, it is also important to carefully control for the proficiency level of the participants in researching the effects of tasks on L2 performance. One means of doing this is to collect baseline data from their native speaking peers. This data is useful in establishing the effects of task demands on language use independent of language proficiency per se and will allow any variation in non-native speech at different levels of proficiency to be better understood. Furthermore, in evaluating L2 task performance, there may be separate, competing factors underlying the successful completion of a task. It is reasonable, for example, to expect that the use of developmentally advanced language will be counteracted to some extent by increased experience and communicative efficiency (White & Robinson, 1995). Task design factors might place increased functional

References

29

demands on all speakers and result in the use of developmentally more advanced language forms for all groups. However, this does not preclude the possibility that higher proficiency speakers who have a broader range of linguistic resources available to them will be able to complete a given set of tasks more simply and efficiently than lower proficiency speakers who are drawing on a more limited range of experience.

3.7 Summary The present chapter has discussed the relationship between referential communication task, L2 performance, and SLA. Tasks were defined in a way that preserves their integrity as tools for promoting incidental SLA processes during meaning- focused L2 communication. The nature of task-based communication was then further clarified by distinguishing between the demands of different types of communication (interactional and transactional) and the demands of different communicative contexts (home and school). Following this, a model of the discourse demands of tasks was outlined together with arguments connecting these demands to different aspects of the developing language system. The limitations of tasks in promoting target levels of L2 use were then discussed together with some specific caveats regarding the role of proficiency in moderating the effects of tasks on L2 production. The next chapter will address recent theories of how tasks might be sequenced into L2 syllabuses based on the cognitive demands for language processing that they make on learners. These theories differ from those outlined in the present chapter in that they attempt to explain the effects of L2 task design factors in terms of general cognitive processes that learners engage in with respect to language while performing tasks rather than with respect to specific aspects of the language system.

References Berman, R. (2008). The psycholinguistics of developing text construction. Journal of Child Language, 35, 775–771. Brown, G., & Yule, G. (1983). Teaching the spoken language. Cambridge, UK: Cambridge University Press. Doughty, C., & Williams, J. (1998). Focus on form in classroom second language acquisition. New York: Cambridge University Press. Ellis, R., Basturkmen, H., & Loewen, S. (2002). Doing focus on form. System, 30, 419–432. Givon, T. (1985). Function, structure and language acquisition. In D. Slobin (Ed.), The crosslinguistic study of language acquisition, Vol. 2: Theoretical issues (pp. 1005–1027). Hillside, NJ: Lawrence Erlbaum. Klein, W., & Perdue, C. (1993). Utterance structure. In C. Perdue (Ed.), Adult language acquisition: Crosslinguistic perspectives: The results (pp. 3–40). Cambridge, UK: Cambridge University Press.

30

3 Referential Communication and L2 Production

Krauss, R., & Weinheimer, S. (1964). Changes in reference phrases as a function of frequency of usage in social interaction. Psychonomic Science, 1, 113–114. Krauss, R., & Weinheimer, S. (1966). Concurrent feedback, confirmation, and the encoding of referents in verbal communication. Journal of Personality and Social Psychology, 4, 343–346. Krauss, R., & Weinheimer, S. (1967). Effect of referent similarity and communication mode on verbal encoding. Journal of Verbal Learning and Verbal Behavior, 6, 359–363. Lambert, C., & Robinson, P. (2014). Learning to perform narrative tasks: A semester-long study of task sequencing effects. In M. Baralt, R. Gilabert, & P. Robinson (Eds.), Task sequencing and instructed second language learning. London: Continuum. Larsen-Freeman, D. (1997). Chaos/complexity science and second language acquisition. Applied Linguistics, 26, 141–165. Long, M., & Robinson, P. (1998). Focus on form: Theory, research, and practice. In C. Doughty & J. Williams (Eds.), Focus on form in classroom second language acquisition (pp. 15–41). Cambridge, UK: Cambridge University Press. Loschky, L., & Bley-Vroman, R. (1993). Grammar and task-based methodology. In G. Crookes & S. Gass (Eds.), Tasks and language learning: Integrating theory and practice (pp. 123–167). Clevedon, UK: Multilingual Matters. Nobuyoshi, J., & Ellis, R. (1993). Focused communication tasks and second language acquisition. ELT Journal, 47, 203–210. Perdue, C. (1993). Adult language acquisition: Crosslinguistic perspectives: The results. Cambridge, UK: Cambridge University Press. Robinson, P. (2011). Second language task complexity, the cognition hypothesis, language learning and performance. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Skehan, P. (2009). Modeling second language performance: Integrating complexity, accuracy, fluency and lexis. Applied Linguistics, 30, 510–532. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output on its development. In S. Gass & C. Madden (Eds.), Input in second language acquisition (pp. 235–253). Rowley, MA: Newbury House. Swain, M. (1995). Three functions of output in second language learning. In G. Cook & B. Seidlhofer (Eds.), Principle and practice in applied linguistics (pp. 125–144). Oxford, UK: Oxford University Press. White, R., & Robinson, P. (1995). Current approaches to syllabus design: A discussion with Ron White. Guidelines, 17, 93–101. Yule, G. (1997). Referential communication tasks. Mahwah, NJ: Lawrence Erlbaum.

Chapter 4

Tasks in L2 Syllabus Design

This chapter discusses current models of L2 task sequencing based on the cognitive demands for L2 processing that they make on learners. Two psycholinguistic models of task-based instruction are compared: The Limited Attention Capacity Hypothesis (Skehan, 1996, 2009, 2014, 2016) and the Cognition Hypothesis (Robinson, 2001, 2011). However, before discussing these models, some background issues, integral to them both, are discussed. These issues relate to theories of attention, speech production, and the cognitive processes of SLA. The chapter then compares the similarities and the differences between the two models together with their implications for L2 instructional planning. The primary concern in both models has been with how learners’ attention is focused during the completion of tasks as reflected by general measures of the fluency, accuracy and syntactic complexity of their L2 production. These three facets of performance have been taken as general indices of learners’ attention to the form and the meaning of the language that they use, and they it is presumably assumed that a balance of attention to these aspects language during L2 use will ultimately relate to balanced L2 development.

4.1 Attention Attention plays an important role in both of models of task performance that will be discussed in the present chapter. Attention is also involved in the initial registration and encoding of phenomena in the usage-based theories of SLA outlined in Chap. 2. It will thus be helpful to briefly discuss theories of attention as they relate to language use.

© Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_4

31

32

4 Tasks in L2 Syllabus Design

4.1.1 Detection According to Tomlin and Villa (1994), attention systems function to reduce and control the influx of information from the senses. Following research by Posner and colleagues (e.g., Posner & Petersen, 1990; Posner & Rothbart, 1992), the authors identify three networks of attention that they argue to have identifiable neurological correlates in the brain: detection, orientation and alertness. Detection refers to the cognitive registration of sensory stimuli in memory so that they are available for subsequent processing (1994: 192–193). Detection thus constitutes the primary function of attention in learning. Without detection, briefly perceived information is not registered in memory and thus does not affect SLA processes. Detection of new information requires attentional resources and can interfere with the processing of other information. Orientation refers to the direction of attention to one type of information to the exclusion of others. Orientation has an inhibiting effect on the detection of non- oriented items and a facilitating effect on the detection of oriented items. In other words, it is more difficult to detect items that are unexpected than it is to detect those that are. Finally, alertness refers to a general eagerness on the part of the learner to deal with incoming stimuli or data. According to Tomlin & Villa, alertness does not directly facilitate detection. It can have positive and negative effects on learning outcomes. It can lead to more rapid detection of high-priority targets but can also result in increased errors if the learner is overeager. Compared to detection, orientation and alertness are distal issues in attention as it relates to SLA. While both may have a facilitating effect on detection, neither is required for detection to occur (Tomlin & Villa, 1994: 198).

4.1.2 Awareness Tomlin and Villa (1994) distinguish these three components of attention (detection, orientation and alertness) from awareness on the part of the learner. Awareness refers to the subjective experience of stimuli and none of the central components of attention – alertness, orientation, or detection – require awareness. Awareness thus requires attention, but attention does not require awareness. Tomlin & Villa argue that information can be detected and made available for subsequent learning processes implicitly and without the learner’s conscious awareness. This is critical for task-based language learning in that learners can detect and register linguistic stimuli for subsequent processing while their attention is focused on meaning (cf. Sect. 2.2 above). In contrast, Schmidt (2001) distinguishes between registration, or detection without awareness, and noticing which he defines as “detection within focal attention accompanied by awareness (Schmidt 2001: 18).” Schmidt claims that noticing is the subjective correlate of attention or awareness at a very low level of abstrac-

4.1 Attention

33

tion, and he equates noticing with what other SLA researchers have referred to as ‘apperception’ (Gass, 1988), ‘detection with selective attention’ (Tomlin & Villa, 1994) and ‘detection plus rehearsal in short term memory’ (Robinson, 1995). Schmidt (1990, 2001), based on his earlier case study research (e.g., Schmidt, 1983; Schmidt & Frota, 1986), argues that learners must attend to and notice any source of variation that makes a difference in meaning for it to be acquired. In other words, for Schmidt, attention is simply a prerequisite to learners becoming aware of ‘gaps’ between the language that they can currently produce and the language that they need to produce. In terms of what learners actually notice, Schmidt limits his notion of noticing to specific linguistic features in the surface structure of utterances. Abstract rules and generalizations based on such examples are more complex processes which imply comparisons across instances and metalinguistic reflection. Turscott and Sharwood-Smith (2011: 500–506) discuss several problems associated with defining and operationalizing noticing. The first has to do with the empirical testability of the hypothesis. While the noticing hypothesis does not claim that all features of a language must be noticed, it provides no basis for determining what needs to be noticed and what does not. The hypothesis also does not make a clear distinction between noticing and understanding, on the one hand, and noticing and simple awareness of input, on the other. In addition, Williams (2005) has provided evidence that L2 learning does occur without awareness, and Hama and Leow (2010) have pointed out the need to consider the role of awareness at different stages of language processing. A key question in theories and research on task-based L2 instruction has been what learners are attending to (with or without conscious awareness) during the performance of tasks, and it has generally been assumed that attention to form and attention to meaning are discreet and separable constructs in this regard both subjectively and empirically: It is generally taken for granted that attention is employed by the learner in one of two ways. It is employed to aid in the comprehension of the meaning of an utterance – attention to meaning; or it is employed to aid in the psycholinguistic processing of the components of an utterance – attention to form (Tomlin & Villa, 1994: 186).

In particular, it has been hypothesized that increased fluency is associated with learners’ attention to meaning while increased complexity is associated with learners’ attention to form. These arguments will be summarized in more detail in Sect. 4.4 below, and common ways in which the constructs of complexity and fluency have been operationalized will be discussed in detail in Chap. 6.

4.1.3 The Output Hypothesis Swain’s (1985, 1995) Output Hypothesis also posits an important role for conscious awareness in SLA. The Output Hypothesis has been central to much of the theory and research related to task-based L2 instruction, particularly as it relates to the

34

4 Tasks in L2 Syllabus Design

notion of designing tasks to push learners’ output as will be discussed in detail in subsequent chapters. Swain (1995) argues that L2 output is essential to the development of accuracy in SLA in three ways: noticing, hypothesis testing, and use of meta-language. In line with Schmidt’s (1990, 2001) Noticing Hypothesis (see Sect. 4.1.2 above) which argues that L2 learning requires both detection and awareness at the level of conscious noticing, Swain claims that the processes involved in L2 production help learners ‘notice the gaps’ between what they intend to say and what they can actually say. In particular, noticing associated with L2 output raises learners’ consciousness of their problems at a more specific level than noticing associated with the processing L2 input. Swain also hypothesizes that learning targets that learners discover in this way are more likely be at a level that they are capable of acquiring than targets identified by teacher or syllabus planners and imposed on learners. Teachers frequently identify and attempt to address problems in students’ output that learners are either unable or unwilling to learn. Problems that are discovered by learners as they attempt to speak, however, constitute a more relevant and meaningful basis for action. The second way Swain (1995) argues that L2 production facilitates SLA is in providing learners with chances to test, evaluate, and revise the hypotheses that they have generated about the language and how it is used. These hypotheses do not have to be formulated explicitly. Learners develop an awareness that a problem exists and consequently attempt to solve it by whatever means are available. Such learner- generated hypothesis testing results in modifications of learner language which ultimately move in the direction of ‘closing the gaps’ between what the learner wants to say and what the learner can currently say. This process of hypothesis testing can require incorporation of new linguistic knowledge or the further consolidation of current knowledge that is still imperfectly mastered. Either way, Swain argues that experimentation with learner-generated hypotheses to address self-set needs during L2 production is crucial to L2 learning. Finally, the third function of output, according to Swain (1995), relates to the use of meta-language connected with learner-identified problems. There will be cases in which noticing and hypothesis testing result in discussion of problematic language forms. In such cases, learners’ proposed solutions to their own problems will be explicitly formulated and expressed. Swain claims that talking about language problems in this way will also facilitate development. Verbalizing problems one is having helps learners better define the problem and to clarify ways of addressing it. The success of such discussions, however, depends on learners having set their own agenda regarding based on self-determined needs. In short, Swain’s (1995) position is that L2 production opportunities are essential for accuracy in using a language because of the level of awareness of problems that it provides together with the corollary benefits of active problem solving by learners that it involves. Information regarding accurate L2 use cannot be imposed on learners in the form of pre-determined learning objectives. Nor can accurate L2 use be achieved through corrective feedback unless the learner is personally and developmentally ready to address the problem, and the feedback results in the learner ‘noticing’ the problem.

4.1 Attention

35

Subsequent research supports the role of output in SLA. Izumi (2002), for example, concludes that pushed output is an ‘internal attention-drawing technique’ and that it may result in three related processes: (1) the detection of formal elements in the input through priming induced by internal feedback, (2) integrative processing of target structures, and (3) the noticing of mismatches between one’s inter-language and the target language. Izumi thus claims that output can serve ‘as an internal priming device for grammatical consciousness-raising’ (Izumi, 2003: 184). Furthermore, Toth (2006) suggests that the act of producing language may force cognitive processes that lead to restructuring of the developing system. Toth claims that the push to syntactically encode what one means to say, in combination with input from others and conscious reflection about language use, drives grammatical development. Erlam, Loewen, and Philp (2009) also argue that learners who engage in L2 output are required to process forms at a ‘deeper level’ than learners who work only on input tasks. In other words, the production of output may necessitate the processing of aspects of language that are not necessary for the processing of input. Finally, Erlam et al. (2009) point out that Levelt’s (1989) model of L1 speech production (see Sect. 4.2 below) provides a theoretical foundation for the output hypothesis in that it posits a feedback system that allows learners’ speech, whether overt or internal, to be fed back to a conceptualizer where it may be monitored (cf., de Bot, 1992; Kormos, 2006). Attention, with conscious awareness, can thus be given to whether the speech is well-formed and appropriate. While intuitive, claims about ‘noticing’ are inherently metaphorical and refer to subjective experiences that are very difficult to operationalize and test empirically. Truscott and Sharwood-Smith (2011), for example, argue that it is doubtful whether subjective experience at the level of noticing constitutes an empirical question at all. The most direct measure of noticing might be concurrent think-aloud protocols or stimulated recalls (e.g., Bowles, 2010; Gass & Mackey, 2000; Jourdenais, 2001). However, the validity of such procedures for identifying how learners attend to language during performance would minimally require controls for prior knowledge and for guessing about what is being targeted (Schmidt, 2001: 20; Robinson, Mackey, Gass, & Schmidt, 2011). Such controls are very difficult to achieve when learners are familiar with linguistic targets though formal study a foreign language and placed in a situation where they are asked questions about what they are thinking by a language teacher or researcher. The present study will thus be limited to the relationship between task design and developmentally advanced language use during task performance.

4.1.4 Resource Pools Crucial to both of the cognitive models of task-based L2 performance discussed later in this chapter is the idea that attention is limited and thus must be strategically allocated. Although early research assumed that attention was a single resource (e.g., Kahneman, 1973; Shiffrin & Schneider, 1977a, 1977b), more recent

36

4 Tasks in L2 Syllabus Design

research on divided attention and dual task performance provides strong evidence that human attention is likely to be divided into what could be referred to metaphorically as ‘resource pools’ which are dedicated to specific functions (e.g., Wickens, 2007). According to this perspective, each pool of attention is still assumed to have a limited capacity, but the allocation of pools to specific functions results in some flexibility in the way that these capacity limitations are manifest in performance. Wickens (2007) argues for a three-way model of attention resources with separate resource pools dedicated to modalities (written, oral), codes (spatial, verbal) and processing stages (perception-working memory, responding). Wickens also claims that these distinctions have a physiological basis in specific regions of the brain. In short, these resource pools allow attention demanding processes to be carried out at the same time more easily if they draw on different resource pools than if they draw on the same resource pool. An example of mode difference might be trying to scan an e-mail while listening to someone on the telephone (different modalites) versus catching the gist of a news announcement on the radio while listening to someone on the telephone (same modality). According to Wickens, completing the latter tasks simultaneously will be more difficult than the former and result in performance decrements because the latter tasks draw on the same pool of attention resources, whereas the former tasks draw on different attention pools. Next, an example of code difference might be trying to scan a visual diagram for information while giving a lecture (different codes) as opposed skimming a written text for the same information while giving a lecture (same codes). Once again, the latter set of tasks will draw on the same resource pool and thus be hypothesized to result in poorer performance than the former which draws on the same attention pool. Finally, an example of differences in processings stages might be responding to one question on a talk while receiving another (different stages) versus receiving two questions on a talk at the same time (same stage). The latter set of tasks will be more demanding on attention than the former due to the attention pools involved. Wickens (2007) predicts that the more any two tasks share common levels of modality, code, and stage, the greater the interference between them during concurrent task performance will be. This has important implications for L2 task design and sequencing. The question relates to whether attention to syntactic complexity and grammatical accuracy during L2 speech production draws on the same pool of attention resources, and thus result in a trade-off in which learners’ attention is allocated to either the accuracy or the complexity of performance, or whether these two aspects of L2 performance can be argued to draw on separate pools of resources and thus allow for attention to be allocated to both the accuracy and the complexity of performance concurrently when learners perform demanding L2 tasks.

