Crosslinguistic Influence and Distinctive Patterns of Language Learning: Findings and Insights from a Learner Corpus 9781783098774

Citation preview

Crosslinguistic Influence and Distinctive Patterns of Language Learning

SECOND LANGUAGE ACQUISITION Series Editors: Professor David Singleton, University of Pannonia, Hungary and Fellow Emeritus, Trinity College, Dublin, Ireland and Dr Simone E. Pfenninger, University of Salzburg, Austria This series brings together titles dealing with a variety of aspects of language acquisition and processing in situations where a language or languages other than the native language is involved. Second language is thus interpreted in its broadest possible sense. The volumes included in the series all offer in their different ways, on the one hand, exposition and discussion of empirical findings and, on the other, some degree of theoretical reflection. In this latter connection, no particular theoretical stance is privileged in the series; nor is any relevant perspective – sociolinguistic, psycholinguistic, neurolinguistic, etc. – deemed out of place. The intended readership of the series includes final-year undergraduates working on second language acquisition projects, postgraduate students involved in second language acquisition research, and researchers, teachers and policy-makers in general whose interests include a second language acquisition component. Full details of all the books in this series and of all our other publications can be found on http://www.multilingual-matters.com, or by writing to Multilingual Matters, St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK.

SECOND LANGUAGE ACQUISITION: 118

Crosslinguistic Influence and Distinctive Patterns of Language Learning Findings and Insights from a Learner Corpus

Edited by Anne Golden, Scott Jarvis and Kari Tenfjord

MULTILINGUAL MATTERS Bristol • Blue Ridge Summit

DOI 10.21832/GOLDEN8767 Library of Congress Cataloging in Publication Data Names: Golden, Anne, editor. | Jarvis, Scott, 1966- editor. | Tenfjord, Kari, editor. Title: Crosslinguistic Influence and Distinctive Patterns of Language Learning: Findings and Insights from a Learner Corpus/Edited by Anne Golden, Scott Jarvis and Kari Tenfjord. Description: Bristol, UK; Blue Ridge Summit, PA: Multilingual Matters, [2017] | Series: Second Language Acquisition: 118 | Includes bibliographical references and index. Identifiers: LCCN 2017014947| ISBN 9781783098767 (hardcover : acid-free paper) | ISBN 9781783098774 (pdf) | ISBN 9781783098781 (epub) | ISBN 9781783098798 (kindle) Subjects: LCSH: Norwegian language—Study and teaching—Foreign speakers. | Norwegian language—Grammar, Generative. | Interlanguage (Language learning)—Norway. | Multilingualism—Norway. | Second language acquisition—Norway. Classification: LCC PD2611 .C76 2017 | DDC 439.8/280071—dc23 LC record available at https://lccn.loc.gov/2017014947 British Library Cataloguing in Publication Data A catalogue entry for this book is available from the British Library. ISBN-13: 978-1-78309-876-7 (hbk) Multilingual Matters UK: St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK. USA: NBN, Blue Ridge Summit, PA, USA. Website: www.multilingual-matters.com Twitter: Multi_Ling_Mat Facebook: https://www.facebook.com/multilingualmatters Blog: www.channelviewpublications.wordpress.com Copyright © 2017 Anne Golden, Scott Jarvis, Kari Tenfjord and the authors of individual chapters. All rights reserved. No part of this work may be reproduced in any form or by any means without permission in writing from the publisher. The policy of Multilingual Matters/Channel View Publications is to use papers that are natural, renewable and recyclable products, made from wood grown in sustainable forests. In the manufacturing process of our books, and to further support our policy, preference is given to printers that have FSC and PEFC Chain of Custody certification. The FSC and/or PEFC logos will appear on those books where full certification has been granted to the printer concerned. Typeset by Nova Techset Private Limited, Bengaluru and Chennai, India. Printed and bound in the UK by the CPI Books Group Ltd. Printed and bound in the US by Edwards Brothers Malloy, Inc.

Contents

Contributors Preface

vii xi

1

Introduction Kari Tenfjord, Scott Jarvis and Anne Golden

1

2

Transfer: An Overview with an Expanded Scope Scott Jarvis

3

The ‘Perfect Candidate’ for Transfer: A Discussion of L1 Influence in L2 Acquisition of Tense-Aspect Morphology Ann-Kristin Helland Gujord

4

On How Polish Learners of Norwegian Render Spatial Prepositions in L2: A Corpus-Based Study of i and på 64 Oliwia Szyman´ska

5

Positive and Negative Transfer in the L2 Adjective Inflection of English-, German- and Polish-speaking Learners of L2 Norwegian Marta Olga Janik

12

29

84

6

Gender Assignment and L1 Transfer in Norwegian Second Language Learners’ Written Performance Silje Ragnhildstveit

110

7

Stranded or Lost? Preposition Stranding in Norwegian Learner Languages Kari Tenfjord and Torodd Kinn

155

v

vi

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

8

Emotions Negotiated in L2 Texts: A Corpus Study of Written Production by Adult Learners on a Norwegian Test Anne Golden

9

Evaluation of Texts in Tests, or: Where is the Dog Buried? Anne Golden, Lars Anders Kulbrandstad and Kari Tenfjord

231

Author Index

272

Subject Index

277

188

Contributors

Anne Golden is a professor of Norwegian as a Second Language and research leader for Theme 1, multilingual competence, at the Center of Multilingualism in Society across the Lifespan (MultiLing) at the University of Oslo, Norway. Her main research is on literacy in a second language, with a focus on vocabulary and in particular metaphors, studying learners’ comprehension, development and practice in and out of school. Transfer studies have also been a recurrent part, using corpus methods and the Norwegian learner corpus ASK. Identity negotiations among migrants are also part of her research using narratives and metaphors as research tools. Ann-Kristin Helland Gujord is an associate professor in second language acquisition in the Department of Linguistic, Literary and Aesthetic Studies at the University of Bergen. Her main research interests fall within the area of grammatical development in adult and child second language acquisition. She has conducted corpus based studies on the effects of crosslinguistic influence on the acquisition of tense-aspect morphology and finite verbal morphology, and on the morphosyntactic realization of the subject in bilingual writing. She has also studied the language use of bilingual children in nursery school. Marta Olga Janik is affiliated with SWPS University of Social Sciences and Humanities in Warsaw, Poland, where she is a lecturer. She obtained her PhD in Norwegian Linguistics from Adam Mickiewicz University in Poznan ´, Poland. Her main area of research is the acquisition of Norwegian as a second language by native speakers of Polish. In her work, Janik examines the written language of learners of Norwegian, and she investigates the influence of diverse factors (linguistic, psycholinguistic and social) that affect Polish learners’ performance in Norwegian, with special attention given to crosslinguistic influence. Scott Jarvis is a professor in the Department of Linguistics at the University of Utah. His research concentrations include crosslinguistic influence and lexical diversity. In the domain of crosslinguistic influence, he has developed frameworks for exploring conceptual transfer and confirming crosslinguistic vii

viii

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

influence in cases where it is difficult to detect or is confounded by competing variables. In the domain of lexical diversity, he has concentrated on developing an adequate, multidimensional construct definition of this phenomenon and an accompanying set of measures that account for how lexical diversity is perceived and how it contributes to the quality of both spoken and written language. Torodd Kinn is a professor of Scandinavian linguistics in the Department of Linguistic, Literary and Aesthetic Studies at the University of Bergen. His wide-ranging research has three foci: grammar, academic writing and onomastics. Grammatical topics include nominal structure, grammar and metaphor, prepositions, present participles, asymmetric ‘V and V’ coordination, and grammaticalization of adverbial, adjectival, and numeral suffixes as well as of quantifiers. His research on academic prose includes work on authorial presence and the author–writer relation, including pronoun use. In onomastics, he has developed innovative methods for the historical study of givenname geography, and he has drawn on genealogical studies to analyze the spread of infrequent given names. Lars Anders Kulbrandstad is Professor of Norwegian in the Faculty of Teacher Education and Natural Sciences at Inland Norway University of Applied Sciences, and Adjunct Professor of Norwegian as a Second Language in the Faculty of Humanities at the University of Oslo. His research activities are mainly in the areas of pedagogical grammar and lexicography, Norwegian as a second language, sociolinguistic perspectives on multilingualism and language attitudes. His recent research has explored the factors that influence raters of Norwegian as a second language, and the extent to which Norway’s tolerance for language variation extends to foreign-accented Norwegian and minority languages. Silje Ragnhildstveit is an associate professor in the Department of Teacher Education at Western Norway University of Applied Sciences. Her area of research is Norwegian as a second language, and more specifically crosslinguistic influence and the acquisition of grammatical gender. Recently, in February 2017, she defended her PhD thesis titled ‘Gender and transfer when Norwegian is a second language: Three corpus-based studies’. Oliwia Szymańska is an assistant professor in the Department of Scandinavian Studies at SWPS Warsaw University of Social Sciences and Humanities. Her research concentrates on foreign and second language acquisition and cross linguistic influence. She also deals with the types of lexical modality typical of Scandinavian languages and investigates the way it is rendered in Polish and English. She wrote her doctoral thesis on conceptual transfer affecting the proper acquisition of prepositional phrases in Norwegian.

Contr ibutors

ix

Kari Tenfjord is a professor of Norwegian as a second language in the Department of Linguistic, Literary and Aesthetic Studies at the University of Bergen. Her research focuses on the development and use of Norwegian as a second language with an emphasis on the acquisition of tense. She is the primary architect behind the Norwegian learner corpus, ASK, which has been designed to facilitate research on crosslinguistic influence in learner language. Her research thus highlights the use of corpus tools in the investigation of transfer and grammaticalization. She is also one of the pioneers in the development of online training materials for teachers of Norwegian as a second language.

Preface

The empirical studies in this book are an outcome of a recently completed, multi-year project titled ASKeladden – A Corpus-Based Approach to L1 transfer in Norwegian Learner Language. The aim of the ASKeladden project was not only to investigate crosslinguistic influence, but also to gain insight into the strengths and weaknesses of the ASK corpus, which we describe in Chapter 1. The ASKeladden project was funded by the Norwegian Research Council1 and the University of Bergen. We also received minor financial support for re-assessing the texts in the ASK in accordance with the Common European Framework of Reference for Languages (CEFR) writing scales. This latter funding was granted by VOX, the Norwegian Agency for Lifelong Learning, which is part of the Norwegian Ministry of Education and Research. The contributors to this book were all members of the ASKeladden project: Kari Tenfjord and Anne Golden were the project leaders, Scott Jarvis was one of the international members of the project network, and Lars Anders Kulbradstad and Torodd Kinn were two of the Norwegian members of the network. Marta Janik and Silje Ragnhildstveit were members of the project team both as Master’s students and later as PhD candidates. Ann-Kristin Helland Gujord and Oliwia Szyman´ska joined the project team as PhD candidates. All four of the former student members have now completed their PhD degrees, and we congratulate them on their fine work in this exciting area of research. Kari Tenfjord, Bergen, Norway Anne Golden, Oslo, Norway Scott Jarvis, Salt Lake City, Utah

Note (1) Project number 185964.

xi

1

Introduction Kari Tenfjord University of Bergen

Scott Jarvis University of Utah

Anne Golden University of Oslo

Background and Orientation As can quickly be seen through a perusal of studies published both historically and even recently in the major journals of applied linguistics, a good deal of what we know about second language acquisition and of the learning of later languages comes from investigations where the target language is English and the learners’ background languages are European. The findings of such studies should of course not be disregarded, but there is a great deal more to be discovered about language acquisition than can be gleaned from patterns produced by European learners of English – or even by English-speaking foreign-language classroom learners of French, German or Spanish, who are also over-represented in the literature. In other words, there is a genuine need for more studies on second language acquisition involving (a) learners of languages other than English and (b) learners who come from non-European backgrounds, especially in cases where (c) the learners are immigrants to the host country and thus experience the language in naturalistic contexts over an extended period of time and with the motivation that comes from a commitment to making the host country their new home. A particularly useful type of study would involve comparisons of a large number of adult immigrant learners from each of several L1 backgrounds – both European and non-European – having comparable levels of proficiency in the language of the host country. The present book offers a number of studies that meet these criteria. The studies included in this book all draw from the ASK1 corpus – a corpus 1

2

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

of texts written by adult immigrant learners of Norwegian at controlled levels of language proficiency. As we will describe in the paragraphs that follow, the ASK corpus includes texts written by L2 learners of Norwegian from 10 different L1 backgrounds, two of which are non-Indo-European (i.e. Somali and Vietnamese), five of which are Indo-European but non-Germanic (i.e. Albanian, Bosnian-Croatian-Serbian, Polish, Russian and Spanish) and the remaining three of which, like Norwegian, are both Indo-European and Germanic (i.e. Dutch, English and German). The composition of these learner groups offers the possibility of a number of intriguing comparisons. Most of the studies in this book examine only a subset of these groups, but the patterns produced by the non-Indo-European groups are given special attention, and inter-L1-group comparisons are a core feature of every study. The studies in this book deal with the acquisition and assessment of a number of disparate features of Norwegian involving vocabulary, grammatical morphology, syntactic constructions and broad categories of errors, but all of the studies in this book are linked to each other through the theme of crosslinguistic influence. Following the common convention of using the term crosslinguistic influence interchangeably with transfer (e.g. Jarvis & Pavlenko, 2008; Odlin, 1989), we characterize the present volume as a book on L1 transfer in the acquisition of Norwegian as a second or additional language. However, in this book we adopt a perspective on transfer that is broader than the goals pursued in most other contemporary research on crosslinguistic influence. Whereas other studies tend to take cases of interL1-group heterogeneity (i.e., differences in the use of the target language by learners who speak different L1s) as merely the starting point for a larger analysis whose goal is to determine whether differences in the learners’ L1s themselves are the direct cause of the inter-group differences in target-language performance (cf. Jarvis, 2000), the studies in the present volume seek first and foremost to identify cases of inter-L1-group heterogeneity and to explore their nature and distribution. Some of the studies in the present book go a step further and adopt Jarvis’ (2000, 2010) methodological framework in order to confirm L1 effects beyond inter-L1-group heterogeneity. However, what characterizes the approach to transfer taken in this book is the recognition that (a) transfer effects are not always immediately traceable to the source language and (b) the unique challenges experienced by learners from different L1 backgrounds are sometimes of tremendous theoretical and practical import regardless of whether those challenges can be shown to be a direct consequence of the learners’ specific L1 knowledge. The ASK corpus is a collection of texts written in an authentic assessment situation by more than 1700 adult learners, resulting in a corpus of about 620,000 words. The texts were written as part of two separate standardized tests of Norwegian as a second language: The Language Test for Adult Immigrants (Språkprøven i norsk for voksne innvandrere) and the

Introduc t ion

3

Test of Norwegian – Advanced Level (Test i norsk – høyere nivå). The former test was designed for learners at an intermediate level of Norwegian proficiency, whereas the latter test is used as a university entrance examination for learners at a higher, pre-academic level. Both tests contain multiple components, only one of which is the essay part that constitutes the content of the ASK corpus. The texts written as part of the intermediate test are mainly expository essays, but they also include narratives and argumentative essays. The texts written as part of the higher-level test, on the other hand, are all argumentative essays. Comparable texts were also collected from a control group of 200 L1 Norwegians, 100 of whom were given some of the same prompts that were used on the intermediate-level test, and the other 100 of whom were given some of the same prompts as were found on the higher-level test. The texts written by L1 Norwegian speakers were produced under the same conditions as the learner texts. There was, nevertheless, one important difference between the two: For the learners, this was a high-stakes test that had consequences for their ability to get a job or to be admitted into a university, whereas for the L1 Norwegian speakers, this was simply an ungraded writing assignment. Although the data in the ASK represent the written language use of different L1 groups in two separate assessment situations – 10 L1 groups at the intermediate level and seven2 at the higher level – the texts in the corpus are fairly homogenous in several respects. First, the texts were all written in similar testing situations and were produced under similar conditions with respect to time constraints, learners’ access to reference tools, task type and so forth. Furthermore, the texts in the ASK were produced by learners who shared the same type of learning context in that they were all learning Norwegian as immigrants to Norway. The available data also provide important information that makes it possible to perform useful comparisons between learners and groups. For example, all of the learners passed their respective test and must therefore be assumed to be at or above the proficiency level associated with a passing score for that test. Although the data were collected before these tests were linked to rating scales of the Common European Framework of Reference for Languages (CEFR), passing scores on the two tests correspond roughly to the B1 and B2 CEFR levels, respectively (Council of Europe, 2001). Also, in 2010 the texts written by the English, German, Polish, Russian, Somali, Spanish and Vietnamese groups were given individual ratings using the CEFR writing scale (Carlsen, 2012), and this information is now included in the metadata for these texts. The ASK also contains additional metadata consisting of self-reported personal information that makes it possible to control for several variables that may affect language learning. In addition to indicating their L1 background and which of the two Norwegian tests they took, the metadata also include information about the participants’ age, sex, years of residence in Norway, which Norwegian courses if any they have taken, how often they use Norwegian

4

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

and in which contexts and what their self-reported level of English proficiency is. Another important feature of the ASK is that all of its texts have been manually coded for errors (Tenfjord et al., 2006) and automatically annotated for parts of speech (POS) and grammatical functions. The ASK is set up with a query engine called Corpuscle (Meurer, 2012), which offers rich search possibilities. Searches can be performed through either a menu-based search, a Corpuscle query-syntax search, or a combination of the two. Because the texts in the ASK were produced in an authentic, high-stakes testing situation, we may assume that they reflect the writers’ best attempts at producing well-written texts. The ASK corpus is therefore highly suitable to the investigation of a range of questions, dealing not only with crosslinguistic influence, but also with how students present themselves in a testing situation and which strategies they use when answering questions about topics ranging from friendship to the advantages and disadvantages of using a mobile phone. The fact that the texts were written on multiple topics does nevertheless present certain challenges to the researcher. Among these challenges is the difficulty of comparing texts that make use of different (though overlapping) sets of vocabulary. Despite these challenges, the ASK is a valuable, rich and well-structured corpus that makes it possible to explore certain questions about language learning that have not yet been addressed at all, and to address other questions that have not yet been fully answered. These are questions whose answers are of great interest to researchers and practitioners of L2 Norwegian, transfer researchers and scholars in the field of learner corpus research.

Organization of the Book This book consists of nine chapters. The present chapter constitutes Chapter 1. The second chapter, written by Scott Jarvis, provides an overview of relevant areas of current transfer research in order to contextualize the chapters that follow it. Chapters 3–9 are all empirical studies that use the ASK as their source of data, and the first six of these (Chapters 3–8) specifically rely on Jarvis’ (2000, 2010) methodological framework for verifying cases of transfer. The final chapter (Chapter 9) does not rely on Jarvis’ methodological framework but rather represents the expanded scope of transfer referred to in Chapter 2; it deals with patterns of language use that are characteristic of learners from specific L1 backgrounds without delving into whether the sources of these patterns can necessarily be identified in learners’ L1s per se. Additional information on each chapter is given in the following paragraphs. In Chapter 2, Scott Jarvis discusses the terminological conventions, theoretical tenets and methodological considerations regarding crosslinguistic

Introduc t ion

5

influence that inform the research designs and analyses of several of the book’s chapters. The core of the chapter is the methodological framework originally proposed by Jarvis (2000) as a way to achieve higher levels of methodological rigor in the investigation of crosslinguistic influence and as a way to ensure that researchers’ claims about the presence or absence of crosslinguistic influence in a sample of learner data are sufficiently supported by relevant types of evidence. The chapter also includes two other key components. One is a discussion of the levels of knowledge and performance at which crosslinguistic influence can occur. This discussion addresses, among other things, the distinction between linguistic and conceptual transfer, how conceptual transfer can be identified and verified, and whether or how conceptual transfer can be investigated in a written learner corpus such as the ASK. A final key component of Chapter 2 is a discussion of the value of expanding the scope of transfer research to include an exploration of the various ways in which learners from different language backgrounds differ in the acquisition and use of the same target language. The author relates this to the methodological framework discussed earlier, and argues that even when there is insufficient evidence to claim that the source of a particular pattern in the learners’ use of the target language is one or more of their specific background languages, there might nevertheless be sufficient evidence to show that the pattern is indeed characteristic of learners with a specific language background. Such a finding may hold great value in and of itself, not least in the practical implications it offers to language learners and teachers. Chapter 3, written by Ann-Kristin Helland Gujord, deals with crosslinguistic influence in the acquisition of temporal morphology. The chapter analyses the use of preterite and perfect morphology in L2 Norwegian texts written by speakers of two non-Indo-European languages: Somali and Vietnamese. Importantly, the learners’ L1s are dissimilar as to how temporal and aspectual distinctions are encoded. The Norwegian perfect construction has no counterpart at all in Somali, but its prototypical function largely corresponds with the way that temporal/aspectual markers are used in Vietnamese. The results of the study show that Somali learners of Norwegian tend to erroneously substitute the preterite and perfect for each other significantly more often than Vietnamese-speaking learners, whereas the Vietnamese learners’ errors in the use of tense and aspect are largely limited to the overuse of the present perfect in preterite contexts. The author argues that the main differences between the Somali and Vietnamese learners are not so much in their levels of accuracy, but instead in the distribution of their errors when referring to past time. This finding aligns with the findings of previous studies that have examined learners of L2 Norwegian, L2 English and L2 Swedish from a variety of L1 backgrounds. Collectively, these studies show that the present perfect is especially challenging for learners whose L1s altogether lack this grammatical category, and is also difficult for learners

6

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

whose L1s have tense/aspect morphology that superficially resembles the L2 perfect construction but is functionally different. In Chapter 4, Oliwia Szyman´ska examines the use of spatial prepositions in L2 Norwegian by learners whose L1s include Bosnian-Croatian-Serbian, English, Polish and Russian. Szyman´ska focuses on learners’ use of the Norwegian prepositions i and på (in and on), examines which factors – including crosslinguistic influence – lead students to choose a certain preposition in a given situation, and discusses whether non-native-like patterns in their use of spatial prepositions might be attributed to conceptual transfer. The results show, among other things, that despite the fact that English, like Norwegian, is a Germanic language, the performance of the English-speaking learners was no better than that of the other groups (who were all speakers of Slavic languages). To probe further into the factors that affect learners’ choices of spatial prepositions, Szyman´ska supplemented her analysis of the ASK data with data from a sentence completion task, where native speakers of Norwegian and Polish learners of the language were asked to supply the correct spatial preposition in a sentence context. The overall results of both analyses were interpreted as showing that the learners’ use of spatial prepositions is affected by L1-specific conceptualization patterns for objects and places (container vs surface), specific vs non-specific referents, and various other types of conceptual distinctions. Chapter 5, written by Marta Janik, deals with crosslinguistic influence in learners’ use of adjective inflection in L2 Norwegian. The L2 writers in this study include native speakers of English, German and Polish. Two types of adjective inflection are examined: (a) adjective–noun agreement markers, which reflect the number, gender and definiteness of the modified noun, and (b) adverbial marking, which gives adjectives adverbial functions through inflectional rather than derivational morphology. One of Janik’s main research questions is whether the English-speaking learners might experience greater difficulties with Norwegian adjective agreement than the other two groups of learners given that adjective inflection does not exist in English at all. Janik also investigates whether the German-speaking learners might be at a disadvantage in comparison to the Polish-speaking learners in relation to adjective agreement in predicative position, where German adjectives remain uninflected but Polish works similarly to Norwegian. The results of the study, which relies on Jarvis’ (2000, 2010) analytical framework, confirm both positive and negative transfer in most of the expected contexts. In Chapter 6, Silje Ragnhildstveit investigates transfer in the domain of grammatical gender by examining second language learners’ use of indefinite articles. Several previous studies have reported that having a gender system in the L1 results in higher levels of accuracy in learners’ patterns of L2 gender assignment. However, Ragnhildstveit challenges this finding with data from ASK texts written by learners whose L1s include English, Dutch, German, Spanish and Vietnamese. The results of the study do not show an overall

Introduc t ion

7

advantage for learners whose L1s have a grammatical gender system (i.e. Dutch, Spanish), nor even for learners whose L1 has a grammatical gender system similar to the target language (i.e. German). The study focuses particularly on the German and Vietnamese groups; German, like Norwegian, has three grammatical genders (masculine, feminine and neuter), whereas Vietnamese lacks grammatical gender but has a classifier system in which nouns are marked for properties such as animacy and geometrical shape. In general, the Vietnamese learners are more likely to omit indefinite articles than the learners from the other groups, but when they actually use articles they are more likely to use them correctly according to gender. This latter result is an example of a transfer effect that is not immediately traceable to Vietnamese. Chapter 7, written by Kari Tenfjord and Torodd Kinn, is a study that examines learners’ use and non-use of preposition stranding in L2 Norwegian. Preposition stranding is a typologically unusual construction among the languages of the world, being found almost exclusively in some Germanic languages, English and the Nordic languages in particular. Norwegian has stranding constructions similar to those found in English, but it also has some stranding constructions that English does not have. The study compares learners of Norwegian whose native languages include English, BosnianCroatian-Serbian, Dutch, German, Polish, Russian and Spanish in order to determine whether the L1 English group shows advantages over the other groups in relation to the correct use of preposition stranding in contexts that are and are not congruent with English. The results suggest an affirmative answer, with some important nuances. Chapter 8, written by Anne Golden, examines the extent to which learners communicate emotions and what types of emotions they express when writing essays on the topic of friendship. The analysis focuses on learners’ use of three categories of emotion-related words: terms denoting emotions (e.g., happy, sad, anger, love), somatic markers of emotion (e.g. ‘It made me cry’), and metaphorical expressions of emotion (e.g. ‘Her feet barely touched the ground’). The results show that the learners used emotion terms quite frequently – especially when compared to the results of a study by Bednarek (2008) on the use of emotions in different genres – and favored adjectives while doing so. Somatic and metaphorical markers of emotions were less used. A more detailed comparison of the Bosnian-Croatian-Serbian, German and Russian groups shows that emotion terms that are either very frequent in Norwegian or represent concepts that have only one major equivalent term expressed in the learners’ L1s, are used more frequently than emotion terms representing concepts that do not align with the L1 or have multiple translation counterparts. Golden argues that this use – and non-use – of emotion concepts in the learners’ texts may be interpreted as a result of crosslinguistic conceptual influence. Finally, the study by Anne Golden, Lars Anders Kulbrandstad and Kari Tenfjord, Chapter 9, addresses a question that lies at the intersection of

8

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

second language acquisition and language testing: Why are certain L1 groups rated lower than others on the same writing composition exam? In order to address this question, the researchers examine all of the texts written by Spanish speakers and Vietnamese speakers that are part of the Språkprøven (intermediate-level test) component of the ASK. These groups were chosen because the texts written by Vietnamese-speaking learners were, on the whole, given lower ratings than those written by Spanish-speaking learners. In this study, the researchers compare the two groups in relation to a number of variables designed to capture both the accuracy and complexity of the texts, and they also examine other global variables (e.g. how the writers presented themselves) that might account for differences in how they were rated. The results show that accuracy combined with text length account for the CEFR ratings of about 80% the texts, whereas the CEFR ratings of the remaining 20% of the essays require a more detailed, nuanced examination of their content and structure.