4.1 Attention

37

4.1.5 Selection (Saliency, Effort, Expectancy & Value) The criteria that determine the information selected and processed from the overwhelming influx of sensory information during L2 performance is crucial to any study of L2 task design. These criteria will also determine what is ultimately learned as a result of performance. According to Wickens (2007), selection during task performance is based on competition. Wickens identifies four features of events that he argues can be used to predict selection when several events compete for attention: salience, effort, expectancy and value (SEEV). The first two of these are physical properties of the environment, including features of tasks and the language code itself. By contrast, the latter two are properties of the knowledge base that learners bring with them to the task. Wickens argues that these four factors operate additively in determining the probability of attending to a given item. Information connected with activated ideas and their associated information are detected and processed in working memory. Short of such processing, stimuli persist in working memory for only a very brief period. Ongoing processing, requiring further attention, is thus necessary for long-term storage in memory. A key question for any theory of task-based L2 learning, of course, is the specific features of tasks and the language code that increase the saliency of events for learners and generate the effort required for the ongoing processing of them. Regarding saliency, Talmy (2008) proposes several factors that determine the ‘distinctness, clarity, and significance’ of entities, on the one hand, and their ‘meldedness, vagueness or ordinariness’ on the other. Two of these factors are particularly important to the design of referential communication tasks. The first relates to the use of referents which are explicitly represented in the lexicon as opposed to those that are not and must be ‘reconstituted’ linguistically based on contextual and background knowledge. According to Talmy (2008: 34), explicit representation in an individual’s lexicon determines the center of his or her gradient of attention. The highest levels of a learner’s attention are directed to referents for which the learner has a lexical item (e.g., a jacket). Lesser levels of attention are then directed toward other concepts in the lexical domain (e.g., lining, sleeves, collar, buttons, lapels). Finally, the lowest level of attention is directed toward abstract concepts associate with the referent (e.g., size, shape, position). Talmy’s (2008) model of salience implies that the tasks learners are likely to accomplish in naturally occurring communication may not draw their attention to aspects of referents not explicitly represented in the lexicon and thus not require them to develop the linguistic means of identifying them precisely. By designing tasks in which referents are very similar and distinguished only be abstract differences may provide one means of drawing learners’ attention to these aspects of referent. However, it will be remembered in this connection that Wickens (2007)

38

4 Tasks in L2 Syllabus Design

model of attention predicts that salience, effort, expectancy and value (SEEV) will work in combination to determine the aspects of referents participants choose to describe. Attention during task performance is likely to be moderated by these other factors producing some variation in task effects between learners.

4.2 Speech Production The modular model of L1 speech production proposed by Levelt (1989, 1999) and adapted to L2 speech production by de Bot (1992) and Kormos (2006) is also crucial to understanding the differences between current models of task-based L2 instruction. The model postulates three primary stages of speech production. The first is a conceptualization stage in which the speaker selects information from world knowledge to include in a message and organizes it into an information structure to create a pre-verbal plan. The second is a complex set of procedures referred to as a formulation stage in which the pre-verbal plan is encoded grammatically and phonetically. Concepts and their relational structure are projected onto a phrase structure driven by lemmas drawn from the speaker’s mental lexicon that have associated semantic, syntactic, morphological, and phonological properties. This phrase structure is then encoded with phonetic and prosodic information to produce a phonetic plan. The third and final stage of speech production is then articulation in which the phonetic plan is buffered and parsed as syllables at the motor level. The essence of this model, as it relates to theories of task-based L2 instruction, is that monitoring can occur at any of these stages. Although each module is hypothesized to work on specific input and generate specific output for the next module, it is claimed that they can operate simultaneously provided that processing in the parallel module is sufficiently automatic. For proficient speakers, formulation may be largely automatic and allow for parallel processing with other modules, whereas conceptualization and monitoring may both require attention and thus require serial processing. For lower proficiency speakers, however, lexical retrieval and grammatical encoding might also require attention and thus produce varying degrees of breakdown in parallel processing (de Bot, 1992; de Jong, 2013; Kormos, 2006).

4.3 Cognitive Processes in SLA Equally central to cognitive theories of task-based L2 performance is the notion that learners must create new knowledge representations for the L2 grammar as well as develop the processing capabilities to produce and comprehend L2 speech in real time (e.g., Tomlin & Villa, 1994: 183). Bialystok (1994) provides a framework based on the cognitive processes of analysis and control which supports these two aspects of L2 development. According to Bialystok, however, the model provides an

4.3 Cognitive Processes in SLA

39

explanation for the acquisition of symbolic knowledge in general (e.g., number, music, maps, language) and posits the minimum number of operations necessary to account for the way in which mental representations develop across the life span with concomitant increases in language proficiency.

4.3.1 Analysis Bialystok (1994) argues that the cognitive process of analysis involves the progressive organization and structuring of mental representations. She postulates three levels of development in these representations: conceptual, formal and symbolic. They differ from one another in the way that information is stored, the explicitness of this information, and the connectivity of this information. The meanings which provide the knowledge base of ordinary thought are initially organized in terms of concepts that relate to specific instances of contextualized world knowledge. This world knowledge is loosely organized and highly implicit with little higher order connectivity. However, with experience in using language for progressively more explicit communication, these concepts gradually evolve into categories of knowledge types which are context independent, explicit, and exhibit higher order interconnectivity. Finally, these formally categories, through uses involving ongoing abstraction, eventually evolve into symbolic representations of content areas which consist of systems of categories, each with its own structure. These symbolic systems are capable of representing, for example, sentences as an abstract system of grammatical relations, places as an abstract system of typographical features on maps, music as an abstract system of tone and rhythm notations on staves, and the relationships between real-world objects in terms of abstract numerical formulae. Knowledge representation evolves through the ongoing cognitive process of analysis and results in change in one’s mental models. Bialystok (1994) compares oral and written language to illustrate how the cognitive process of analysis is related to language use. Language in the oral mode is generally limited in the range of functions that it can achieve compared to language in the written mode. Meanings are usually conceptually derived based on specific referents, and a range of contextual support is typically available to aid interpretation. In the written mode, however, meanings are formally derived based on properties of the text rather than the context. These differences in language use have important implications for acquisition. Instruction aimed at advanced development of L2 proficiency will require learners, whether speaking or writing, to formally derive the meanings they wish to express through the language code rather than relying on context and loosely organized conceptual representations as in oral language. Analysis is the cognitive process by which specific knowledge is restructured and developed into progressively more abstract representations. It will be argued in Chap. 5 that the demands of differentiating very similar referents provides one means of requiring learners to engage in analysis and formally derive the meanings that are required to arrive at task outcomes.

40

4 Tasks in L2 Syllabus Design

4.3.2 Control In contrast, Bialystok (1994) argues that the process of control provides the basis for the development of fluency in the ability access and use knowledge. Whereas analysis relates to representation, control is concerned with processing. Bialystok relates her discussion of this aspect of language development to a limited-capacity model of attention, arguing that control allows the learner to reduce the attention required to use knowledge in real time. Automating rule-based processing allows learners to selectively direct attention to where it is needed. Consequently, learners are able to activate the specific representations that are relevant for the task at hand as well as simultaneously activate other representations in order to solve new problems. Control thus allows learners to direct their performance more efficiently and increase their fluency. Practice in the sense of the repetition of tasks and versions of tasks at a given level of communicative demand will thus be an essential complement to the performance of tasks at progressively higher levels of communicative demand. Bialystok’s (1994) theory, based on the complementary processes of analysis and control, thus attempts to account for cognitive evolution as it pertains to knowledge representation. She argues that the processes of analysis and control themselves will remain the same across all levels of development regardless of the differences in knowledge representation at different stages. Furthermore, the outcomes of analysis and control processes are not independent of one another. Rather, the development resulting from one process supports the development resulting from the other. More analyzed representations provide the foundation for increased accessibility, and increased accessibility in turn provides the foundation for ongoing analysis. Finally, performance variation, whether synchronic or diachronic, can be at least partially explained by inappropriate representation or imperfectly developed control.

4.4 Cognitive Models of Task-Based L2 Instruction There are two closely-related psycholinguistic rationales for the use of tasks as a basis for organizing L2 instruction. The first is the Limited Attention Capacity Hypothesis (Skehan, 1996, 2009, 2016), and the second is the Cognition Hypothesis (Robinson, 2001, 2011). Both of these models provide a rationale for the notion of pushed output in instructed L2 acquisition. Both models also relate tasks to the cognitive processes of L2 production and acquisition via variation in the complexity, accuracy and fluency (CAF) of learners’ performance. The Limited Attention Capacity Hypothesis relates task to the cognitive processes of conceptualization, encoding and monitoring (Levelt, 1989), whereas the Cognition Hypothesis relates tasks to the cognitive processes of analysis and control (Bialystok, 1994). The essential argument that these models have in common is: (1) complexity, accuracy and fluency (CAF) are goals for L2 instruction; (2) factors in the design

4.4 Cognitive Models of Task-Based L2 Instruction

41

and implementation of communicative tasks channel attention predictably during L2 production; (3) some of these factors (individually or in combination) will ease the performance demands of tasks resulting in increased fluency of performance; (4) others will increase the performance demands of tasks requiring learners to push their L2 production with a concomitant increase in the syntactic complexity of their speech; and, (5) it is possible to use tasks to guide L2 production and maintain balanced development across a course of L2 instruction. However, the two models make different predictions regarding grammatical accuracy. The Limited Attention Capacity Hypothesis argues that attention to complexity and accuracy draw on the same pool of attention and thus will tend to trade-off with one another during performance with learners paying attention to one or the other, but rarely both. In contrast, the Cognition Hypothesis argues that attention to complexity and accuracy draw on different pools of attention and thus that learners can pay attention to both of these aspects of language simultaneously.

4.4.1 The Limited Attention Capacity Hypothesis Skehan (1996, 2016) argues that attention is essential for learning, and that as the capacity to pay attention to things while learning is limited, L2 learners will not be able to meet the demands of real-time communication and pay attention to novel L2 forms at the same time. He predicts a trade-off during task performance between the fluency and the accuracy of learners’ L2 production, on the one hand, and the complexity of their L2 production, on the other. According to Skehan, learners’ capacity for dual-mode processing must be developed and maintained throughout the instructional process if L2 learners are to progress to target norms of competence. In other words, learners must retain the ability to employ both rule-based and exemplar- based processing during performance. He argues that rule-based processing is characterized by complex production, a slower speech rate, reformulations, hesitations, and redundancies, whereas exemplar-based processing is characterized by fluent production of known forms and avoidance of new and partially mastered ones. Balanced L2 instruction might thus meet the needs of learners’ limited attention capacity by providing continuous cycles of such analysis and synthesis. The demands of task sequences might alternatively focus learners’ attention on complexity in experimenting with new forms (inter-language restructuring) and using known forms fluently and accurately (inter-language stabilization). Learners might thus retain the ability to “operate a dual-mode system in which well-organized exemplars are available to respond to real-time pressures, but a rule-based system can still be accessed when the need for precision or creativity arises” (Skehan, 1996, p. 49). In addition, Skehan (2009) argues, drawing on Levelt’s (1989) model of L1 speech production, that there is a ‘natural tension’ between L2 accuracy and complexity. L2 complexity, he claims, is related to the conceptualization stage of speech production. Task demands related to this stage determine the extent to which l earners must allocate attention to creating more or less complex pre-verbal messages which,

42

4 Tasks in L2 Syllabus Design

in turn require more or less complex language to express them. On the other hand, L2 accuracy and fluency are related to the formulation stage of Levelt’s model and are concerned with access to the lexicon and syntax required to realize pre-verbal messages. Task demands related to this stage determine the extent to which learners must allocate attention to accessing the language required by the pre-verbal message. In other words, proficient L2 speakers will have sufficiently developed and automatic access to lexicon and syntax so as to require very little attention and allow parallel processing during the conceptualization and formulation of messages, whereas less proficient speakers will lack such access and consequently need to allocate more attention to retrieving and formulating the language itself. Performance breakdown on L2 tasks, according to Skehan, is thus a consequence of attention not being available for formulator operations parallel to conceptualizer operations. In cases where complexity and accuracy have been shown to increase simultaneously, Skehan argues that task factors relating to distinct stages of L2 speech production (i.e., conceptualization and formulation) are operating additively to ease demands for attention during performance and to overcome the constraints of limited attention capacity. Finally, while not denying the existence of multiple attention pools which may support some types of dual task performance, Skehan (2011) claims that research in this area has been concerned with concurrent performance of distinct tasks (e.g., monitoring gauges while talking on a radio), and he does not see such a distinction applying to conceptualization and formulation processes during language production which will draw on the same pool of attention. In terms of the implications of the Limited Attention Capacity Hypothesis for L2 instruction, Skehan (2009) argues that task design factors relating to the conceptualization and formulation stages of speech production can be manipulated additively to either ease the overall attention demands of tasks, and thus promote fluent and accurate use of current language resources during performance, or to increase the attention demands of tasks, and thus promote more complex speech involving novel language resources. In this way, tasks might be used to organize instruction in a way that alternates performance over time in order to promote balanced, dual-mode processing and continued L2 development to the level of target norms so that one aspect of performance will not dominate at the expense of the others. Finally, Skehan (2016) argues that factors imposed on performance during the implementation of tasks in the classroom such as the provision of pre-task planning time and opportunities to practice tasks repeated across the syllabus will have more impact on learners’ L2 performance than factors in the design of tasks. An important asssumption of this theory, of course, is whether balanced performance on tasks in a classroom does, in fact, result in balanced acquisition or change in learners’ L2 competence over time.

4.4.2 The Cognition Hypothesis Robinson’s model (2011) posits a broader range of task design factors to account for L2 performance on tasks. Robinson argues that three types of factors and the interactions between them must be taken into account in order to predict the fluency,

4.4 Cognitive Models of Task-Based L2 Instruction

43

accuracy and complexity of learners’ language during task performance. These are: (1) cognitive factors related to the complexity of the tasks themselves (both those that direct and those that disperse cognitive resources), (2) interactive factors related to the implementation of tasks in the classroom (both those relating to the participation demands of tasks as formatted for classroom use and those relating to participant demands resulting from grouping decisions), and (3) learner factors related to individual differences between task participants (both those relating to differences in ability and those related to differences in affective response). According to Robinson, cognitive factors in task design will affect the complexity of tasks for all learners regardless of context or individual differences, whereas interactive and affective factors may further increase or decrease the difficulty of tasks depending of the specific context in which the tasks are performed and the specific learners who perform them. He thus advocates using cognitive factors as the sole basis for task sequencing at the syllabus level as these factors are arguably under the syllabus designer’s control and, all other factors being equal, can be expected to have predictable effects on the L2 performance of all learners regardless of their background or the context in which they perform the tasks. Cognitive factors in task design are thus the primary variables in Robinson’s model of L2 syllabus design, and he makes a distinction between two types of factors affecting the cognitive demands of tasks: (a) those that direct cognitive resources to specific features of task-essential content and the language needed to express it (e.g., the number of elements in the task, the temporal & spatial displacement of elements, the need to make the reasoning connecting elements explicit, and the need to make multiple perspectives on elements explicit), and (b) those that disperse these resources over several aspects of task-essential content, language and the larger performance context simultaneously (e.g., the number of steps in the task, the need to multi-task, or the lack of time to plan for the task). Robinson argues that task factors which direct learners’ attention to specific aspects of language will increase their effort and attention during the conceptualization of the messages that they are to communicate and that this will result in speech that is both more complex and more accurate, but also less fluent. Thus, according to Robinson, performance breakdown is not a result of limited attention and the inability to parallel process, as Skehan (2009) argues, but rather, following Neumann (1987) and Sanders (1998), a breakdown in ‘action-control’ and the amount of effort that learners must put into message conceptualization in response to the cognitive demands of the task. In Robinson’s terms, increasing task complexity as determined by cognitive factors in task design, “promotes greater effort at controlling production and more vigilant monitoring of output” as well as “increased noticing and improved uptake of task relevant input” (Robinson, Mackey, Gass, & Schmidt, 2011, p. 255). Robinson (2010) also proposes relatively explicit recommendations for the implementation of the Cognition Hypothesis in L2 syllabus design. He argues that balanced development of CAF can be maintained by increasing the cognitive demands of tasks in a three-stage cycle over a course of L2 instruction which he refers to as SSARC (Simplify-Stabilize, Automatize and Restructure-Complexify). The first of these stages consists of the simplification of tasks in terms of both

44

4 Tasks in L2 Syllabus Design

resource directing and resource dispersing demands in order to allow successful performance and stabilize learners’ current L2 resources. The second stage consists of automatizing these resources through practice on versions of the tasks which gradually increase in terms of resource dispersing task demands until these demands match the natural performance demands of un-simplified target tasks with respect to these factors. Finally, the third stage consists of increasing resource directing demands while again initially reducing resource dispersing demands in order to require learners to analyze and restructure the resources which they have automatized in the previous stage in line with the demands of these slightly more complex versions of the task. According to Robinson, this cycle will continue until the learner can complete full versions of the target tasks. The SSARC model of L2 task sequencing thus provides a second rationale for alternating task demands in order to promote balanced L2 development over a course of instruction so that fluency does not dominate at the expense of continued syntactic development or vice versa.

4.5 Summary This chapter has provided an overview of the theories on which current cognitive approaches to task-based L2 instruction are based together with a comparison of the similarities and differences between the two primary cognitive models of task-based L2 instruction and their implications for L2 instructional planning. A summary of theories on the role of attention in the registration and processing of language, speech production, and the cognitive processes of analysis and control in SLA provided the necessary background for understanding the two cognitive models of task- based performance. The next chapter will motivate and define referent similarity in relationship to similar variables that have been discussed in the TBLT literature. Chapter 6 will then define and motivate the dependent variables used to measure the effects of referent similarity on L2 production.

References Bialystok, E. (1994). Analysis and control in the development of second language proficiency. Studies in Second Language Acquisition, 16, 157–168. Bowles, M. (2010). The think aloud controversy in second language research. London: Routledge. De Bot, K. (1992). A bilingual production model: Levelt’s ‘speaking’ model adapted. Applied Linguistics, 13, 1–24. De Jong, N. (2013). Levelt’s model of speech production and comprehension. In P. Robinson (Ed.), The Routledge encyclopedia of second language acquisition (pp. 379–382). London: Routledge.

References

45

Erlam, R., Loewen, S., & Philp, J. (2009). The roles of output-based and input-based instruction in the acquisition of L2 implicit and explicit knowledge. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, & H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing and teaching (pp. 241–261). Bristol, UK: Multilingual Matters. Gass, S. (1988). Integrating research areas: A framework for second language studies. Applied Linguistics, 9, 198–217. Gass, S., & Mackey, A. (2000). Stimulated recall methodology in second language research. London: Routledge. Hama, M., & Leow, R. (2010). Learning without awareness revisited: Extending Williams (2005). Studies in Second Language Acquisition, 27, 269–304. Izumi, S. (2002). Output, input enhancement and the noticing hypothesis. Studies in Second Language Acquisition, 24, 541–577. Izumi, S. (2003). Comprehension and production processes in second language learning: In search of the psycholinguistic rationale of the output hypothesis. Applied Linguistics, 24, 168–196. Jourdenais, R. (2001). Cognition, instruction and protocol analysis. In P. Robinson (Ed.), Cognition and second language instruction (pp. 354–375). Cambridge, UK: Cambridge University Press. Kahneman, D. (1973). Attention and effort. Englewood-Cliffs, NJ: Prentice-Hall. Kormos, J. (2006). Speech production and second language acquisition. Mahwah, NJ: Routledge. Levelt, W. (1989). Speaking from intention to articulation. Cambridge, MA: MIT Press. Levelt, W. (1999). Producing the language: A blueprint of the speaker. In C. Brown & P. Hagoort (Eds.), The neurocognition of language (pp. 83–122). New York: Oxford Press. Neumann, O. (1987). Beyond capacity: A functional view of attention. In H. Heuer & A. Sanders (Eds.), Perspectives on perception and action (pp. 361–394). Berlin, Germany: Springer. Posner, M., & Petersen, S. (1990). The attention system of the human brain. Annual Review of Neuroscience, 13, 25–42. Posner, M., & Rothbart, M. (1992). Attentional mechanisms in conscious experience. In M. Milner & A. Rugg (Eds.), Foundations of neuropsychology series (pp. 91–112). New York: Academic. Robinson, P. (1995). Attention, memory and the ‘noticing’ hypothesis. Language Learning, 45, 283–331. Robinson, P. (2001). Task complexity, cognitive resources and second language syllabus design. In P. Robinson (Ed.), Cognition and second language instruction (pp. 287–318). Cambridge, UK: Cambridge University Press. Robinson, P. (2010). Situating and distributing cognition across task demands: The SSARC model of pedagogic task sequencing. In M. Putz & L. Sicola (Eds.), Cognitive processing in second language acquisition: Inside the learner’s mind (pp. 243–268). Amsterdam: John Benjamins. Robinson, P. (2011). Second language task complexity, the cognition hypothesis, language learning and performance. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Robinson, P., Mackey, A., Gass, S., & Schmidt, R. (2011). Attention and awareness in second language acquisition. In S. Gass & A. Mackey (Eds.), The Routledge handbook of second language acquisition (pp. 247–267). London: Routledge. Sanders, A. (1998). Elements of human performance. Mahwah, NJ: Lawrence Erlbaum. Schmidt, R. (1983). Interaction, acculturation and the acquisition of communicative competence: A case study of an adult. In N. Wolfson & E. Judd (Eds.), Sociolinguistics and language acquisition (pp. 137–174). Rowley, MA: Newbury House. Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 11, 129–158. Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3–32). Cambridge, UK: Cambridge University Press. Schmidt, R., & Frota, S. (1986). Developing basic conversational ability in a second language: A case study of an adult learner of Portuguese. In R. Day (Ed.), Talking to learn: Conversation in second language acquisition. Rowley, MA: Newbury House.

46

4 Tasks in L2 Syllabus Design

Shiffrin, W., & Schneider, R. (1977a). Controlled and automatic human information processing I: Detection, search, and attention. Psychology Review, 84, 1–66. Shiffrin, W., & Schneider, R. (1977b). Controlled and automatic human information processing II: Perceptual learning, automatic attending and a general theory. Psychology Review, 84, 121–190. Skehan, P. (1996). A framework for the implementation of task-based learning. Applied Linguistics, 17, 38–62. Skehan, P. (2009). Modeling second language performance: Integrating complexity, accuracy, fluency and lexis. Applied Linguistics, 30, 510–532. Skehan, P. (2011, September). Tasks, conditions and characteristics: Understanding the influences upon task performance. Plenary at JACET 50th commemorative international convention, Seinan University, Fukuoka, Japan. Skehan, P. (2014). Processing perspectives on task performance. Amsterdam: John Benjamins. Skehan, P. (2016). Tasks versus conditions: Two perspectives on task research and their implications for pedagogy. International Review of Applied Linguistics, 36, 34–49. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output on its development. In S. Gass & C. Madden (Eds.), Input in second language acquisition (pp. 235–253). Rowley, MA: Newbury House. Swain, M. (1995). Three functions of output in second language learning. In G. Cook & B. Seidlhofer (Eds.), Principle and practice in applied linguistics (pp. 125–144). Oxford, UK: Oxford University Press. Talmy, L. (2008). Aspects of attention in language. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 27–38). New York: Routledge. Tomlin, R., & Villa, V. (1994). Attention in cognitive science and second language acquisition. Studies in Second Language Acquisition, 16, 183–203. Toth, P. (2006). Processing instruction and a role for output in second language acquisition. Language Learning, 56, 319–385. Truscott, J., & Sharwood-Smith, M. (2011). Input, intake and consciousness: The quest for a theoretical foundation. Studies in Second Language Acquisition, 33, 497–528. Wickens, C. (2007). Attention to the second language. International Review of Applied Linguistics, 45, 177–191. Williams, J. (2005). Learning without awareness. Studies in Second Language Acquisition, 32, 465–491.

Chapter 5

Referent Similarity

This chapter defines referent similarity as factor in L2 task design. It begins by outlining Rosenberg & Cohen’s (1964) model of referent identification processes and how this model relates to early research on the relationship between the similarity of referents and variation in L1 English speech. More recent work on similarity and L1 speakers of French and Italian is then considered. As yet, there has been no research on referent similarity in TBLT although the need for such research has frequently been pointed out (e.g., Long, 2015; Yule, 1997). Referent similarity is then disambiguated from another construct discussed in the TBLT literature: the number of elements involved in completing a task (Robinson, 2011b). It is argued that these two variables represent separate underlying constructs and relate to separate processes. The chapter concludes with an operational definition of referent similarity as factor in L2 task design together with a model of its predicted effects on L2 speech production.

5.1 Referent Identification Rosenberg and Cohen (1964) provide a model of the process of referent identification which has played a central role in subsequent research on the role of referent similarity in L1 speech production. After establishing word association norms for a set of synonyms in an undergraduate population in the United States, Rosenberg and Cohen conducted a study in which an independent sample of subjects from the same population were given pairs of synonyms with one word marked as the target referent and asked to supply a one-word clue that would help a future listener identify the target word from the pair. This clue could not resemble the target referent in sound or spelling. Based on their results, they argue that in completing this task speakers must: (1) sample a response from their repertoire of word associations for the referent at a rate commensurate with the word’s position in the frequency distribution, (2) compare that response to each of the stimulus words in the synonym set (referent © Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_5

47

48

5 Referent Similarity

and non-referent), (3) decide whether the sampled word is suitable to uniquely identify the referent in the pair or whether that word must be rejected, and (4) if rejected, sample another word from lower on the frequency distribution, again at a rate based on the item’s position in the frequency distribution. In other words, the time and cognitive effort required to complete the task was directly related to rarity of the words within the participants’ population that the task required. Rosenberg and Cohen conclude that the four processes of sampling and comparison that they identify are repeated until a word is found that is judged by the speaker to adequately identify the target referent in the pair. The key processes in Rosenberg and Cohen’s model of referent identification are thus sampling and comparison.

5.1.1 Frequency Rosenberg and Cohen’s (1964) model implies that common words from the top of the frequency distribution for a given population will be adequate for unambiguously identifying targeted referents in pairs of dissimilar synonyms, but that these conventional words will fail to be adequate as the pairs of synonyms become more similar, and speakers will be forced to sample words from further and further down the frequency distribution. The result will be a preference for frequent, conventional terms in dissimilar sets of referents and rarer, unconventional terms in similar sets of referents. The implications of this model for L2 task design and pedagogy are readily apparent. If tasks are designed in which target referents are similar to the surrounding items, learners might be forced to push their L2 output either lexically or syntactically. In the first case, learners might attempt to locate a word that is more specific to differentiate the item. This might be done by referring to a dictionary, asking the teacher or a classmate for the L2 equivalent of an L1 word, or putting forth the time and effort to try and recall a word from memory. If a lexical item that adequately distinguishes the targeted referent proves to be unavailable, however, learners might be forced to push their L2 resources syntactically by using more general, abstract lexical items and then further specifying their meanings through the use of syntactic modifiers (e.g., adjectives, prepositional phrases, relative clauses). Krauss and Weinheimer (1967) investigated the effects of increased referent similarity on noun phrases complexity using tasks on which single lexical items to adequately distinguish similar referents were unlikely to be available. Based on a set of 24 graded color chips, they created sets in which each target color appeared once in a set of four similar colors (adjacent on the scale) and once in a set of four dissimilar colors (six steps apart from one another on the scale). The task required speakers to describe the four chips so that a listener might be able to order them correctly. They found that a sample of 30 secretaries at Bell Telephone Laboratories used significantly longer noun phrases (number of words minus initial articles) in identifying the target colors in similar sets than they did in identifying them in dissimilar sets. These findings are consistent with Rosenberg and Cohen’s (1964)

5.1 Referent Identification

49

model of the processes of referent identification. When the colors appeared in dissimilar sets, speakers would have first sampled a popular, conventional term such as a single word (e.g., blue, red) from the top of the frequency distribution for that color and found it sufficient to distinguish the target color unambiguously from the other three colors. However, in similar sets the speaker would be required to select longer, more qualified names which are lower in frequency (e.g., light red, a dark shade of green). Unfortunately, the authors do not provide specific examples of noun phrases that were used by the speakers in their study, nor do they provide information about the frequency of the lexis used in the target population. Hupet, Seron, and Chantraine (1991), however, in a study of the referential communication of adult L1 speakers of French, provide considerably more insight into the nature of referential similarity and the linguistic processes by which it is addressed. They investigate the effects of two aspects of referent similarity on the number of words and turns used to identify referents as well as the perspective taken on the referent (holistic or segmental) over multiple trials. Based on population norms for sets of Tangram figures,1 they created four sets of figures that isolated referent codability (i.e., the difficulty of naming, or expressing linguistically, the referent’s critical differences in relation to the non-referents) and referent discriminability (i.e., the difficulty of distinguishing the referent’s critical features in relation to those of the non-referents). The task required pairs of visually separated interlocutors to sequence one of four sets of eight Tangram figures six times in different orders. The sets isolated the factors of codability (C) and discriminability (D), in a repeated-measures design. The four task configurations were: (1) C+/D+, (2) C+/D-, (3) C−/D+, and (4) C−/D-. Findings revealed that the primary variable was codability: C+ conditions consistently resulted in fewer words and turns per referent than C- conditions. Discriminability had less of an effect on these measures: D+ conditions resulted in fewer words and turns than D- conditions when the C effect was controlled. In other words, the ordering of the tasks from easy to difficult was C+/D+, C+/D-, C−/D+, C−/D- with a much greater difference between C+/D+ and the other conditions. The analysis of the perspective taken on the referents in Hupet et al.’s study revealed an important distinction in the linguistic process of referent identification. The C+ conditions resulted in more holistic perspectives on the referents (i.e., the entire referent is identified with a word or phrase), whereas the Cconditions resulted in segmental perspectives on the referents (i.e. individual parts of the referent is described in turn). Furthermore, the D condition had a negligible effect on the perspective taken on the referents. However, the D- condition was found to result in the use of more comparative structures than the D+ conditions. Comparison is a key cognitive process in referent identification according to Rosenberg & Cohen’s (1964), and the results of Hupet et al. (1991) lend some support to the idea that there may be specific linguistic correlates to the processes they identify. 1 Collections of simple two-dimensional shapes that can be combined to create a broad range of figures in silhouette which can be interpreted to suggest various postures, activities, attitudes or positions.

50

5 Referent Similarity

An interesting additional finding by Hupet et al. (1991) was that the effects of both dimensions of similarity on learners’ speech largely dissipated on repetition of the task with the same interlocutor. In other words, a pair of participants would use longer, more elaborate segmental descriptions initially until a shared means of identifying a given referent had been established and then use that term or expression to identify the referent holistically on repeated reference. This indicates that the problem was finding a shared means of identification on a post hoc basis as Yule (1997) suggests. The findings of Hupet et al. on the effects of task repetition thus lend support to the idea mentioned above that speakers prefer lexical means of referent identification and resort to syntactic means only until the necessary distinction is lexicalized. In a pair of more recent studies of the effect of referential similarity on L1 speech, Barbieri and Iozzi (2007) and Iozzi and Barbieri (2009) further elaborate the specific strategies that Italian children use to overcome lexical deficiencies. In both studies, they found that easy-to-identify differences resulted in the use of more conventional labels, whereas difficult-to-identify referents resulted in the use of more non-conventional referential strategies (e.g., analogy, metonymy, functional description, literal-segmental description and super-ordinate noun use). These studies thus provide some insight into the range of linguistic devices that speakers use to overcome referential problems. Furthermore, they confirmed that these strategies are used more frequently when speakers lack a conventional term for a referent. Their findings thus provided additional cross-linguistic support for the primary role of codability in referential similarity postulated by Hupet et al. (1991), at least in initial references to an object, as well as for the general preference for holistic perspectives over segmental perspectives. However, the findings of Barbieri and Iozzi (2007) and Iozzi and Barbieri (2009) also provide support for the hypothesis that there will be a competition between lexical and syntactic means of making critical distinctions in referent identification. Based on their findings, it can be argued that lexical means of making critical distinctions are preferred, when available, but, when lacking, syntactic means come into play. In summary, cross-linguistic evidence on L1 production suggests that the linguistic variation involved in more or less difficult referent identification is likely to include: (1) the frequency of the lexical items used as established by norms for the specific population being studied and for the specific set of referents being studied (Barbieri & Iozzi, 2007; Iozzi & Barbieri, 2009; Rosenberg & Cohen, 1964), (2) the length, or complexity, of the noun phrases used (Hupet et al., 1991; Krauss and Weinheimer, 1967), (3) the use of explicit comparative structures (Hupet et al., 1991), and (4) differences in the decriptive strategies used (e.g., analogy, metonymy, functional description, literal-segmental description, and super-ordinate noun use) (Barbieri & Iozzi, 2007; Iozzi & Barbieri, 2009). Unfortunately, no research exists to date on the effects of referent similarity in L2 task on L2 speech production although the need for such research has been pointed out in the TBLT literature (Long, 2015; Yule, 1997).