Contributions, Limitations and Future Directions The present book offers a number of important contributions to ongoing work in crosslinguistic influence and learner corpus research. First, it confirms the findings of previous studies while at the same time offering valuable new and nuanced findings regarding the impact of the L1 on later-learned languages. These new insights have been made possible by the large and unique nature of the ASK corpus, which is one of the very few large-scale learner corpora dealing with a target language other than English, and which also stands out with respect to the variety and number of L1 groups it encompasses. A second important contribution of the book stems specifically from the fact that it deals with Norwegian learner language; the findings brought to light throughout this volume will serve as important baseline findings for future studies of Norwegian as a second language, and will also serve as a rich resource for teachers and learners of Norwegian as a second language. The many other empirical and practical contributions of the present volume can be seen in the details of its individual chapters, which show inter alia (a) how learners’ L1s affect the way they approach the learning situation (e.g. whether they tend to be risk-takers, and whether they tend to take a holistic versus analytic approach to learning; see the chapter by Ragnhildstveit), (b) how subtle differences between the L1 and L2 can be overlooked by learners and can lead them to think that the two systems are the same (see the chapter by Helland Gujord) and (c) how perceived similarities provide leaners with certain advantages but can also mislead them into thinking that there is nothing new to learn. Besides its empirical, theoretical and pedagogical contributions, the present volume also contributes to the interdependent and developing

Introduc t ion

9

relationship between learner corpus research and second language acquisition. The studies in the present volume can be seen as studies that are theoretically and empirically grounded in the field of second language acquisition while being methodologically anchored in the tools and conventions of learner corpus research. Insofar as this book demonstrates the value of these tools and conventions, we believe that one of the important achievements of the book is its integration of learner corpus research more firmly into the established methods of the field of second language acquisition. Above all, the use of corpus tools and conventions has made it possible to compile, tag and analyse a large number of relatively infrequent forms and constructions produced freely and spontaneously by a large number of learners in openended and relatively natural language tasks, unlike the experimental elicitation tasks that have traditionally been used to investigate such features of language (see, e.g. Bardovi-Harlig, 2000; Ellis, 1994). The findings that come out of such large-scale analyses help move the field of second language acquisition forward, and the strong theoretical orientation and methodological standards adopted by each of the studies in this volume help add rigor and depth to the body of research dealing with learner corpora. The studies contained in this volume do of course have certain limitations, many of which are related to the ASK itself. The texts included in the ASK were written in a formal testing situation in response to prompts that were chosen by the test creators rather than by the learners themselves. We recognize that this carries consequences for the naturalness and authenticity of the language found in the texts. Another limitation is that the ASK consists only of L2 Norwegian; it would be ideal if it also contained texts in the learners’ L1s on the same topics and written under the same conditions by the learners themselves or by other native speakers of those languages who were as comparable as possible to the learners whose texts are currently included in the ASK. This would allow for useful comparisons of patterns in both the L1 and L2 in order to verify whether the L2 patterns have indeed been carried over from the L1. Another limitation of the ASK is that the topics written about are not equally balanced across L1 groups, and some topics do not have sufficient representation among any of the groups. These and other limitations of the ASK are limitations that apply to most learner corpora, and we hope that our contribution to the ongoing discussion of these limitations will lead to promising new solutions in the future. One of the solutions we recommend for the future is, of course, to supplement the existing learner corpus data with comparable L1 data and to perform the types of comparison recommended by Jarvis (2000, 2010) in order to determine the extent to which the L2 patterns really do reflect L1 tendencies. Another recommendation for future research that makes use of the ASK or other learner corpora is to follow the example of Szyman´ska (this volume) and augment the corpus data with supplemental tasks that allow the researcher to verify the patterns found in the corpus data in more controlled

10

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

contexts and to triangulate the factors that govern these patterns. Asking the writers of the texts in the corpus (or comparable learners) to reflect on the similarities they perceive (or assume to exist) between concepts in their L1 and the target language (as Golden did with respect to emotion concepts in this volume) will also increase the potential of research related to crosslinguistic conceptual influence. Research that deals with features of the L2 that have counterparts in some but not all of the learners’ L1s, and where the degree of L1-L2 congruence varies across those L1s that do have these features, is of particular interest for future research. Grammatical gender (see the study in this volume by Ragnhildstveit) is especially noteworthy in this regard, and in this area it would be valuable in the future to examine gender assignment and gender agreement not just in the article system, but also in learners’ use of adjectives that modify the target nouns. Future research would also benefit by looking more carefully at potential frequency effects that might result in different patterns of performance with words, morphemes or syntactic constructions that are highly frequent versus those that are less frequent. Studies like those in the present volume nearly always leave valuable leftovers that are beyond the scope of the current investigation but are worth returning to in the future. Each study in this book contains many such leftovers, and here we will mention just a couple: We noted earlier that the study in this volume by Golden, Kulbrandstad and Tenfjord found that the CEFR ratings for about 80% of texts examined were predictable based on a combination of the grammatical accuracy of those texts and how long they were. This is an intriguing finding, but it also leaves 20% of the texts still unaccounted for; we believe that some of the most interesting findings regarding the factors that affect CEFR ratings are yet to be discovered in a future study that delves deeply into those leftover texts. Among the many factors that affect CEFR ratings might be the writers’ style and some of the content they chose to include or not to include in their essays. We hope that these suggestions and the many others provided throughout the chapters of this volume will lead to valuable new avenues of discovery for ourselves as well as for those who share our passion for exploring and solving the mysteries of second language acquisition – especially as they relate to the distinctive patterns produced by learners from differing language backgrounds.

Notes (1) ASK is an acronym for the three constituent morphemes in the Norwegian term andrespråkskorpus (second language corpus). The ASK is available to registered users at the following website: http://clarino.uib.no/ask/ask.) (2) Three of the L1 groups at the higher level are incomplete: The ASK contains only 24 texts written by L1 Albanian learners, five by L1 Vietnamese learners and seven by L1 Somali learners from the higher level test.

Introduc t ion

11

References Bardovi-Harlig, K. (2000) Tense and aspect in second language acquisition: Form, meaning, and use. Language Learning 50 (1), xi–491. Bednarek, M. (2008) Emotion Talk Across Corpora. New York: Palgrave. Carlsen, C. (2012) Proficiency level: A fuzzy variable in computer learner corpora. Applied Linguistics 33 (1), 1–24. Council of Europe (2001) Common European Framework of Reference for Languages: Learning, Teaching, Assessment. Cambridge: Cambridge University Press. Ellis, R. (1994) The Study of Second Language Acquisition. Oxford: Oxford University Press. Jarvis, S. (2000) Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning 50 (2), 245–309. Jarvis, S. (2010) Comparison-based and detection-based approaches to transfer research. EUROSLA Yearbook 10, 169–192. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York/London: Routledge. Meurer, P. (2012) Corpuscle: A New Corpus Management Platform for Annotated Corpora. Amsterdam: Benjamins. Odlin, T. (1989) Language Transfer: Cross-linguistic Influence in Language Learning. Cambridge: Cambridge University Press. Tenfjord, K., Hagen, J.E. and Johansen, H. (2006) The hows and whys of Coding Categories in a Learner Corpus (or ‘How and why an error-tagged Learner Corpus is not ipso facto one big Comparative Fallacy’). Rivista di Psicolinguistica Applicata (RiPLA) VI (3), 198–208.

2

Transfer: An Overview with an Expanded Scope Scott Jarvis University of Utah

What is Transfer and Why is it Important? Transfer refers to the ways in which a person’s knowledge of one language can affect his or her learning, knowledge and use of another language (cf. Jarvis & Pavlenko, 2008: 1). This term is often used interchangeably with crosslinguistic influence – or CLI – and this is a convention that will be followed throughout this chapter and the entire book. Although some scholars have implied that the term transfer as it relates to language learning originated in the behaviorist paradigm in the early 1900s (e.g. Gass, 2013: 79–85), the truth is that transfer and its translation equivalents in other languages were already being used to refer to CLI well before this time and in a way that was completely independent of behaviorism (see Odlin & Yu, 2016: 5–6). Even so, some researchers are unhappy with the term transfer because they feel it implies that something is literally being moved from one place to another (e.g. Corder, 1983: 92). Although most researchers today do not view CLI as involving the literal transfer of forms, structures, meanings or functions from one part of a language learner’s brain to another, the original meaning of the Latin term transferre (trans- = ‘across’, ferre = ‘to carry’) is still relevant to many cases of CLI, where language learners appear to carry over certain patterns of language use from one language into their use of another language (cf. Odlin & Yu, 2016: 3–4). Other types of CLI – such as the enhanced rate of learning a similar language – do not reflect this common meaning of transfer, but even in these cases it can be preferable to use transfer rather than crosslinguistic influence for reasons of convenience. It is certainly less cumbersome to say lexical transfer, for example, than it is to say lexical crosslinguistic influence, and this should not be a problem as long as researchers clearly define the phenomenon they are investigating. Although some researchers have associated transfer more or less exclusively with the influence of learners’ native language (L1) on their acquisition of a second language (L2) (for a discussion of this, see De Angelis & Dewaele, 12

Transfer: An Over v iew w ith an E xpanded Scope

13

2011: vii), most researchers today use transfer, crosslinguistic influence, and CLI interchangeably to refer to the influence of any language(s) a person knows on his or her learning, knowledge or use of any other language(s) (see Jarvis & Pavlenko, 2008). This is what transfer will mean here, too, though admittedly the primary focus throughout the present book is indeed L1 influence on learners’ use of an L2. Importantly, L2 in this book is defined as referring to an additional language learned after the L1, which in many cases is not chronologically the second in the series of languages a given learner has learned or attempted to learn. Linguists, philosophers, philologers, anthropologists, psychologists and language practitioners have been interested in language transfer for about as long as such professions have existed – and well before second language acquisition became its own recognized field of research (see, e.g. Janse, 2002; Odlin & Yu, 2016: 5–6). Because CLI has been researched for so long, some researchers have felt that there is nothing new to be discovered.1 However, this is not true. For one thing, a number of carefully conducted empirical studies of CLI published since 2000 have debunked earlier claims about the nature of CLI and especially about the areas of language use that were previously believed to be impervious to transfer, such as morphology (see Jarvis & Odlin, 2000; Kaivapalu & Martin, 2007; Luk & Shirai, 2009; Murakami & Alexopoulou, 2016), syntax (Bohnacker, 2006; Meriläinen, 2010) and conceptual meaning (Pavlenko, 2011; Stutterheim & Nüse, 2003). Another important point is that transfer research is old only in terms of how long it has been of interest to scholars, not in terms of the theories that it draws from or the empirical methods it makes use of; like many other areas of research, transfer research has been developing at a rapid rate both theoretically and methodologically. Empirical research on CLI has been generating a steady stream of new, intriguing questions to follow up on. Additionally, new theories and empirical findings in other domains of research are also providing a constant flow of new questions to be explored and new hypotheses to be tested concerning, for example, the implications of linguistic relativity for second language acquisition and use (e.g. Bylund & Athanasopoulos, 2014; Jarvis, 2016a; Odlin, 2005), the ways in which multiple languages are stored in and processed by a single brain (e.g. Cook, 2003; Fabbro, 2002), what happens in the brain when learners receptively process structures in an L2 that are either similar to or different from corresponding structures in the L1 (e.g. Jeong et al., 2007; Trude & Tokowicz, 2011) and which types of differences between the L1 and L2 result in the greatest challenges for L2 acquisition (e.g. Tokowicz & MacWhinney, 2005). The value of transfer research to scholars, practitioners and laypeople is ultimately tied to how it helps answer questions that are central to learning itself: How does the brain make use of prior knowledge when learning something new, how do similarities between what learners know and what they are learning affect the learning process and how do learners recognize similarities to begin with?

14

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Where Can You Find CLI, and How Can You Be Sure it is CLI? Odlin (2003) has rightly pointed out that there is essentially no area of language use where CLI does not occur. Finding it, however, can sometimes be a challenge because even though it does affect ‘all linguistic subsystems including pragmatics and rhetoric, semantics, syntax, morphology, phonology, phonetics, and orthography’ (2003: 437), it does not occur in any these areas of language use at all times or under all conditions. The odds of encountering CLI in learners’ language use are often greatest when (a) the target language is related to a language the learners have already mastered (e.g. Kellerman, 1983), (b) the learners are using a language feature (i.e. a form, function, structure or meaning) that is prototypical or unmarked (e.g. Kellerman, 1978), (c) the feature is frequent in the learners’ L1 (e.g. Poulisse, 1999) or another previously acquired language and (d) the learner is at an appropriate stage of development to make use of that feature (e.g. Ellis, 2015: 131–132). There are a number of other factors, too, that combine to determine the likelihood of encountering CLI in any given sample of learner data (see Jarvis & Pavlenko, 2008: 174–210). Beyond the specific linguistic subsystems and contexts in which CLI might occur, there is also the question of what types of analysis will be the most successful in detecting its occurrence. One particular distinction that is relevant to this discussion is the distinction between the examination of individual cases of CLI versus the analysis of group tendencies (see Odlin, 2014). To be sure, group tendencies are made up of individual cases just as forests are made up of trees, but sometimes the forces that result in differences between forests are easier to see and confirm than are the forces that affect the specific characteristics of individual trees. The reason for this – as it applies to CLI – is that individual cases are often not uniquely identifiable as CLI. That is, there are often multiple plausible explanations for why a particular learner used or did not use a particular target-language feature at a specific moment in time. Consider, for example, the following sentence produced by a 14-year-old Swedish-speaking learner of English, who produced this sentence as part of an essay describing a Charlie Chaplin film: The police came and girl was afraid. The lack of a determiner before the word girl in this sentence is ungrammatical, but is CLI the cause of this error? Could it not just as easily reflect a lack of proficiency or a lack of control? The learner who produced this sentence used an article before the word police in the same sentence and also correctly used an article 17 out of the 19 times she referred to the girl in her essay, so perhaps this particular error was merely a blip that should be ignored, similar to the types of performance errors that Chomsky (1965) has described as occurring even in the speech of native speakers but which do not reflect their actual linguistic competence. Or, perhaps this

Transfer: An Over v iew w ith an E xpanded Scope

15

error reflects a deliberate use of simplified language, similar to what is commonly found on street signs, in text messages and in news headlines such as ‘Accidental missile launch kills boater’ (a headline on cnn.com on 1 July 2016). Again, individual instances of language use often have multiple plausible explanations, which can make it difficult to identify CLI confidently without taking group-level patterns into consideration. Turning to group-level considerations, if the error in the above example were caused by influence from the learner’s L1 Swedish, then we should expect other L1 Swedish learners of English who are the same age (14 years), have the same amount of English instruction (2.5 years) and were given the same task (a written narrative of an eight-minute segment of a Chaplin film) to produce the same type of error. Indeed, this is what we find – but only to a small degree. In a sample of 68 learners with the same profile (see Jarvis, 2002), five of them omitted an obligatory determiner at least once when referring to the female protagonist in the film. So, this raises the following question: Is five out of 68 learners – i.e. 7% of the sample – enough to say that this is a pattern that is representative of Swedish-speaking learners as a group? Probably not. Another important group-level consideration is whether the omission of determiners is something that learners from all language backgrounds do. If it is, then this would be a universal tendency in second language acquisition that might have nothing to do with learners’ language backgrounds. One way to find out in relation to the present example is to compare the 68 Swedish speakers’ performance on this task with that of a comparable group of learners who speak a different native language, such as Finnish. The data I have from 46 Finnish speakers who performed the same task and are at an equivalent level of English proficiency to the Swedish speakers show that 35 of the Finnish speakers – i.e. 76% of the learners in this sample – omitted determiners when referring to the girl. This is considerably larger than the 7% of Swedish speakers who did so and is clearly indicative of a strong within-group tendency for the Finnish speakers. Furthermore, the differences between the Finnish group and the Swedish group are significant (see Jarvis, 2002), which means that we have both within-group and betweengroup evidence that it is actually the Finnish speakers rather than the Swedish speakers whose L1 elevates their likelihood of omitting determiners in this context. In previous work (Jarvis, 2000, 2010a, 2016b; Jarvis & Pavlenko, 2008), I have referred to within-group tendencies and between-group differences as intra-group homogeneity and inter-group heterogeneity. These are important constructs that underlie common statistical tests such as the t-test and analysis of variance (ANOVA). When differences between groups are significantly greater than differences within groups, then we can be quite certain that those between-group differences are generalizable to the populations from which the groups were drawn (as long as the groups really are

16

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

representative of those populations). Given enough of the right types of evidence, we can also conclude that the factor or set of factors that distinguishes the groups as belonging to separate populations is directly or indirectly responsible for their differences in performance. Importantly, there are two levels of analysis here: first, the determination of whether the differences between groups are significant and, second, the determination of which factor or set of factors is responsible for those differences. In the example given earlier, the two groups were L1 Finnish speakers and L1 Swedish speakers. If the only difference between these groups is simply what their native language is, then we can be certain that their L1s are responsible for the differences in their performance in relation to the use of English determiners. However, despite researchers’ best attempts at recruiting comparable groups, it is rare if not impossible to find two groups of learners anywhere whose differences really are limited only to what their L1s are. In the present example, the Finnish speakers and Swedish speakers were both citizens of Finland and therefore more or less unicultural (cf. Ringbom, 1987), and they were also at roughly equivalent levels of English proficiency (Jarvis, 2010b), but they began learning English at different ages (age 9 for the Finns, age 11 for the Swedes), had studied English for different numbers of years (6.5 years for the Finns, 2.5 years for the Swedes), had had different English teachers, had used different English textbooks and so forth (Jarvis, 1998). Thus, because there are multiple factors beyond the learners’ L1s that distinguish them from each other, it is necessary to bring in additional evidence in order to determine whether the observed differences in the groups’ performances really are a function of their L1s. In other words, the purpose of this extra step is to attempt to rule out the possibility that a factor other than the L1 is responsible for the differences between groups. The most direct and relevant way to do this is to verify that the targetlanguage feature that is suspected of being affected by a particular L1 has a counterpart in that L1. This can be done on the basis of what the researcher already knows about the L1 or on the basis of what has been written about it in existing grammars and empirical studies. However, because language use tends to vary by task and context (e.g. Tarone & Parrish, 1988), the best way to do this is to collect performance data from native speakers of the L1 using the same task that was used to collect data in the target language. The L1 control group should of course be as comparable as possible to the learner group who performed the task in the target language. When comparable data are collected in both the L1 and L2, the researcher can verify whether patterns produced in the L2 can indeed be regarded as reflections of how the learners would have performed the same task in their L1. The data I have from Finland include narrative essays written in Finnish and Swedish by native speakers of these languages who had the same profiles as the learner groups and who performed the same task that they did. The results of the L1 groups show that all 41 (100%) of the participants in the L1 Finnish group

Transfer: An Over v iew w ith an E xpanded Scope

17

omitted determiners at least once when referring to the female protagonist, whereas only one out of 40 – i.e. 3% – of the participants in the L1 Swedish group did so. These patterns are very similar to the patterns found in the corresponding learner groups, which increases the likelihood that the Finnish-speaking learners’ tendency to omit determiners and the Swedishspeaking learners’ tendency to use determiners in this context are predilections carried over from their L1s. In previous work (Jarvis, 2000, 2010a, 2016b; Jarvis & Pavlenko, 2008), I have referred to similarities between learners’ use of their L1 and L2 as crosslanguage congruity. When combined with intra-group homogeneity and inter-group heterogeneity, cross-language congruity increases the positive evidence for CLI, helps rule out competing explanations for the pattern in question, and helps pinpoint the exact source of the effect. In contrast to intra-group homogeneity and inter-group heterogeneity, which are used to show that the pattern in question is more or less group-specific, crosslanguage congruity shows that the pattern is more or less source-language specific (i.e. L1-specific in the case of L1 influence). Both group specificity and source-language specificity are important to the investigation of CLI because, when they are used in combination, this greatly reduces the number of competing explanations that might account for the pattern in question. Wherever possible, though, competing explanations should be eliminated in the design phase of a study by making sure that (a) groups of participants are truly representative of their identified populations, (b) the different groups are comparable in all possible ways except in relation to the independent variable(s) (e.g. CLI) that are being investigated and (c) the different groups are given the same tasks in the same conditions. It will probably never be possible to eliminate all potential intervening variables through the research design, but this should be the goal. In the example used throughout this section (using data from Jarvis, 2002), the research design made it possible to rule out differences between groups that might otherwise be attributable to differences in culture, L2 proficiency and task conditions. However, some differences (e.g. age, years of instruction) were unavoidable, and their effects could be satisfactorily ruled out only through a detailed examination of group specificity and source-language specificity in learners’ use and non-use of determiners in a specific context. In summary, it is important to verify that the area of learners’ performance that is suspected of being influenced by CLI is indeed characteristic of (via within-group similarities) and distinctive of (via between-group differences) learners representing a particular language background, and that their performance is furthermore motivated by the corresponding features or conventions of their background language(s). These types of analysis need to be performed in conjunction with an attempt to eliminate as many other potential differences between groups as possible. Ultimately, however, what determines whether learner performance can be ascribed to CLI is how

18

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

successfully the researcher has been able to rule out competing explanations. Where no other plausible explanation exists, very little evidence for CLI is needed, but where many exist, a more rigorous method is required. These are the methodological principles that inform the studies included in this volume. The major challenge for these studies and most investigations of CLI that make use of learner corpora is the general unavailability of L1 data.

Frontiers in CLI CLI research has recently entered a new era of sophistication characterized by both intensive and extensive investigations revolving around questions and hypotheses that could not previously be addressed due to the lack of necessary knowledge and tools. Indeed, many of these questions and hypotheses had not even been conceived until our theoretical understanding of language, cognition and the brain reached a critical stage of development, and also not until recent technological advances opened our eyes to the range of questions that could even be asked. In most cases, of course, the theoretical and methodological advances that have propelled CLI research forward have not been the creations of CLI researchers themselves; rather, CLI researchers have looked to other areas of research and even to other fields in order to gain new sources of inspiration, new insights and new solutions to both old and new problems related to the nature of CLI and the forces that govern its occurrence. Importantly, CLI researchers have shown a high level of skill and commitment in their efforts to learn from other areas of research and in their efforts to adapt and apply the things they have learned to the investigation of CLI. It is often difficult to separate theoretical advances from empirical and methodological advances, but some of the most notable expansions to our theoretical understanding of CLI have come from recent research on the differential effects of crosslinguistic similarities versus crosslinguistic differences on second language acquisition and use, the effects of prior knowledge and experience on present performance in a second language, the relationship between prior knowledge and language aptitude and the effects of prior patterns of conceptualization on present performance in another language. I will say just a few words about each of these areas, where important theoretical and empirical advances have taken place in recent years. Regarding the differential effects of similarities versus differences, these can clearly be seen in a variety of empirical contexts, including those involving normal spoken or written L2 production, hemodynamics and electrical activity in the brain and learners’ performance on psycholinguistic tests. A study by Jeong et al. (2007) provides an excellent example of the effects of crosslinguistic similarities on blood-flow patterns in the brain. In this study, two groups of language learners were recruited: Korean speakers who had

Transfer: An Over v iew w ith an E xpanded Scope

19

learned both English and Japanese to comparable levels of proficiency, and Chinese speakers who had learned both English and Japanese to comparable levels of proficiency. The learners were tested one at a time: They were placed inside an fMRI (functional magnetic resonance imaging) machine, where they listened to sentences in both of the target languages. All of the sentences included an agent, a transitive verb, a direct object and a spatial expression (e.g. ‘John met the teacher at school’). Importantly, the structure of the English sentences was more similar to Chinese than it was to Korean, whereas the structure of the Japanese sentences was more similar to Korean than it was to Chinese. The fMRI results showed, as expected, that the Chinese-speaking learners exhibited higher levels of blood flow to a specific area of the brain when listening to Japanese than when listening to English, whereas the Koreans exhibited the opposite pattern. Importantly, blood is what fuels the activation of neurons and neural pathways, so the fact that the learners showed less blood flow when processing the more similar L2 suggests that learners of a similar language are able to make use of already existing neural pathways, which require less fuel (and correspondingly less mental energy) to activate. Other research has shown that crosslinguistic similarities are not always beneficial to language learning, and crosslinguistic differences are not always detrimental. Instead, similarities can sometimes have negative consequences, as in the case of false cognates (Jarvis, 2009), and differences can sometimes have positive consequences, as in the case of novelty effects (Kleinmann, 1977). Importantly, Ringbom (2007) and Tokowicz and MacWhinney (2005) have pointed out that there are two important classes of crosslinguistic differences: those that involve points of comparison and contrast between two languages (e.g. English and Spanish both have definite articles, but English uses the same definite article for both singular and plural nouns whereas Spanish uses different articles for singular and plural), and those in which one language has a feature that has no identifiable counterpart in the other (Spanish has grammatical gender but English has no counterpart to this). Tokowicz and MacWhinney investigated the effects of these two types of differences with English-speaking university students who were studying Spanish as a foreign language. The researchers found that even though the learners did better on a grammaticality judgment test with contrasting structures (i.e. Spanish singular and plural definite articles) than with unique structures (i.e. Spanish grammatical gender), their electrical brain activity showed that they began acquiring intuitions about the unique structures earlier than the contrasting structures. Thus, it appears that some types of crosslinguistic differences can be more conducive to implicit learning than others. Importantly, unique structures are impervious to one critical effect that makes crosslinguistically similar structures so easy to process and contrasting structures so challenging: the priming effect of previous knowledge and experience. Research involving lexical decision tasks with masked

20

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

priming – i.e., a task where a participant is shown a ‘prime’ word for less than 100 milliseconds before it is covered with a ‘mask,’ after which the participant is shown a target word and asked to decide as quickly as possible whether the target word is a real word – has shown that prime words that are orthographically, phonologically and/or semantically similar to the target word facilitate the lexical decision task (i.e. help participants decide more quickly whether the target word is a real word). This is true for both intralingual and interlingual primes (see, e.g. Dikstra, 2005). Crucially, the priming effect is not limited to words, but also extends to larger syntactic constructions. For example, a study by Shin and Christianson (2009) investigated the ability of Korean-speaking learners of English to repeat English sentences that were presented to them aurally. The target sentences involved double-object constructions presented in either of two possible orders (e.g. ‘The lawyer gave the child a gift’ vs ‘The lawyer gave a gift to the child’). Immediately after the participants heard the target sentence but before they repeated it, they also heard a Korean prime sentence that was unrelated in meaning to the target sentence but included a double-object construction of its own. The results showed that the syntactic ordering of the double-object constructions in the participants’ recalled sentences was significantly affected by the order of arguments in the Korean prime sentences. Some researchers have suggested that findings like this might not be limited to the laboratory; rather, the L1 and any other language(s) that a person has used frequently and recently might produce an overall priming effect on the way that person uses the target language (e.g. Jackson & Ruf, 2017), or even on the ways he or she perceives and categorizes experience (e.g. Thierry et al., 2009), which will also have an effect on language use (Jarvis, 2016a). Priming effects are typically seen as fleeting, short-term facilitating effects of immediately prior experience that result in the cross-activation of related mental representations. There is probably a natural continuum that runs from such immediate and fleeting effects of prior experience to the more entrenched and enduring effects of prior experience and prior knowledge that occur even in the absence of a priming condition. For convenience, I will refer to the entrenched effects of prior experience and prior knowledge simply as prior knowledge. One of the most intriguing characteristics of prior knowledge is its ability to disguise itself as innate ability. For example, research on short-term memory using the digit-span task has shown that English speakers can typically recall between five and nine digits accurately in the correct order. A person’s digit span is generally assumed to be an index of an innate and immutable memory capacity. However, research by Ericsson and Kintsch (1995) has shown that, with sufficient prior knowledge of digit sequences, people can be trained to increase their digit spans substantially – even beyond 80 digits. Prior language knowledge can have similar effects, and it has been shown that learners of a target language that is similar to a language they have already mastered (including their L1) tend to be vastly

Transfer: An Over v iew w ith an E xpanded Scope

21

superior to learners of a distant or unrelated language in terms of both their rate of acquisition and the ultimate levels of proficiency they achieve (e.g. Ringbom, 2007). Not only this, but there is also strong evidence that learners of a distant or unrelated language are significantly better at processing, retaining and acquiring those features of the target language that are congruent with their L1 than they are of incongruent features (e.g. Odlin, 2016; Paribakht, 2005). Empirical evidence also suggests that crosslinguistic similarity – or relevant prior knowledge – tends to be a better predictor of L2 processing and retention than is either short-term memory capacity or working memory capacity (Jarvis et al., 2013). Thus, prior knowledge as it relates to crosslinguistic similarities appears to be an exceedingly important part of the aptitude that a learner brings to the task of acquiring a new language. The prior knowledge that affects the acquisition and use of an L2 is not limited to prior linguistic knowledge. In previous work (e.g. Jarvis & Pavlenko, 2008), I have emphasized the distinction between linguistic knowledge and conceptual knowledge, and accordingly also the distinction between linguistic transfer and conceptual transfer. Linguistic transfer involves the influence of a person’s language-specific knowledge of one language on his or her acquisition or use of another language. Linguistic transfer thus includes phenomena commonly referred to as phonological transfer, morphological transfer, syntactic transfer and some types of lexical transfer. Conceptual transfer, on the other hand, is CLI that originates not from a person’s knowledge of linguistic forms, structures or rules, but from the ways a person has learned – as a speaker of a particular language and as a member of a particular discourse community – to attend to, perceive, interpret, construe, conceptualize, categorize and refer to experience. As discussed in Jarvis (2016a), conceptual transfer includes phenomena that range from differences in the conceptual meanings that learners from different language backgrounds express when performing the same target-language production task, to differences in target-language comprehension or production that can ultimately be traced to language-specific differences in learners’ extralinguistic patterns of attention, perception, categorization and memory. Stated differently, conceptual transfer covers the range of L2 phenomena associated with Thinking for Speaking (e.g. Slobin, 1996) on one end, and Linguistic Relativity (e.g. Bylund & Athanasopoulos, 2014; Odlin, 2005) on the other. The research discussed in the preceding paragraphs represents some of the exciting new developments in CLI research that have expanded our theoretical understanding of what CLI means, the mechanisms through which it operates, what its effects are and where and how it might be detected. Tied to these theoretical developments are a number of new procedures, methods, tools and technologies that have likewise led to new discoveries about and new ways of researching CLI. Some of these methodological developments have already been mentioned, including brain-imaging technologies for measuring blood flow and electrical activity, psycholinguistic tools and

22

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

techniques for measuring priming effects and the perception of crosslinguistic similarities, and cognitive tools and techniques used for measuring aptitude, attention, perception, conceptualization, categorization and memory. Beyond these, perhaps the most important methodological advances in CLI research have involved the adoption of sophisticated data mining tools and practices (see, e.g. Jarvis, 2011). These include tools and techniques that allow for the collection, organization, storage, tagging, rapid retrieval and efficient analysis of massive amounts of data, as can be found in the form of learner language corpora such as the ASK and the ICLE (Granger et al., 2009), as well as in the form of huge databases of L2 test performance such as the Dutch as a Second Language state exam in The Netherlands (Schepens et al., 2013). One area of CLI research involving both data mining and learner corpora – and which has been particularly active since about 2005 – is commonly referred to as native-language identification (NLI) (Jarvis & Paquot, 2015). NLI uses computer-automated, machine-learning systems called classifiers that attempt to learn which features (e.g. words, phrases, grammatical constructions, errors) in samples of target-language production tend to be associated with which L1 backgrounds. The final stage in this type of research is often a testing phase, where the classifier applies the classification model it has ‘learned’ to new samples of learner data in an attempt to ‘predict’ what the learners’ L1s are. Depending on the specific classifier that has been selected, the language features that are fed into it, and the number of L1s it has to choose among, NLI tends to result in surprisingly high levels of L1 classification accuracy. The highest L1 classification accuracy achieved so far (relativized to the number of L1s involved in the analysis) is perhaps that reported by Jarvis et al. (2013), whose classifier identified the correct L1 affiliations of 920 (83.6%) of the 1100 texts in the testing sample of the TOEFL11 learner corpus (a corpus of over 12,000 essays written by learners of English from 11 different L1 backgrounds), and it did so in less than a second. The technology and statistical methods that make this possible are fascinating, but the real value of NLI to CLI research is seen in its ability to gauge the prevalence and magnitude of CLI in learner corpora, and in the speed and efficiency with which it is able to identify the combinations of targetlanguage features that reflect group-characteristic and group-distinctive tendencies. As mentioned earlier, when L1 groups differ in relation to more than just their L1s, then additional evidence – especially performance congruity between learners’ use of the L1 and the target language – will be required to reduce the likelihood that the performance differences between L1 groups could have been caused by something other than their L1s. So, NLI is often not sufficient by itself as a method for investigating CLI, but it is one of the most powerful methods for gathering certain types of evidence for L1 influence – namely, evidence of group specificity. In the future, it should be possible to adapt NLI methodology to the investigation of the influence of languages other than the L1.

Transfer: An Over v iew w ith an E xpanded Scope

23

NLI is a valuable way of investigating CLI in learner corpora, but it is important to recognize that learner corpora are an extremely important resource to CLI research even when they are not analysed through machinelearning protocols such as those that underlie NLI. Even when they involve more traditional analyses of CLI, such as those found in this book, learner corpora and the metadata and search tools they include allow CLI researchers to examine larger amounts of data in more detail and in less time, and to perform larger and more complex group comparisons and analyses of the main effects and interactions of numerous variables, than has previously been possible (see, e.g. Granger et al., 2015). It is clear that learner corpora are a valuable domain for research on linguistic transfer, but can they also be used to investigate conceptual transfer? As mentioned earlier, conceptual transfer refers to a continuum of related phenomena that range from language-specific ways of expressing conceptual meaning to language-specific ways of attending to and creating mental representations of experience, whether that experience is real or imagined (Jarvis, 2016a). Learner corpora should certainly be amenable to investigations of whether learners from different language backgrounds differ in terms of the conceptual meanings they express in the same contexts. However, the usefulness of doing this depends a great deal on the nature of the corpus, and one very important limitation of learner corpora in relation to the investigation of conceptual transfer is that they tend to consist of essays written on a wide variety of topics that make it difficult to find similar contexts across multiple texts. Moreover, even the texts written on the same topic tend to be openended in such a way that it is difficult to find different learners referring to the same precise contexts where useful comparisons of conceptual meaning could be made. Researchers interested in investigating conceptual transfer in a learner corpus should therefore attempt to locate or compile a corpus of language samples produced by learners who are all given the same task and topic. In order to ensure that the contexts of meaning really are comparable across texts, it is perhaps best to use visual stimuli, such as a film description task (e.g. von Stutterheim, 2003), a film retell task (e.g. Jarvis, 2000) or a picture description task (e.g. Sanchez, 2011). Alternatively, researchers could investigate conceptual transfer while using learner corpora as only one part of a multi-part study, where differences in the expression of meaning observed between different groups of learners in a typical learner corpus are explored in more detail through a second phase of the study that targets the same learner populations and determines whether the observed differences really do have conceptual origins and really do reflect the tendencies of specific language populations (see Jarvis, 2016a for more information on relevant methods for investigating conceptual transfer). In short, a learner corpus can be a valuable domain for investigating conceptual transfer, but this requires that the learner corpus (a) be designed and compiled with this purpose in mind, or (b) be supplemented with additional data that meet the

24

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

requirements for this type of research. It is relevant to point out that there was an early interest in using the ASK to investigate Thinking for Speaking patterns and conceptual transfer, but it was realized that the ASK is not properly designed for this type of research. This can still be achieved through the use of the right types of supplemental data, but this has not yet been attempted on a comprehensive scale (but see Golden, this volume).

Expanding the Scope In past work (e.g. Jarvis, 2000, 2010a), I have emphasized the conditions and types of evidence needed to verify CLI and to identify its sources with high levels of confidence. In more recent work (Jarvis, 2012, 2016b), I have also argued that the scope of transfer research extends beyond those relatively few studies that achieve full methodological rigor, to all of the additional studies that expand our understanding of the potential causes and effects of CLI, and which point to cases where there exists some evidence for it and where it might be confirmed through additional investigation. Here, I want to re-emphasize these points regarding the scope of transfer research and the range of studies that make positive contributions to it. I also want to propose that the scope of transfer research be expanded slightly beyond this to include studies designed to explore and identify the distinctive patterns of target-language behavior exhibited by learners from different language backgrounds, even when such studies are not designed to isolate the sources of these distinctive patterns. There are at least two reasons for proposing that studies of this type be included within the scope of CLI research. The first reason is that these types of studies carry a great deal of intrinsic value that – depending on the audience – can be indistinguishable from the value of traditional CLI studies. The findings and implications of such studies are especially valuable to language teachers, language testers and language learners, for whom a knowledge of the language learning challenges they will likely face and the solutions for overcoming these challenges are far more important than the specific origins of the challenges. Stated differently, for language practitioners, studies that successfully link language learners’ behaviors to the features of a specific source language have essentially the same thrust as those studies that document distinctive patterns without attempting to trace those patterns back to their origins. Returning to the example given earlier of Finnish speakers’ tendency to omit articles in English, what really matters to language practitioners is information about which learners tend to have this problem, what the exact nature and extent of the problem is, and especially how to overcome it – regardless of where the problem stems from. A second reason not to exclude such studies from the scope of transfer research is that – as mentioned earlier – CLI does not always have linguistic

Transfer: An Over v iew w ith an E xpanded Scope

25

sources, and studies designed with broader perspectives in mind might be capable of discovering some of the important consequences that being a member of a particular discourse community has for learning another language. These are consequences that might not be directly tied to the linguistic (e.g. phonological, lexical, grammatical) inventories of learners’ background languages, but might instead be tied to the language habits and conventions of the discourse communities learners come from – for example, conventions related to how particular entities, qualities, relationships and events are construed and expressed (e.g. Jarvis, 2016a). These are CLI-related consequences that might otherwise be overlooked by researchers who focus narrowly on the linguistic inventories of learners’ background languages without taking into consideration the ways those languages are actually used in real communicative contexts. Studies designed to investigate the distinctive patterns of learners from different language backgrounds and which are not constrained by pre-existing assumptions about where such distinctive patterns might be found, could play a very important role in the discovery of heretofore overlooked cases of, for example, conceptual transfer and pragmatic transfer. It might be argued that most NLI studies, which seek to discover patterns of target-language use that are distinctive of particular L1 backgrounds, are of exactly the type of study I have been referring to in this section. I believe that they are, and the value of such studies to CLI research should be clear. They bring to light the most reliably recurrent patterns found in the language use of learners who share a particular language background, they provide valuable and actionable information to language practitioners about the strengths and challenges that learners from specific backgrounds face, and they provide a wealth of valuable fodder for follow-up analyses designed to confirm and isolate the sources of those distinctive patterns (cf. Jarvis, 2012). Importantly, these are characteristics of all of the empirical studies contained in the present volume, all of which explore the distinctive patterns of different groups of learners of L2 Norwegian, and all of which fall within this expanded scope of CLI research.

Note (1) This is a sentiment that I have heard expressed privately by a few scholars, but it has perhaps not appeared in the published literature since Krashen (1983).

References Bohnacker, U. (2006) When Swedes begin to learn German: from V2 to V2. Second Language Research 22 (4), 443–486. doi:10.1191/0267658306sr275oa Bylund, E. and Athanasopoulos, P. (2014) Linguistic relativity in SLA: Toward a new research program. Language Learning 64, 952–985. doi:10.1111/lang.12080 Chomsky, N. (1965) Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.

26

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Cook, V. (ed.) (2003) Effects of the Second Language on the First. Clevedon: Multilingual Matters. Corder, S. (1983) A role for the mother tongue. In S.M. Gass and L. Selinker (eds) Language Transfer in Language Learning (pp. 85–97). Rowley, MA: Newbury House. De Angelis, G. and Dewaele, J.-M. (eds) (2011) New Trends in Crosslinguistic Influence and Multilingualism Research. Bristol: Multilingual Matters. Dikstra, T. (2005) Bilingual visual word recognition and lexical access. In J.F. Kroll and A.M.B. De Groot (eds) Handbook of Bilingualism: Psycholinguistic Approaches (pp. 179– 201). Oxford: Oxford University Press. Ellis, R. (2015) Understanding Second Language Acquisition (2nd edn). Oxford: Oxford University Press. Ericsson, K.A. and Kintsch, W. (1995) Long-term working memory. Psychological Review 102 (2), 211–245. Fabbro, F. (2002) The neurolinguistics of L2 users. In V. Cook (ed.) Portraits of the L2 User (pp. 199–218). Clevedon: Multilingual Matters. Gass, S.M. (2013) Second Language Acquisition: An Introductory Course. New York: Routledge. Granger, S., Dagneaux, E., Meunier, F. and Paquot, M. (2009) The International Corpus of Learner English. Handbook and CD-ROM (Version 2). Louvain-la-Neuve, Belgium: Presses universitaires de Louvain. Granger, S., Gilquin, G. and Meunier, F. (eds) (2015) The Cambridge Handbook of Learner Corpus Research. Cambridge: Cambridge University Press. Jackson, C.N. and Ruf, H.T. (2017) The priming of word order in second language German. Applied Psycholinguistics 38 (2), 315–345. Janse, M. (2002) Aspects of bilingualism in the history of the Greek language. In J. Adams, M. Janse and S. Swain (eds) Bilingualism in Ancient Society: Language Contact and the Written Text (pp. 332–390). Oxford, UK: Oxford University Press. Jarvis, S. (1998) Conceptual Transfer in the Interlingual Lexicon. Bloomington, IN: Indiana University Linguistics Club Publications. Jarvis, S. (2000) Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning 50 (2), 245–309. Jarvis, S. (2002) Topic continuity in L2 English article use. Studies in Second Language Acquisition 24 (3), 387–418. Jarvis, S. (2009) Lexical transfer. In A. Pavlenko (ed.) The Bilingual Mental Lexicon: Interdisciplinary Approaches (pp. 99–124). Bristol: Multilingual Matters. Jarvis, S. (2010a) Comparison-based and detection-based approaches to transfer research. EUROSLA Yearbook 10, 169–192. Jarvis, S. (2010b, February) Learning and Acquiring English in Finland both Inside and Outside of School. Keynote address delivered at the Ainedidaktiikan symposiumi (Symposium on Content Didactics) held at the University of Helsinki, Finland, February 12, 2010. Jarvis, S. (2011) Data mining with learner corpora: Choosing classifiers for L1 detection. In F. Meunier, S. De Cock, G. Giquin and M. Paquot (eds) A Taste for Corpora. In Honour of Sylviane Granger (pp. 127–154). John Benjamins. Jarvis, S. (2012) The detection-based approach: An overview. In S. Jarvis and S.A. Crossley (eds) Approaching Language Transfer Through Text Classification: Explorations in the Detectionbased Approach (pp. 1–33). Bristol: Multilingual Matters. Jarvis, S. (2016a) Clarifying the scope of conceptual transfer. Language Learning 66 (3), 608–635. Jarvis, S. (2016b) The scope of transfer research. In L. Yu and T. Odlin (eds) New Perspectives on Transfer in Second Language Learning (pp. 17–47). Bristol: Multilingual Matters. Jarvis, S., Bestgen, Y. and Pepper, S. (2013) Maximizing classification accuracy in native language identification. In: Proceedings of the Eighth Workshop on Innovative Use of NLP for Building Educational Applications (pp. 111–118). Atlanta: Association for Computational Linguistics.

Transfer: An Over v iew w ith an E xpanded Scope

27

Jarvis, S., O’Malley, M., Jing, L., Zhang, J., Hill, J., Chan, C. and Sevostyanova, N. (2013) Cognitive foundations of crosslinguistic influence. In J. Schwieter (ed.) Innovative Research and Practices in Second Language Acquisition and Bilingualism (pp. 287–308). Amsterdam: Benjamins. Jarvis, S. and Odlin, T. (2000) Morphological type, spatial reference, and language transfer. Studies in Second Language Acquisition 22 (4), 535–556. doi:10.1017/S0272263100004034 Jarvis, S. and Paquot, M. (2015) Native language identification. In S. Granger, G. Gilquin and F. Meunier (eds) The Cambridge Handbook of Learner Corpus Research (pp. 605– 627). Cambridge: Cambridge University Press. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York/ London: Routledge. Jeong, H., Sugiura, M., Sassa, Y., Yokoyama, S., Horie, K., Sato, S., Taira, M. and Kawashima, R. (2007) Cross-linguistic influence on brain activation during second language processing: An fMRI study. Bilingualism: Language and Cognition 10 (2), 175. doi:10.1017/ S1366728907002921 Kaivapalu, A. and Martin, M. (2007) Morphology in transition: Plural inflection of Finnish nouns by Estonian and Russian learners. Acta Linguistica Hungarica 54 (2), 129–156. Kellerman, E. (1978) Giving learners a break: Native language intuitions as a source of predictions about transferability. Working Papers on Bilingualism 15, 59–92. Kellerman, E. (1983) Now you see it, now you don’t. In S.M. Gass and L. Selinker (eds) Language Transfer in Language Learning (pp. 112–134). Rowley, MA: Newbury House. Kleinmann, H. (1977) Avoidance behavior in adult second language acquisition. Language Learning 27, 93–107. Krashen, S. (1983) Newmark’s ‘Ignorance Hypothesis’ and current second language acquisition theory. In S.M. Gass and L. Selinker (eds) Language Transfer in Language Learning (pp. 135–153). Rowley, MA: Newbury House. Luk, Z.P. and Shirai, Y. (2009) Is the acquisition order of grammatical morphemes impervious to L1 knowledge? Evidence from the acquisition of plural -s, articles, and possessive’s. Language Learning 59 (4), 721–754. Meriläinen, L. (2010) Language Transfer in the Written English of Finnish Students. Joensuu, Finland: University of Eastern Finland. Murakami, A. and Alexopoulou, T. (2016) L1 influence on the acquisition order of English grammatical morphemes: A learner corpus study. Studies in Second Language Acquisition 38, 365–401. doi: http://dx.doi.org/10.1017/S0272263115000352 Odlin, T. (2003) Cross-linguistic influence. In C. Doughty and M. Long (eds) Handbook of Second Language Acquisition (pp. 436–486). Oxford: Blackwell. Odlin, T. (2005) Crosslinguistic influence and conceptual transfer: What are the concepts? Annual Review of Applied Linguistics 25, 3–25. doi:10.1017/S0267190505000012 Odlin, T. (2014) Rediscovering prediction. In Z. Han and E. Tarone (eds) Interlanguage: Forty Years Later (pp. 27–46). Amsterdam: Benjamins. Odlin, T. (2016). Language transfer and the link between comprehension and production. In L. Yu and T. Odlin (eds) New Perspectives on Transfer in Second Language Learning (pp. 207–225). Bristol: Multilingual Matters. Odlin, T. and Yu, L. (2016) Introduction. In L. Yu and T. Odlin (eds) New Perspectives on Transfer in Second Language Learning (pp. 1–16). Bristol: Multilingual Matters. Paribakht, T.S. (2005) The influence of first language lexicalization on second language lexical inferencing: A study of Farsi-speaking learners of English as a foreign language. Language Learning 55, 701–748. Pavlenko, A. (ed.) (2011) Thinking and Speaking in Two Languages. Bristol: Multilingual Matters. Poulisse, N. (1999) Slips of the Tongue: Speech Errors in First and Second Language Production. Amsterdam: Benjamins.

28

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Ringbom, H. (1987) The Role of the First Language in Foreign Language Learning. Clevedon: Multilingual Matters. Ringbom, H. (2007) Cross-linguistic Similarity in Foreign Language Learning. Clevedon: Multilingual Matters. Sanchez, L. (2011) ‘Luisa and Pedrito’s dog will the breakfast eat’: Interlanguage transfer and the role of the second language factor. In G. De Angelis and J.-M. Dewaele (eds) New Trends in Crosslinguistic Influence and Multilingualism Research (pp. 86–104). Bristol: Multilingual Matters. Schepens, J., Van der Slik, F. and Van Hout, R. (2013) Learning complex features: A morphological account of L2 learnability. Language Dynamics and Change 3 (2), 218–244. doi:10.1163/22105832-13030203. Shin, J.-A. and Christianson, K. (2009) Syntactic processing in Korean-English bilingual production: Evidence from cross-linguistic structural priming. Cognition 112 (1), 175–180. Slobin, D. (1996) From ‘thought and language’ to ‘thinking for speaking’. In J. Gumperz and S. Levinson (eds) Rethinking Linguistic Relativity (pp. 70–96). Cambridge: Cambridge University Press. Stutterheim, C. and Nüse, R. (2003) Processes of conceptualization in language production: Language-specific perspectives and event construal. Linguistics 41, 851–881. Tarone, E. and Parrish, B. (1988) Task-related variation in Interlanguage: The case of articles. Language Learning 38 (1), 21–44. DOI: 10.1111/j.1467-1770.1988.tb00400.x Thierry, G., Athanasopoulos, P., Wiggett, A., Dering, B. and Kuipers, J. (2009) Unconscious effects of language-specific terminology on pre-attentive colour perception. Proceedings of the National Academy of Sciences 106, 4567–4570. Tokowicz, N. and MacWhinney, B. (2005) Implicit and explicit measures of sensitivity to violations in second language grammar: An event-related potential investigation. Studies in Second Language Acquisition 27, 173–204. Trude, A.M. and Tokowicz, N. (2011) Negative transfer from Spanish and English to Portuguese pronunciation: The roles of inhibition and working memory. Language Learning 61 (1), 259–280. doi:10.1111/j.1467-9922.2010.00611.x von Stutterheim, C. (2003) Linguistic structure and information organization: The case of very advanced learners. EUROSLA Yearbook 3, 183–206.

3

The ‘Perfect Candidate’ for Transfer: A Discussion of L1 Influence in L2 Acquisition of Tense-Aspect Morphology Ann-Kristin Helland Gujord University of Bergen

Introduction The focus of this chapter is on the role of learners’ L1 in a domain in which language-specific factors have traditionally received less attention: L2 acquisition of temporal morphology. Research on how learners acquire temporal expressions in a second language is fairly extensive within second language acquisition (SLA). A lot of research has been conducted on the acquisition of tense aspect morphology in particular (Bardovi-Harlig, 2000; Shirai, 2009; Mitchell et al., 2013). However, the research in this area has revolved around the universal aspects of the developmental process and the identification of stages of acquisition. Although this research has certainly provided important insight into L2 acquisition of temporal expression, the role of language-specific factors has received less attention. The L1 has not been regarded as a significant factor, and the effect of the L1 on the L2 acquisition process has not been studied systematically (Collins, 2004; Shirai, 2009). Accordingly, the knowledge we have about how L2 learners acquire temporal markers in an L2 stems predominantly from universalistic-oriented research aiming to identify the common features and paths. For instance, this applies to the ESF project1 on L2 acquisition of temporal expressions by adult immigrants (Dietrich et al., 1995). Although the question of transfer was considered in the ESF project, and it was designed to compare interlanguages from different L1 groups (Dietrich et al., 1995: 3), the researchers concluded that ‘there is no significant SL [source language] influence in the acquisition of temporality’ (Dietrich et al., 1995: 278). Other examples are found in 29

30

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Bardovi-Harlig’s (2000) extensive survey of research on temporality in an L2. The sequences of acquisition found in learners’ emergent tense and aspect systems in several Germanic L2s were emphasized, and the findings of the relevant studies led to the conclusion that ‘the target language exerts a much greater influence in the acquisition of morphology than a learner’s first language’ (2000: 419). Moreover, the studies as a whole do not reveal a ‘significant L1 effect’ on the acquisition of temporal expressions (2000: 411). Rather, the acquisition of tense and aspect is developmentally constrained. The conclusions of the ESF project and Bardovi-Harlig’s synthesis of the existing literature contrast with the findings from several studies that will be presented in this chapter, including the current one. The results of the present study will show that Vietnamese and Somali learners of Norwegian exhibit systematic differences in their grammatical encoding of past time (Gujord, 2013, 2015), and that these differences may be accounted for in terms of L1 influence. The structures addressed are the preterite and the present perfect. Further support to a finding that seems to be consistent across several studies of L2 acquisition of tense in Norwegian is added by the current study: systematic differences in the acquisition of tense in Norwegian are shown by learners from different L1 backgrounds, and this L1 influence is revealed particularly in the learners’ encoding of present perfect contexts. The hypotheses in the current study not only build on previous findings, but also on L1-L2 contrastive relations. Three very typologically different languages are included in the current transfer investigation, which involves an exploration of the differences and similarities in past time encoding in Norwegian and the L1s, both in terms of how the languages contrast, as well as how the different L1 groups compare in their reference to past events in Norwegian. Special attention will also be received by the present perfect category in the section analysing the contrastive relations between Norwegian, Somali and Vietnamese. The reasons for this are the semantic complexity that is associated with the present perfect category, as well as the findings from studies of L2 Norwegian and L2 English that suggest that L2 acquisition of the present perfect, when certain L1-L2 relations exist, seems to result in L1 influence. These findings will be presented in the following literature review.