5.1 Referent Identification

51

5.1.2 Comparison In addition to sampling language from lower in a given frequency distribution, Rosenberg and Cohen’s (1964) model of referent identification also posits comparison as a key process contributing to the difficulty of tasks. While no empirical studies seem to have been conducted which examine the effects of referent similarity on comparison per se, with the exception of the serendipitous findings reported by Hupet et al. (1991) discussed in the previous section, several studies have investigated the relationship between comparison and successful referential communication (Asher & Oden, 1976; Asher & Parke, 1975; Camaioni & Ercolani, 1988; Girbau & Boada, 2004; Whitehurst & Sonnenschein, 1978). Camaioni and Ercolani (1988), for example, in a study of 64 Italian children speaking their native language, investigated the relationship between independent measures of children’s ability to compare objects and their ability to perform effectively on referential communication tasks. First, in a pre-test consisting of two measures of comparison ability given to a sample of 150 children, it was found that 7-year-olds significantly outperformed 5-year-olds on both tests. The ability to compare thus seems to be a variable connected with a stage of development separating these two age groups. They then selected a sub-sample of 64 children in four groups based on their age and comparison ability (i.e., young-low, young-high, old-low, old-high) and asked them to complete two referential communication tasks, adopted from Krauss and Glucksberg (1969), in which a speaker was required to describe one referent so that a listener could select it from a context consisting of the target referent and five distracters. One of these tasks was based on familiar content (six dogs) and the other was based on unfamiliar content (six abstract figures). Speaker’s responses were divided into one of three categories: (1) contrastive when the message uniquely identified the target referent, (2) ambiguous when the message identified two or more items in the referential context, and (3) uninformative when the message identified all six items in the referential context. Correlation coefficients showed pretest scores of comparison ability to be significantly related to the number of contrastive messages produced by learners in both of the age groups. In subsequent analyses of variance (ANOVAs), both the age factor and the comparison ability factor had significant effects on the number of contrastive messages and the number of uninformative messages. The results were mixed, however, for the effect of these factors on the number of ambiguous messages. The authors speculate that this was due to experimentation by the children in a stage of developmental transition. On the basis of their results, Caimoni and Ercolani conclude that, “while all good communicators are able to compare adequately, poor communicators are just not able to make the necessary distinctions” (p. 412). As the correlations in Caimoni & Ercolani’s study reveal, the ability to effectively make comparisons in a language is highly predictive of successful referential task performance.

52

5 Referent Similarity

An important question raised by Camaioni and Ercolani (1988) relates to a typical phenomenon in their data: how is it that messages classified as ‘poor’ in the sense that they are either ambiguous or uninformative were often still adequate for a particular pair of children in a particular situation to identify the target referent? In other words, it is very likely that there is more involved in successful referential communication than the formulation of a contrastive response. Several variables connected with the interlocutor, for example, need to be taken into consideration in the design of referential communication studies. The first is the availability of concurrent feedback. The similarity effect observed in Krauss and Weinheimer’s (1967) study (Sect. 5.1.1 above), for example, was consistent across monologue and dialogue formats of their color identification task, but there was a significant interaction between referent similarity and the task format with the difference in phrase length being more pronounced in the dialogue format. As we shall see in the next section, Michel (2010) observed similar effects for L2 speakers of Dutch on monologue and dialogue versions of her task. The reason for this may be related to the availability of concurrent feedback. Such feedback on the adequacy of messages throughout a performance allows the speaker to elaborate or reformulate messages until they are effective for a given listener. Such signals may be as simple and inconspicuous as a non-verbal gesture of uncertainty. Deutsch and Pechmann (1982) demonstrate that even such minimalistic hints of problems in the efficacy of the message can often be enough to prompt an adequate reformulation. Similar effects are reported for L2 self-repair by Nobuyoshi and Ellis (1993). The interlocutor in a task thus has an important role in redirecting and restructuring the message. Other important interlocutor factors that have been shown to have significant effects on observed variance in language production are the speaker’s familiarity with the interlocutor (Fussell & Krauss, 1989), the interlocutor’s level of development (Deutsch & Pechmann, 1982, Whitehurst & Sonnenschein, 1978), and the amount of task-relevant background that the speaker assumes that the interlocutor to possess (Isaacs & Clark, 1987). Finally, it must be remembered that different interlocutors may have very different interpretations of the cost and rewards involved in choosing a wrong answer (Clark & Brennan, 1991; Gergle, Kraut, & Fussell, 2004; Thompson, 2009), and that these costs and rewards are likely to determine their tolerance for ambiguity and willingness to guess during a given speaker’s performance of a task. In any study of the effects of tasks on L2 production, it will thus be essential to control for both the interlocutor and the concurrent feedback available to speakers across task performances in order to avoid confounding this range of variables with the effects of the task factor being studied.

5.2 The Number of Elements The popularity of the Cognition Hypothesis (Robinson, 2011a, b) among TBLT researchers has produced a renewed interest in the number of elements that must be considered in completing tasks. Robinson argues that the number of elements that

5.2 The Number of Elements

53

need to be considered in arriving at a task outcome will direct learners’ attention to specific aspects of the language during performance and push the complexity and accuracy of their output (Chap. 4, Sect. 4.4.2). Some studies of the effects of this variable on L1 and L2 speech production have been D. Ellis (2009), Ford and Olson (1975), and Michel (2010). The classic study of the effect of the number of elements in a task on language use is Ford and Olson (1975). The authors investigated the combined effect of the number of attributes that must be distinguished and the number of alternatives from which these attributes must be distinguished on the length of the noun phrases produced by their participants. Forty L1 speakers of English in two groups (4.5-year- olds and 7-year-olds) completed fifteen referential communication tasks at five levels of difficulty. Three versions of the task were completed at each of the five levels. The task required the children to describe simple figures that varied in terms five binary attributes: shape (triangle/circle), size (big/little), brightness (black/ white), stripes (with/without), and dots (with/without). The materials set was created by randomly choosing fifteen shapes to be used as target referents for each of the fifteen versions of the task. Additional shapes were then selected and added to each target to create a referential context that required the appropriate number of attributes to be distinguished. In Level 1, shapes were selected that only differed from the target on one key feature. This level thus only required one attribute to be taken into account in order to identify the target. In Level 2, on the other hand, shapes were selected that required the speakers to take two critical features into account. This process was repeated until all five critical features, as summarized above, had to be taken into account in order to identify the target referent on the most difficult versions of the task. The number of alternatives or distractors also varied across levels from two to eight depending on the number of figures required to create the necessary contrasts. Analysis of variance (ANOVA) revealed significant effects for the age factor (4.5-year olds vs. 7-year-olds) on the length of noun phrases produced, but the number of elements in the task (combined attributes and alternatives) had no effect on the length of the noun phrases that these forty children produced. Length of noun phrase was thus found to reflect age-related development, but not the immediate demands of the tasks that learners were completing. The authors note that both age groups consistently provided more information than necessary to identify the target referent in the easier conditions. They conclude that children’s descriptions at both age levels reflected the cumulative or inferred alternatives in the task sequence as a whole rather than reflecting the immediate demands of each task. In interpreting the results of repeated-measures studies of task effects, it is thus important to remember that performance on a given task in isolation might differ from the performance the same tasks when part of an experimental task set. In a more recent study, D. Ellis (2009) takes the novel approach of investigating the effect of the number of elements in a task on the amount of verbal embedding in the speech of 36 university-level native speakers of English. As will be discussed in Chap. 6, research on referential communication has identified the noun phrase as the locus of referentially related language variation. TBLT researchers, by contrast, have tended to focus on variation in embedding when measuring the effects of tasks

54

5 Referent Similarity

on syntactic complexity in terms of amount of verbal embedding in utterances (Lambert & Kormos, 2014). D. Ellis (2009) employed a 2×3 repeated-measures design. The first factor was task at two levels (narrative and instruction) and the second factor was the number of elements at three levels. The narrative task, set in a country where people drive on the right side of the road, required the description of a car crash which involved a left-hand turn in oncoming traffic. The narrations were based on a sequence of four pictures with toy cars and pedestrians that illustrated what had taken place. Participants were asked to narrate the sequence of events and to give their opinion on who they thought was at fault and why they thought so. As the author’s purpose was to test the predictions of Robinson’s Cognition Hypothesis and Skehan’s Limited Attention Hypothesis (D. Ellis, 2009: 17), it is unclear whether he intended the latter opinion element in the task to be an additional step in the task (a resource dispersing variable, Chap. 4, Sect. 4.3.2), an additional reasoning dimension of the task (a resource directing variable, Chap. 4, Sect. 4.3.2), or a separate task altogether. However, this factor was a constant across all three levels of the narrative task allowing the compound event to be compared across conditions. The independent variable was operationalized by configuring the first level to involve only two cars, the second level to involve three cars and a pedestrian, and the third level to involve four cars and a pedestrian grouping. The instruction task, in contrast, required speakers to give directions based on an authentic map during a simulated telephone conversation. The first level involved a route on a single road, the second level involved a route on four roads, and the third level required the speaker to create an original route at the same time as they described it. It is thus unclear how many roads the third version of the instruction task actually involved. It is also unclear what the purpose of the multi-task dimension (a resource dispersing variable in Robinson’s model, Chap. 4, Sect. 4.3.2) could have been in the third level of the task. However, D. Ellis reports that three experienced teachers were able to order the three versions of each task type in terms of increasing complexity with 100% agreement. Analysis of variance (ANOVA) revealed no significant difference in the amount of verbal embedding in participants’ speech across the three levels on either the narrative or the instruction task separately or in combination. Very interestingly, however, Ellis found significant differences for task type: the narrative tasks resulted in more verbal embedding than the instructional tasks. This is consistent with Yule’s (1997, Chap. 3, Sect. 3.4) prediction that narrative tasks require inherently more use of verbal syntax than instructional tasks. Based on the author’s analysis and discussion of his data, however, it is not possible to gain any insight into the specific formal features and processes that may have distinguished the language used on the instruction and narrative tasks. The author simply concludes that “different task types clearly elicit different features” (D. Ellis, 2009: 23). Michel (2010) also studied the effect of the number of elements in a task on measures of the complexity, accuracy and fluency of the L2 production of forty-four university-age learners of Dutch. The opinion task used in the study required participants to match male-female couples based on personal profiles and explain the

5.3 The Similarity Between Elements

55

reasoning behind their decisions. Participants were presented with pictures of each person together with a six-point personal profile which included age, nationality, area of study, reading habits, learning background, and learning goals. Participants were then asked to choose two people from the available options who they believed would make the best couple and support their decision by explaining why their chosen couple was best and why the other possible combinations were not as good. The simple version consisted of four possible combinations, and the complex version consisted of nine possible combinations. Michel found no effects for the number of elements on the observed measures of the syntactic complexity (i.e., words per clause, clauses per speech unit), the grammatical accuracy, or the fluency of the participants’ speech. Nor did she find any differences between the two levels of the task on learners’ perceptions of task difficulty or on their confidence, frustration, stress or motivation while completing the respective versions of the task. While there was a significant difference in the overall number of words that the participants used to complete the two versions of the task, the author reports that qualitative analysis of the data revealed that these differences were a function of the number of terms and proper names in the task input. In short, previous research on the effects of the number of elements that must be considered in arriving at a task outcome has focused on various aspects of the language code (noun phrases, verbal embedding, CAF). This research has also considered the language use of both L1 and L2 speakers of English. However, no significant effects on the syntactic complexity or the accuracy of language used to identify referents have been documented in any of these studies. TBLT researchers might thus start to look for other ways of operationalizing referent identification in L2 task design.

5.3 The Similarity Between Elements Referent similarity might be confused with the number of elements in a task. Tversky (1977), for example, outlines a set-theoretic approach to measuring similarity in which referents are represented as collections of features. Essentially, concepts or objects will be similar if they have many features in common and few attributes by which to distinguish them. In this sense, the understanding of similarity is the result of a process of feature matching. The empirical support for this notion of similarity is based on human categorization research (Reed, 1972; Rosch & Mervis, 1975; Rosch, Mervis, Grey, Johnson, & Boyes-Braem, 1976) which has shown that fewer differentiating features result in faster similarity judgments. Although Tversky (1977) is concerned with the role of similarity in the difficulty of categorizing items rather in the difficulty of identifying them linguistically, if one were to apply his notion of similarity to the design of referential communication tasks, one would end up with a factor very similar to the number of elements in a task. A target referent would be similar to other referents if it had many features in common with them and few features unique to it. Conversely, a referent would be

56

5 Referent Similarity

dissimilar to other referents if it had few features in common with them and many features distinguishing it from them. What this quantitative approach to the elements in tasks does not account for, however, is the possibility that referents distinguishing one item from another may not be equal in terms of the linguistic demands that identifying them places on speakers. As seen in the empirical work on referent identification by Barbieri and Iozzi (2007), Hupet et al. (1991), and Iozzi and Barbieri (2009) summarized in Sect. 5.1 above, features can be easily identifiable in that there is conventional terminology by which they can be encoded (e.g., popular designs, materials, or patterns), or they can be difficult to identify in that no clear means of referring to a distinguishing feature is available to the speaker (e.g., colors, textures, positions, orientations or attitudes). Hupet et al. (1991), for example, were able to reliably operationalize the distinction between the codability of critical elements in their Tangram task and the discriminability of these elements. Furthermore, the authors demonstrated that although these two task design factors may not have been completely independent with respect to the dependent variables that they observed (i.e., the number of words and turns used), they did have differential effects on specific aspects of L2 production (i.e., the use of comparative structures). It is thus possible that some referents are inherently more difficulty to identify linguistically than others. The quantitative variables proposed by Tversky (1977) to account for difficulties in completing human categorization tasks, and by Robinson (2011b) to account for L2 task demands, do not take into account the possibility of such differences between task elements. As seen in Sect. 5.2 above, studies of the number of elements in tasks on language production (e.g., D. Ellis, 2009; Ford & Olson, 1975; Michel, 2010) have operationalized this variable as a collection of more or less equivalent features that must be taken into account in completing the task. In other words, these tasks could be expected to result in a linear sequence of feature matching similar to Tversky’s (1977) model of the process of formulating similarity judgments during categorization. As Michel’s (2010) study demonstrates, this approach in fact resulted in sub- task routines being repeated in response to the increased demands of the task rather than the syntactic restructuring of the language used to meet these demands. In this sense, it could be argued that what has commonly been researched as the number of elements in tasks has actually operationalized a variable closer to what Robinson (2011b) refers to as the number of steps in a task (Sect. 4.4.2). It will be remembered that this variable in Robinson’s model of task demands this is not a resource directing variable, but a resource dispersing variable. Robinson does not predict that such variables increase syntactic complexity, but rather that they increase fluent, automatic access to current L2 resources (c.f., Lambert, Kormos, & Minn, 2017). If an elements factor is to function as a resource directing variable in Robinson’s model, it will need to be defined and operationalized in a way that will push further analysis during the process of speech production (Sects. 4.2 and 4.3) rather than the repetition of recently encoded concepts. Additional support for the importance of qualitative differences in referent features is provided by recent research on the modeling of category intuitiveness.

5.4 The Present Study

57

Pothos, Perlman, Bailey, Kurtz, Edwards, Hines, et al. (2011) investigated the effect of four aspects of two-dimensional visual stimuli in free categorization tasks based on both concrete, real-world visuals (illustrations of a spider) and on abstract, context-neutral visuals (dot patterns). The four aspects of the materials they investigated were: (1) the number of element clusters, (2) the size of elements clusters, (3) the separation between element clusters, and (4) the tightness within element clusters. The effect of these four aspects of stimuli was observed on the frequency with which 160 participants categorized these objects into groups based on each of these dimensions. The results showed that the three significant factors, in order of their effect sizes, were: (1) the internal tightness of the clusters, (2) the proximity of the clusters, and (3) their size on the page. The number of clusters in the task was found to have no effect on category intuitiveness. These findings point to the conclusion that exclusively quantitative approaches to conceptualizing the elements in referential communication tasks may mask important aspect of both the cognitive and linguistic demands that task elements place on speakers. Based on their findings, Pothos et al. conclude that people implicitly and instinctively look for kinds of features based on the categories and linguistic labels available to them. This is a conceptual as well as a linguistic process that reflects the speakers’ total experience of the world rather than being limited to competence in their L1 or in additional languages. Malt, Sloman, Gennari, Shi, and Wang (1999) provide some insight into what this process may involve. They distinguish between two stages in object identification. The first is recognition that the object has properties in common with entities stored in memory, which results in the encoding of the item in an internal representation system, and the second is connecting the object with words. Based on their data, in which speakers from the three language groups (English, Chinese, and Spanish) described common containers, the authors conclude that although the speakers all showed substantially different patterns of naming for the container sets, they nevertheless perceived the similarities between them in much the same way. These findings draw attention to the fact that the availability of parsimonious items in the mental lexicon of a speaker can create considerable variation in the encoding of a concept (cf., Talmy, 2008; Chap. 4, Sect. 4.1.5 on competition for attentional selection).

5.4 The Present Study In the present study, referent similarity will refer to the inherent difficulty of identifying a critical feature distinguishing a referent from its context, regardless of the number of other elements that may co-occur in this context. Figure 5.1 provides a model of the effects of referent similarity on L1 and L2 production based on previous research. On tasks consisting of dissimilar referents, routine responses are likely to be sufficient to complete the task. These will generally consist of single nouns or very

58

5 Referent Similarity

Dissimilar Referents

Conventionalized Response

Similar Referents

Lexical Solution

Syntactic Solution

Concrete Lexis Simple NPs

Abstract Lexis Complex NPs

Fig. 5.1 Model of hypothesized effects of increased referent similarity on production

simple noun phrases. The result will be a conventionalized expression which is in line with the speaker’s current language resources as they relate to the functional demands of the task. On tasks consisting of similar referents, however, analyzed solutions are likely to be required. When speakers are able to access an appropriate term during this process of analysis (e.g., by referring to a dictionary, requesting help, or recalling a term), a lexical solution to the problem will be preferred, resulting in the use of specific, concrete lexical items and relatively simple syntax. When analysis reveals such a term to be lacking, however, a syntactic solution to the referential problem will be required. In this case, a number of strategies may be adopted (e.g., literal-segmental description, functional description, analogy, etc., see Sect. 5.1.1 above), but the result will be the use of more general, abstract lexical items and concomitant use of more complex noun phrase syntax. The operationalization of these constructs will be discussed in more detail in Chap. 6. For the present, an example will suffice. If a learner wants to predicate narrow to the perceptually salient and simple concept of lapels in describing a jacket, but lacks the term or an appropriate synonym, she might be forced to use a more abstract term (e.g., things). To compensate for the lack of the term lapels in narrow lapels, she may produce something along the lines of: some kind of long narrow triangular things on either side of the jacket near the center. The chapter that follows will investigate how tasks might be designed to push learners beyond routine, unanalyzed responses to describing common target referents and producing more advanced form-function mappings in identifying these referents.