Evidence for Transfer in L2 Acquisition of Temporal Morphology Although it is probably fair to say that the role of learners’ L1 has been downplayed in research on temporal morphology, this is not true for the Norwegian SLA research. On the contrary, L1 influence has been an important issue in many of the studies of how learners acquire grammatical categories in Norwegian, and temporality is the domain in which the issue of

The ‘Per fec t Candidate’ for Transfer

31

transfer has been explored most deeply (Golden et al., 2007). These transfer studies have been comparison-based and founded on contrastive analyses of the L1 and L2, and often the Norwegian interlanguages of learners of remote language backgrounds have been explored. Vietnamese and Turkish are the L1s that have been studied most frequently in L2 Norwegian (Golden et al., 2007). Moreover, these studies have been largely meaning-oriented or concept-oriented in approach; they are not primarily structural/formal analyses, but place a good deal of emphasis on the underlying meaning of grammatical categories when exploring how temporal marking in the learners’ L1 affects the acquisition of the tense system of Norwegian.2 This line of research is closely related to the SLA milieu at the University of Bergen. The main findings of these studies will be outlined, starting with Tenfjord’s (1997) doctoral thesis, which has been an important impetus for the study of transfer in L2 acquisition of temporal morphology conducted in Bergen. Tenfjord’s (1997) dissertation is a longitudinal case study of four Vietnamese pupils’ grammaticalization of the preterite and the present perfect in Norwegian. The data included audio recordings; the first recordings started immediately after the pupils arrived, and the last recordings were done after the pupils had been in Norway for between five and six years (Tenfjord, 1997: 134). Tenfjord predicted that (1) the interlanguages of the Vietnamese learners would be characterized by a lack of the preterite and (2) the present perfect would emerge as a grammatical category before the preterite in the interlanguages. The observed sequence of the acquisition of verb morphology in Germanic languages is contradicted by the latter hypothesis (Bardovi-Harlig, 2000: 419),3 where the preterite is predicted to emerge before the present perfect.4 However, according to Tenfjord’s functionalistic view of language use and language learning, the perfect category has properties that make it more important to express during the early stages of acquisition than the redundant preterite category. This hypothesis was supported by the data. Even though highly individual paths towards the tense system in the target language were displayed by the learners, only 21% of the obligatory contexts for the preterite were inflected by the learners as a group. In contrast, the perfect was morphologically expressed in as many as 60% of the contexts where it would be the appropriate form. These quantitative results were further supported by qualitative analyses, such as productivity analyses. The learner’s tendency to begin using the present perfect before the preterite applied only to the basic perfect function, which Tenfjord regarded as the resultative perfect (1997: 218). Tenfjord interpreted her findings in terms of L1 influence and functionalism (1997: 237). As for L1 influence, Tenfjord argued that the Vietnamese learners may have relied on their L1 when acquiring the basic perfect function in Norwegian because there a similarity between the temporal/aspectual markers of Vietnamese and the resultative function of the perfect category in Norwegian exists. This will be further elaborated on in the section discussing contrastive relations.

32

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Tenfjord’s longitudinal study broke ground for several cross-sectional transfer studies of written interlanguage data, which focused specifically on the present perfect category. In a study of German learners’ (N = 16) choices of past tense forms in Norwegian, Karrer (1999) observed that German learners had problems with the perfect-preterite distinction in Norwegian. In translations of a German text into Norwegian, she found that the preterite and the present perfect were used by the learners in a manner that corresponds to the distributional pattern of the two forms in their German L1. The German present perfect and Norwegian present perfect are similar in form, but not in function. Whereas the German perfect has expanded into the territory of the preterite and is used as a marker of general past, the Norwegian perfect cannot be used in combination with definite time reference under any circumstances. According to Karrer, the functions of the perfect category were transferred by the German learners from their L1 to the IL, which resulted in inappropriate uses of the forms in Norwegian. A similar pattern was also detected in Randen (1999). She investigated the uses of the present perfect by Russian students enrolled in Norwegian courses in Russia (N = 7), and her analyses were based on the Russian students’ email correspondence with Norwegian students at a university in Norway. In Russian, each verb has to be inflected for the imperfect or perfective aspect, and Randen’s study aimed at investigating whether – and if so how – the Russian speakers’ knowledge of these aspectual distinctions affected their use of the Norwegian perfect category. Randen adopted Tenfjord’s explanation of the present perfect in Norwegian as being primarily a marker of aspect and not tense. Randen further argued that the various uses of the Norwegian present perfect5 are all tied to the perfective aspect because they all express termination (Randen, 1999: 68). Randen’s analysis revealed that Russian learners use the perfect in Norwegian predominantly to express terminated actions. In fact, in 97% of the Russian learners’ uses of the present perfect in Norwegian, the forms referred to events or conditions that started in the past and reached an endpoint. Furthermore, Randen’s interpretation of her findings indicated that Russian learners make associations between the perfective aspect in their Russian L1 and the (resultative) function of the present perfect in Norwegian because they restricted their uses of the present perfect to the uses that correspond to the Russian perfective aspect, and because they use the Norwegian perfect correctly in contexts in which they would use the perfective aspect in their L1. For instance, a tendency that was exhibited by the Russian learners was to encode contexts for the extended now-perfect in Norwegian incorrectly, and to use the present instead of the present perfect (e.g. ‘*I live in Bergen for seven years now’ instead of ‘I have lived in Bergen for seven years now’). In Russian, the perfective aspect would not appear in such contexts where the situation that started in the past still holds. Further evidence for L1 influence in the use of the preterite and present perfect in Norwegian is found in Moskvil (2004), Helland (2005) and Janik

The ‘Per fec t Candidate’ for Transfer

33

(2010). The same type of information as the present study was used by these studies: written texts produced in response to an official test of Norwegian as a second language. Moskvil (2004) examined Turkish learners’ use (N = 51) of the preterite and the perfect in Norwegian in written texts.6 Moskvil compared the distribution she found in the Turkish learners’ texts to the distribution from a comparison group of texts produced under the same conditions by Vietnamese learners (N = 9). Helland (2005) conducted a similar study, but used texts from Vietnamese learners (N = 36) as the primary material and texts written by Turkish learners as the comparison group (N = 15). Both studies found that the distributions of the preterite and the perfect in the two L1 groups were distinct in that Turkish learners, whose L1 lacks a perfect category, displayed a stronger tendency for non-appropriate use of the perfect in preterite contexts when compared to the Vietnamese learners. On the other hand, a more frequent target-like use of the perfect was exhibited by the Vietnamese learners, a finding that underscores that of Tenfjord (1997). Janik (2010) investigated transfer in Polish learners’ use of the present perfect and the preterite in Norwegian. She found that the Polish learners (N = 100), who lack a perfect category in their L1, had more problems in distinguishing the preterite from the perfect in Norwegian as compared to learners whose L1s have a perfect category similar in form to the one in Norwegian (English N = 100 and German N = 100). In a study of the acquisition of interrogative clauses and verb morphology in Swedish – a language very similar to Norwegian – Philipsson (2007) observed that Somali learners of Swedish (N = 12) have more problems with the present perfect form when compared to other L1 groups (Iraqi Arabic N = 12 and Persian learners N = 12). The studies surveyed have observed patterns that appear to be L1-specific. Furthermore, several of these studies have found that differences between the L1 groups often involve an overuse of the present perfect in preterite contexts. Importantly, this is a pattern that is not exclusive to studies of L2 Norwegian. For instance, Polunenko (2004) observed a similar pattern to that of Randen (1999) in an investigation of Russian-speaking learners of L2 English (N = 120). Similar to Randen, Polunenko found that her Russian participants overused the present perfect in the L2. She attributed this to their interlingual identification between perfective aspect in L1 Russian and the present perfect in the L2. Likewise, Collins (2002, 2004) and Ayoun and Salaberry (2008) found that French learners of English frequently overuse the present perfect in preterite contexts, but for a different reason from Russian learners of English. The explanation for French speakers’ overuse of the present perfect is similar to Karrer’s (1999) explanation of German learners’ problems with the present perfect-preterite distinction in Norwegian: Formal similarities accompanied by semantic (and functional) differences lead to L1 effects. The passé composé (=simple past) in French is structurally similar to the present perfect in English even though it has a very different function (see Collins, 2002).

34

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

As discussed by Bardovi-Harlig (2000: 419), the findings of studies on the order of emergence of verb morphology in several Germanic L2s suggest that a common path is followed by L2 learners regardless of L1 background. Nevertheless, the studies of L2 Norwegian, L2 Swedish and L2 English reviewed in this section show that the perfect is more challenging than the preterite for learners from some L1 backgrounds. In particular, L1 influence has been observed when the L1 does not have a perfect category at all (e.g. Janik, 2010; Moskvil, 2004; Philipsson, 2007; Polunenko, 2004; Randen, 1999), and also when there are formal or conceptual differences or similarities between a temporal category in the L1 and a perfect category in the L2 (e.g. Ayoun & Salaberry, 2008; Collins, 2002, 2004; Helland, 2005; Karrer, 1999). In summary, the studies reviewed here lead to the conclusion that learners’ L1s seem to play a role in the acquisition of temporal morphology. However, there are methodological challenges connected to several of these studies which warrant a new study. Firstly, in several of these studies, a control group of learners with another L1 background is not included. Secondly, in some of these studies, inferential statistics is not applied at all. Even though the doctoral theses and the master theses have generated important insight into transfer, there remains a problem of generalizability; the amount of data collected has simply been too small. This is not a uniquely Norwegian ‘problem’. The problem of study design and generalizability in transfer studies is one of the objections that Jarvis (2000) puts forth in his article discussing methodological issues in transfer research. Finally, the lack of empirically validated contrastive analyses is another problem in transfer studies in general. Granger (2003: 18) holds that a lot of previous contrastive studies and contrastive statements are ‘largely intuition-based’ and not empirically founded. In view of this discussion, it might also be that not enough attention has been paid by SLA researchers in general to methods for comparing the L1 and the L2. However, the methodological challenges identified here are accounted for in Jarvis’s (2000) methodological framework for transfer studies. The current study predicts systematic differences in the Vietnamese and Somali learners’ use of present perfect and the preterite in Norwegian based on the earlier findings outlined in the present section and contrastive analyses of how past time in the learners’ L1 is encoded. However, contrary to several of the previous studies, the analysis of transfer in the current study departs from a traditional contrastive analysis in that the present study is empirically grounded and adheres to Jarvis’s (2000) methodological framework.

The Present Study Approach The approach taken in the current study of L1 influence is the comparison-based approach. It relies on comparisons of the Norwegian interlanguage

The ‘Per fec t Candidate’ for Transfer

35

performances of two different L1 groups, comparisons of the encoding of time in three different language systems (Vietnamese, Somali, Norwegian), and comparisons of the grammatical encoding in the learners’ Norwegian interlanguage performance with the system of their L1s (Vietnamese or Somali). The method for identifying L1 effects meets several of the criteria set forward in Jarvis’s (2000) framework. The investigation of intra-L1-group homogeneity (criterion 1) and inter-L1-group heterogeneity (criterion 2) entails comparisons of the interlanguage performances of the Vietnamese-speaking learners and the Somali-speaking learners. Effect 1, intra-L1-group homogeneity, is identified if a detected pattern of grammatical encoding is sufficiently similar in the interlanguage performances of speakers who share the same L1 background. Effect 2, inter-L1 group heterogeneity, is found if the detected pattern of grammatical encoding in one L1 group is sufficiently dissimilar to the pattern of the other L1 group. Effects 1 and 2 were tested statistically by means of the Mann-Whitney U test. The group’s internal behaviour is compared to the differences in behaviour between the groups by this test (see the Statistics section). Consequently, a significant result indicates that the observations within one group are sufficiently similar, and are sufficiently dissimilar from the observations in the other group, to establish that the groups behave in reliably different ways with respect to the measured variable. The application of criterion 3, intra-L1-group cross language congruity, is not completely aligned with how Jarvis described cross-language congruity. The current study does not have access to the informants’ use of past marking in their L1 since only the texts that they have written in their L2, Norwegian, is in the possession of the current study. However, the current study does have empirically validated contrastive data of the encoding of time in Somali and Vietnamese as will be described in the Contrastive relations section. Furthermore, the results from the contrastive analyses strongly inform the specific predictions raised in the current study. Hence, the hypotheses are presented after the Contrastive section. In this section, the rationale for expecting a different outcome in the learners’ use of past morphology in Norwegian will be given based on translation data and careful analyses and comparisons of past time encoding in Norwegian, Vietnamese and Somali.

The data The current study is based on written performances by 161 adult test takers of an official test of Norwegian for immigrants. Of the 161 texts in the sample, 73 were written by native speakers of Vietnamese and 88 were written by native speakers of Somali. The Vietnamese and Somali learners constitute a mixture of immigrants comprising a variety of social and educational backgrounds, language skills, ages etc. Some of this variety is controlled for by means of a form with personal information that the test takers

36

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

filled out when taking the test.7 Table 3.10 in the Appendix shows how the 161 informants are distributed across various categories of personal information based on their background. The table reveals that the informants are rather similar when it comes to age, time of residence in Norway, length and duration of L2 instruction, and degree of practising Norwegian and socializing with Norwegians. The differences in the background variables between the two L1 groups include gender, reported levels of English knowledge, and educational background. However, analyses in Gujord (2013: 233–239) show that the difference in educational background and English skills between the Vietnamese and Somali learner group cannot account for the differences that evolve when the L1 is taken into account. The texts were extracted from the Norwegian learner corpus, ASK. They are essay compositions written as part of the Language test for adult immigrants (Språkprøven i norsk for voksne innvandrere). This test measures Norwegian language abilities at the intermediate level, and is aimed towards a language level that is sufficient for managing everyday life in the Norwegian community (Carlsen, 2012: 9).8 The test takers were asked to write a short text on a given topic that either related to personal experience, required a description of an event, or asked for a viewpoint on a subject of general, public interest (e.g. smoking, upbringing, equality of the sexes etc.). The test takers were allowed to choose between several different prompts, and thus their topics varied significantly. Topical diversity is a challenge for a study like the current one because the variable ‘writing topic’ strongly influences the temporal perspectives in the text, and therefore also determines the grammatical encoding that researchers can observe in the text (Gujord, 2013: 166; see also Golden, 2012). Furthermore, the prompt that asked the candidates to write about the future elicited the highest number of responses among the texts written by Vietnamese and Somali test takers. Hence, of the 99 Vietnamese and 97 Somali texts available in the corpus, 35 (26 Vietnamese and 9 Somali) of these could not be included in this study because they did not have contexts for the preterite or the present perfect. In addition, the considerable variety in the length of the texts made it challenging to compare them. As will be seen in the presentation of the analysis and results, there is great variation within the data set. However, this is strongly connected to the fact that the data was not elicited in a controlled research environment; the data comprised written texts that were not produced for a specific research goal, but were elicited for the purpose of evaluating learners’ language skills in Norwegian.

Contrastive relations The current transfer investigation deals with the semantic domain of temporality and the grammatical domain of morphology, and tense and aspect are the two fundamental grammatical categories encoding temporal

The ‘Per fec t Candidate’ for Transfer

37

notions. Tense is a category of the verb (Klein, 2009: 40) and is traditionally defined as ‘grammaticalised expressions of location in time’ (Comrie, 1985: 9). Tense is a deictic relation between a situation9 and the time of the utterance. Furthermore, within the canonical view of tense, the time of the utterance is taken to be the reference point. Contrary to the grammatical category of tense, aspect is a non-deictic category and does not express reference to time as tense does. Instead, it conveys different types of temporal perspectives that focus on different parts of the situation, such as the beginning of a situation or the completeness of a situation. Comrie’s definition of aspect is a classic reference: ‘aspect is different ways of viewing the internal temporal constituency of a situation’ (Comrie, 1976: 3, 5).

Past Time in Norwegian Norwegian is a tense-prominent language; neither aspect nor mood is grammaticalized. Instead, these notions are conveyed through lexical and syntactic devices. Together with the preterites in the Nordic language, the Norwegian preterite category is a typological peculiarity because of the fact that it only refers to the past without differentiating between aspects (Hammarberg & Viberg, 1977: 139). Hence, when compared to languages with several categories of grammatical aspect and tense, fewer distinctions are marked grammatically on the verb in Norwegian. However, the Norwegian language does grammaticalize temporal distinctions through a perfect category, which is lacking in many languages with morphologically rich verbal systems. Accordingly, L2 learners of Norwegian have to figure out the appropriate contexts for two past forms, the preterite and the perfect, in order to encode temporal information in a target-like manner. As indicated in the review of previous studies, it is seemingly not an easy task to know when to use the perfect and when to use the preterite for some learner groups. However, the notion that the perfect is a particularly challenging category to acquire is also supported by typological studies. The perfect category is characterized as a complex category across languages. It is one that is more complex than its related past category, the preterite (e.g. Comrie, 1976, 1985; Dahl, 1985, 2000; Bybee, 1985; Bybee et al., 1994; Bybee & Dahl, 1989), which according to Dahl, can be described as a tense form without any controversy because it does not carry any aspectual functions (Dahl, 1985: 116). On the other hand, the perfect has continued to be the subject of much debate, particularly regarding the issue of whether or not the perfect category is a category of tense or aspect. This discussion reflects the complexity of the category as well as its shifting and notoriously unstable nature (Bybee & Dahl, 1989; Lindstedt, 2000). Lindstedt (2000) has surveyed the general characteristics of the perfect and its diachronic development and current uses in

38

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

several languages have been discussed. One of Lindstedt’s main arguments is that the perfect category has shown a common path of grammaticalization in many different European languages from originally having aspectual constructions10 to developing more general tense functions, as well as more modal functions later on. The development of the additional functions has been accompanied by a change in the nature of the time content because the perfect becomes less aspect-like and more tense-like as the form develops new functions (Lindstedt, 2000: 369). Accordingly, the type of perfect that is closest to the historical sources has more to do with aspectuality than time reference. The ‘central and prototypical meaning’ of the perfect that Lindstedt describes as ‘expressing current relevance of a past situation’ (2000: 378) is encoded by this type of perfect. Lindstedt’s prototypical perfect, which he labels current relevance perfect, corresponds with Comrie’s (1976: 56) resultative perfect in which ‘a present state is being referred to as being the result of some past situation’. The aspectual character of the Norwegian perfect form is also a central topic in Tenfjord (1997) and is an integral part of her contrastive analysis of Norwegian and Vietnamese. The resultative perfect in Norwegian is regarded by Tenfjord as being a category of aspect and not of tense, whereas the preterite, on the other hand, is a category of tense (Tenfjord, 1997: 112). Uses of present perfect forms11 in the current study will be classified as either prototypical perfect (PP) or secondary perfect (SP). This distinction is influenced by Lindstedt’s analysis of European perfect forms, but there is not a one-to-one relationship between Lindstedt’s current relevance perfect and the current study’s prototypical perfect. This is because prototypical perfect in the current study’s categorization includes the experiential perfect.12 According to Lindstedt, experiential perfect is the first stage in the grammaticalization process of perfect categories in languages. Experiential perfect is slightly more tense-like than current relevance perfect, but Lindstedt underscores that the distinction between these two types of perfect is not immediately clear. Verb phrases containing a current relevance or experiential perfect can have elements of both (Lindstedt, 2000: 369). Both current relevance perfect and experiential perfect describe a completed situation in the past, which would make it challenging to distinguish between them in the interlanguage data. However, the distinction between completeness and non-completeness is highly relevant for understanding the contrastive relationship between Vietnamese and Somali. The analysis of PP (prototypical perfect) in the present study has been conducted based on the following description of what characterizes a prototypical perfect: a PP refers to a terminated event which took place in the past and which has consequences for the current state of affairs. Termination is a criterion, and a PP cannot have inclusive reference to the present time. Uses of the present perfect in the texts that do not qualify as prototypical are lumped together in a rather broad group that the present study refers to as

The ‘Per fec t Candidate’ for Transfer

39

secondary perfect (SP). This is a much more heterogeneous category than the PP category as it comprises various additional functions of the Norwegian present perfect, such as the extended now. The uses of the present perfect forms are classified based on a contextual interpretation of what type of content the forms encode. Below, 1 and 2 are examples of uses of the present perfect in the texts classified as prototypical, whereas 3 and 4 are examples of uses which have been classified as secondary. 1) Jeg ha-r les-t en bok som het-er Sinddbads verden I have-PRS AUX read- PST PTCP a book which be call-PRS Sinddbad’s world 2) Verden er ikke lenger så stor som den ha-r vær-t world be.PRS not any more so big as it have-PRS AUX be- PST PTCP

3) Jeg ha-r alltid ønsk-et å jobb-e i hjemmebasert i kommunen I have-PRS AUX always wish- PST PTCP to work-INF in home care services 4) Nå ha-r jeg vær-t i Norge i tre år now.ADV. have-PRS AUX I be- PST PTCP in Norway in three years

The present perfect uses in sentences 1 and 2 refer to past events or conditions that are finished. Hence, they are classified as PP. The present perfect forms in 3 and 4 are classified as SP because they refer to events or conditions that started in the past but are not terminated. For instance, it is possible to imagine sentence 3 being rephrased as follows: ‘I have always wanted, and still want, to become a nurse’. Sentence 4 has an inclusive reference to present time because it describes conditions that apply at the time of utterance. The subsequent sections address the question of how past time is expressed linguistically in the learners’ L2s, with a particular focus on how the time content encoded in the present perfect and preterite in Norwegian is conveyed in Vietnamese and Somali. However, since reference grammars can only serve as a secondary source of contrastive data (Bybee et al., 1994: 32; Gujord, 2013: 74), translation is used in the current study as a method of obtaining primary data about the encoding of time in the informants’ L1s. Two native speakers of Vietnamese and two native speakers of Somali have translated The perfect questionnaire (TPQ) developed by Dahl (1985, 2000) and Lindstedt (2000). This translation questionnaire is used to collect information about the perfect category and related categories, such as the preterite in Norwegian. TPQ consists of 88 contexts with 151 sentences. It is oriented towards semantic content and function and is based on universal features of crosslinguistic categories expressing concepts all natural languages are presumably able to express. Hence, the opportunity to observe how the same meaning is conveyed in the three languages is provided to the researcher.

40

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Past Time in Vietnamese The typological distance between Vietnamese and Norwegian is vast. Vietnamese verbs are tenseless, and temporal reference is not marked linguistically because it is accessible through the basic time of the contexts.13 Yet, in cases where the situation refers to ‘time different from this basic time’ (Thompson, 1965: 209), particular time markers must be used in order to clarify the temporal frame. Of these time markers, the markers đã and rôi merit special attention in the present contrastive analysis because they express past time. In the literature, đã is described as a tense marker that is used when the temporal reference is different from the basic time and is not accounted for in the context (e.g. Thompson, 1965; Ngo & Tran, 2001). Rôi is usually not presented as a marker of tense, but of aspect, and is commonly described as expressing ‘a terminated action or condition’ (Thompson, 1965: 112). Although the Vietnamese system for time marking is different from European languages such as Norwegian, where each sentence must express temporal reference morphologically, an important similarity relation between the Vietnamese temporal/aspectual markers and the resultative function of present perfect in Norwegian was revealed by Tenfjord (1995, 1997). This is a crucial point in her work, which provided the impetus for the current analysis of đã and rôi. Tenfjord argued that ‘the perfect in Norwegian may be looked upon as an aspectual category … the so-called past tense markers in Vietnamese mark anteriority and function in a way similar to the perfect in Norwegian’ (1995: 236). Tenfjord’s claims were rooted in a contrastive analysis, and not in empirical data. However, her claim will be explored in the current study based on the Vietnamese translations of TPQ. There are 60 contexts for the Norwegian present perfect in the questionnaire. Of these, đã and/or rôi appear 42 times. However, đã occurs with a higher frequency than rôi (30 versus 20). There seem to be three factors explaining why neither đã nor rôi occurs in the 18 remaining contexts for the Norwegian perfect as exemplified in the sentences14 from the questionnaire in Table 3.1.15 Table 3.1 Vietnamese translations without d¯ã and/or rôi in TPQ Nr. 4

Question: You MEET my sister (at any time in your life up to now)? Ha-r du have-PRS AUX you

Nr. 14

noen gang ever

mø-tt meet- PST PTCP

min my

søster? sister?

Ba.n gæ.p chi. gái cûa tôi chÜa you meet older sister of I not yet [It is morning. A wakes up, looks out of the window and sees that the courtyard (or the streets) is wet.] A: It RAIN during the night. Det it

ha-r has-PRS AUX

regne-t rain – PST PTCP

i natt last night

The ‘Per fec t Candidate’ for Transfer

41

Table 3.1 (Continued)

tÒi Nr. 56

hôm quá tÓt mÜa night day over good rain [A has just seen the king arrive. The event is totally unexpected.] A. The king ARRIVE! Kongen the king

ha-r has-PRS AUX

Ông vua sir king

t§i arrive

kome-tt arrive- PST PTCP

kìa there

Firstly, the presence of a time marker would be redundant in sentence 4 because the sentence is a question formulated with the negative chÜa, which means ‘not yet’. According to Ngo and Tran (2001), interrogatives usually refer to the past when they are placed at the end of sentences. Hence, chÜa adds temporal information to the discourse since it expresses that something has not happened yet. All the Vietnamese translations that contain this negative marker chÜa, as well as the present perfect in the Norwegian translation of the same sentence, never include đã and rôi.16 The presence of a time marker is also redundant in sentences such as sentence 14 because the presence of an adverbial or lexical expression clarifies the temporal frame. Finally, đã and rôi do not appear in sentences such as sentence 56, where it is clear from the context that the speaker is describing a present time situation: the king is (unexpectedly) coming in the moment of speaking. Table 3.2 shows some of the contexts for the Norwegian present perfect in the perfect questionnaire in which either đã and/or rôi occur. Table 3.2 Vietnamese translations with đã and/or rôi in TPQ Nr. 2

[A: It seems that your sister never finishes books.] B: (That is not quite true.) She READ this book (=all of it). Hun ha-r les-t denne she has-PRS AUX read- PST PTCP this

Nr. 3

Vâng có chÎ âÿ Çã džc yes exist sister that TM read [Question: Is the king still alive?] No, he DIE. Nei no

han he

không no sir

er be.PRS

ông Ãy that

boken book

quy‹n sách này CLF book this

død dead Çã TM

ch‰t die

rÒi. TM (Continued)

42

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 3.2 (Continued) Nr. 30

Nr. 30 [A: Don’t talk so loud! You’ll wake the baby.] Answer: He WAKE up already. Han ha-r allerede våkne-t he has-PRS AUX already wake up- PST PTCP

Nr. 36

Nó thÙc dÆy rÒi he wake up TM [A has been talking to B about C’s personal tastes.] B: You MEET her (sometime) as you know all that? A: Yes, I MEET her, so I know Ja jeg ha-r yes I have-PRS AUX

Nr. 49

Nr. 59

mø-tt henne så jeg kjenn-er henne meet- PST PTCP her so I know-PRS her

Vâng tôi g¥p cô ta rÒi yes I meet she TM [A is still living in this town] I LIVE here for seven years.

nên so

tôi I

biet know

Jeg ha-r bo-dd her i syv år I have-PRS AUX live-PST PTCP here for seven years t`ôi sông ªÇây ÇÜ®c bay næm rÒi. I live here for seven year TM A comes from the kitchen where he has just seen the sad remains of the cake. He tells B what he assumes happened.] The dog EAT our cake! Hunden ha-r the dog have-PRS AUX Con chò an dog eat

spis-t kaka eat-PST PTCP the cake

bành cûa cake of

chûng tôi we

vår our

rÒi TM

¯Dã and rôi appear in various types of contexts. However, in the majority of cases (36 of the total 42 cases), đã and/or rôi occur in a context which the present study classifies as PP (nr. 2, nr. 3, nr. 30 and nr. 36). Of these 36 contexts, 20 of them correspond to current relevance perfect/resultative perfect (e.g. nr. 2) and 16 are contexts for the experiential perfect (e.g. nr. 36). ¯Dã and rôi appear in contexts that involve SP in Norwegian in only five cases. This observation can be made in sentence 49 where the situation described still holds, as well as in sentence 59 where rôi seems to have an evidential function. As for the translations of the 70 sentences on the questionnaire where the preterite is the appropriate form in the Norwegian translation, the past time marker đã is used by one of the Vietnamese informants once out of all the 70 contexts. However, the usage of đã in these contexts has little to do with temporality. Rather, đã is used in order to emphasize something, as in sentence 24 in Table 3.3.

The ‘Per fec t Candidate’ for Transfer

43

Table 3.3 The Vietnamese translation of sentence 24 in TPQ Nr. 24

[Question: Do you know what remarkable event TAKE PLACE in 1550?] Answer: In that year, our town BE FOUNDED I det året in that year Vào in

ble byen become.PRT the town

næm Çó year that

vår grunnlagt our found

thành phÓ chúng tôi town many I

Çã

ÇÜ thành lÆp get found

The Vietnamese translation data provides information that is important for the present transfer study. Firstly, empirical support for Tenfjord’s (1997) prediction that there exists a parallel between the system of encoding past time in Norwegian and the use of temporal/aspectual markers in Vietnamese is added by the findings. Secondly, additional information about the semantic overlap between the Norwegian present perfect and đã and rôi in Vietnamese is also provided by the analysis. Tenfjord’s (1997) assumption concerned resultative perfect. However, the Vietnamese translation data show that đã and rôi frequently appear in contexts with the experiential perfect as well. In fact, it is suggested by the present analysis that it is completeness that governs the distribution of đã and rôi. Completeness is a common semantic point between the current relevance perfect/resultative perfect and the experiential perfect, and in 86%17 of the 42 contexts for the Norwegian perfect in which đã and/or rôi appear, the situation described refers to a completed action in the past. Finally, the analysis also clearly shows that there exists a zero contrast between Vietnamese and what the preterite in Norwegian grammaticalizes: purely past reference. When past time is marked by means of time markers in the Vietnamese translation data, it takes place in contexts describing a completed situation in the past.

Past Time in Somali Somali has a very complex grammar, and the basic structure for most suffixed verbs in Somali is this: [ROOT + LEX + AGR + INFL]. The morphological analysis of the Somali sentence wày keentay (‘She brought (it)’) in (5) below illustrates the basic structure18: 5) way waa=ay she.DM=CPRO.3.SG.F

keentay keen-t-ay bring-3.SG.F-PST SIMPLE

Indeed, Somali has a very complicated morphology compared to Norwegian. Somali verbs carry information about tense, aspect and mood. This information is expressed mainly by affixes, but also by vowel alternations and accentual patterns. According to Saeed (1993: 85), three tenses, three aspects and six moods can be identified in Somali. There are three different categories that encode the notion of pastness (Saeed, 1993): past

44

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

simple, past progressive and past habitual. The past simple in Somali is used for completed actions in the past, the past progressive denotes actions in process in the past, and the past habitual expresses repeated or habitual actions in the past (1993: 74–85). The perfect does not exist as a grammatical category in Somali. According to Saeed (1993: 77), ‘the past-simple in Somali includes both the ‘still-going on’ meaning of the English present perfect, and the ‘complete’ meaning of the English simple past’. This implies that the past simple in Somali will cover some of the semantic aspects of the Norwegian present perfect category, and most likely content encoded in the prototypical perfect because it expresses completed situations in the past. Table 3.4 shows how the 60 contexts for the Norwegian present perfect are encoded in Somali. As expected, the past simple in Somali will, in the majority of cases, be used in contexts where Norwegian native speakers use the present perfect, but contexts for the Norwegian present perfect are coded by means of other past forms as well. Table 3.4 Somali translations of sentences in TPQ Nr. 1

[A: I want to give your sister a book to read, but I don’t know which one. Are there any of these books that she READ already?] B: Yes, she READ this book. Ja hun yes she

Nr. 2

ha-r les-t denne has-PRS AUX read- PST PTCP this

boken book

Haa, iyadu way akhrid-ay buugan yes she DM read-PST SIMPLE book [A: It seems that your sister never finishes books.] B: (That is not quite true.) She READ this book (=all of it). Hun ha-r les-t denne boken she has-PRS AUX read- PST PTCP this book

Nr. 4

Iyadu way akhrid-ay buugan yes she read-PST SIMPLE book Question: You MEET my sister (at any time in your life up to now)? Ha-r du noen gang have-PRS AUX you ever

Nr. 6

mø-tt meet- PST PTCP

Adigu ma la kulant-aa walaashay? you Q with meet-PRS GEN my sister [Question: Do you know my sister?] Answer: Yes, I MEET her (so I know her). Ja Yes

jeg ha-r I have-PRS AUX

Haa, iyadda waan yes she DM

mø-tt henne meet- PST PTCP her

la kulm-ay with meet-PST SIMPLE

min søster? my sister?

The ‘Per fec t Candidate’ for Transfer

45

Table 3.4 (Continued) Nr. 7

[Can you swim in this lake? (=Is it possible for anybody to swim in this lake?)] Answer: Yes, at least I SWIM in it several times. Ja jeg ha-r yes I have-PRS AUX

Nr. 48

Haa, uguu jaraan dhowr jeer waan ku dabaash-ay yes at least several times DM in swim-PST SIMPLE [She is still watching television! How long she DO that?] Answer: she WATCH (it) for three hours. Hun she

Nr. 49

ha-r have-PRS AUX

se-tt på TV watch- PST PTCP on TV

ha-r have-PRS AUX

bo-dd live- PST PTCP

her here

timar hours

i for

syv år seven years

Anigu, halkan waxaan ku nool-a toddoba sanadood I here DM live-PRS GEN seven years [A is visiting a town she used to live in several years ago; now she lives somewhere else.] A. I LIVE here, so I know every street here. Jeg ha-r bo-dd her, I have-PRS AUX live-PST PTCP here

Nr. 63

i tre for three

Saddex saacadood ayey daawan-ayse three hours she watch-PRS PROG [A is still living in this town] I LIVE here for seven years. Jeg I

Nr. 51

i det minste svøm-t her flere ganger at least swim- PST PTCP here several times

så so

jeg kjenn-er hver eneste gate I know-PRS every street

Halkan, aadan waan u kala aqaan-aa wayo waan ku noolan jirey here every street DM know-PRS GEN because DM live-PST HAB [A tells what she has heard from her father. Nothing shows that she would believe it.] A. My father TELL me that when he BE a child, schools BE better than nowadays. Min far ha-r My father have-PRS AUX han var liten, he be.PRT young abahay father my marki he

fortalt tell.PST.PTCP

var skolen be.PRT the school

meg me

at da that when

bedre enn i dag better than today

wuxuu iiga sheeke-eyey in DM me to about tell-PST PROG that

uu CPRO

yaraa, be.child.PST SIMPLE

wanaagsan-aayeen be-PST SIMPLE

iskuuladu schools

ka better

siday manta yi-hiin the way today be-PRES GEN

46

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

In 77% (46 occurrences) of the sentences with a Norwegian present perfect, the past simple occurs in the Somali translation (sentences 1, 2, 6 and 7). These are prototypical contexts for the Norwegian present perfect. In seven sentences, the perfect contexts are translated by means of the present general in Somali (e.g. sentences 4 and 49). The latter example is interesting. The past and the present are combined by the temporal frame given in sentence 49. Since Saeed remarks that the past simple in Somali also includes the ‘still-going on’ meaning of the English present perfect, we should expect the Somali past simple to occur here as well, which it does not. Finally, in the remaining seven sentences in which a present perfect form is required in Norwegian, there are two occurrences of the past progressive in the Somali translation (e.g. nr. 63): one of the past habitual (sentence 51), and one occurrence of the present progressive (sentence 48). The analysis of the Somali translation data shows that the prototypical perfect in Norwegian must, in the great majority of cases, be rendered by the simple past in Somali, which is a general past. However, in some contexts for the Norwegian perfect form, usually in secondary contexts, the temporal information must be encoded by means of other forms in Somali, such as a present general form, a present or past progressive form, or a past habitual form.

Hypotheses The hypotheses specify how the effects of L1 influence are predicted to emerge in the texts written by the learners in the Vietnamese learner group and in the Somali learner group. These are based on previous findings and the L1-L2 contrastive relations involved. Several of the studies referred to in the literature review not only point to L1 influence, but also to the fact that the perfect category is subject to influence from the tense and aspect system in the learner’s L1 in many cases, and moreover, that this transfer is particularly evident in the encoding of perfect contexts and often shows up as overuse of the perfect in preterite contexts. The findings from the contrastive analysis further warrant the specific hypotheses in the current study. Firstly, the Somali language does not encode the distinction that exists between the preterite and the present perfect in Norwegian. Contexts that require the present perfect in Norwegian are usually rendered through the past simple in Somali. In particular, the past simple in Somali encodes the content encoded in the PP in Norwegian. In order to use the Norwegian past forms correctly, Somali speakers have to notice that the Norwegian preterite does not cover all the functions of the past simple in Somali. They must also notice that they need an additional category, the Norwegian present perfect, to express the functions not covered by the Norwegian preterite. Hence, it can be expected that the Somali learners will have difficulties in the distribution of the two forms in accordance with the target language rules. By comparison with Somali, the contrastive data from Vietnamese point to the existence of a semantic parallel between the present

The ‘Per fec t Candidate’ for Transfer

47

perfect in Norwegian, and the use of the time markers đã and rôi in Vietnamese. Whereas the translation data indicate that contexts for the preterite in Norwegian will not be marked by a time marker in Vietnamese, in many cases, when the present perfect in Norwegian describes a situation completed in the past, the presence of đã or rôi, or both, is required in a Vietnamese translation. To sum up, the following are the hypotheses to be tested in the present study: (1) The Vietnamese-speaking learners will use the present perfect correctly more frequently than the Somali-speaking learners in contexts in which a completed situation in the past is described. (2) The Somali-speaking learners will have a higher degree of incorrect use of the preterite than the Vietnamese-speaking learners in contexts where Norwegian requires the present perfect, and will also show a higher degree of incorrect use of the present perfect in preterite contexts.

Coding procedure Two types of analysis were required in order to test the hypotheses: an analysis of the learners’ correct encoding of preterite contexts and contexts for the present perfect where a completed situation in the past was referred to (PP), as well as an analysis of the learners’ incorrect encoding of preterite and present perfect contexts of all types. Hence, for the purpose of performing these analyses, each of the clauses that comprise the individual texts were analysed and coded as described below. Firstly, a context analysis, which generated the identification of obligatory contexts for use of past morphology in Norwegian, was conducted based on internal properties of the clause (e.g. time expressions) as well as through contextual inference based on the parts of the discourse in which the clause occurs. These 161 texts generated 1934 obligatory contexts for the preterite and 235 obligatory contexts for present perfect. Next, the learners’ morphological marking of the temporal content in the clauses was analysed. These grammatically encoded clauses were coded as either present, preterite, present perfect or past perfect depending on which of these forms occurred in the clauses. Furthermore, clauses were coded as correct if the temporal morphological form used in a clause (preterite or present perfect) encoded the identified past context. In cases with no correspondence between the past context identified in the clause (e.g. a preterite context), and the temporal morphological form that occurs in the clause (e.g. a present tense form), the clause was coded as incorrect encoding, and the type of error distribution was registered. The analysis of the grammatical encoding in the clauses generated 1712 correct uses and 222 incorrect uses of the preterite, and 178 correct uses and 57 incorrect uses of the present perfect. The uses of the present perfect were also coded as PP or SP in accordance with the principles accounted for in the Contrastive section.

48

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Statistics For testing differences between the L1 groups, the Mann-Whitney U test is used, a standard non-parametric test for identifying differences in the group medians of two independent groups. If a significant result is detected, additional steps will be taken. This is because the distributions for the variables measuring correctness rates are highly left-skewed because many texts obtain a 100% score, while the rest are scattered from 0–99%. However, the distributions tend to take the reversed form for some variables measuring error rates. Hence, if a significant difference is detected, and more than 30% of the texts obtain a 100% score or a 0% score, a post hoc chi-square test will be conducted. The chi-square test tests if the overall significance revealed in the initial testing is due to a significant difference between the groups in the proportion of texts having a 100% value or a 0% value for the quality measured. Effect sizes will also be reported. The effect sizes for the Mann-Whitney U test and the Wilcoxon signed rank test are indicated by r, which is calculated from the Z statistics from the U test by dividing the Z statistics by the square root of N (Larson-Hall, 2010: 377). For the chi-square test on a 2 × 2 table, Cramer’s V is used as a measure of effect size (Larson-Hall, 2010: 237).

Results Analysis of correct encoding The proportion of correct use of PP19 is reported by Table 3.5, which is perfect in (obligatory) contexts for completed situations in the past. For instance, it can be ascertained from Table 3.5 that there are 1.9 contexts for the PP in the Vietnamese texts, on average, and that the Vietnamese learners inflect these correctly in 90.4% of the cases (mean). However, if the median is studied, it can be seen that 100% is the middle value in the data set, and from the last row, it is found that 26 out of the 51 Vietnamese texts inflect contexts for the PP in Norwegian correctly in all cases. The gap between the mean and the median is an indication that the distribution is skewed. It is important to keep in mind that a feature that does not occur very often in the data set is not only being compared now, but that proportions in two rather small groups are also being compared. A notable difference is seen in correct uses of the present perfect in the predicted direction: Prototypical perfect is used correctly more frequently in the Vietnamese group (mean 90.4%) than in the Somali group (mean 80.8%), and the highest proportion of texts with 100% correct use of PP is found in the Vietnamese group (26 versus 21, or 51% versus 36%). However, a marginally significant result for the difference in medians between the groups (U = 439.0, z = −1.687, p = 0.09) and a small effect size (r = 0.2) are revealed by a Mann-Whitney U test.

The ‘Per fec t Candidate’ for Transfer

49

Table 3.5 Comparison of the L1 groups’ proportions of correct use of prototypical perfect (PP) Correct use of PP

Vietnamese Mean Median Std.d. Texts with 100%/texts total Somali Mean Median Std.d. Texts with 100%/texts total

No. of contexts for the present perfect

% of correct uses of PP

1.9 2.0 1.1

90.4 100.0 23.4 26/51

2.2 2.0 1.7

80.8 100.0 29.2 21/58

A broader picture of the correctness rates in past contexts in the Vietnamese and Somali texts can be found in Table 3.6, which reports the proportion of correct uses of the present perfect and the preterite in all contexts. Table 3.6 Comparison of the L1 groups’ proportions of correct uses of perfect and preterite contexts

Vietnamese Mean Median Std.d. Texts with 100%/texts total Somali Mean Median Std.d. Texts with 100%/texts total

Correct use of present perfect

Correct use of preterite

No. of contexts

% of correct uses

No. of contexts

% of correct uses

1.8 1.0 1.1

83.4 100.0 33.3 40/53

13.6 6.0 13.1

87.1 100.0 21.5 32/63

2.1 1.0 1.6

72.5 100.0 39.9 42/66

15.2 13.0 12.6

85.1 93.3 18.4 24/71

It can be seen from Table 3.6 that a pattern similar to the one observed in Table 3.5, which presented the proportion of correct encoding of the contexts for PP, is found. Also, the proportion of texts with 100% correct use of the present perfect is higher in the Vietnamese group (40 out of 53) than in the Somali group (42 out of 66). A one-tailed Mann Whitney U test finds that

50

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

this difference is not significant (U = 1514.5, z = −1.533, p = 0.1) and that the effect size is small (r = 0.1). As for preterite contexts, the percentage of correct uses of the preterite by both groups is rather similar (87.1 and 85.1), and the difference between the proportions of correct uses in preterite contexts is also of no significance (U = 1904.5, z = −1.537, p = 0.1, small effect size, r = 0.1).

Analysis of incorrect encoding The next step is to analyse L1 differences in incorrect encoding of past contexts. This part of the analysis concerns those texts with clauses in which the temporal context was not encoded correctly due to the use of a preterite or a perfect in an inappropriate context (see the section describing the coding procedures). The two types of error distribution are presented in Table 3.7. Table 3.7 Proportions of preterite and perfect forms incorrectly used in place of each other

Vietnamese Mean Median Std.d. Min-Max Texts with 0%/ texts total Somali Mean Median Std.d. Min-Max Texts with 0%/ texts total

Preterite incorrectly occurring in a prs. perfect context

Prs. perfect incorrectly occurring in a preterite context

No. of forms

% incorrect uses

No. of forms

% incorrect uses

12.5 6.0 13.1 1–51

1.8 0.0 8.9 0.0–50.0 64/67

1.9 2.0 1.1 1-5

4.1 0.0 13.4 0.0–66.7 46/51

13.2 10.0 11.2 1–46

8.2 0.0 21.0 0.0–100.0 58/77

2.2 0.0 1.7 1–10

10.3 0.0 23.1 0.0–100.0 45/58

A systematic difference between the two L1 groups in their error distribution is indicated by Table 3.7. Firstly, the incorrect use of the preterite in perfect contexts appears mainly in Somali texts. This type of error is found in only three Vietnamese texts (64 of 67 texts have no occurrences), but is found in 19 Somali texts. This difference is significant (U = 2066.0, z = −3.285, p = 0.0005) and the effect size is medium (r = 0.3). Because more than 30% of the texts obtain a 0% score (see the Statistics section), a post hoc chi-square test was performed showing that there is a significantly higher proportion of Somali texts with this particular incorrect distribution (χ² = 11.291, p = 0.001),

The ‘Per fec t Candidate’ for Transfer

51

compared to Vietnamese texts, and that the effect size of this difference is medium (Cramer’s V = 0.3). Secondly, the proportion of perfect forms occurring incorrectly in contexts for the preterite is also higher in the Somali group. This difference was tested for significance by means of the same stepwise approach, which produced a significant result (U = 1292.5, z = −1.752, p = 0.04 with a small effect size, r = 0.2). A significantly higher proportion of Somali texts than Vietnamese texts were shown by a post hoc chi-square test to exhibit an incorrect distribution of the present perfect in preterite contexts (χ² = 3.130, p = 0.04), and that the size of this effect is small (Cramer’s V = 0.2). This statistically significant pattern of incorrect encoding is illustrated by Figure 3.1. The bars represent individual texts, and the y-axis refers to the number of misuses of the present perfect and the preterite in place of each other per text. Clearly, the encoding of past time in Somali texts is characterized by this type of error. The grey bars, representing Somali texts, dominate (31 out of total 39 texts): 5 4 3

Vietnamese

2

Somali

1 0

Figure 3.1 Texts with incorrect uses of the preterite and the present perfect in place of each other by L1

The incorrect use of the prototypical present perfect is scrutinized by the final analysis and an L1 difference seems to be present in Table 3.8: Table 3.8 Comparison of the L1 groups’ proportions of incorrect use of PP

Vietnamese Mean Median Std.d. Texts with 0%/texts total Somali Mean Median Std.d. Texts with 0%/texts total

No. of perfect forms

% of incorrect uses of PP

1.9 2.0 1.1

9.6 0.0 23.3 26/51

2.2 2.0 1.7

19.2 0.0 29.2 21/58

52

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Very different trends between the two groups in their incorrect encoding of PP are shown by Table 3.8. The Somali informants use the PP (19.2) incorrectly more often than the Vietnamese do (9.6). The dispersion is vast, and Table 3.9 provides information about the individual variation: Table 3.9 Comparison of the L1 groups’ incorrect PP use in relation to the number of texts with different proportions of errors P used incorrectly

0.0 ≤25.0 ≤50.0 ≤75.0 ≤99.0 100.0 total N

Frequency of texts Vi (N = 51)

So (N = 58)

45 1 3 1 0 1 51

45 3 7 1 0 2 58

It can be seen from Table 3.9 that in the Somali group the number of texts with incorrect use of the PP (13) is larger than the number of texts with incorrect use of the PP (6) in the Vietnamese group. This suggests an L1-related difference in the frequency of incorrect use of the PP. The observed tendency for Somali texts to have more incorrect use of the PP than Vietnamese texts was confirmed to be marginally significant by means of a two-tailed Mann-Whitney U test (U = 439.0, z = −1.687, p = 0.09) and the effect size was small (r = 0.2).

Summing up: Outcome of the hypothesis predicting L1 effects Hypothesis 1 was not supported. The evidence for the present perfect in contexts describing completed situations in the past being used correctly more frequently in Vietnamese texts than in Somali texts was simply not solid enough because only a marginally significant differences and small effect size were detected. The analysis of correct encoding of all types of present perfect contexts also showed that the Vietnamese learners did not use the present perfect in Norwegian correctly more often than the Somali learners. However, Hypothesis 2 was confirmed. The analysis of the incorrect encoding revealed systematic differences between the L1 groups. The Somali-speaking learners used the preterite incorrectly in present perfect contexts more often than the Vietnamese-speaking learners did, and the difference was significant at a very robust alpha level (p = 0.0005) and with a medium effect size (Cramer’s V = 0.3). As for the incorrect use of the present perfect in preterite contexts, the frequency of this type of incorrect distribution was also higher in the

The ‘Per fec t Candidate’ for Transfer

53

Somali group. However, this was a slightly weaker result than the difference in incorrect use of the preterite in present perfect contexts (p = 0.04, r = 0.2). The analyses of incorrect uses of the present perfect in prototypical contexts indicated that contexts describing a completed situation in the past are more challenging for the Somali-speaking learners than for the Vietnamese-speaking learners (marginally significant difference, p = 0.09, and a small effect size, r = 0.2). These results will be discussed in the following section.

Discussion The ‘perfect candidate’ for transfer It seems that the L1 difference in the encoding of past time revealed in the current study is not so much about correctness, but instead constitutes a difference in distributional patterns that emerge as a particular type of incorrect encoding. The analysis supports one of the specific hypotheses for transfer: Somali-speaking learners use the preterite and the present perfect incorrectly in place of each other more often than Vietnamese-speaking learners do. In particular, the Somali-speaking learners’ overuse of the preterite in Norwegian is solidly documented. However, even though the Somali learners have more difficulty distinguishing the preterite from the present perfect, we do not have evidence for claiming that the Vietnamese-speaking learners overall use the present perfect category more successfully than the Somali learners do. Surely, the detected L1 effects in the current study are first and foremost a matter of tense-marking errors: the Somali learners, whose L1 does not grammaticalize the content of the present perfect category, have more problems distinguishing the preterite from the present perfect form. The clearest proof of this is the high frequency of incorrect use of the preterite in present perfect contexts. This result aligns with previous findings presented earlier from studies of L2 Norwegian, L2 English and L2 Swedish of learners of different L1 backgrounds (Somali, Turkish, French, Russian, Polish), that show that the present perfect category poses challenges when L2 learners are not familiar with the category from their L1 or when there are formal or conceptual differences or similarities between a tense-aspect category in the L1 and the perfect category in the L2.