5.5 Summary The present chapter has defined and operationalized the referent similarity as it relates to current theories and research on task-based L2 instruction. A model of the relationship between this factor and the demands that it can be expected to make on

References

59

language production has also been provided. The next chapter will further investigate how referent similarity might impact language production and define the dependent measures to be used in the study that follows in Chaps. 7 and 8.

References Asher, S., & Oden, S. (1976). Children’s failure to communicate: An assessment of comparison and egocentrism explanations. Developmental Psychology, 12, 132–139. Asher, S., & Parke, R. (1975). Influence of sampling and comparison processes on the development of communication effectiveness. Journal of Educational Psychology, 67, 64–75. Barbieri, M., & Iozzi, L. (2007). Production and comprehension of analogies in preschool children’s referential communication. Psychology of Language and Communication, 11, 3–21. Camaioni, L., & Ercolani, A. (1988). The role of comparison activity in the development of referential communication. International Journal of Behavioral Development, 11, 403–413. Clark, H., & Brennan, S. (1991). Grounding in communication. In L. Resnick, J. Levine, & S. Teasley (Eds.), Perspectives on socially shared cognition (pp. 127–149). New York: APA Books. Deutsch, W., & Pechmann, T. (1982). Social interaction and the development of definite descriptions. Cognition, 11, 159–184. Ellis, D. (2009). The relationship between task complexity and linguistic complexity: An analysis of L1 speaker production. Unpublished manuscript of a paper presented at TBLT2009, Lancaster University, UK. Ford, W., & Olson, D. (1975). The elaboration of the noun phrase in children’s descriptions of objects. Journal of Experimental Child Psychology, 19, 371–382. Fussell, S., & Krauss, R. (1989). Understanding friends and strangers: The effects of audience design on message comprehension. European Journal of Social Psychology, 19, 509–525. Gergle, D., Kraut, R., & Fussell, S. (2004). Language efficiency and visual technology: Minimizing collaboration effort with visual information. Journal of Language and Social Psychology, 23, 491–517. Girbau, D., & Boada, H. (2004). Accurate referential communication and its relation with private and social speech in a naturalistic context. The Spanish Journal of Psychology, 7, 81–92. Hupet, M., Seron, X., & Chantraine, Y. (1991). The effects of the codability and discriminability of the referents on the collaborative referring procedure. British Journal of Psychology, 82, 449–462. Iozzi, L., & Barbieri, M. (2009). Preschoolers’ use of analogies in referential communication. First Language, 29, 192–207. Isaacs, E., & Clark, H. (1987). References in conversation between experts and novices. Journal of Experimental Psychology: General, 116, 26–37. Krauss, R., & Glucksberg, S. (1969). The development of communication: Competence as a function of age. Child Development, 40, 255–266. Krauss, R., & Weinheimer, S. (1967). Effect of referent similarity and communication mode on verbal encoding. Journal of Verbal Learning and Verbal Behavior, 6, 359–363. Lambert, C., & Kormos, J. (2014). Complexity, accuracy and fluency in task-based research: Toward more developmentally-based measures of second language acquisition. Applied Linguistics, 35, 607–614. Lambert, C., Philp, J., & Nakamura, S. (2017). Learner-generated content and engagement in L2 task performance. Language Teaching Research, 21, 665–680. Long, M. (2015). Second language acquisition and task-based language teaching. Oxford, UK: Wiley Blackwell.

60

5 Referent Similarity

Malt, B., Sloman, S., Gennari, S., Shi, M., & Wang, Y. (1999). Knowing versus naming: Similarity and the linguistic categorization of artifacts. Journal of Memory and Language, 40, 230–262. Michel, M. (2010). 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 (pp. 141–174). Amsterdam: John Benjamins. Nobuyoshi, J., & Ellis, R. (1993). Focused communication tasks and second language acquisition. ELT Journal, 47, 203–210. Pothos, E., Perlman, A., Bailey, T., Kurtz, K., Edwards, D., Hines, P., et al. (2011). Measuring category intuitiveness in unconstrained categorization tasks. Cognition, 121, 28–100. Reed, S. (1972). Pattern recognition and categorization. Cognitive Psychology, 3, 382–407. Robinson, P. (2011a). Task-based language learning: A review of the issues. Language Learning, 61(Suppl. 1), 1–36. Robinson, P. (2011b). Second language task complexity, the cognition hypothesis, language learning and performance. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Rosch, E., & Mervis, C. (1975). Family resemblances: Studies in the internal structure of categories. Cognitive Psychology, 7, 573–603. Rosch, E., Mervis, C., Grey, W., Johnson, D., & Boyes-Braem, P. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382–439. Rosenberg, S., & Cohen, B. (1964). Speakers’ and listeners’ processes in a word-communication task. Science, 145, 1201–1203. Talmy, L. (2008). Aspects of attention in language. In P. Robinson & N. Ellis (Eds.), Handbook of cognitive linguistics and second language acquisition (pp. 27–38). New York: Routledge. Thompson, W. (2009). A game-theoretic model of grounding for referential communication tasks. Unpublished PhD dissertation, Northwestern University, Evanston, IL, USA. Tversky, A. (1977). Features of similarity. Psychological Review, 84, 327–352. Whitehurst, G., & Sonnenschein, S. (1978). The development of communication: Attribution variation leads to contrast failure. Journal of Experimental Child Psychology, 25, 490–504. Yule, G. (1997). Referential communication tasks. Mahwah, NJ: Lawrence Erlbaum.

Chapter 6

Referent Similarity and L2 Production

This chapter defines ten dependent measures argued to be associated with referent similarity based on developmental L1 research. These include a multi-faceted model of noun phrase complexity, explicit comparative structures, and later-emerging relative clause forms. These variables have been shown to relate to language use by speakers at different developmental levels and reflect the types of language variation expected in different conditions of referent similarity based on the theoretical and empirical research outlined in Chaps. 3, 4 and 5.

6.1 Noun Phrase Complexity Ravid and Berman (2010) define the noun phrase (NP) in terms of an obligatory head noun plus five optional modifiers: (1) determiners (articles, demonstratives, and quantifiers), (2) adjectives (single adjectives or adjective phrases), (3) nominal modifiers (compounded nouns, genitives), (4) prepositional phrases, and (5) relative clauses. They operationalize the NP developmentally in terms of five measures. Four of these relate to syntax and the fifth relates to lexical selection: (1) length of NP in number of words, (2) number of modifier tokens, (3) number of modifier types, (4) number of nouns subordinated to the head noun, and (5) the abstractness of the head noun. The authors point out that this approach to NP analysis is limited to the internal syntax of NPs. This avoids double-counting nouns in complex phrases, but ignores the internal complexity of the dependent phrases and clauses within NPs such as adjective phrases, prepositional phrases and relative clauses. The internal complexity of these dependent structures is reflected indirectly in the overall length of the NP in words and in the number of lexical nouns governed by the head noun of each phrase. However, supplementary analysis of the complexity of these specific structures is possible. In the study of referent similarity reported in the chapters which follow, for example, the internal structure of relative clauses is

© Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_6

61

62

6 Referent Similarity and L2 Production

analyzed independently of general NP complexity as developmentally-relevant variation in relative clause structures (see Sect. 6.3 below). Using their model of NP complexity, Ravid and Berman (2010) analyzed the L1 production of 96 participants (48 English speakers and 48 Hebrew speakers) in four age groups (9–10, middle childhood; 12–13, pre-adolescence; 16–17, adolescence; 20–30, adult) who completed four tasks based on the topic of ‘problems between people’. After receiving the same video stimulus of an unresolved problem between people, they were asked to: (1) tell a story based on one of their own experiences with a problem between people (narrative task, spoken mode); (2) write a story based on one of their own experiences with a problem between people (narrative task, written mode); (3) give a short speech on the topic of problems between people (expository task, spoken mode); and (4) write an essay on the topic of problems between people (expository task, written mode). Analysis of variance (ANOVA) showed significant effects for the age factor on all five NP criteria, and post hoc analyses revealed: (1) significant differences between the adult group and the three younger groups for both length and diversity of modifiers; (2) significant differences between the two youngest groups, the adolescent group, and the adult group for both number of modifiers and number of lexical nouns governed by the head; and finally (3) significant differences between all four of the age groups for abstractness of head nouns. In addition, mean length of clause (MLC) was analyzed as a baseline sentential unit of analysis for comparison. Although there were significant results for the age factor on MLC, these results were due to differences in the written mode only, and post hoc analysis revealed mixed results across groups. Thus, although MLC in words increased in tandem with internal NP complexity, the noun phrase was a more sensitive measure of syntactic development across modes, genres and language types. Ravid and Berman (2010) conclude that elaborated lexical NPs provide a valid measure of developing syntax and of text construction abilities across genres and modalities from the school years into early adulthood. Their study thus provides an empirical basis for measuring developmentally-advanced or later-emerging language production, and they argue for the validity of the noun phrase (NP) as platform for developing syntax across genres, modalities and language types. They claim that there are several advantages to this approach: As constructions, NPs are mid-level in size, smaller than a clause or sentence, but easily extendable beyond a single word. Moreover, they are involved in both clause-internal structure (as subject or as object of a verb or preposition) and inter-clause syntactic linkage (in use of relative clauses). As units of linguistic processing, NPs have clearly defined syntactic boundaries and typically fixed ordering of internal elements, not accessible to syntactic extraction. In consequence, they impose relatively little burden on short-term memory and hence on learning, as reflected in the fact that they are well formed from early on (Ravid & Berman, 2010, p. 4).

Furthermore, based on their empirical findings, the authors argue that NPs manifest a gradual increase in complexity over time. From the one-word, pragmatically- motivated constituents, NPs are progressively elaborated by means of more abstract and categorical heads as well as with a larger number of modifiers with various

6.2 Comparative Structures

63

types of dependencies within them. These developments in NPs take place in line with developments in lexical repertoire, syntactic proficiency and communicative competence. In particular, the authors stress that lexical semantics plays a key role in the development of NP constructions. Thus, their model of NP complexity is argued to be a late-developing area of syntax that is sensitive to more or less advanced language use. As L1 and L2 development have been argued to be comparable in terms of the essential developmental processes and learning mechanisms, but differing in L1 entrenchment and motivation (see Sect. 2.5), Ravid and Berman’s model of NP complexity seems to be potentially very well suited to measuring the effects of referent similarity on developmentally-related aspects of L2 production.

6.2 Comparative Structures As seen in Chap. 5, the importance of comparison is integral to theories of referent identification in L1 (Rosenberg & Cohen, 1964) and in L2 (Yule, 1997), and the ability to compare has also been shown to correlate highly with successful L1 referential communication ability on description tasks (Camaioni & Ercolani, 1988), and on tasks high in referent similarity demands (Hupet, Seron, & Chantraine, 1991). A similar role for comparison might be expected in L2 learners’ language use on tasks high in referent similarity demands. In order to verify whether this is the case, however, it is necessary to operationally define comparative structures. According to Stassen (1984), comparison can be defined as, “a mental act by which two objects are assigned to a position on a predicative scale” (p. 686). Stassen (2006) argues that this notion implies three components: (1) a scale (usually encoded as a gradable predicate), (2) an object of comparison (usually encoded as a noun phrase) that represents the thing that is compared, and (3) a standard of comparison (also usually encoded as noun phrase) that represents the thing to which the object of comparison is compared. There are two primary forms of comparison in English. The first establishes a comparison of equality (A is as X as B), and the second establishes a comparison of inequality (A is more/less X/X-er than B). In these structures, the scale of comparison is represented by ‘X’, the object NP is represented by ‘A’ and forms the subject of each sentence, and the standard NP is represented by ‘B’ and follows the comparative particle. Comparison in English is thus encoded through two linguistic devices. The first is a specific comparative particle (than, as) that accompanies the standard NP, and the second is either an affix on the comparative adjective (-er) or a special adverb (more/less) that, depending on the length in the number syllables of the comparative adjective, modifies this adjective. Thus, the items as, more, less (or -er) establish the ordering of the referents, whereas than and as mark the standard against which an object is compared. It may also be possible to quantify degrees of comparison. Sapir (1944) relates comparison to the broader concept of gradability. He discusses four levels of gradability in language. The first is what he refers to as implicitly gradable but ungraded. This refers to the range of things that can be compared (e.g. the sleeve). The second

64

6 Referent Similarity and L2 Production

Table 6.1 Three types of comparison in English Type 1 Type 2 Type 3

Implicit Explicit Quantified

One NP Two NPs Two NPs + adverb of degree

The right sleeve is wide The right sleeve is wider (than the left sleeve) The right sleeve is 20% wider (than the left sleeve)

is implicitly graded by quantification. This refers to a gradable concept that is used with an explicit quantifier (e.g., the sleeve that is two centimeters wide). The third is quantified by implicit grading (e.g., the wide sleeve). As will be discussed below, Kennedy (2005: 691–692) refers to expression exemplifying this degree of comparison as “vague predicates” because they depend highly on contextual factors (i.e., what is meant by ‘wide’ can vary greatly depending on the context in which the expression is used). Finally, the fourth degree of comparison identified by Sapir (1944) is explicitly graded and implicitly quantified (e.g., the right sleeve is wider than the left sleeve). Although Sapir’s typology stops here, its obvious logical extension would be a fifth degree, explicitly graded and explicitly quantified (e.g., the right sleeve is 20% wider than the left sleeve). For the purpose of practical oral discourse analysis, Sapir’s typology of comparative structures would allow three levels of explicitness in comparison to be distinguished. These could be referred to as implicit, explicit and quantified (see Table 6.1 for examples). Finally, at the functional level, Kennedy (2005, 2006) argues that the ability to establish orderings among objects according to some property that they possess is basic to human cognition. Kennedy further elaborates the semantic nature of grading and comparison, arguing that the former process is prerequisite to the latter and can be implicit or explicit. He argues that gradable predicates map their arguments onto abstract representations which constitute implicit norms or standards of comparison. The use of gradable predicates implicitly fixes the value of an argument to norms which vary depending on contextual factors. For example, the statement ‘the right sleeve is wide’ is vague in that its truth value is relative to a shared context on the part of the speaker and the listener. Comparative structures, on the other hand, function to explicitly fix the value of the argument expressed by the predicate in a particular relation to the comparative standard (e.g., the right sleeve is wider than the left sleeve). However, as noted above, it could also be argued that when the functional demands of a description task require even more precision on the part of the speaker, the degree of difference might be specified to reduce ambiguity even further (e.g., the right sleeve is 20% wider than the left sleeve). In making comparisons in which the precise degree of difference between referents is not essential, L2 learners might simply juxtapose independent clauses in what Givon (1985) refers to as the pragmatic mode and produce something along the lines of ‘the left sleeve is big, and the right sleeve is small.’ However, in comparing referents on a task in which the referential context creates a functional demand to make the degree of difference between the two referents explicit, the same learner might be pushed to produce discourse in what Givon refers to as the syntactic mode, and produce something along the lines of ‘the left sleeve is about

6.3 Relative Clause Structures

65

20% larger than the right sleeve’. In this way, factors in the design of tasks (e.g., similar items which all have a larger left sleeve than right sleeve) may result in increased exposure to explicit lexical items and syntactic devices, contributing to their representation in the developing network of implicit L2 form to function connections as argued in Chap. 2. This type of variation in the use of comparative structures during task performance is closely related to the discussion of effective L1 referential communication by Camaioni and Ercolani (1988, Sect. 5.1.2) as well as to the theory of the role of referential communication in SLA outlined by Yule (1997, Sect. 3.1). Furthermore, Hupet et al. (1991, Sect. 5.1.1) report that L1 speakers of French used more comparative structures in conditions high in referent similarity than they did in conditions low in referent similarity. Comparison may thus be a specific structure that is connected to the type of variation that is expected across conditions of differing referent similarity as well as being closely connected to referential communication ability. The use of comparative structures could also been seen as a form of noun phrase complexity as it involves specifying the relationship between two such structures on a single quality of relevance to them both. The frequency with which participants produced comparative structures of equality and inequality was thus included as a sixth dependent variable in the present study which represents a specific rather than a general measure of syntactic complexity.

6.3 Relative Clause Structures Relative clause structures have been shown empirically to represent a late-emerging aspect of English syntax. A pair of studies by Nippold and colleagues (Nippold, Hesketh, Duthie, & Mansfield, 2005; Nippold, Ward-Lonergan, & Fanning, 2005) which build on work by Loban (1976), Scott and Windsor (2000) and Berman and Verhoeven (2002), compared the L1 production of 180 and 120 native speakers of English, respectively, in three age groups: (1) 11 years, childhood; (2) 17 years, adolescence; and (3) 20–29 years, young adult. Comparing three types of clausal subordination (relative, adverbial, and nominal), Nippold and colleges found no effect for age on the overall amount of subordination in speech or in writing. In fact, the results confirmed a plateau in the use of subordination in early adolescence at around the age of 11 years (cf., Rubin, 1982; Rubin & Piche, 1979). However, post- hoc analyses revealed that the relative clause was the only type of subordination that evinced an age-related increase into early adulthood (20–29 years). Nominal subordination, on the other hand, particularly with early-emerging mental state verbs such as think, was the predominant form of subordination in discourse of the youngest learners (11 years). General measures of the overall amount of subordination in language production may thus mask important developmentally-related variation (Lambert & Kormos, 2014). Relative clauses can be understood in terms of their internal or their external syntax. External syntax refers to the function of the clause within the superordinate

66

6 Referent Similarity and L2 Production

Table 6.2 Developmental typology of relative clauses (Diessel & Tomasello, 2005) 1

S-Relative

Subject of intransitive verb

2

A-Relative

Subject of transitive verb

3

P-Relative

Direct object of verb

4

IO-Relative

Indirect object of verb

5

OBL-Relative

Object of a phrasal verb

6

GEN-Relative

Possessive

The boy who played The part which is on the left* The man who saw Peter The line which divides the two sides* The boy who the girl teased The area which the triangle encloses* The girl who Peter borrowed the ball from The edge I took the measurement from* The dog that the cat ran away from The place that the line is coming from* The part that the stiches are on* The woman whose cat caught a mouse The sleeve whose cuff touches the body*

*Examples of inanimate referents provided based on data from the present study

syntactic unit (e.g., the sentence) whereas the internal syntax refers to the syntactic function of the gapped item within the clause (see Table 6.2 for examples). In terms of external syntax, Diessel and Tomasello (2005), based on the findings of Diessel (2004), argue that the vast majority of relative clauses produced by children are attached to either an isolated head noun (e.g., The X which is Y), and frequently the predicate nominal of a copular clause (e.g., There is an X which is Y). Diessel and Tomasello argue that the early emergence of the latter form of relative clause results from the pragmatic usefulness of these structures for children who tend to refer to entities in the immediate context. Copular constructions involve predication to referents identified deictically. Furthermore, they are relatively simple in comparison to other relative clauses in that they contain only a single proposition. As can be seen in Table 6.2, other types of relative clauses typically contain two or more propositions. Diessel and Tomasello (2005) then investigate the difficulty of the internal syntax of six types of relative clauses for 5-year-old L1 English speakers based on error rates in comprehending and reproducing these clauses. These six relative clause types are summarized in Table 6.2 with examples of both animate referents and inanimate referents. The results of the study confirmed that relative clauses in copulative constructions resulted in significantly fewer errors than relative clauses directly following nouns. In addition, error rates increased based on the gapped element in the relative clauses. The most accurate were subject of intransitive verbs (S-Relatives), followed by subject of transitive verbs (A-Relatives) and then by direct object relatives (P-Relatives). There were no statistically significant differences between direct object relatives (P-Relatives), indirect object relatives (IO-Relatives) and oblique relatives (OBL-Relatives). Possessive relatives (GEN-Relatives) were excluded from analysis as there were too few instances of their correct production. Based on their findings, Diessel and Tomasello (2005) conclude that the difficulty of relative

6.4 Lexical Abstractness

67

clauses can be best explained by a multi-faceted construct which they refer to as ease of access and that the primary factors involved are frequency and similarity. While they do not deny that simplicity (number of referents) and distance (working memory demands) might affect the difficulty of relative clauses, they argue that the effect of these factors was “limited and must have been relatively small.” (p. 896). Finally, Diessel and Tomasello (2005) argue that the difficulty learners experience in comprehending and producing relative clauses relates directly to their level of language development. In other words, the relative clause types which cause more comprehension and production difficulties for learners are argued to be more advanced or later-emerging than those that cause fewer of these difficulties. The authors thus provide a heuristic for measuring developmentally more advanced use of relative clauses which Nippold, Hesketh, et al. (2005), Nippold, Ward-Lonergan, et al. (2005) show to be a late emerging aspect of English syntax. It is therefore worthwhile to consider the use of relative clause types produced by speakers in order to understand the effects of referent similarity on developmentally-related language variation. The six types of relative clause structures discussed by Diessel and Tomasello (2005, see Table 6.2) were thus considered in the study discussed in Chaps. 7 and 8.

6.4 Lexical Abstractness Another late-emerging aspect of language use that was mentioned in connection with Ravid and Berman’s (2010) model of NP complexity in Sect. 6.1. is lexical abstractness. In the broadest sense, ‘abstractness’ refers to an idea or a concept that is not associated with a specific instance of an event or an object. The opposite is ‘concreteness’. In linguistic studies which have sought to establish quantitative values for the concreteness of English nouns, for example, concreteness has been defined in terms of the degree to which it is possible to experience a given referent with the senses (Gilhooly & Logie, 1980; Paivio, Yuille, & Madigan, 1968; Spreen & Schultz, 1966). Similarly, in L1 developmental research, Nippold and colleagues have defined abstract entities as those that cannot be “easily seen, heard or touched” (Nippold, Ward-Lonergan, et al., 2005): 130). In computational linguistics research, on the other hand, McNamara and colleagues define abstract lexical items as having “few distinctive feature and few attributes that can be pictured in the mind” (McNamara, Louwerse, Cai, & Graesser, 2005: 8). For the purpose of the present discussion, abstract will be used to refer to the ease of experiencing an entity with the senses as this was the basis for establishing the metrics that will be discussed in the final paragraph of this section below. The lexicon that L1 speakers of English employ has been show to become more abstract until at least the age of thirty (Nippold, Hesketh, et al., 2005; Nippold, Ward-Lonergan, et al., 2005). Nippold and colleagues compared the L1 production of 180 and 120 native speakers of English, respectively, in three age groups:

68

6 Referent Similarity and L2 Production

Table 6.3 Early and late emerging terms (Nippold, Hesketh, et al., 2005; Nippold, Ward- Lonergan, et al., 2005) Verbs Early Say Think Know Feel Tell Ask

Late Predict Conclude Interpret

Adverbial conjunctions Early Late So Technically Also Personally Then Furthermore Well In conclusion Nevertheless For this reason On the other hand Consequently

First of all Finally Meanwhile Conversely Similarly Rather Next

(1) 11 years, childhood; (2) 17 years, adolescence; and (3) 20–29 years, young adult. They found significant differences in the abstractness of the lexis used in each of these age groups. In addition to using more abstract nouns (e.g., longevity, kindness, respect), L1 speakers of English used more meta-linguistic and cognitive verbs as well a more adverbial conjunctions with age. Table 6.3 compares verbs and adverbial conjunctions that Nippold and colleagues mention as being early versus later emerging exemplars of each of these lexical classes. These finding on the use abstract lexis among L1 speakers of English are also supported by an earlier set of findings by Nippold and colleagues (Nippold, Hegel, Sohlberg, & Schwarz, 1999). Comparing the performance of 240 L1 English speakers in four age groups (12 years, 15 years, 18 years, and 29 years), they found that the ability to define abstract entities is incomplete in late adolescence and improves markedly in early adulthood. In particular, they found that older learners tend: (1) to mention an appropriate super-ordinate category to which the word belongs, (2) to provide more unique features of the referent, and (3) to provide the positive and negative connotations of the word. Command of abstract lexical items and the ability to explain abstract features of referents may play a key role in the completion of the tasks involving high levels of referent similarity. The MRC Psycholinguistics Database (Coltheart, 1981) provides concreteness values for a range of common English lexical items derived from human ratings and developed and expanded over a number of years (see Gilhooly & Logie, 1980; Paivio et al., 1968; Toglia & Battig, 1978). These values are represented as integers between 100 and 700 in which higher numbers represent more concrete terms and lower numbers more abstract terms. For examples, some common concrete nouns in the present study were collar (622), buttons (613) and sleeve (587), whereas some common abstract nouns were top (435), side (394) and part (339). The MRC values were used to address the head noun abstractness component of Ravid and Berman’s (2010) model of NP complexity.

References

69

6.5 Summary This chapter has motivated and operationalized the dependent measures that will be used to estimate the effects of referent similarity on oral language production in the present study. Provided that L2 development can be argued to parallel L1 development, as was argued in Chap. 2, noun phrase complexity, the use of explicit comparative structures, and the use of later-emerging relative clause structures, might provide suitable means of measuring developmentally-relevant L2 variation related to referent similarity. These measures have the advantage of being relatively easy to identify reliably, and empirical support for their sensitivity to the developmental level of speakers has been established in cross-linguistic L1 research. The specific ways in which these variables were coded will be discussed in the Analysis section of Chap. 7.

References Berman, R., & Verhoeven, L. (2002). Cross-linguistic perspectives on the development of text- production abilities: Speech and writing. Written Language and Literacy, 5, 1–43. Camaioni, L., & Ercolani, A. (1988). The role of comparison activity in the development of referential communication. International Journal of Behavioral Development, 11, 403–413. Coltheart, M. (1981). The MRC psycholinguistics database. Quarterly Journal of Experimental Psychology, 33A, 497–505. Diessel, H. (2004). The acquisition of complex sentences. Cambridge, UK: Cambridge University Press. Diessel, H., & Tomasello, M. (2005). A new look at the acquisition of relative clauses. Language, 81, 882–906. Gilhooly, K., & Logie, R. (1980). Age-of-acquisition, imagery, concreteness, familiarity, and ambiguity measures for 1,944 words. Behavior Research Methods & Instrumentation, 12, 395–427. Givon, T. (1985). Function, structure and language acquisition. In D. Slobin (Ed.), The crosslinguistic study of language acquisition, Vol. 2: Theoretical issues (pp. 1005–1027). Hillside, NJ: Lawrence Erlbaum. Hupet, M., Seron, X., & Chantraine, Y. (1991). The effects of the codability and discriminability of the referents on the collaborative referring procedure. British Journal of Psychology, 82, 449–462. Kennedy, C. (2005). Semantics of comparatives. In K. Brown (Ed.), Encyclopedia of language and linguistics (Vol. 2, 2nd ed., pp. 690–694). Oxford, UK: Elsevier. Kennedy, C. (2006). Modes of comparison. Proceeding of CLS, 43. Lambert, C., & Kormos, J. (2014). Complexity, accuracy and fluency in task-based research: Toward more developmentally-based measures of second language acquisition. Applied Linguistics, 35, 607–614. Loban, W. (1976). Language development: Kindergarten through grade twelve. Urbana, IL: National Council of Teachers of English. McNamara, D., Louwerse, M., Cai, Z., & Graesser, A. (2005). Coh-Metrix version 2.0 indices. Retrieved December, 2011, from http://cohmetrix.memphis.edu Nippold, M., Hegel, S., Sohlberg, M., & Schwarz, I. (1999). Defining abstract entities: Development in pre-adolescents, and young adults. Journal of Speech, Language, and Hearing Research, 42, 473–481.

70

6 Referent Similarity and L2 Production

Nippold, M., Hesketh, L., Duthie, J., & Mansfield, T. (2005). Conversational versus expository discourse: A study of syntactic development in children, adolescents, and adults. Journal of Speech, Language, and Hearing Research, 48, 1048–1064. Nippold, M., Ward-Lonergan, J., & Fanning, J. (2005). Persuasive writing in children, adolescents, and adults: A study of syntactic, semantic and pragmatic development. Language, Speech and Hearing Services in Schools, 36, 125–138. Paivio, A., Yuille, J., & Madigan, S. (1968). Concreteness, imagery and meaningfulness values for 945 nouns. Journal of Experimental Psychology Monograph Supplement, 76, 1–25. Ravid, D., & Berman, R. (2010). Developing noun phrase complexity at school age: A text- embedded cross-linguistic analysis. First Language, 30, 3–26. Rosenberg, S., & Cohen, B. (1964). Speakers’ and listeners’ processes in a word-communication task. Science, 145, 1201–1203. Rubin, D. (1982). Adapting syntax in writing to varying audiences as a function of age and social cognitive ability. Journal of Child Language, 9, 497–510. Rubin, D., & Piche, G. (1979). Development in syntactic and strategic aspects of audience adaptation skills in written persuasive communication. Research in the Teaching of English, 13, 293–316. Sapir, E. (1944). Grading, a study in semantics. Philosophy of Science, 11, 93–116. Scott, C., & Windsor, J. (2000). General language performance measures in spoken and written narrative and expository discourse of school-age children with language learning disabilities. Journal of Speech, Language, and Hearing Research, 43, 324–339. Spreen, O., & Schultz, R. (1966). Parameters of abstraction, meaningfulness, and pronouncability for 329 nouns. Journal of Verbal Learning and Verbal Behavior, 5, 459–468. Stassen, L. (1984). The comparative compared. Journal of Semantics, 3, 143–182. Stassen, L. (2006). Comparative constructions. In K. Brown (Ed.), Encyclopedia of language and linguistics (Vol. 2, 2nd ed., pp. 686–690). Oxford, UK: Elsevier. Toglia, M., & Battig, W. (1978). Handbook of semantic word norms. New York: Erlbaum. Yule, G. (1997). Referential communication tasks. Mahwah, NJ: Lawrence Erlbaum.

Chapter 7

Methods

This chapter describes the design of an empirical study of the effects of referent similarity on the speech of L1 and L2 speakers of English. The chapter outlines the purpose of the study, the experimental design employed, background information concerning the participants, the tasks used to collect the data, and the data collection and analysis procedures employed. The aim throughout is to provide a sound basis for the replication of the study, in whole or in part, in future research.  The following chapter will discuss the findings for referent similarity on English use.

7.1 Purpose of the Study The study investigates the relationship between referent similarity and English nominalization processes in the oral descriptions of both native and non-native speakers at the advanced and intermediate levels. The purpose of the study is to determine the efficacy of referent similarity as a task design factor used to increase or decrease the functional demands for explicitness and precision that tasks place on speakers in line with their developing abilities to complete them. In other words, the study investigates how tasks might be designed to push learners’ language resources within a task-based approach to second language (L2) instruction. The speech of native speakers was included in the study to establish a baseline of the demands of the tasks independent of variation related to L2 proficiency (Long, 2015: 239). The study investigates both general and specific measures of nominalization (Robinson, 2011a, 2011b). The general measures relate to the five aspects of noun phrase (NP) complexity identified by Ravid and Berman (2010, Sect. 6.1), and the specific measures relate to the use of comparative structures (Sect. 6.2) and later-emerging relative clause structures (Sect. 6.3). Three research questions are addressed: 1. Is there a significant relationship between referent similarity and the use of complex noun phrases, comparative structures, and relative clauses? © Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_7

71

72

7 Methods

2. Is there a significant relationship between the specific referent being described and the use of complex noun phrases, comparative structures, and relative clauses? 3. Is there a significant relationship between language proficiency and the use of complex noun phrases, comparative structures, and relative clauses? Based on the theoretical and empirical support summarized in the previous chapter of this volume, the following three hypotheses were formulated with respect to these questions: 1. Referent similarity will be positively related to linguistic complexity during task performance. Identifying target referents which are more similar to non- target referents in tasks will result in the use of more complex noun phrases, more comparative structures, and more late-emerging relative clauses than identifying target referents which are less similar to non-target referents. 2. The referent being identified in a given task will impact lexical selection in completing tasks, but it will not significantly impact measures of syntactic complexity. Although speakers’ selection of more or less abstract lexical items may be influenced the demands of specific tasks, the demands for precision imposed by the level of referent similarity across tasks will determine the complexity of the syntax that they use in conjunction with this lexis. 3. Proficiency will be positively related to linguistic complexity. Higher proficiency speakers will use more complex noun phrases, more comparative structures, and more late-emerging relative clauses in completing the tasks than lower proficiency speakers.

7.2 Design The study employed a three-way full-factorial design. Similarity was a within- subjects factor at three levels (control, concrete and abstract). Referent was a second within-subject factor at three levels (jacket, boots and umbrella). Finally, proficiency was a between-subject factor at three levels (native, advanced non-native and intermediate non-native). Advanced non-native speakers had recent institutional TOEIC scores of 810–975 (BAND A), whereas intermediate non-native speakers had recent institutional TOEIC scores of 685–795 (BAND B) (Educational Testing Service, 2008). These bands are roughly equivalent to the C and B levels of the CEFR, respectively (Lambert, Kormos, & Minn, 2017). All participants thus completed nine versions of a description task. The same three referents were described at all three levels of referent similarity (see Appendix 1 for details). The nine versions of the task were given in 18 treatment orders for a balanced Latin Square design (Shuttleworth, 2009).

7.3 Participants

73

7.3 Participants 7.3.1 Native Speakers of English The native speakers of English who provided baseline data on the demands of the tasks used in the study were recorded individually over a four-month period. These speakers were from Great Britain, Australia, the United States, and New Zealand. They were all undergraduate exchange students studying for either a semester or a year at one of the three universities in Japan from which the non-native participants in the study were sampled. Three of the native speaking participants were female, and the remaining 15 were male. Table 7.1 summarizes the regional and educational background of the 18 native speakers in the study. This sample of native speakers represents a relatively broad range of national, regional and academic backgrounds. In terms of language skills, six of the participants reported having native or near-native proficiency in languages other than English. These languages included Cantonese, Mandarin, Vietnamese, Cambodian, French and Welsh. Furthermore, all 18 participants reported having studied languages in addition to Japanese. They were all studying Japanese intensively at the time that the data for the study was collected. These languages included French, Spanish and German.