Types of transfer The two L1s in question in the current study are both very different from the Norwegian language, but they are distinguished from Norwegian in very different ways. Somali is a highly synthetic language with rich verb morphology. Yet, the semantic distinction that exists between the preterite and the present perfect in Norwegian is not encoded in Somali because a perfect category does not exist. In contrast, Vietnamese verbs are tenseless,

54

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

and temporal reference is not marked linguistically because it is accessible through the basic time of the contexts. However, the Vietnamese translations of the perfect questionnaire empirically document the existence of a semantic parallel between temporal/aspectual markers in Vietnamese and the present perfect category in Norwegian. In sum, as a consequence of these very different L1-L2 relations, Vietnamese-speaking and Somalispeaking learners of Norwegian exhibit different types of linguistic and conceptual knowledge when learning Norwegian, which in turn indicates that different types of transfer process could be going on in the Vietnamese and Somali learners. This can help explain why Vietnamese-speaking and Somali-speaking learners have a different degree of success with the present perfect in Norwegian. Starting with the Vietnamese transfer, this is not a clear-cut example of linguistic transfer, such as Collins’ (2002, 2004) and Karrer’s (1999) studies for instance, where the learners will obviously find a target language structure (the present perfect in English and in Norwegian) that is formally similar to an L1 structure (the passé composé in French and the present perfect in German). Another possible interpretation of the process taking place could be suggested: conceptual transfer. This reasoning holds another type of understanding of what the identified similarity relation between Norwegian and Vietnamese encompasses. Namely, that this is an example of a similarity in conceptual distinctions underlying the linguistic encoding which can be found between languages. The distinction grammaticalized through the present perfect in Norwegian usually provides information about a present result or a present state of affairs. Accordingly, it can be argued that the conceptualization of time in Norwegian entails two fundamental distinctions: the past-nonpast distinction grammaticalized in the preterite–present opposition, and the past-present result distinction grammaticalized in the contrast between the preterite and the present perfect. According to Jarvis and Pavlenko, speakers develop sensitivity towards conceptual distinctions that are encoded in the language they are exposed to through the language socialization process (2008: 73). Norwegians learning their L1 are tuned to different distinctions than are native speakers of languages with other sets of grammatical categories, one of them being the fundamental distinction between pure reference to the past (the preterite) and reference to a present state resulting from a past event (the present perfect). In this case, despite the enormous contrast between the Norwegian language system and the Vietnamese language system, there would be a similarity in conceptualization between the languages in that both languages direct their speakers to pay attention to the specific distinction in time accounted for here, which in Norwegian requires the use of a present perfect tense form, and which in Vietnamese requires the use of đã and rôi. Accordingly, the Vietnamese speakers are possibly supported by their L1 when acquiring the present perfect form in Norwegian. Remember that the Vietnamese speakers have less

The ‘Per fec t Candidate’ for Transfer

55

trouble using the present perfect correctly in prototypical contexts, a result which aligns with the reasoning given here because the present perfect in Norwegian has the closest parallel to đã and rôi in Vietnamese. A reasonable interpretation of the L1 effects on the encoding of present perfect contexts in the Somali group is that the Somali learners overuse the preterite in present perfect contexts because they make intralingual identifications between the Somali general past and the Norwegian preterite. This is also evident in the analysis of the incorrect use of the prototypical perfect, which is marginally and significantly more prevalent in the Somali data than in the Vietnamese data. The Somali-speaking learners have more problems with PP because this type of perfect is often used to code contexts that would be candidates for the Somali general past, a form which resembles the preterite in Norwegian. This result may serve as an indication that the Somali learners are making interlingual identifications between the preterite in Norwegian and the general past in Somali because a similarity relation exists. The Somali learners probably understand at some level that there exists an underlying similarity between these languages in that both languages, by means of different types of verb inflection, supply information about when the situation referred to took place, and moreover, that this information is something which all sentences require. However, this correspondence between Norwegian and Somali exists at a rather abstract level, and according to Ringbom’s (2007) types of contrastive relations,20 it is probably at a level that is too abstract for the learner to even reflect upon consciously and be able to exploit in learning the L2: The zero relation does not mean that the learner finds nothing at all that is relevant to L1 as the learning progresses. There are, after all, some linguistic universals common to all languages. But the level of abstraction is too high that an average learner cannot easily notice features that a totally different TL [target language] has in common with L1. The zero relation merely means that items and patterns in the TL at early stages of learning appear to have little or no perceptible relations to the L1 or any other language the learner knows. The learner’s L1 may lack the concepts necessary to perceive fundamental distinctions in the TL. (Ringbom, 2007: 6, author’s own emphasis added) If we consider the fact that a perfect category is missing in Somali, the last line in bold in Ringbom’s quote is particularly interesting regarding the possibility of describing the relation between Norwegian and Somali encoding of time as a zero relation. This opens for interpreting the transfer effects in the Somali data as conceptual instead of linguistic, or possibly both. It could be argued that Somali speakers are not socialized through their L1 to pay attention to this particular distinction when encoding events. This possibly makes them less sensitive than Vietnamese learners of Norwegian to

56

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

the distinction captured in the Norwegian present perfect. To put it differently, at this particular point in the encoding of time in Norwegian, the Somali learners might face a greater challenge in restructuring their conceptualization of time than the Vietnamese learners, who have probably developed a stronger awareness of this particular distinction, which enables them to encode present perfect contexts more easily and maybe also earlier than Somali learners do. This interpretation of the difference in incorrect use of the preterite and encoding of present perfect contexts as a matter of conceptual transfer is supported by the new theoretical accounts of the relation between cognition, language and crosslinguistic influence as well as by findings generated from studies within these new frameworks (e.g. von Stutterheim & Carroll, 2006). The L1 effects documented in the Somali texts can be understood in the framework of Slobin’s thinking for speaking hypothesis as well as Jarvis and Pavlenko’s perspective on conceptual transfer because the L1-L2 differences analysed concern not only structural differences, but also conceptual differences that underlie the forms and structures. In addition, temporal distinctions are abstract relations and not visible distinctions that can be observed by the eye. According to Slobin (1993, 1996), Kellerman (1995), and the more recent theoreticians exploring the relation between language and cognition, linguistic categories coding such notions are the ones that learners typically have a hard time restructuring when learning a new and different language. Furthermore, Jarvis and Pavlenko consider ‘L2 learners’ failure to mark temporality in accordance with the language-specific temporal system of the target language’ (2008: 142) as one of the possible outcomes of transfer originating at the conceptual level. In that case, if we bring in Jarvis’s (2000) perspective that conceptual transfer results from something very different than a process of transference based on intralingual identifications between L1 items and L2 items, the transfer effects observed in the Somali texts in the current study may also be interpreted as a matter of L1 influence arising as an inert outcome.

Conclusion As cited in the introduction, Bardovi-Harlig (2000) did not find strong evidence for claiming the L1 to be an important variable in the L2 acquisition of temporality. The principal reason for this claim is believed by this study to be found in how studies of temporality have typically been designed and carried out: they have largely been designed to reveal common paths in language development, and the learner’s L1 has mostly been treated as a secondary variable. It is my view that the current study has shown that in order for transfer effects to be identified, the analysis must rely on systematic investigations of group internal and group external behaviour in transfer studies as well as detailed empirically validated analyses of L1-L2 relations. These are

The ‘Per fec t Candidate’ for Transfer

57

requirements that are captured in Jarvis’s (2000) framework for transfer studies, and which the current transfer study has used. However, BardoviHarlig did add that transfer effects might be found ‘in the details’, and it is believed that this is what is actually demonstrated in the current study. The transfer effects observed emerge from a close scrutiny of how the different temporal contexts are encoded in the learners’ interlanguages, and not merely from a comparison of correctness rates between the L1 groups. The present study has demonstrated L1-specific patterns in learners’ use of the present perfect and preterite in L2 Norwegian. From these patterns, which have been documented in previous studies as well, the perfect rises as a ‘perfect candidate’ for the study of transfer in L2 acquisition of temporal morphology.

Notes (1) The European Science Foundation Project (the ESF project) studied language acquisition in adult immigrants living in Western Europe. One of the research teams investigated temporality in a second language (Dietrich et al., 1995). (2) According to Bardovi-Harlig (2000), one can distinguish between two approaches to the acquisition of temporal expressions: the meaning-oriented (or concept-oriented or semantically-oriented) approach and the form-oriented (including Sato’s 1990 form-tofunction studies) approach. The meaning-oriented approach investigates how a concept, e.g. pastness, is expressed through pragmatic, lexical and/or morphological devices. Form-oriented studies start at the other end and investigate the distribution of particular forms, and ask ‘how and where it is used by the learners’ (BardoviHarlig, 2000: 11). The Norwegian tense-aspect studies share features of both these approaches; however, they are characterized by their emphasis on the underlying meaning of the forms. (3) According to Bardovi-Harlig (2000: 419), findings from studies of the order of emergence of verb morphology in several Germanic L2s (Swedish, Dutch, English, German, French, Spanish and Italian) lay the basis for posing the following acquisitional sequences for tense and aspect morphology: present (default form) – simple past (the past – nonpast distinction is acquired first) – present perfect – past perfect. (4) Bardovi-Harlig attributes this observation to the morphosyntactic complexity of the perfect category (Bardovi-Harlig, 2000: 180–182). (5) This does not apply to the extended now perfect in Norwegian as in Jeg har bodd i Bergen i syv år (‘I have lived in Bergen for seven years’). According to Randen, this use of the present perfect is more related to imperfective aspect (Randen, 1999: 68). (6) 20 of the 51 Turkish informants in Moskvil’s study did not pass the test. The texts in Helland (2005) and Janik (2010) were all written by informants who passed the test. However, Moskvil’s analysis did not reveal important differences in error distribution between the fail-texts and the pass-texts (Moskvil, 2004: 103). (7) See Gujord (2013: 145–146) for detailed information about the distribution of the 161 informants across various categories related to their personal backgrounds. (8) The texts are also linked to the Common European Framework of Reference for Languages, CEFR. However, this level placement is not taken into account here because the CEFR level does not influence the results presented in this chapter (Gujord, 2013: 172). (9) The term situation is used in the same manner as Comrie (1985: 5): as a general term for events, processes and states. (10) The historical sources of the perfect in many European languages, including Norwegian, are two resultative constructions (Bybee & Dahl, 1989; Lindstedt,

58

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

2000): a possessive construction have plus a participle, as exemplified in Norwegian in (a), and a copula construction be plus the participle of the main verb (b): (a) Ek hefi brefit skrifit > Jeg har skrevet brevet (I have written the letter) (b) Gestirnir eru farnir > Gjestene er reist (The guests are gone) (11) The past perfect form is excluded from the investigation because it occurs very infrequently in the data (only 44 contexts and 35 uses of the form). (12) ‘The experiential perfect indicates that a given situation has held at least once during some time in the past leading up to the present’ Comrie (1976: 59). (13) The basic time is ‘the time which has been made clear in the context up to that point’ (Thompson, 1965: 209). (14) The numbering of the sentences in Table 3.1 corresponds to the original numbering in the perfect questionnaire available in the appendix in Gujord (2013: 342–348), which is downloadable. (15) See Appendix for abbreviations. (16) There is only one exception. However, the informant explains that đã in that specific context is used for emphasis (Gujord, 2013: 95). (17) (20 current relevance + 16 experiential SPs = 36 out of 42) (18) Apparently, this is a rather consistent system; however, the morphological pattern in Somali is very difficult to see because of a rich system of lexical affixes that can be added to many categories, such as verbs (Saeed, 1993: 22). This affixation process leads to sound changes that make it challenging to see past the fusion of forms and the change of lexical meaning that occurs after an affix is added to the root (Saeed, 1993: 22). (19) Since PP (prototypical perfect) and SP (secondary perfect) are mutually exclusive categories, it is sufficient to analyse only one of the categories. (20) Ringbom (2007) outlines three types of relations which he believes can be found between languages: similarity relation, contrast relation and zero relation. A similarity relation exists if ‘an item or pattern in the TL is perceived as formally and/or functionally similar to a form or pattern in L1’ (Ringbom, 2007: 5). A contrast relation is found when the learner is aware that a pattern or item in the L2 differs significantly from the L1, but knows nevertheless that there are also some underlying similarities between them (Saeed, 1993: 6). A zero relation is described as existing when the learner does not assume the L1 knowledge to be significant for the learning task.

References Ayoun, D. and Salaberry, R.M. (2008) Acquisition and English tense-aspect morphology by advanced French instructed learners. Language Learning 58 (3), 555–595. Bardovi-Harlig, K. (2000) Tense and aspect in language acquistion: Form, meaning and use. Language Learning 50 (1), xi-491. Bybee, J. (1985) Morphology: a Study of the Relation between Meaning and Form. Amsterdam: Benjamins. Bybee, J.L. and Dahl, Ö. (1989) The creation of tense and aspect systems in the languages of the world. Studies in Language 13 (1), 51–103. Bybee, J., Perkins, R. and Pagliuca, W. (1994) The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Carlsen, C. (2012) Proficiency level – a fuzzy variable in computer learner corpora. Applied Linguistics 33 (1), 1–24. Collins, L. (2002) The role of L1 influence and lexical aspect in the acquistion of temporal morphology. Language Learning 52 (1), 43–94.

The ‘Per fec t Candidate’ for Transfer

59

Collins, L. (2004) The particulars on universals: A comparison of the acquisition of tenseaspect morphology among Japanese and French-speaking learners of English. Canadian Modern Language Review 61, 251–274. Comrie, B. (1976) Aspect: an Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Comrie, B. (1985) Tense. Cambridge: Cambridge University Press. Dahl, Ö. (1985) Tense and Aspect Systems. Oxford: Blackwell. Dahl, Ö. (2000) Tense and Aspect in theLanguages of Europe. Berlin: Mouton de Gruyter. Dietrich, R., Klein, W. and Noyau, C. (1995) The Acquisition of Temporality in a Second Language. Amsterdam: J. Benjamins. Golden, A., Kulbrandstad, L.I. and Tenfjord, K. (2007) Norsk andrespråksforskning – utviklingslinjer fra 1980 til 2005. Nordand 2 (1), 5–41. Golden, A. (2012) Metaphorical expressions in L2 production: The importance of text topic in corpus research. In F. MacArthur, J.L. Oncins-Martínez, M. Sánchez-García and A.M. Piquer-Píriz (eds) Metaphor in Use: Context, Culture, and Communication (pp. 135–148). Amsterdam: John Benjamins Publishing Company. Granger, S. ( 2003) The corpus approach: a common way forward for contrastive linguistics and translation studies? In S. Granger, J. Lerot and S. Petc-Tyson (eds) Corpus-based Approaches to Contrastive Linguistics and Translation Studies (pp. 17–29). Amsterdam: Rodopi. Gujord, A.K.H. (2013) Grammatical encoding of past time in L2 Norwegian. The roles of L1 influence and verb semantics, Doctoral dissertation. Faculty of Humanities, University of Bergen. (downloadable here: https://bora.uib.no/handle/1956/6867) Gujord, A.K.H. (2015) L2 acquisition of temporality: Findings from a corpus based study of the grammatical encoding of past time. BeLLS 6, 63–86. See https://bells.uib.no/ index.php/bells/article/view/809 Hammarberg, B. and Viberg, Å. (1977) The place-holder constraint, language typology, and the teaching of Swedish to immigrants. Studia Linguistica 31 (2), 106–163. Helland, A.K. (2005) I møte med eit tempusprominent språk: ei undersøking av mellomspråka til vietnamesiske norskinnlærarar Encounter with a tense prominent language: a study of Vietnamese speaking learners of Norwegian], Unpublished MA thesis. Faculty of Humanities, University of Bergen. Janik, M.O. (2010) En polakk i norsk fortid. Om feil de polskspråklige norskinnlærerne gjør i bruk av perfektum og preteritum [A Pole in the Norwegian Past. About errors produced by Polish speaking learners of Norwegian regarding perfect and preterite]. Folia Scandinavia 11, 87–100. Jarvis, S. (2000) Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning 50 (2), 245–309. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York: Routledge. Larson-Hall, J. (2010) A Guide to doing Statistics in Second Language Research using SPSS. New York, London: Routledge. Karrer, I.U. (1999) Falske venner?: en studie av preteritums-og perfektumsformene i tyske språkinnlæreres norske mellomspråk [False friends? A study of German speaking learners of Norwegian regarding the forms used in preterite and perfect tense]. Unpublished MA thesis. Faculty of Humanities, University of Bergen. Kellerman, E. (1995) Crosslinguistic influence: Transfer to nowhere? Annual Review of Applied Linguistics 15, 125–150. Klein, W. (2009) How time is encoded. In W. Klein and P. Li (eds) The Expression of Time (pp. 39–82). Berlin: Walter de Gruyter. Lindstedt, J. (2000) The perfect – aspectual, temporal and evidential. In Ö. Dahl (ed.) Tense and Aspect in the Languages of Europe (pp. 365–383). Berlin: Mouton de Gruyter. Mitchell, R., Myles, F. and Marsden, E. (2013) Second Language Learning Theories. London: Routledge.

60

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Moskvil, M.E. (2004) Temporalitet i morsmål, målspråk og mellomspråk, Unpublished MA thesis. Faculty of Humanities, University of Bergen. Ngo, B.N. and Tran, B.H. (2001) The Vietnamese language learning framework. Journal of Southeast Asian Language Teaching 10, 1–24. Philipsson, A. (2007) Interrogative clauses and verb morphology in L2 Swedish: Theoretical interpretations of grammatical development and effects of different elicitation techniques, Doctoral dissertation. Centre for Research on Bilingualism, Stockholm University. Polunenko, A. (2004) English past tense forms in Russian’s speakers oral and written production. Unpublished MA thesis. Department of Linguistics, Ohio University. Randen, G.T. (1999) Aspektualitet: en sammenlignende studie av norsk, russisk og russiske informanters norske mellomspråk. Unpublished MA thesis. Faculty of Humanities, University of Bergen. Ringbom, H. (2007) Cross-linguistic Similarity in Foreign Language Learning. Clevedon: Multilingual Matters. Saeed, J.I. (1993) Somali Reference Grammar. Kensington, Md.: Dunwoody Press. Sato, C.J. (1990) The Syntax of Conversation in Interlanguage Development. Tübingen: Gunter Narr. Shirai, Y. (2009) Temporality in first and second language acquisition. In W. Klein and P. Li (eds) The Expression of Time (pp. 167–193). Berlin: Walter de Gruyter. Slobin, D.I. (1993) Adult language acquisition: A view from child language study. In C. Perdue (ed.) Adult Language Acquisition: Cross-linguistic Perspectives (pp. 239–252). Cambridge: Cambridge University Press. Slobin, D.I. (1996) From ‘thought and language’ to ‘thinking for speaking’. In J.J. Gumperz and S.C. Levinson (eds) Rethinking Linguistic Relativity (pp. 70–96). Cambridge: Cambridge University Press. Tenfjord, K. (1995) On the acquisition of tense… In M. Kalin and S. Latomaa (eds) Nordens språk som andraspråk 3: tredje forskarsymposiet i Jyväskylä 24.-25.3.1995. Jyväskylä universitet: Högskolornas språkcentral. Tenfjord, K. (1997) Å ha en fortid på vietnamesisk: en kasusstudie av fire vietnamesiske språkinnlæreres utvikling av grammatisk fortidsreferanse og perfektum [To have a past in Vietnamese: A case study of the developement of past reference and perfect tense by four Vietnamese speaking learners], Unpublished doctoral dissertation. Faculty of Humanities, University of Bergen. Thompson, L.C. (1965) A Vietnamese Grammar. Seattle: University of Washington Press. von Stutterheim, C. and Carroll, M. (2006) The impact of grammatical temporal categories on ultimate attainment in L2 learning. In H. Byrnes, H. Weger-Guntharp and K. Sprang (eds) Educating for Advanced Foreign Language Capacities (pp. 40–53). Georgetown: Georgetown University Press.

The ‘Per fec t Candidate’ for Transfer

Appendix List of abbreviations 1 2 3 AGR ADV AUX CLF CPRO DM F FOC INF INFL LEX M ROOT PST PRS PRT PTCP PRS GEN PRS PRF PRS PROG PST HAB PST PRF PST PROG PST SIMPLE PL SG TM Q

first person second person third person agreement adverb(ial) auxiliary classifier clitical pronoun declarative marker feminine focus word infinitive inflection lexical affix masculine verb root past present preterite participle present general present perfect present progressive past habitual past perfect past progressive past simple plural singular time marker question marker

61

62

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 3.10 Overview of personal information about the informants

Gender Age group

Years of residence in Norway (before taking the test)

English skills

Educational background

Current status

Amount of L2 instruction (hours)

female male not reported less than 25 25–34 35–44 45–55 more than 55 less than 1 up to 2 up to 3 up to 4 up to 5 5 and more not reported none basic intermediate advanced not reported elementary high school higher educ. other not reported working studying applying to jobs other not reported 201–400 401–500 501–850 851–1500 1501–2000 2001–3000

Vietnamese (N = 73)

Somali (N = 88)

65 8 1 18 39 12 2 1 4 37 16 1 3 12 10 2 41 16 4 3 7 21 39 3 4 12 14 12

26 62 0 23 47 14 4 0 4 43 23 8 4 6 5 0 15 50 18 2 25 34 23 4 5 15 39 17

31 5 6 6 32 19 5 0

12 6 13 14 29 20 3 3

The ‘Per fec t Candidate’ for Transfer

63

Table 3.10 (Continued)

Duration of L2 instruction (months)

Do you practice Norwegian? (outside the classroom) Do you socialize with Norwegians?

not reported less than 6 6–12 13–24 25–36 37 or more no answer never seldom daily no answer yes no

Vietnamese (N = 73)

Somali (N = 88)

8 1 20 34 6 4 0 1 40 32 3 56 14

19 2 16 38 10 3 3 0 40 45 5 72 11

4

On How Polish Learners of Norwegian Render Spatial Prepositions in L2: A CorpusBased Study of i and på Oliwia Szyman´ska SWPS Warsaw University of Social Sciences and Humanities

Introduction The aim of this study is to investigate the way Polish learners of Norwegian render spatial prepositions in the target language. Inspired by the framework of conceptual transfer, this study builds on patterns observed during prior analyses suggesting that native speakers of Polish and Norwegian may conceptualize space slightly differently. According to Jarvis and Pavlenko (2008), space, being the fabric of our daily experience, is one of the domains that are especially resistant to conceptual restructuring. Evidence for this can certainly be found in language learners’ use of prepositions. The purpose of the present study is to examine learners’ use and distribution of two basic spatial prepositions in Norwegian, på and i, which similarly to English on and in express support and containment. These two prepositions are believed to signify the first spatial relations a human being can differentiate between (Bowerman & Choi, 2003). In addition, they are listed among the most frequent Norwegian words, so learners of this language encounter them already in the early stages of acquisition. The present study is partly motivated by my own background as a student of Norwegian and as a teacher of Norwegian to Polish students, which has given me frequent encounters with the difficulties associated with the proper distribution of på and i, even though these prepositions have close counterparts in Polish (na and w). It is well known that the idiomatic use of prepositions differs across languages, but the discrepancies between the usages of close translation 64

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

65

equivalents can be bewildering to learners especially during the early stages of acquisition. In my earlier work on this topic (Szyman´ska, 2010), focus was placed on pinpointing reasons why mastering these two basic Norwegian prepositions is so difficult for Polish-speaking learners. In addition, an attempt was made to determine potential cues that motivate the use of a wrong preposition in the target language. Although one might claim that the use of the wrong preposition is simply a matter of direct L1 influence on the target language, i.e. Polish w rendered as Norwegian i, and Polish na rendered as på, the reality may in fact be far more complex, and may involve conceptual factors that are not at all transparent to the speaker. That is, there may exist underlying influences that lead the speaker to the use of a certain preposition, and cause the speaker to believe that the use of any other preposition would be erroneous or even illogical. Prepositions are so challenging to learners of Norwegian that their use and distribution are dealt with in virtually every course relating to Norwegian as a second or foreign language. Learners are not only bothered by the lack of similarities between Norwegian prepositions and corresponding structures in their native languages, but also encounter a general lack of predictable patterns in the target language, as there appears to be a fair amount of variability even among native Norwegian speakers concerning which preposition is the most appropriate in certain contexts, which is partly due to the dialectal manifoldness of Norwegian, or to be more precise, the manifoldness of two separate written forms of Norwegian, namely, Bokmål and Nynorsk. A major objective for the present study is therefore the identification of some of these potential sources of difficulty.

Literature Review and Theoretical Framework The concept of space and the way it is encoded and expressed in languages has been investigated by many. The question has been approached from different perspectives, including a mathematical and schematic perspective (e.g. Cienki, 1989; Croft & Cruse, 2004; Kracht, 2002, 2006, 2008; Zwarts, 2008), a semi-schematic perspective with a strong cognitive component (e.g. Coventry & Garrod, 2004; Coventry & Guijarro-Fuentes, 2008; Herskovits, 1985, 1986, 1988; Jackendoff, 1983; Talmy, 1978a, 1983, 2006) and a purely cognitive point of view (e.g. Levinson, 1996, 1997, 2003a, 2003b; Levinson & Wilkins, 2006; Saint-Dizier, 2006; Vandeloise, 1991, 2006). Scholars working within all three perspectives regard the analysis of prepositional usage and related spatial morphology as being central not only to an understanding of how space is represented in language, but also to how it is represented in human cognition. Levinson and Wilkins (2006: 1) state that ‘spatial cognition is a fundamental design requirement for every mobile species with a fixed territory or home base. And there is little doubt that it plays a central role in human thinking

66

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

and reasoning.’ One of the questions that guide the present study is whether speakers of different languages conceptualize spatial relationships differently, and whether such differences might have consequences for how people express spatial relationships in a second language. The present study points in the direction of conceptual transfer proposed by Jarvis and Pavlenko (2008), and will attempt to follow their methodology for detecting this type of transfer. However, the linguistic approach adopted can be described as problemdriven, more than data- or theory-driven. Contrary to purely linguistic treatments of learners’ use of prepositions, the present study will follow a more cognitive approach to explaining the distribution of prepositions in learners’ language use. The erroneous use of prepositions by Polish-speaking learners of other languages has been looked at from many perspectives. However, there are rather few studies that have focused on the core, spatial meanings of prepositions. Cienki (1989) investigated the general semantics of prepositions in English, Polish and Russian. Sysak-Boron´ska (1980) compared the spatial system in Polish with the one in English. Weinsberg (1973) identified differences between spatial prepositions in Polish, German and Romanian. However, none of these studies focused on how Polish learners performed in the target languages, but rather highlighted the lack of equivalency in certain spatial phrases, emphasizing that the relationship between translation equivalents in, for example, Polish and English is not one-to-one. Sysak-Boron´ska also dealt somewhat with hypothetical reasons behind these dissimilarities, pointing out matters that turn out to be partly applicable to Polish vs Norwegian, which I will mention later in the present paper. Another study, Bluszcz (1987), examined spatial relations in Polish speakers’ use of Czech and Slovak. Parandowski (2002) examined mistakes in use of prepositions that Polish learners make in English. He presented a tentative classification of errors and presented their underlying causes where the way the learners describe a situation is taken into account. As far as I know there has been only one corpus-based analysis of the crosslinguistic (ir)regularities between Polish and Scandinavian languages. This was a study by Gawron´ska et al. (2006), who looked at the use of in, on and through in English, Swedish, Polish and Russian. Although the study seemed relevant due to the similarities between Swedish and Norwegian as far as the distribution of spatial preposition is concerned, the focus of the study was first and foremost on the collostructural strength of prepositional phrases and their frequency in the analysed languages, and all types of prepositional phrases were considered. One of the findings from that study that is relevant to the present investigation deals with the conceptualization of objects as surfaces versus containers. Gawron´ska noted that whereas ‘world’ is categorized as a surface in Polish, this is not the case in Russian, Swedish (and Norwegian) or English, where it is treated as a container and is therefore accompanied by the preposition ‘in’. Another observation made by the researcher is that ‘way/road’ is

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

67

two-dimensional in Swedish (and Norwegian) and English (both in a spatial and metaphorical sense – vara på väg, be on the way). In Polish, on the other hand, it is three-dimensional when conceptualized from a traveler’s perspective (Polish: ‘Jestem w drodze’ = literally ‘I’m in the way’ meaning ‘I’m on the way’) and two-dimensional otherwise. The literature and studies that had been done so far did not provide a profound analysis of specifically the two prepositions in question. They all considered prepositions as such, and the occurrence of patterns when a Slavic and a Scandinavian language come into contact were not displayed by any. However, some conceptual traces are provided by the literature, as it was Sysak-Boron´ska, Gawron´ska and Cienki who pointed out differences in the categorization of places between Slavic and Germanic language and Malcher (2011) who compared the performance in distribution of i and på among German, English and Spanish learners of Norwegian. The correct distribution of in and on is seemingly a rather frequent object of investigation. Ijaz (1986), who investigated a broader range of spatial prepositions, carried out a study based on two types of tests; semantic relatedness test and sentence completion. Five groups of advanced English learners with different L1 backgrounds were presented with these tests. In the first part the participant were supposed to map similarities between the prepositions, whereas in the sentence completion task they were presented a set of sentences with one word missing and asked to decide which of the given words fit the context. The study was semantic in nature as the given answers depended on how participants understood the meaning of the target words. The aim of the study was to compare the performance of non-native speakers with the performance of native speakers, with the intention of identifying whether non-native speakers had semantic structures of these words similar to those of native speakers. Both similarities and differences between the performance of native speakers and non-native speakers were shown. In the first test, 49.3% of the results of non-native speakers differed from those of native speakers, whereas in the sentence completion test 5 out of 28 answers between the two groups were significantly different. A closer analysis of non-native speakers’ performance displayed influence by their first language, which was concluded by Iliaz in the following way: ‘native language conceptual patterns appear to be powerful determinants of the meaning ascribed to L2 words and they seem to be very rigid and difficult to permeate’ (1986: 447). Another study on acquisition of spatial relations in L2, this time a longitudinal one, was carried out by Becker and Carroll (1997). The study lasted 30 months and analysed the progress that learners of three different target languages, English, German and French made in acquiring spatial relations in L2. An analysis is made of the oral data together with a wide range of factors that can determine specifics of individual source/target language pairings as well as the general characteristics of all cases of acquisition. Most importantly, a basic system for expressing spatial relations

68

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

seemingly common to all learners irrespective of their language background is identified by the authors. My own previous work (Szyman´ska, 2010) concentrated on Polish learners’ performance in Norwegian as their target language. Their performance was compared with that of Russian, Bosnian-Croatian-Serbian and English learners to determine whether significant differences might be found between the learner groups in relation to accurate usage of Norwegian prepositions. This was a corpus-based study with material from the ASK corpus where 800 texts (200 texts per L1 group, with 100 texts at each of two proficiency levels) were analysed. In addition to an investigation of learners’ error rates, the intention was to discover patterns in Polish that might account for Polishspeaking learners’ use of certain prepositions in L2 Norwegian. To this end a specially designed study was required. The scope of analysis was narrowed down to 31 nouns denoting places of daily relevance that tend to appear, depending on the context, in prepositional phrases with either i or på. Thirty native speakers of Norwegian from various backgrounds were presented with a fill-in-the-blank-test containing 127 phrases with the prepositions removed. Thirty Polish learners whose proficiency in Norwegian was assessed at the B1 level (intermediate) on the Common European Framework of Reference for Languages (CEFR) scale were then presented with the same test. The study showed that although the overall error rate was not that high, there were some phrases that were transferred directly from Polish, resulting in errors in the L2. Some of the errors appeared to be motivated by different categorizations of place, for example involving locations that tend to be construed by Norwegian speakers as containers but by Polish speakers as surfaces, or the other way around. Interestingly, the study also showed that although some prepositional phrases could be transferred directly into Norwegian and remain correct, the learners often chose other Norwegian prepositions that were not grammatical in the given context.

Prepositions in Polish and Norwegian In both languages, all three frames of reference (absolute, relative and intrinsic) are applicable and used with more or less the same frequency, which makes the perception of space rather similar. An absolute frame of reference is the most original and probably applicable frame as an object is defined in relation to cardinal directions, i.e. North, South, East and West, or fixed bearings such as seacoast, upriver/downriver, uphill/downhill (e.g. the lake is to the south of the house). When it comes to a relative frame of reference, the location of an object is expressed in relation to the viewpoint of the perceiver and position of another object, e.g. he girl sits next to the bus stop. An intrinsic frame, on the other hand, is a binary spatial relation in which the location of an object is defined in relation to a part of another object (its side, back, front, etc.), e.g. the car is in front of the house.

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

69

In general, native speakers of Polish and Norwegian understand and express spatial relations in a very homogenous way, i.e. using prepositions. As mentioned before, one can even talk about rough translation equivalents of Polish prepositions in Norwegian, w being i and na being rendered by på. A Polish prepositional phrase consists of a preposition and a noun with a suffix representing an appropriate case. With spatial prepositions in Polish, the noun is predominantly inflected into locative case, but genitive and accusative are also sometimes possible. In Norwegian, on the other hand, which lacks nominal case and case suffixes, a prepositional phrase is formed by a preposition and a bare noun (along with possible determiners, modifiers and suffixes representing number and definiteness). The fact that Polish is a case language and Norwegian is not might seem like a trivial difference when it comes to the choice of appropriate spatial prepositions. The present study nevertheless explores the possibility that this difference might have profound consequences for how Polish-speaking learners use Norwegian prepositions. Framework/research questions the theoretical point of departure for the present study is Slobin’s (1996: 89) claim that ‘each native language has trained its speakers to pay different kinds of attention to events and experiences when talking about them. This training is carried out in childhood and is exceptionally resistant to restructuring in adult second-language acquisition.’ Jarvis and Pavlenko (2008: 115) have referred to this phenomenon as conceptual transfer, ‘hypothesizing that certain instances of CLI [crosslinguistic-influence] in a person’s use of one language are influenced by conceptual categories acquired through another language.’ Jarvis and Pavlenko suggest that there are eight domains of conceptual categories that are particularly difficult for adults to restructure while learning another language; one of these domains is space. As mentioned before, previous research provides some indications that negative forward transfer from L1 Polish to L2 Norwegian in the domain of spatial reference might not simply be a matter of structural or even semantic transfer (i.e. transfer related to differences in the structural and functional distribution of spatial prepositions in Polish and Norwegian), but might also be driven by language-specific patterns regarding how spatial relations are perceived and projected. Bearing this in mind, the following questions are addressed: (1) How do Polish learners of Norwegian distribute the prepositions i and på? (2) Which areas are most prone to mistakes? (3) Which phrases (preposition + noun) have the highest accuracy ratio – are there any particular phrases that Polish learners manage to master? (4) Are there any clear transfer instances, i.e. errors that are most probably caused by transfer from L1?

70

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Method My analysis of the way Polish learners of Norwegian distribute prepositions in the target language consists of two main data sources, used in a slightly different way and for two different purposes. Nevertheless, both are corpus-based and provide both quantitative and qualitative information.

Study 1: ASK The first analysis was based on corpus data extracted from the ASK corpus at the University of Bergen, Norway. As the corpus includes texts representing two proficiency levels (in 2010 there were two levels: intermediate and advanced) and 10 different language backgrounds, it was possible to investigate learners’ use of prepositions from a broader perspective. Although the ASK corpus provides metadata about the texts’ authors, it only gives a rough self-evaluation of their competence in other languages, so this issue has not been addressed in the analysis even if it might play a certain role in their L2 performance. Following the methodology for detecting transfer described by Jarvis and Pavlenko (2008), data written by informants with four different mother tongues were selected in order to analyse the data for intragroup homogeneity, intergroup heterogeneity and similarities between L1 and L2 performance. In total, the quantitative analysis was based on 800 texts – 200 texts for each language group, 100 from the lower and 100 from the higher level. Three Slavic languages (Polish, Russian and Bosnian-Croatian-Serbian) and English make up the L1. Russian and Bosnian-Croatian-Serbian (BCS) were chosen in order to determine whether any substantial differences exist in the use of Norwegian prepositions by speakers of different Slavic languages. The social background of the Slavic informants is very diverse – they are students, manual labourers and spouses of Norwegian partners, but mostly rather lowly educated. Norwegian is spoken only at work with some, while also at home with others. Work, in this case, is their main motivation for learning the language. The reason for also including English was that this language and its prepositional system are closely related to Norwegian, and this fact makes it possible to investigate whether certain patterns of prepositional use in L2 Norwegian are merely an artefact of the distance and differences between Norwegian and Slavic languages, or whether they may be more attributable to universal principles. The learners’ use of i (in) and på (on) formed the focus of the data analysis, with an emphasis on how many times i was erroneously used in place of på, and how many times på was erroneously used instead of i. To perform this search tools were used, the fact that the corpus had been error-tagged with metadata that showed not only which prepositions were used erroneously, but also what the correct forms should have been helped. In addition to i and

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

71

på, occurrences of til (to) erroneously used in place of either i or på, as well as erroneous uses of på and i in contexts where til should have been used were also analysed. The overall occurrence of i and på in the Polish corpus was 4150 (5.3% relative frequency), 3956 (5.4%) in the Russian corpus, 3917 (5.3%) in the Bosnian-Croatian-Serbian (BCS) corpus and 4258 (5.6%) in the English corpus. The error rate was relatively high, with 58 errors among the 595 relevant prepositions used by Polish participants at the intermediate level, and 48 errors among the 1176 instances of i and på used by Polish participants at the advanced level. These figures, however, reflect learners’ uses of i and på in general rather than specifically spatial usages. A specific analysis of spatial usage could only be accomplished by examining each text individually with the purpose of extracting exclusively spatial use. This revealed a scarcity of spatial contexts in the data; most likely due to the fact that there was only one exam task (i.e., one which dealt with the use of cellular phones in public) that specifically elicited descriptions of space and the use of spatial prepositions. Designing a separate task to provoke usage of spatial prepositions, would most likely have resulted in the data providing a significantly larger scope of exploration. Limiting the data in this way resulted in fewer than 30 tokens for each category. The data were therefore not robust enough to perform analyses of intragroup homogeneity, intergroup heterogeneity, and crosslinguistic congruity between learners’ spatial markers in their L1 and L2. The analysis did nevertheless provide one clear and surprising result: The English comparison group, which was expected to score much better than Slavic learners due to the close relationship between English and Norwegian, did not display a better mastery of Norwegian prepositions. As mentioned before, the occurrence of relevant tokens was low, but the results show very clearly that the English speakers do not have a lower number of errors than the Slavic learners (see Tables 4.1–4.6). A comparison was made of the average number of erratic uses of spatial prepositions made by learners with a Slavic background to those with an English background. The difference occurred

Table 4.1 Results from ASK corpus, i used instead of på or til, intermediate level i corrected Intermediate Total Number of Errors level number of errors occurring in to på i tokens relevant contexts POLISH RUSSIAN BCS ENGLISH

595 753 643 611

58 (9.7%) 46 66 60

22 (3,7%) 19 26 27

X-squared = 0.0014, df = 1, p-value = 0.9699

18 (81.8%) 5 (26.3%) 10 (38.5%) 23 (85.2%)

i corrected Omissions to til

1 (4.5%) 8 (17.3%) 13 (50%) 3 (1.1%)

3 (13.6%) 1 (5.2%) 3 (11.5%) 1 (3.7%)

72

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 4.2 Results from ASK corpus, i used instead of på or til, advanced level Advanced level

i corrected Total Number of Errors number of errors occurring in to på i tokens relevant contexts

POLISH

1176

48

i corrected Omissions to til

12

11 (91.7%) 0

17 (58.6%) 8 (27.5%) 4 (13.7%)

RUSSIAN

1234

69

29

BCS

1117

61

20

ENGLISH

1182

62

19

9 (45%)

5 (25%)

15 (78.9%) 1 (5.2%)

1 (8.3%) 6 (30%) 3 (15.7%)

X-squared = 0, df = 1, p-value = 0.9947

Table 4.3 Results from ASK corpus, på used instead of i or til, intermediate level på corrected på corrected Omissions Intermediate Total Number of Errors to til level number of errors occurring in to i på tokens relevant contexts POLISH

529

32

12

RUSSIAN

417

31

16

11 (68.7%) 3 (18.7%)

3 (25%)

6 (50%)

2 (12.5%)

BCS

454

41

22

15 (68.1%) 2 (9%)

5 (22.7%)

ENGLISH

483

54

31

27 (87%)

2 (6.4%)

2 (6.4%)

3 (25%)

X-squared = 0.0081, df = 1, p-value = 0.9282

Table 4.4 Results from ASK corpus, på used instead of i or til, advanced level Advanced level

på corrected på corrected Omissions Total Number of Errors to til number of errors occurring in to i på tokens relevant contexts

POLISH

726

31

7

RUSSIAN

685

36

15

12 (80%)

BCS

743

31

10

5 (55%)

ENGLISH

759

46

9

5 (55.5%) 2 (22.2%)

X-squared = 3e-04, df = 1, p-value = 0.9854

5 (71.4%) 1 (14.2%)

2 (28.6%)

2 (13.3%)

1 (6.7%)

3 (30%)

2 (20%) 2 (22.2%)

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

73

Table 4.5 Results from ASK corpus, til used instead of på or i, intermediate level til corrected til corrected Omissions Intermediate Total Number of Errors to i level number of errors occurring in to på til tokens relevant contexts POLISH

376

66

3

2 (66.6%)

1 (1.5%)

0

RUSSIAN

335

31

1

0

0

1 (100%)

BCS

302

36

2

2 (100%)

2 (100%)

0

ENGLISH

386

62

6

6 (9,70%)

6 (100%)

0

X-squared = 0.0043, df = 1, p-value = 0.9476

Table 4.6 Results from ASK corpus, til used instead of på or i, advanced level Advanced level

til corrected til corrected Omissions Total Number of Errors to i number of errors occurring in to på til tokens relevant contexts

POLISH

748

42

2

0

0

2 (100%)

RUSSIAN

690

65

4

2 (50%)

2 (50%)

0

BCS

658

56

0

0

0

0

ENGLISH

837

77

5

4 (80%)

0

1 (20%)

X-squared = 0.0011, df = 1, p-value = 0.9736

not to be statistically significant, which means that the two groups do not differ in respect of erratic distribution of spatial prepositions i and på. The results of the Chi-square test are presented below each table.

Study 2: Fill-in-the-blank test Given the limited nature of the corpus data, it was necessary to collect additional information about learners’ knowledge and use of Norwegian prepositions through the use of a fill-in-the-blank test. The test, designed by myself, consisted of 127 Norwegian sentences that contained highly frequent spatial phrases involving the use of i or på together with any of 31 nouns reflecting spaces encountered in everyday life, such as a pharmacy, public space, working place, street, hill, library and so forth. The sentences in the fill-in-the-blank test were complete except that the preposition was omitted and a gap left in its place. As Norwegian prepositions sometimes reflect a distinction between concrete and abstract contexts (specific vs

74

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

non-specific), every noun on the test was used in both a concrete and an abstract context (e.g. på flyet vs i flyet – on the plane, *in the plane; på biblioteket vs i biblioteket – *on the library, in the library). This allowed me to examine how consistent Norwegians are in making this distinction, and, if they are consistent, whether Polish-speaking learners of Norwegian have also mastered it.

Stage 1 (L1 Norwegian) Data collection with the fill-in-the-blank test were divided into two stages. During the first stage, 30 adult native speakers of Norwegian with various social, dialectal and professional backgrounds were recruited. Recruitment of participants was carried out among university employees, students, their friends and relatives as well as my own acquaintances. Eventually, a group of 30 native speakers of Norwegian agreed to participate in the study and complete the test. The test required participants to complete each of the 127 sentences by supplying the missing preposition. Participants’ responses showed, however, that for many of the sentences, more than one preposition was possible, and several of the participants commented afterwards that it was often difficult to decide which was the right preposition for a given context. While analysing the Norwegians’ responses to the fill-in-the-blank-test, their agreement ratio for each sentence was tested in order to determine which contexts were most dependent on individual preferences – in relation to the proper choice of prepositions – even for native Norwegian speakers. The results, reported in a later section of this paper, show a good deal of disparity in the native speakers’ choices. Since the choice of preposition is to some extent individual for Norwegian native speakers, three research questions were formed: (1) What is the distribution of i and på by Norwegians speakers and to what extent is there a difference between concrete and abstract uses? (2) Are there any phrases that are used with only one of the prepositions? (3) In which contexts are i and på mostly used?

Stage 2 (L2 Norwegian) The same fill-in-the-blank test was presented to a group of 30 adult native speakers of Polish who learned the target language in Norway and were at least at the B1 level of language proficiency according to the CEFR scale. Participants came from different regions of Poland (seven from Northern Poland, six from the West, five from the East, eight from the South, four from central Poland). They had various educational background (18 of them were manual workers with basic education, whereas the rest had a university degree) and professional backgrounds (four shop assistants, six cleaners, three babysitters, five carpenters, four nurses, five students, one lawyer, two customer service employees). The test they were given was the same as was used in Stage 1. As before, the participants were asked to fill in the gaps with whatever preposition they considered to be the most suitable.

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

75

Again, many commented on the difficulty of mastering prepositions in Norwegian in general, but all of them handed in the test. This time the participants offered no second alternatives. Perhaps owing to the high level of consistency with which spatial relationships are expressed in their mother tongue, the Polish participants felt that in each case there could be only one preposition that fit the context. As before, the participants’ competence in other foreign languages was not taken into consideration because they did not provide this information.

Stage 3 (L1 Polish) A new fill-in-the-blank test was drafted with the aim of looking more closely at how much consistency there actually is in L1 Polish spatial expressions, and in order to determine whether the use of prepositions by Polish speakers in L2 Norwegian is directly motivated by corresponding choices in L1 Polish. The new test contained sentences in Polish that were very similar to those in the original Norwegian version of the test. As with the Norwegian test, the Polish edition contained phrases from everyday life that refer to spatial relationships. Once again, 30 adult participants, this time in Poland, were recruited, all of whom were native speakers of Polish representing diversity in terms of regional (four from the North, eight from the Centre, seven from Western Poland, five from the East and six from the South), educational (15 with basic education, 15 with higher education) and professional backgrounds (two accountants, three teachers, four cooks, five nurses, two shop assistants, three students, four cleaners, one mechanic, four manual workers, two freelancers). The diversity of the participants’ backgrounds did not have any effect on their responses; the results showed near-absolute agreement in the prepositions chosen by the participants. This fact will be dealt with at more length in the following section as it may reflect a way of thinking that is transferred by Polish speakers into their use of the target language.

Results and Discussion Agreement and accuracy The results from Stage 1 show a high level of variability in the L1 Norwegian speakers’ choice of prepositions. The L1 Norwegian participants supplied the same preposition in 58% of all cases, just as it was in the original source, whereas in the remaining 42% they supplied more than one possible preposition or filled in the gaps with a different preposition than in the original version. Among other things, this shows that Norwegian speakers’ choice of prepositions is not determined by strict criteria, such as whether the spatial relationship in question is abstract or concrete. In Stage 2, the Polish-speaking learners of Norwegian achieved 48.5% correctness

76

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

(compared to the original text), whereas in 51.5% of cases the results varied. In Stage 3, the native Polish speakers showed disagreement in only 1.5% of all cases, which resulted from a misunderstanding of the context, as their answers did not make any sense. If one looks at the two native groups and their results in their L1s, one can see that there is a substantial difference in how they view the proper use of prepositions, and more importantly in how they understand the meanings of prepositions. As mentioned before, the context in Polish cues the speaker towards only one preposition, and any other would result in a mistake. How could it be that native speakers of Norwegian have so much difficulty in deciding what the correct preposition is in a given context? The majority of the Norwegians examined admitted that they regarded the task to be difficult. And, regarding the many cases where native Norwegian speakers consider there to be multiple acceptable alternatives, what does this mean for someone acquiring Norwegian as a second language? A learner might know the rules and meanings of prepositions as presented in grammar books and dictionaries, but actually using these prepositions in real contexts might be extremely challenging. Learners trained by their mother tongue to associate only one preposition with a given noun might find this to be particularly challenging. It turns out that the choice between på and i is not just a question of support versus containment as seen through the eyes of a native Polish speaker; native Norwegian speakers clearly rely on far more cues than this simple distinction.

Phrases that favour one of the prepositions The preceding quantitative analysis has shown that Norwegians are much more eager to use på than i, although, as mentioned before, the fill-in-the-blank test included an equal number of both concrete and abstract uses, which theoretically speaking should be distinguished by the use of different prepositions. The 2160 (72%) cases of på used by Norwegians and only 1365 (27.7%) cases of i showed that Norwegian speakers clearly do not make a categorical distinction between concrete and abstract spatial contexts through their choice of prepositions (there were also 166 cases where respondents claimed that both i and på were applicable, in spite of the clear context). On this basis one can say that på is much more frequent than i, and there are only a few phrases where på is not considered by native speakers to be appropriate: i.e. those involving the nouns barnehage, trær, ansikt, hage and kiosk (kindergarten, trees, face, garden and kiosk). The increasing tendency to use på instead of i has been commented by Golden et al. (2014), although there is no clear explanation for this phenomenon. In the Polish speakers’ responses on the Norwegian test, på was used 2476 (63%) times and i 1336 (34%) times, there were only two cases where a participant supplied both prepositions. These results are quite similar to the Norwegian group. However, there is a variation from the Norwegian group in the way the prepositions are distributed.

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

77

As Polish w (in) seems to be more frequent than Norwegian i, there was an expectation that there would be many errors involving an overuse of i. This turned out to be true in 13 of 31 cases (41.9%). In these cases, phrases were transferred directly from Polish yielding incorrect uses of i in Norwegian. I believe that the use of på in Norwegian by native speakers of Polish is more a question of living in the Norwegian environment and encountering everyday life, as the use of na with places and institutions in Polish is rather limited. There are a few examples in Polish where a noun referring to a closed space is used in a prepositional phrase headed by na (on), but these cases usually have historical explanations. For example, in Polish one is on the train station (na stacji), or on the post office (na poczcie), but this is truly motivated by the previous character of these places, which consisted of many parts or posts. For most other nouns referring to enclosed spaces, Polish is consistent in the use of w (in), whereas Norwegian varies between i and på (in and on).

Differences in conceptualizing objects and places As mentioned before, according to the cognitive framework, the choice of preposition is driven by assumptions about the nature of the object the accompanying noun refers to. Nouns referring to objects that are perceived as surfaces would normally be preceded by på, whereas nouns referring to a container-like object would attract i. Why, then, do native speakers of Norwegian and Polish differ in their choice of i versus på given that the two languages seem similar in relation to how objects are categorized, i.e. either as a container or as a surface? It turns out that there are substantial crosslinguistic differences in terms of how objects and places are perceived as containers or surfaces, and this makes it difficult for learners to know which preposition to use in a given context even if they have a perfect understanding of the meanings that each preposition represents. Although one can memorize certain word clusters (e.g. prepositional phrases), it may be impossible for learners to see things exactly as they are seen by native speakers. Figure 4.1 shows that a native speaker of Polish sits on the sofa (na sofie), or on a tree (na drzewie), lives on the second floor (na 2. pie˛trze) and is *on the desert (na pustyni), and *on the world (na s´wiecie), whereas a native speaker of Norwegian sits in the sofa (i sofaen), in a tree (i treet), lives *in the second floor (i 2. etasje) and is in the desert (i ørkenen) and in the world (i verden). Polish appears to have stricter criteria for categorizing something as a container. It requires full containment, such as four walls that limit the space of the enclosure, whereas Norwegian allows for a great deal more variation. This is reflected in the Polish-speaking learners’ unanimous use of på with the word for tree, and this pattern corresponds exactly with the native Polish speakers’ unanimous use of Polish na in the corresponding sentences on the Polish test. This suggests that Polish speakers may perceive trees more as surfaces instead of as containers. Norwegians on the other hand did not

78

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Figure 4.1 Phrases with i and på a)

Norwegian

b)

Polish

Norwegian

c)

Polish

Norwegian

Polish

Figure 4.2 Sofa, tree and floor conceptualized as a container (in Norwegian) and a surface (in Polish)

have any doubts about using i (in) in the same context, suggesting that they perceive trees more as containers instead of as surfaces. Sysak-Boron´ska (1980) points to the same fact, claiming that Polish has stricter criteria for container relationships than English does. Such discrepancies across languages lead to numerous errors in learners’ reference to everyday spatial relations. Although it is probably rather rare that this kind of error results in a serious misunderstanding, it does nevertheless interfere with effective communication and the learner’s desire to be viewed as a competent member of the target speech community.

Inaccessible double concept (specific vs non-specific referent) Another interesting fact that differentiates Norwegian from Polish and is reported by many teachers (e.g. from Adam Mickiewicz University in Poznan´) as being out of the students’ reach, at least at the beginner level, is

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

79

the so-called double concept. As already mentioned, Norwegians distinguish linguistically between place as a concept and place as a concrete space. Generally speaking, på is more frequent in spatial expressions referring to places as concepts, whereas i is more frequent in spatial expressions with concrete referents. According to the data, it is sometimes difficult even for Norwegians to choose the right preposition, if they to do it deliberately, indicating that it must be even more complicated for Polish learners to firstly learn to understand how to mark this distinction, and afterwards to know whether one’s choice was correct, given the lack of consistency in the input from L1 Norwegian speakers, as their mother tongue trained them to only one possibility and this is what they probably expect to find in the target language too. Below is an illustration of how i and på alternate with the same nouns depending on whether those nouns refer to specific (concrete) or a non-specific referents, which I call abstracts. I kaféen har vi et bredt utvalg av påsmurte baguetter og rundstykker. (concrete) In the café we have a wide range of baguettes and rolls. Skal vi gå og sette oss på en kafé? (abstract, concept) Shall we go and sit in (literally: on) a café? Hvordan finner jeg den beste plassen i flyet? (concrete) How do I find the best seat on (literally: in) the plane? Statistikken viser at nordmenn flørter mer på flyet enn både italienere og franskmenn. (abstract, concept) The statistics show that Norwegians flirt on the plane more often than Italians and Frenchmen.

Minor differences expressed by a preposition The third source of difficulty involving the choice between i and på by Polish learners of Norwegian is in expressing fine differences and nuances. Again, although native speakers of both Polish and Norwegian perceive the differences, Polish speakers are not trained to mark the differences through the use of prepositions as such differences are expressed in a descriptive way. In Norwegian a simple change of preposition can convey a lot of extra information, sa in: Han sitter i sofaen. (He’s sitting in the sofa.) > The sofa is prototypical and so is his way of sitting (relaxed, comfortable). Han sitter på sofaen. (He’s sitting on the sofa.) > The sofa might be hard and his way of sitting may be somewhat uncomfortable, anticipating.

80

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

In Polish, a certain choice of words or a description of the sofa would be necessary to achieve the same result. There are only few cases where a preposition can be used in the same way as it is used in Norwegian, as in the following examples: lez˙eć w łóz˙ku (to lie in bed) – under the bedcover lez˙eć na łóz˙ku (to lie on bed) – on the bedcover siedzieć w trawie (to sit in the grass) – the grass is long siedzieć na trawie (to sit on the grass) – the grass is rather short but otherwise such nuances are expressed by different language tools, such as adverbs: Norwegian: Anna sitter i sofaen. Polish: Anna siedzi wygodnie na sofie. English: Anna sits comfortably on the sofa.) Norwegian: Anna sitter på sofaen. Polish: Anna siedzi spi˛eta na sofie. English: Anna sits uneasily on the sofa.). In Norwegian, on the other hand, the use of prepositions to mark such nuances of meaning is much more common, and this is yet another factor that makes it difficult for L2 learners of Norwegian to test their hypotheses about which prepositions should accompany which nouns. How can you rely on the native speaker input if they tend to use different prepositions with the same noun. If learners are unaware of the nuances of meaning prepositions can convey, they are likely to feel disoriented when choosing a preposition.