7.3.2 Non-native Speakers of English All 36 of the non-native participants in the study were L1 speakers of Japanese. Nearly all Japanese who attend university have had six years of mandatory English in secondary school following the structural syllabus of the Japanese Ministry of Education, and several of them supplement this instruction with cram school classes aimed at improving scores for university entrance examinations and conversation school classes aimed at gaining oral proficiency in the language. The 36 Japanese learners in the present study were all Japanese nationals, primarily born and raised in the surrounding prefectures of the area of southern Japan in which the present study was conducted. However, three of the participants were originally from Kanagawa, Shizuoka, and Hiroshima in the north, and one had immigrated from China when she was young. They were all very successful English learners by local standards. The programs from which they were sampled are the primary English programs in the region, attracting the highest-achieving English learners from the region who do not enter universities in Tokyo or the Kansai region of Japan. Tables 7.2 and 7.3 summarize the background of the advanced-level and intermediate-level learners of English, respectively. In terms of foreign language learning experience, 22 of the participants had studied a language besides English for either one or two years as a requirement for their courses of studies. These languages included French, Spanish, Chinese, and Korean.

NZ

USA

Australia

Nationality UK

Region of origin Reading Reading Manchester Yorkshire London Wales Plymouth Brisbane Brisbane Brisbane Sydney Sydney Tasmania New York Washington Illinois North Carolina Auckland

Age 23 30 22 22 28 22 22 20 21 26 20 21 23 24 24 27 22 28

Gender M M M M F F M M M F M M M M M M M M

Year of study 3 3 3 3 3 3 3 3 3 4 2 2 3 2 4 4 4 na

Table 7.1 Background of native-speaking participants Academic major Japanese Japanese Anthropology Japanese Education Business Art History Japanese Intnl Relations Business Mathematics Law Cmpt Science History Interpretation Japanese Business na

Home university Oxford-Brookes Cardiff Oxford-Brookes Oxford-Brookes Oxford-Brookes Cardiff Oxford-Brookes Queensland Queensland Queensland U of Tech,Sydney U of Tech,Sydney U of Tasmania CUNY U of Kitakyushu U of Hawaii, Hilo UNC, Greensboro Japanese Center

Years of L2 Japanese 2.50 2.50 3.00 2.50 3.00 2.50 4.00 8.00 3.00 14.00 4.00 4.00 3.00 0.30 1.00 3.00 0.30 0.50

Other languages French; German None None French French; Spanish Welsh None Chinese French None Vietnamese Mandarin None Italian French; Spanish None Cambodian, Mandarin Cantonese, Mandarin

74 7 Methods

Home region Kagoshima Fukuoka Kanagawa Fukuoka Fukuoka Fukuoka Oita Fukuoka Kagoshima Shizuoka Nagasaki Kagoshima Fukuoka Fukuoka Hiroshima Fukuoka Fukuoka

Age 19 29 28 22 28 23 20 22 21 19 19 20 21 19 19 23 20

G F F F F M F F F F M M F F F M F F

Year of study 2 5 5 4 2 4 3 4 3 2 2 3 4 2 2 4 3

Academic major English English Lit Law Intnl Relations English Intnl Relations English English Lit English English English English Economics English English English English

Home university Kitakyushu Fukuoka Keio Kitakyushu Kitakyushu Kitakyushu Kitakyushu Seinan Kitakyushu Kitakyushu Kitakyushu Kitakyushu Seinan Kitakyushu Kitakyushu Seinan Kitakyushu

Table 7.2 Background of advanced non-native participants TOEIC score 975 975 950 940 885 880 875 875 860 850 835 830 830 820 815 815 810 Llama D 35 40 40 30 40 50 45 45 55 45 30 35 30 45 45 50 35

Llama F 40 80 80 40 90 70 40 50 60 80 60 80 70 10 80 50 80

Years abroad 5.00 2.00 8.00 0.75 0.50 0.75 5.00 0.75 1.00 0.10 3.00 0.75 1.50 0.05 None 0.80 0.10

Years of L2 English 8 17 16 10 8 10 9 10 9 8 8 9 10 8 8 10 9

Years of additional language study Vietnamese (1) Spanish (4) French (2) Chinese (2); Spanish (2) None Korean (2.5); French (0.5) Chinese (1) None Spanish (1) Spanish (1) None None Korean (0.5); French (1) None None Chinese (2) None

7.3 Participants 75

Home region Okayama Fukuoka Fukuoka Heilongjiang Kagoshima Oita Kitakyushu Kumamoto Yamaguchi Fukuoka Fukuoka Wakayama Kagawa Fukuoka Kitakyushu Fukuoka Oita Ehime Fukuoka

Age 18 23 20 25 19 19 22 20 20 22 21 21 20 19 18 22 19 21 21

Gender F F F F F F F M M F M F F F F M F F F

Year of study 1 4 3 5 2 2 4 2 2 3 4 3 2 2 1 5 2 3 3

Academic major English English English Comp. Cult. English English Business English English English Intrnl. law English English English English Comp. Lit. English English English

Home university Kitakyushu Seinan Seinan Kitakyushu Kitakyushu Kitakyushu Seinan Kitakyushu Kitakyushu Kitakyushu Seinan Seinan Kitakyushu Kitakyushu Kitakyushu Kitakyushu Kitakyushu Kitakyushu Kitakyushu

Table 7.3 Background of intermediate non-native participants TOEIC score 795 795 795 785 785 775 770 765 760 760 760 750 745 740 740 735 730 725 685 Llama D 35 35 55 25 35 45 50 20 25 35 35 55 55 10 35 35 30 40 35

Llama F 30 70 50 50 10 30 50 30 80 20 70 10 30 80 60 100 60 50 70

Years abroad 0.15 1.65 0.10 None 1.00 None 1.00 1.00 1.00 0.65 1.10 0.05 0.05 None None 0.50 None 0.60 1.00

Years of L2 English 7 10 9 6 8 8 10 8 8 9 10 9 8 8 7 11 8 9 9

Years of additional language study None Korean(1); Chinese (1) French (2) Chinese (native) None None Korean (1) Spanish (2) Spanish (1) Chinese (2) Chinese (1); Spanish (0.2) Chinese (2) None None None French (2) French (1); Korean (0.2) Chinese (1) Spanish (1)

76 7 Methods

7.3 Participants

77

Only two of the participants reported having experience in learning a language besides English beyond this basic requirement. One participant had studied Vietnamese for a year in Vietnam, and the other had completed a second major in Spanish while studying in the United States. The remaining 14 participants had never studied a foreign language except English. This is not to say, however, that the participants had not had international experience. Seven of the learners had lived abroad from one to five years in countries including Canada, Australia, Vietnam, Bulgaria, Costa Rica, the US and the UK, and 15 had studied abroad from between six months and two years in the US, Canada, Australia, New Zealand and the UK. Five others had travelled abroad or done homestays for periods of up to one month. In addition to experience abroad, five learners reported engaging in extra-curricular activities related to English learning including club activities, cram schools, and part-time jobs teaching English. In fact, only three of the 36 participants in the study reported that they had had no English learning experiences outside of the school curriculum. In addition to collecting the data on educational background, learning experiences and proficiency level, the non-native speakers who participated in the study were also given two computer-based tests of foreign language aptitude (see Tables 7.2 and 7.3). These tests were adopted from the suite of LLAMA language aptitude tests (Meara, 2005a, 2005b). They were selected because they were the most closely connected to the skills involved in completing referential communication tasks of the type used in the study. Furthermore, the items on these tests are not dependent on a particular language for their input and are thus well-suited for use with learners from non-English L1 backgrounds (Meara, 2005c). The first of these tests was the LLAMA_D test. According to Meara (2005c: 8), LLAMA_D is designed to test learners’ ability to recognize short stretches of spoken discourse in an unknown language. Based on work by Service (1992), Service and Kohonen (1995) and Speciale, Ellis, and Bywater (2004), Meara (2005c) argues that this is a key skill in language learning ability. The basic argument is that if one can recognize repeated patterns, one will be more likely to recognize words in a language when they are heard for a second time. Meara (2005c) argues that this skill is particularly important to vocabulary acquisition and affix recognition during the processing of oral input. It may also be important in repeating chunks of language verbatim during an oral monologue task such as the ones used in the present study. On the LLAMA-D test, test takers first hear ten words in an unknown language separated by clear pauses. The sequence is heard only once. It cannot be paused, nor can any notes be taken. The test takers then hear twenty words, one at a time, and must decide whether each occurred in the sequence or not. Test takers can play the twenty items at their own pace, but each item can only be heard once each. A choice (yes or no) must be made in each case before the next item can be played. Test takers receive immediate aural feedback on the accuracy of their choices in the form of a positive or a negative beep sound. To control for prior experience with the words heard, the program’s audio files are based on a dialect of a Native American language spoken in Northwest British Columbia (Meara, 2005c). However, the actual sounds in the program are phonetic realizations of the original words generated by

78

7 Methods

a speech engine, and are thus argued to be different from any language that test takers internationally are likely to know unless they have extensive familiarity with the languages of this region. As it relates to performance on the referential communication tasks used in the present study, the abilities measured by the LLAMA_D test can be argued to play an important role in syntactic priming. Speakers who are able to retain phonetic traces of previous utterances may be more likely to recycle them, in part or in whole, in subsequent speech over longer periods of time than speakers who cannot retain these traces as well. The ability to retain and incorporate previous words and phrase within subsequent structures could affect performance with respect to the syntactic measures used in the present study. For example, modifiers and other forms could be recycled more frequently resulting in longer noun phrases. If there were significant differences between groups in language aptitude as measured by the LLAMA_D test, this factor might interact with proficiency level and confound any findings regarding its effect on L2 production. The second test of language aptitude that was given to the non-native speakers was the LLAMA_F test. This is a test of the ability to make grammatical inferences. The LLAMA_F test uses simple pictures to illustrate the targeted grammatical relations and thus requires no L1 input (Meara, 2005c: 18). In order to support this format, the version of the test used focuses primarily on L2 agreement features. Test takers were given 300 seconds (5 minutes) to study the grammar of an unknown language. They do this by comparing 20 sentences in the unknown language which are paired with pictures illustrating the relationships each sentence expresses. They can move between the examples freely during this period reviewing or previewing each pair as they see fit. They are also allowed to take as many notes as they like during this period which they will be able to use during the subsequent test. A timer is available at the top of the screen so that they can adjust their strategies accordingly. Following this five-minute study period, 20 test items are presented. These consist of a picture illustrating a grammatical relationship and two sentences in the target language. The test taker must choose the sentence that correctly describes the picture. Again, they receive immediate aural feedback on the correctness of each of their responses in the form of a positive or a negative beep sound. In relation to the present study, the abilities measured by the LLAMA_F test can be argued to play an important role in inferring syntactic solutions to tasks high in referent similarity. In such tasks, suitable lexical items are often unavailable, even to native speakers of the language, and speakers are forced to restructure their syntactic resources during real-time communication in order to provide more precise and explicit descriptions of items than typically required for effectively identifying them. Speakers who are better able to infer such syntactic solutions in the target language may be inherently more effective communicators than those who lack this ability regardless of their proficiency level. If there were significant differences between the proficiency groups in aptitude as measured by the LLAMA_F test, this factor could be expected to confound the results obtained.

7.3 Participants

79

Table 7.4 Comparative experience of the participants Age

University study

L2 study

L3 study

Years abroad

Group Native Advanced Intermediate Native Advanced Intermediate Native Advanced Intermediate Native Advanced Intermediate Advanced Intermediate

N 18 17 19 18 17 19 18 17 19 18 17 19 17 19

Mean 23.61 21.88 20.53 2.94 3.18 2.79 3.39 9.82 8.53 1.44 1.21 1.28 1.77 .518

SD 2.97 3.37 1.78 0.80 1.07 1.18 3.18 2.65 1.23 3.03 1.38 1.41 2.24 .522

Min 20 19 18 1 2 1 0.30 8 6 0.00 0 0 0 0

Max 30 29 25 4 5 5 14.00 17 11 10.00 4 6 8 1.65

Range 10 10 7 3 3 4 13.70 9 5 10.00 4 6 8 1.65

Table 7.5 Proficiency and aptitude comparisons of the two non-native groups Proficiency Llama-D Llama-F

Group Advanced Intermediate Advanced Intermediate Advanced Intermediate

N 17 19 17 19 17 19

Mean 872 758 40.88 36.32 62.35 50.00

SD 56.32 28.55 7.55 12.00 21.37 25.17

Min 810 685 30 10 10 10

Max 975 795 55 55 90 100

Range 165 110 25 45 80 90

t

Sig.

7.525*

.000

0.187

.272

0.124

.575

*Significant at .05 level

Table 7.4 provides descriptive statistics comparing the background, educational level and experience of the three groups of participants in the study. The means in Table 7.4, if taken in comparison with the data in Tables 7.1, 7.2 and 7.3, demonstrate that the three groups were roughly comparable in terms of the background, educational level and learning experience. The two primary differences between the groups were in experience learning their current L2 and in the amount of time that they had spent in the target culture. Both groups of Japanese had spent a considerably longer time learning English than the native speakers of English speakers had spent learning Japanese. In addition, the advanced non-native speakers had spent considerably more time abroad on average than the intermediate non-native speakers. To confirm that the two groups of non-natives differed in L2 proficiency, but not in aptitude, their TOEIC, LLAMA-D, and LLAMA-F scores were compared using two-way t-tests for independent samples. The results of these tests are summarized in Table 7.5.

80

7 Methods

The t-tests reported in Table 7.5 reveal that the two groups did, in fact, differ significantly in L2 of proficiency, T (23.116) = 7.525, p < .05, and that they did not differ on either aspect of foreign language learning aptitude: LLAMA-D, T (34) = 0.187, p > .05; LLAMA-F, T (34) = 0.124, p > .05. The two groups of non-native speakers thus differed in terms of proficiency, but can be assumed to be approximately equal in terms of L2 aptitude. These differences in proficiency might be accounted for by differences in exposure among other possible factors. It will be remembered from the discussion of Table 7.4 that the advanced non-native speakers of English had spent more time abroad on average than the intermediate speakers.

7.4 Materials The nine oral description tasks used to elicit the data for the study consisted of picture-based information gaps. On each of the nine tasks, the speaker saw one drawing depicting the target referent, and the listener saw four drawings depicting the target referent and three distractors. The speaker was required to describe each of the target referents in enough detail for the listener to identify it from the four alternatives. However, to facilitate the process of elaboration and push the speakers’ linguistic resources, the listener was only allowed to prompt the speaker for more information using expressions such as ‘tell me more’ or ‘I have several like that’ (Yule, 1997). The objects that were described in the nine tasks described were jackets, pairs of boots, and umbrellas. Each of these referents were described three time at different levels of referent similarity. The complete task set and instructions are provided in Appendix 1. The tasks were formatted so that the speaker saw only the target referent and the interlocutor saw the target and the three distractors. We might imagine a claim at a lost and found office where a person who has lost an item must be able to convincingly describe it in order to be able to claim it while not being allowed to see the items that have been turned in. By not being allowed to see the referential context from which the target referent must be distinguished, speakers were required to work through progressively more precise formulations of their descriptions based on the full range of their previous experience together with their personal and linguistic resources. In other words, the tasks were formatted to tap into inferred referential context (Yule, 1997) and gain insight into how the demands of tasks relate to the broader range of resources that are brought to bear in completing the tasks (Schmid, Verspoor, & MacWhinney, 2011: 39; Sect. 1.1). In the first level of referent similarity (control), the three target referents were presented in the context of four alternatives which were distinct items. The differences between them would thus be lexicalized (jacket, boots, umbrella) for all but the lowest proficiency learners of a second language. As this level required only a bare head noun to successfully arrive at the task outcome, this condition was

7.4 Materials

81

intended as a control to see how much speaker elaborated descriptions regardless of the demands of the task (cf. Ford & Olson, 1975; Sect. 5.2). In the second level of referent similarity (concrete), the same three target referents were presented in the context of four alternatives which were the same items (jackets, pairs of boots, umbrellas). Each of these items differed in terms of conventional features (color, number, pattern or presence/absence of specific features). This level thus required some qualification beyond a bare head noun to successfully arrive at the task outcome. In the third level of referent similarity (abstract), the same three referents were again presented in the context of four alternatives which were the same items (jackets, pairs of boots, umbrellas), but this time the items were identical in terms of their conventional features, and differed only in terms of abstract details relating to the relative size, shape, position and orientation of their component parts, making them difficult to distinguish even for proficient speakers of the language. Conventional descriptions were inadequate to arrive at the task outcome in Level 3. It became necessary to experiment with more precise and explicit means of description (see Appendix 1 for illustrations; Yule, 1997). In terms of the level of precision required, the intention was that the first condition, in which referents were dissimilar, would require a level of precision which was roughly equivalent to descriptive demands that L2 learners might face in the situations in which they typically use the oral language in context-rich interactions (e.g., Please bring my jacket when you come; Where did you buy your boots?; You can borrow this umbrella). The second condition, in which referents were similar in conventional ways, was intended to require a level of precision which was roughly equivalent to the more difficult demands that L2 leaners might face only intermittently in everyday oral communication (e.g., Could you show me the jacket with the little stitches on the cuffs; I wore the heelless boots that have furry stuff around the top; Mine is the small white umbrella that has the three black stripes on it). Finally, the third condition, in which referents differed only in abstract, unconventional ways, was intended to push speakers beyond the level of precision and elaboration that they were accustom to providing in describing the target items and provide them with exposure to communicative demands they would only be faced with rarely when the need for more explicit communication arises (e.g., The irregular cloud-shaped material on the top part of the left boot is about half the size of the entire boot, and the height of the boot from the very top, which probably comes up to about the lower calf area on the wearer, to the bottom of her heel, is about ten percent greater on the left boot than it is on the right boot). Such demands provide L2 learners with needed exposure to a broader range of linguistic resources within the task-based syllabus to give them some headroom in completing tasks of a given type flexibly and naturally as proficient speakers complete them. It will be remembered from Chap. 2 (Sect. 2.2.1) that unlike L1 speakers for whom task frequency is not a problem, L2 learners typically suffer from a general slowdown in the development of cue strength on less frequent tasks as a lack of task frequency (MacWhinney, 2001: 74).

82

7 Methods

7.5 Procedures A series of posters and fliers were circulated throughout departments and classes at three universities in the area of southern Japan in which the study was conducted. While the posters varied slightly, they all offered payment for participation in a study focusing on the English language use of native speakers and Japanese students. Requirements for participation were current enrollment in university, and, in the case of non-native speakers, a recent TOEIC test score of 730 or higher. A classified advertisement was also run on a local English-language website specifying the same information and requirements. Qualified respondents were accepted in the order in which they applied. Each participant completed the task set individually in a single session. The participant completed the task set in the role of the speaker with the researcher in the role of the listener. The purpose was to control for the range of interlocutor variables such as the concurrent feedback provided and the listener’s tolerance for ambiguity in selecting target referents. The researcher intentionally avoided visual contact with the speakers during the tasks so that they were required to explicitly grammaticalize relevant distinctions rather than relying on non-verbal and deictic communication (e.g., drawing shapes with fingers). While listening to the descriptions, the researcher did not provide any corrective feedback or input on what to describe or how to describe it. When participants’ descriptions lacked enough factual information to positively eliminate the three distractors and identify the target referent, the researcher responded with, ‘tell me more’ or ‘there are several like that’ in order to prompt each participant to describe the item in more detail based on inferred context and their own linguistic resources. When a speaker’s description was adequate for unambiguously selecting one of the four items, the researcher responded with simple expressions such as, “okay” or “I’ve got it.” This was a means of providing controlled concurrent feedback to the speakers on the adequacy of their messages while limiting their performances to monologues for ease of analysis. The researcher continued this process until all 54 participants positively identified the target referent for all nine tasks (486 task performances total). The criterion used for task success was that, in each case, a third person who listened to the recordings should be able to eliminate all three distractors and positively identify the target regardless of familiarity with the tasks or the procedures. Before the task performances began, both the native speakers and the non-native speakers completed a simple background questionnaire (see Appendix 1). After the tasks were completed, the non-native speakers completed the two computer-based tests of foreign language learning aptitude (see Sect. 7.3.2 above). These tests required 30 minutes. The time allocated to each data collection session was thus 60 minutes for native speakers and 90 min for non-native speakers.

7.6 Analysis

83

7.6 Analysis The database for the study consisted of 486 oral task performances (nine task performances by 54 speakers), the background questionnaire for all participants, and TOEIC, LLAMA_D and LLAMA_F scores for the non-native speakers.

7.6.1 Transcription Pruned transcriptions (Ellis & Barkhuizen, 2005) were made of the 486 task performances (approximately 63,000 words). In other words, filled pauses, verbatim repetitions, hesitations, false-starts or self-repairs were not included, and the transcripts consisted of the final or repaired version of each sentential utterance produced. All of the performances were initially transcribed by native speakers of English from the United Kingdom and then individually verified in conjunction with the recordings and adjusted for consistency. The first type of adjustment made to the transcripts related to the identification of conjoined clauses. A pervasive phenomenon in the non-native data was the conjunction ‘and’ used in sentence-initial or sentence-final position. In the vast majority of cases, clauses of this type were transcribed as separate sentences unless the subject of the second clause was elided. In cases where two independent clauses were connected by a conjunction (and, but, or, etc.) and were clearly under the same intonational contour, they were transcribed as a single sentence and separated by a comma rather than a full stop. The final versions of the transcripts were carefully edited for consistency in this respect. The second adjustment related to topic-comment structures in Japanese speakers of English as a Second Language (e.g., The left collar, the top part is slightly longer than the right). These structures are more common in Japanese than in English and result in considerable L1 transfer. Transcribers were asked to set topics off with a comma as in the English expression speaking of (e.g., In terms of the sleeves, the right one curves out slightly). However, they frequently misinterpreted the topical element as a false start, and failed to transcribe it. The final transcripts were carefully edited for consistency in this respect.

7.6.2 Noun Phrase Coding The database contained 9,649 noun phrases headed by lexical nouns. The list of noun phrases for each performance was pasted into an Excel spreadsheet. The spreadsheet automatically counted the words in each cell. The number of determiners, adjective phrases, nominal modifiers, prepositional phrases and relative clauses were then counted and entered into separate adjacent cells. Subordinated lexical

84

7 Methods

nouns were marked with an asterisk in the original cell. The details of how these elements were identified are provided in the noun phrase coding manual in Appendix 2. Excel formulas were then used to tally the number of words, modifier types, modifier tokens and subordinated lexical head nouns in each noun phrase. The head nouns of each phase as well as any relative clauses were then typed or pasted into separate adjacent cells. The data were subsequently sorted by head noun and MRC concreteness values (Coltheart, 1981; Sect. 6.4) were pasted in for the head nouns for which they were available. MRC concreteness values were available for 87.6% of the lexical head nouns used by the participants on the task set used in the study. Based on the procedures outlined in the noun phrase coding manual, a research assistant was trained and inter-rater agreement was established on each of the noun phrase variables. After initial training, 5% of the database was independently coded by both the researcher and the research assistant and initial agreement was found to be 97% or higher on each of the seven variables (noun phrases identification, determiners, adjectives, nominals, prepositional phrases, relative clauses, and s ubordinate nouns). After resolving differences to 100% agreement on each variable, another 20% of the database was independently coded and inter-rater agreement was once again found to be 97% or higher agreement on each of the seven measures. This was deemed acceptable, and the research assistant coded the remainder of the noun phrases in the database.

7.6.3 Comparative Structures Only well-formed comparative structures of inequality (e.g., X is bigger than Y) and equality (e.g., Y is as big as X) were coded. This approach ignored two additional forms of comparison which were relatively common in the data. The first was implicit comparison based on contextualized graded predicates (e.g., X is big, see Sapir, 1944; Kennedy, 2006, Sect. 6.2), and the second was comparison through simple juxtaposition of independent clauses with a coordinated conjunction (e.g., Y is very big, but X is only a little big). However, these forms did not include explicit grammatical comparison. It was thus decided that the analysis should be limited to the use of well-formed comparative structures expressing equality and inequality. The structures that were prominent in the data but not analyzed are summarized in Table 7.6. Eliminating the types of semantic rather than grammatical comparison summarized in Table 7.6 resulted in nearly perfect inter-rater agreement by the researcher and a research assistant coding the full database independently. Differences between the two raters were due to structures containing grammatical errors and were easily resolved to reach 100% agreement. Finally, in order to control for text length, the number of explicit comparative structures produced in each task performance was divided by the total number sentential units produced in each performance.

85

7.6 Analysis Table 7.6 Structures excluded from the analysis of comparisons 1 2 3 4

5 6

Different to Same as Equal to Same (as a copulative predicate) In comparison compared with in correlation to -est more Half, twice, three times (unless modifying explicit comparisons)

Table 7.7 Relative clause types and token frequencies in the database

Type A

X is different to Y X is the same as Y X is equal to Y X and Y are the same X is big in comparison X is big in comparison with Y X is big in correlation to Y X is the biggest X is more extended X occupies half of Y X is three times of Y

Subject gapped relatives with intransitive verbs Type B Subject gapped relatives with transitive verbs Type C Direct object gapped relatives Type D Indirect object gapped relatives Type E Oblique object gapped relatives with phrasal verbs Type F Genitive gapped relatives

286 68 99 28 44 3

7.6.4 Relative Clause Types Of the 9,649 noun phrases produced by the 54 speakers in performing the nine versions of the task, 528 (5.47%) contained one or more relative clauses. These relative clauses were coded based on their internal syntax into six types following Diessel and Tomasello (2005). Table 7.7 provides a summary of the raw frequency of each of the six types of relative clauses in the database. The relative frequencies of the different relative clause structures in the database were highly comparable with those reported by Diessel and Tomasello (2005). As the last three types of relative clauses in Table 7.7 occurred only intermittently across the database, and the vast majority of performances had zero values on these variables, these three types were combined for analysis. While this still resulted in relatively few cases in the final category, the combined value was comparable to values for other relative clause types. Inter-rater agreement for the identification of relative clauses was established in conjunction with the coding of noun phrase complexity in Sect. 7.6.2 above. As the identification of the types of relative clauses based on the gapped element in their internal syntax is relatively objective, the coding was done by the research assistant after an initial meeting in which the examples provided by Diessel and Tomasello (2005: 902–903) were discussed and applied to a range of examples of relative

86

7 Methods

clauses in the database for the present in the study. The researcher subsequently verified the coding for all 528 of these clauses for accuracy. Finally, in order to control for differences in text length, the total number of relative clauses produced in each performance was divided by the total number of noun phrases produced in the performance.

7.6.5 Statistical Analyses The main analysis consisted of a three-way MANOVA using SPSS19 for Windows to compare main effects for task condition (similarity), referent (item described), and group (proficiency level) on the five aspect of noun phrase complexity in the model proposed by Ravid and Berman (2010). As will be discussed in Chap. 8, separate statistical analysis had to be used for the use of comparative structures and the use of relative clauses, respectively, as the distributions of scores on these variables were not normal.

7.7 Summary This chapter has outlined the purpose and design of the study used to test the effects of referent similarity in task design on the oral discourse of L1 and L2 speakers of English. It has also provided the details of how the data that will be discussed in the next chapter was transcribed and coded for noun phrase complexity, comparative structures, and relative clauses. Chapter 8 discusses the results of the statistical analyses used to test the hypotheses summarized at the beginning of the present chapter (Sect. 7.1).

References Coltheart, M. (1981). The MRC psycholinguistics database. Quarterly Journal of Experimental Psychology, 33A, 497–505. Diessel, H., & Tomasello, M. (2005). A new look at the acquisition of relative clauses. Language, 81, 882–906. Educational Testing Service. (2008). TOEIC test data and analysis 2007: Number of examinees and scores in FY2007. Tokyo: The Institute for International Business Corporation, TOEIC Steering Committee. Ellis, R., & Barkhuizen, G. (2005). Analyzing learner language. Oxford, UK: Oxford University Press. Ford, W., & Olson, D. (1975). The elaboration of the noun phrase in children’s descriptions of objects. Journal of Experimental Child Psychology, 19, 371–382. Kennedy, C. (2006). Modes of comparison. Proceeding of CLS, 43. Lambert, C., Kormos, J., & Minn, D. (2017). Task repetition and second language speech processing. Studies in Second Language Acquisition, 39, 167–196.

References

87

Long, M. (2015). Second language acquisition and task-based language teaching. Oxford, UK: Wiley Blackwell. MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and second language instruction (pp. 69–90). Cambridge, UK: Cambridge University Press. Meara, P. (2005a). Llama_D: A test of phonetic memory. Retrieved March, 2012, from http://www. lognostics.co.uk/tools/llama/ Meara, P. (2005b). Llama_F: A test of grammatical inferencing. Retrieved March, 2012, from http://www.lognostics.co.uk/tools/llama/ Meara, P. (2005c). LLAMA language aptitude tests: The manual. Retrieved March, 2012, from http://www.lognostics.co.uk/tools/llama/ Ravid, D., & Berman, R. (2010). Developing noun phrase complexity at school age: A text- embedded cross-linguistic analysis. First Language, 30, 3–26. Robinson, P. (2011a). Task-based language learning: A review of the issues. Language Learning, 61(Suppl. 1), 1–36. Robinson, P. (2011b). Second language task complexity, the cognition hypothesis, language learning and performance. In P. Robinson (Ed.), Second language task complexity: Researching the cognition hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins. Sapir, E. (1944). Grading, a study in semantics. Philosophy of Science, 11, 93–116. Schmid, M., Verspoor, M., & MacWhinney, B. (2011). Coding and extracting data. In M. Verspoor, K. de Bot, & W. Lowie (Eds.), A dynamic approach to second language development: Methods and techniques (pp. 39–54). Amsterdam: John Benjamins. Service, E. (1992). Phonology, working memory & foreign language learning. Quarterly Journal of Experimental Psychology, 45a, 21–50. Service, E., & Kahonen, V. (1995). Is the relationship between phonological memory and foreign language learning accounted for by vocabulary acquisition? Applied PsychoLinguistics, 16, 155–172. Shuttleworth, M. (2009). Counterbalanced measures design. Retrieved November, 2011, from Experiment Resources: http://www.experiment-resources.com/cpimterbalanced-measuresdesign.html Speciale, G., Ellis, N., & Bywater, T. (2004). Phonological sequence learning and short-term store capacity determine second language vocabulary acquisition. Applied PsychoLinguistics, 25, 293–321. Yule, G. (1997). Referential communication tasks. Mahwah, NJ: Lawrence Erlbaum.

Chapter 8

Results

This Chapter begins with a summary of the tests used to screen the data on each of the dependent variables to determine which statistical analyses were appropriate. The screening procedures consisted of confirming the normality of score distributions, testing for outliers, and confirming the homogeneity of the variance on each of the variables. It was found that the five variables connected with the model of NP complexity were suitable for a full factorial analysis using a MANOVA model. Comparative structures had to be collapsed across one factor to normalize the distribution and variance and then analyzed using a two-way factorial ANOVA model. Relative clause use was analyzed using non-parametric tests of the effects of each factor separately and subsequently corrected for the number of comparisons. In other words, the research hypotheses (Sect. 7.1) had to be tested in three parts as they related to each set of dependent measures.

8.1 Preliminary Data Screening Three methods were used to determine the normality of the score distributions on each of the dependent variables in the study: (1) Shapiro-Wilk tests of normality (p > .05) (Razali & Wah, 2011, 2; Shapiro & Wilk, 1965) individual z-score tests of skew and kurtosis through the comparison of their respective values to their standard error (p > .05) (Cramer, 1998; Cramer & Howitt, 2004; Duane & Seward, 2011), and finally visual inspection of histograms, normal Q-Q plots, and box plots for each of the variables. The methods used to determine homogeneity of variance, on the other hand, were a series of Levene’s F tests (Field, 2009, Martin & Bridgman, 2012) (p > .05) together with simple comparisons of within-group differences in standard deviation (Howell, 2009) to confirm univariate homogeneity of variance, and Box’s M test (p > .005, Huberty & Petoskey, 2000), which compares variance- covariance matrices, to determine multivariate homogeneity of variance. Although the value of Box’s M test is debated due to its sensitivity to sample size and the © Springer Nature Singapore Pte Ltd. 2019 C. Lambert, Referent Similarity and Nominal Syntax in Task-Based Language Teaching, https://doi.org/10.1007/978-981-13-3089-6_8

89

90

8 Results

robustness of the Pillai’s trace statistic provided by MANOVA (Field, 2009: 604– 605), it was used as a supplement to the series of Levene’s F tests. The preliminary screening of the data revealed that three sets of variables in the database were not suitable for analysis with the MANOVA model. The first was the participants’ performance on the Level 1 version of the task. Preliminary screening revealed that participants’ score distributions on these three versions of the task varied significantly from normal as measured Shapiro Wilk test (p  .05). This was probably due to the fact that these structures did not occur on several versions of the tasks at all and were relatively infrequent when they did. In addition, many of these distributions showed significant differences in variance between groups as measured by a series of Levene’s F tests (p > .05). When the variables were collapsed across referents, however, the score distributions proved to be normal and the variance on the distributions connected with each variable did not differ significantly. Univariate analysis of variance (ANOVA) could thus be used to test the research hypotheses concerning the participants’ use of comparative structures in relation to similarity and proficiency level. This analysis was then supplemented by non-parametric tests to determine if the referent factor had statistically significant effects on the use of comparative structures. Finally, the relative clause data, even when collapsed for the six relative clause types, was not normally distributed. Based on the results of Shapiro Wilk tests of normality (p > .05) and visual inspection of histograms, normal Q-Q plots, and box plots, it was unlikely that the data in many of these distributions could be corrected through transformation or combination across factors. Participants’ production of relative clauses thus had to be analyzed for each of the independent factors separately with non-parametric statistical tests. As the data for only the Level 2 and Level 3 versions of the task were compared, the effects of the task condition factor were analyzed using a series of Wilcoxon Signed Rank tests for paired samples. The between-groups differences in proficiency level, however, were analyzed using a

8.1 Preliminary Data Screening

91

Kruskal-Wallis test for independent samples followed by post hoc Mann-Whitney tests. Finally, the referent effect was analyzed using a Friedman’s non-parametric ANOVA test for dependent samples followed by post hoc Wilcoxon Signed Rank tests. These non-parametric tests allowed the research hypotheses concerning relative clause use to be tested in spite of the non-normal distribution of the scores. The data for the five noun phrase complexity variables in the model, however, proved to be approximately normally distributed without significant differences in variances between the three groups. However, an outlier analysis following the outlier labeling rule (Hoaglin, Iglewicz, & Tukey, 1986; Tukey, 1977) with a g-value of 2.2 (Hoaglin & Iglewicz, 1987) still revealed three scores which greatly affected the normality of their respective distributions. In all three cases, the original data were re-examined and simple coding errors were detected. Following these corrections, Shapiro-Wilk tests (Razali & Wah, 2011; Shapiro & Wilk, 1965) showed no significant effects for any of the distributions (p

Referent Similarity and Nominal Syntax in Task-Based Language Teaching

Recommend Stories

Idea Transcript

Helpful Links

Smile Life

Get in touch