Conclusion The results of the present study show that the proper distribution of Norwegian spatial prepositions i and på does, in fact, pose a serious challenge for native speakers of Polish. The difficulties appear to be caused by a few factors motivated by cognition. On the one hand, habits learned from the L1, such as a relatively strong language norm in Polish, result in a resistance to exceptions. In Polish, there is often only one preposition that can be used with a given noun denoting a place or object where something can be situated. This fact is reflected in the patterns of responses found in the Polish participants’ tests. In nearly every case, for each gap they supplied only one preposition, whereas the Norwegian participants, who have more varied language norms and also a great degree of dialectal variation, showed a tendency to provide two or more prepositions, although they were instructed to choose the most suitable one.

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

81

Differences in how objects and places are conceptualized pose yet another obstacle in the acquisition of Norwegian as a second language. L1 Polish, which has stricter conceptual criteria regarding whether something can be viewed as a container, seems to reinforce Polish-speaking learners’ prepositional choices in L2 Norwegian. The distinction between specific and non-specific places in Norwegian also lacks transparency to a Polish speaker, all the more so as Norwegians themselves seem to be inconsistent in marking this distinction. Some phrases would never appear with one of the prepositions, whereas the other would follow the institution rule (Hagen, 2000). This inconsistency and unpredictability is rather misleading to any learner of Norwegian. Beyond these complications is the matter of nuances of meaning – which are also often marked through differential prepositional use in L1 Norwegian. The differences might be clear and understood to some extent by L2 learners, but I believe that Polish-speaking learners of Norwegian have very limited access to know what types of nuances of meaning can be expressed, and how one marks these nuances, e.g. the way one sits on the sofa expressed by applying a certain preposition. Of course, I do not claim that the importance of such fine differences is great. However, the nature of such differences and the challenges they pose for language learners seem to point to one of the central implications of conceptual transfer, which is that certain elements of an L2 might never be fully mastered because the learner may never become fully aware of the subtle and complex characteristics they represent. One can memorize rules and phrases, but the door leading to cognition and understanding matters in a deeper way remains locked. Studies like the present one can sensitize teachers who work with Polish-speaking learners of Norwegian to certain recurrent errors. Being aware of what lies behind certain choices, teachers can better understand and explain why Norwegian works the way it does and provide their students with better insights into the inner workings of the target language. Figuratively speaking, doing so allows teachers to give their students Norwegian eyeglasses that help them see and categorize space in a slightly different way.

References Becker, A. and Carroll, M. (eds) (1997) The Acquisition of Spatial Relations in a Second Language. Amsterdam: Benjamins. Bluszcz, A.J. (1987) Relacje przestrzenne w polskich, czeskich i słowackich konstrukcjach z wyra˙zeniami przyimkowymi. [Spatial relations in Polish, Czech and Slovak structures with prepositional phrases]. Katowice: Wydawnictwo US´. Bowerman, M. and Choi, S. (2003) Space under construction: Language-specific spatial categorization in first language acquisition. In D. Gentner and S. Goldin-Meadow (eds) Language in Mind: Advances in the Study of Language and Thought. Cambridge: MIT Press. Cienki, A.J. (1989) Spatial Cognition and the Semantics of Prepositions in English, Polish and Russian. München: Verlag Otto Sagner. Coventry, K.R. and Garrod, S.C. (2004) Saying, Seeing and Acting. The Psychological Semantics of Spatial Prepositions. Hove/New York: Psychology Press.

82

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Coventry, K.R. and Guijarro-Fuentes, P. (2008) Spatial language learning. In P. Robinson and N.C. Ellis (eds) Handbook of Cognitive Linguistics and Second Language Acquisition (pp. 114–138). New York/London: Routledge. Croft, W. and Cruse, D. (2004) Cognitive Linguistics. Cambridge: Cambridge University Press. Gawron´ska, B., Nikolaenkova, O. and Erlenddson, B. (2006) A corpus based analysis of English, Swedish, Polish and Russian prepositions. In: ISCA Tutorial and Research Workshop on Experimental Linguistics. Conference Paper, 137–140. Golden, A., Mac Donald, K. and Ryen, E. (2014) Norsk som fremmedspråk. Grammatikk (4th edn) [Norwegian as a foreign language. Grammar]. Oslo: Universitetsforlaget. Hagen, J.E. (2000) Norsk Grammatikk for andrespråkslærere. [Norwegian grammar for second language teachers]. Oslo: Ad Notam Gyldendal. Herskovits, A. (1985) Semantics and pragmatics of locative expressions. Cognitive Science 9 (2), 341–378. Herskovits, A. (1986) Language and Spatial Cognition. An Interdisciplinary Study of the Prepositions in English. Cambridge: Cambridge University Press. Herskovits, A. (1988) Spatial expression and the plasticity of meaning. In B. Rudzka-Ostyn ´ (ed.) Topic in Cognitive Linguistics. Amsterdam/Philadelphia: John Benjamins. Ijaz, I.H. (1986) Linguistic and cognitive determinants of lexical acquisition in a second language. Language Learning 36 (4), 401–451. Jackendoff, R. (1983) Semantics and Cognition. Cambridge, MA: MIT Press. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York: Routledge. Kracht, M. (2002) On the semantics and locatives. Linguistics and Philosophy 25 (2), 157–232. Kracht, M. (2006) Directionality selection. In P. Saint-Dizier (ed.) Syntax and Semantics of Prepositions. Dodrecht: Springer. Kracht, M. (2008) The fine structure of spatial expressions. In A. Asbury, J. Dotlačil, B. Gehrke and R. Nouwen (eds) Syntax and Semantics of Spatial Prepositions. Amsterdam/ Philadelphia: John Benjamins. Levinson, S.C. (1996) Relativity in spatial conception and description. In J.J. Gumperz, and S.C. Levinson (eds) Rethinking Linguistic Relativity (pp. 177–202). Cambridge: Cambridge University Press. Levinson, S.C. (1997) From outer to inner space: Linguistic categories and non-linguistic thinking. In J. Nuyts, and E. Pederson (eds) Language and Conceptualization (pp. 13–45). Cambridge: Cambridge University Press. Levinson, S.C. (2003a) Space in Language and Cognition. Explorations in Cognitive Diversity. Cambridge: Cambridge University Press. Levinson, S.C. (2003b) Language and mind: Let’s get the issues straight! In D. Gentner and S. Goldin-Meadow (eds) Language in Mind: Advances in the Study of Language and Thought (pp. 25–46). Cambridge, MA: MIT Press. Levinson, S.C. and Wilkins, D. (2006) Grammars of Space. Explorations in Cognitive Diversity. Cambridge: Cambridge University Press. Malcher, J. (2011) Jeg liker å teffe folk i café. Man må nyter de fine tingene på verden! Preposisjoner og morsmålstransfer – en korpusbasert studie med i og på i fokus. [Prepositions and mother tongue transfer – a corpus based study on ‘in’ and ‘on’]. M.A. Thesis, University of Oslo. Saint-Dizier, P. (2006) Introduction to the syntax and semantics of prepositions. In P. Saint-Dizier (ed.) Syntax and Semantics of Prepositions. Dodrecht: Springer. Slobin, D.I. (1996) From ‘thought and language’ to ‘thinking for speaking’. In J. Gumperz and S. Levinson (eds) Rethinking Linguistic Relativity. Cambridge: Cambridge University Press.

On How Polish Lear ners of Nor wegian Render Spat ial Preposit ions in L2

83

Sysak-Boron´ska, M. (1980) The Spatial System in Polish and English: Prepositions of Direction and Location. Katowice: Uniwersytet S´la˛ski. Szyman´ska, O. (2010) Konseptualisering av rommet hos polske norskinnlærere – utfordringer og diagnoser. En korpusbasert studie med i og på i fokus. [On how Polish learners of Norwegian structure space – challenges and diagnoses. A corpus-based study on ‘i’ and ‘på’]. PhD thesis. Universitet i Poznan´. Talmy, L. (1978a) Figure and ground in complex sentences. In J. Greenberg, C. Ferguson, and E. Moravcsik (eds) Universals of Human Language, Vol. 4, Stanford University Press, Palo Alto. Talmy, L. (1983) How language structures space. In L. Talmy (ed.) Towards Cognitive Semantics, Vol. 1. Cambridge, MA: The MIT Press. Talmy, L. (2006) The representation of spatial structure in spoken and signed language. In M. Hickmann and S. Robert (eds) Space in Languages – Linguistic Systems and Cognitive Categories (pp. 207–238). Amsterdam/Philadelphia: John Benjamins. Vandeloise, C. (1991) Spatial Prepositions: A Case Study from French. Chicago: University of Chicago Press. Vandeloise, C. (2006) Are there spatial prepositions? In M. Hickmann and S. Roberts (eds) Space in Languages - Linguistic Systems and Cognitive Categories. Amsterdam/Philadelphia: John Benjamins. Weinsberg, A. (1973) Przyimki przestrzenne w j˛zyku e polskim, niemieckim i rumun´skim. [Spatial prepositions in Polish, German and Romanian.] Wrocław: Ossolineum. Zwarts, J. (2008) Priorities in the production of prepositions. In A. Asbury, J. Dotlačil, B. Gehrke and R. Nouwen (eds) Syntax and Semantics of Spatial Prepositions. Amsterdam/ Philadelphia: John Benjamins.

5

Positive and Negative Transfer in the L2 Adjective Inflection of English-, German- and Polish-speaking Learners of L2 Norwegian Marta Olga Janik SWPS Warsaw University of Social Sciences and Humanities

Introduction Adjectives in Norwegian have one of the most complex inflections in the language and, after verbs, are inflected for the largest number of categories – four (whereas verbs are inflected for five categories (Hagen, 2008: 82)). Inflection in general (also adjective inflection) is characteristic of synthetic and agglutinative languages, although Norwegian is generally not regarded as a primarily synthetic language with its relatively few inflectional categories (for instance Kulbrandstad (2005: 98) places Norwegian around the middle of the analytic-synthetic axis mainly because of derivation and composition, though he mentions also inflection). The inflectional suffix on Norwegian adjectives gives information about the number, gender, definiteness and degree of comparison of the adjective, so one suffix may carry up to four grammatical functions, as in, for instance: (1) et stort hus (a big-SG,N,INDEF,POS house)1 Table 5.1 shows the distribution of adjective suffixes in Norwegian.

84

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

85

Table 5.1 Adjective suffixes in Norwegian Adjective Attribute

gender

Predicate

singular

indef. form

def. form

indef. & def. form

masculine feminine neuter plural

Ø Ø -t -e

-e -e -e -e

Ø Ø -t -e

However, in some cases, and always in the case of plural and definite adjectives, the forms of the inflectional morphemes converge so that the suffixes are the same for all genders, e.g.: (2) (3) (4) (5) (6) (7) (8) (9) (10)

den høye blokka (the high-SG.F.DEF. block) den høye bygningen (the high-SG.M.DEF. building) det høye huset (the high-SG.N.DEF. house) høye blokker (high-PL.INDEF. blocks)2 høye bygninger (high-PL.INDEF. buildings) høye hus (high-PL.INDEF. houses) de høye blokkene (the high-PL.DEF. blocks) de høye bygningene (the high-PL.DEF. buildings) de høye husene (the high-PL.DEF. houses)

Adjectival inflectional morphemes also undergo form convergence in other conditions, some of which involve sets of adjectives commonly referred to as irregular adjectives. Hagen (2002: 119) mentions five classes of irregular adjectives, three of which show convergence with respect to their inflected forms. Convergence of forms is, moreover, also present in the comparative and superlative forms of Norwegian adjectives, and it is more often encountered in predicative position than in attributive position (there are several adjectives which in predicative position can be considered as idioms, especially when they occur together with a preposition (Hagen, 2002: 363)). Although there are not many adjectival inflectional suffixes in Norwegian, the convergence of forms and the dialectal variation (in many dialects in Norway, e.g. in Bergen, adjectives are often not inflected in predicative position (Papazian & Helleland, 2012; Vigeland, 1995)) make the acquisition of adjective inflection in Norwegian difficult. Even advanced learners often have difficulties in inflecting Norwegian adjectives. This is surprising because adjective inflection is introduced in language courses quite early (e.g. in Ellingsen & Mac Donald (2004) På vei, a course book for beginners (A1–A2 level): chapter 6), and is assumed to be one of the more basic areas of Norwegian grammar that are expected to be acquired at a fairly early stage.

86

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

The present study investigates whether the learners’ L1s have any influence on the process of acquiring adjective inflection in L2 Norwegian, and if they have – what kind of influence (positive or negative) it is. The focus here is put on errors made in adjective inflection, and the assumption is that the fewer errors the learners produce, the better they have acquired adjective inflection. Three L1s (English, German and Polish) are included in the investigation, which have three different patterns of adjective inflection, where only the Polish inflectional pattern is similar to the Norwegian one. The main hypothesis is that learners speaking L1s with an adjective inflection pattern similar to the Norwegian one, will produce fewer errors in L2 Norwegian adjective inflection, because their L1s facilitate their acquisition of this pattern in the target language.

Framework Crosslinguistic influence Second language acquisition is a long and complex cognitive process, which is affected by several factors. The most often named and therefore perhaps most important factors are age, crosslinguistic influence, the linguistic environment, individual mental processes, language aptitude, the learner’s motivation, and various social factors (Odlin, 1989). The present chapter focuses mainly on the effects of crosslinguistic influence, which is defined as ‘the influence of a person’s knowledge of one language in that person’s knowledge or use of another language’ (Jarvis & Pavlenko, 2008: 1). There is now widespread agreement among transfer researchers that crosslinguistic influence is a phenomenon that can surface in several ways (Odlin, 1989: 36; Odlin, 2003: 442). The most typical of them are positive transfer (facilitative effects of L1 knowledge), negative transfer (constraining effects of L1 knowledge, seen in the L2 in the form of production errors, underuse, overuse or misinterpretation) and differing lengths of acquisition (Odlin, 1989: 36). The present chapter gives attention to both production errors (negative transfer) and positive transfer. Other types of transfer, including avoidance, underuse and overuse, will nevertheless also be given due attention where relevant. It is said that learners transfer structures from one language to another when they believe those structures are transferable and similar in the two languages (Kellerman, 1979; Odlin, 2003: 454). Under such conditions, learners make interlingual identifications, which are defined by Odlin (2003) as mental associations that reflect learners’ judgements that a feature in the native language is similar to the corresponding feature in the target language. Learners’ actual knowledge of the L2, their perceived distance between the L1 and the L2 and their intuitions of what is transferable or not can affect their

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

87

judgements of crosslinguistic similarity (Kellerman, 1979). These factors are not properties of the target language itself, but are subjective. Although learners’ proficiency level can help them to perceive the objective closeness (or distance) of the source and target language more accurately, their perceptions may in many cases reflect incorrect assumptions about the similarities and differences between the languages in question (Pienemann et al., 2005a: 95). One important type of constraint on L1 transfer should be mentioned here, namely the effects of processability, which are described as a factor that can constrain L1 transfer and also override the effects of typological distance (Pienemann et al., 2005b: 147). According to Pienemann’s Processability Theory (1998), language learners are able to produce structures in the L2 only if they are capable of processing the procedures that those structures entail (Pienemann, 1998: 1). He also claims that the mental procedures needed to process grammatical information form a hierarchy from less complex to more complex, and that they are acquired in that specific order (Pienemann, 2003: 689): (1) (2) (3) (4) (5)

lemma access the category procedure the phrasal procedure the S-procedure the subordinate clause procedure, if applicable.

Adjective inflection in Norwegian entails the last three levels of the hierarchy. Thus, when adjectives occur in attributive position, the inflection is restricted to a single phrase (requiring only the phrasal procedure), but when adjectives occur in predicative position, the agreement inflection encompasses the whole clause (requiring the S-procedure) or even the entire sentence (requiring the subordinate clause procedure). Consequently, according to Processability Theory, learners can be predicted to acquire the ability to produce agreement inflections with attributive adjectives before they acquire the ability to produce agreement inflections with predicative adjectives. Identifying instances of crosslinguistic influence can often be a difficult task as there are usually several language learning processes that operate simultaneously. Selinker (1992: 263) maintains that persistent errors are typically the result of multiple factors, and language transfer is always one of them. To be able to reveal crosslinguistic influence, which many times can be ‘so subtle or so obscured by other factors that it cannot be detected simply by looking at the data, no matter how carefully’ (Jarvis & Pavlenko, 2008: 27), one needs to follow a principled method for verifying crosslinguistic influence on the basis of multiple types of evidence (Jarvis, 2010: 178). Jarvis (2000) and Jarvis and Pavlenko (2008) have proposed such a method, which builds on three types of evidence: intragroup homogeneity, intergroup heterogeneity and crosslinguistic performance congruity. More recently, Jarvis (2010) has introduced a

88

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

fourth type of evidence (intralingual contrasts) and has described the four types of evidence using simpler and more heuristic terminology (Jarvis, 2010: 174): within-group similarities, between-group differences, between-language differences and within-language similarities. The two first types of evidence can be either quantitative or qualitative, whereas the two last types are primarily qualitative since they build on a comparison of certain features in the languages in question. Jarvis (2010: 182) and Jarvis and Pavlenko (2008: 35) give the following explanation of the four types of evidence. (1) within-group similarities – show that the behaviour is common for a representative group of individuals with the same L1 (or who know the same combination of languages); (2) between-group differences – show that the behaviour is specific to a certain L1 group, whereas it is not characteristic of a different L1 group whose L1 differs as far as the feature in question is concerned; (3) between-language similarities – show that a language user’s behaviour in one language is motivated by the user’s behaviour in the other language; (4) within-language differences – show how a language user deals with a set of structures in the target language that are not equally congruent with corresponding structures of the source language. Methodological rigour requires all four types of evidence to be examined, even when it is possible for one of them to be substituted by another (Jarvis, 2010: 182–183). Moreover, the ideal study would compare performances by the same participants in both the L1 and L2. However, this is often unfeasible, especially when a researcher has only corpus data at their disposal. Jarvis and Pavlenko (2008: 49) indicate that a contrastive analysis relying on accurate (preferably empirically based) descriptions of the languages concerned is an appropriate alternative in such cases. The present study is an investigation of transfer in learners’ use of Norwegian adjective inflection, which is part of the language’s grammatical morphology. The possibility of transfer in morphology and syntax has often been called into question, but Jarvis and Odlin (2000: 553) and Odlin (2003: 478) has concluded that the empirical evidence available in the relevant literature clearly shows that ‘transfer is evident in all linguistic subsystems’ (Odlin, 2003). The present study offers further opportunity for evaluating this claim in relation to an area of grammatical morphology that, to my knowledge, has not previously been examined for potential crosslinguistic effects. It is worth noticing that adjective agreement is often described as redundant and devoid of meaning. Rønhovd (1993: 44) claims that it only has a syntactic function and is used to show the adjective’s connection to the headnoun (Kulbrandstad, 2005: 151), and to correctly associate the adjectives with the relevant head-noun (Bhat, 1994: 4). Because of its redundancy, it might be possible for adjective agreement not to be subject to the same crosslinguistic

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

89

effects as more meaning-driven areas of language. At this point, there is only one study examining crosslinguistic influence in adjective inflection in L2 Norwegian: Janik (2016) – a doctoral thesis focused on the acquisition of adjective-noun agreement by L1 Polish speakers, which also investigates crosslinguistic influence in learners’ performance (based on the same data set as the current study). However, there are some studies examining the process of acquiring adjective inflection in other languages (e.g. Axelsson’s (1994) doctoral thesis on the acquisition order for adjective inflection in Swedish by learners with Finnish, Polish and Spanish L1 backgrounds), one of which also concerns Norwegian (Glahn et al., 2001, where the main focus is on Pienemann’s Processability Theory (1998)).3 One of these is also Lichtman’s (2009) study on the effect of distance between the adjective and the headnoun in the L2 Spanish of L1 English speakers. She shows that learners’ knowledge of adjective agreement is not automatized at all levels of distance at the same time, but with increasing proficiency in L2 Spanish it becomes more native-like. Lichtman’s findings seem to support the claim that L2 learners have cognitive limitations that prevent them from processing all grammatical functions at the same time at the beginning of their L2 acquisition, but their processing capacities expand as they become more proficient in L2 (which also lends support to Pienemann’s Processability Theory (1998)). The starting point for the present investigation is a set of predictions for L1 influence that are motivated by similarities and differences between the Norwegian adjective inflection system and the corresponding systems of English, German and Polish.

Adjective inflection in English, German, Polish and Norwegian The three groups of participants chosen for the study are especially interesting since their native languages – English, German and Polish – show different patterns in adjective inflection. (Note: In the present chapter, the term adjective inflection refers only to adjective agreement and adverbial marking of adjectives, but not gradation.) Therefore, the differences that the speakers of the three L1s show in their performance in L2 Norwegian might be an indication of transfer. Among the languages, Polish is most similar to Norwegian as far as adjective inflection is concerned. Adjectives, both in Polish and Norwegian, must obligatorily agree with the noun in both attributive and predicative position. In both languages, adjectives show agreement with nouns in relation to three functional categories: number (both languages), gender (both languages), definiteness (Norwegian) and case (Polish). An additional similarity between these languages in relation to adjectival marking is that both use a productive morpheme for using adjectives adverbially (Hagen, 2008: 111–122; Wróbel, 2001: 127–128). Adverbial marking of adjectives is, of course, also present in English (e.g. slow + -ly). Apart from adverbial marking, there are no other forms of adjective inflection in English

90

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 5.2 Adjective inflection in Norwegian, English, Polish and German

Agreement

Feature

Norwegian

English

Polish

German

adj.inflection in attributive pos.

ü

û

ü

ü

adj.inflection in predicative pos.

ü

û

ü

û

adverbial

ü

ü

ü

û

(Leech & Svartvik, 1991: 172–176). (Note: Adverbial marking of adjectives in both English and Polish is typically regarded as a derivational process, but it is treated in this study as a matter of inflection because this is how it is regarded in Norwegian.) German, on the other hand, has no adverbial marking. However, German does have adjective agreement in attributive position, but not in predicative. German adjectives in attributive position must obligatorily agree with their head nouns in relation to four functional categories: number, gender, definiteness and case (Helbig & Buscha, 1984: 299–355). Table 5.2 shows to what extent the four languages are similar to each other as far as adjective agreement is concerned.

Hypotheses The preceding short contrastive description of the adjective systems in Norwegian, English, Polish and German motivate the following predictions concerning the effects of L1 influence on learners’ performance with respect to L2 Norwegian adjective inflection: (1) L1 English speakers will exhibit negative transfer in their use of adjective agreement because English does not have adjective agreement at all, whereas Norwegian does. L1 English speakers will nevertheless exhibit positive transfer in their adverbial marking because this kind of inflection is present in both English and Norwegian; (2) L1 Polish speakers will exhibit positive transfer in all areas of adjective inflection because the patterns of adjective inflection are mostly the same between Polish and Norwegian; (3) L1 German speakers will exhibit negative transfer in their use of adverbial marking and adjective agreement in predicative position because German does not have these kinds of adjective inflection. However, L1 German speakers will exhibit positive transfer in adjective agreement in attributive position because this type of adjective agreement is present in both German and Norwegian. As discussed earlier, learners’ processing capacity might also influence their performance. According to Pienemann’s (1998) Processability Theory, the phrasal procedure (in this study: inflecting attributive adjectives for agreement)

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

91

is easier to carry out and is acquired earlier than the S-procedure and the subordinate clause procedure (in this study: inflecting predicative adjectives for agreement). If true, this effect is universal and should apply equally to all L1 groups. However, it could also combine with L1 influence in such a way that its effects are stronger and longer-lasting in some groups than it is in others. Processability constraints combined with L1 influence might be particularly disadvantageous for L1 German speakers in relation to Hypothesis 3.

Method The data used in this study come from the ASK corpus – The Norwegian Language Learner Corpus, which is described at length in the second chapter of the book. The data for the present study include 600 texts (100 texts per each L1 group, i.e. English, German and Polish, for each of the two test levels: Språkprøven and Høyere Nivå which test language proficiency on respectively B1 and B2 levels4). The number of texts is thus balanced across L1 groups and learners’ proficiency levels. The participants are adult learners of L2 Norwegian. The youngest participants are aged 17, and the oldest 63. The average age is 31 years. Most of the participants (79%) are at B1, B1/B2 and B2 proficiency level, 5 with less than 5% of them at A2 and C1 level. The rest are at a level between A2 and B1 (9%), and between B2 and C1 (9%). They have in most cases (91%) finished secondary school as a minimum. The majority of the participants (77%) use Norwegian every day, and have taken a test to confirm their language proficiency. They are representatives of various professions and had been living in Norway from less than 1 year to more than five (on average 1.8 years) before they took the test. The participants produced the texts in a test situation. The range of topics they wrote about was wide: occupation and education, traffic, family and children, culture and language learning, politics and economy, environment, health, equal rights and more. The choice of topics was quite balanced across the groups, and in each of the texts adjectives were used. Data analysis began with the use of corpus search tools to identify and extract all occurrences of adjectives in the three L1 groups. The queries were as follows: [pos = ‘adj’] \ \\ :: language = ‘engelsk’ [pos = ‘adj’] \ \\ :: language = ‘tysk’ [pos = ‘adj’] \ \\ :: language = ‘polsk’ The process of identifying and extracting adjective occurrences was complicated by the fact that words tagged as adjectives in the ASK include not only adjectives, but also other parts of speech, included in the adjectival class either by mistake (which was occasionally the case with homonyms, e.g.

92

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

få = ‘few’, ‘to get’) or as a result of using a wider definition of an adjective than is used in this study. The parts of speech tagging in the ASK corpus was done by an automatic tagger for Norwegian bokmål (called The Oslo-Bergen tagger) based on The Norwegian Reference Grammar (NRG) (Norsk referansegrammatikk), so the tagging protocol for the ASK reflects the definition of adjectives in NRG, where adjectives are primarily defined as ‘words which first of all can be gradated6’ (either inflectionally by using suffixes -(e)re for comparative and -(e)st for superlative or syntactically by using particles mer|meir = ‘more’ and mest = ‘most’) (Faarlund et al., 1997: 22). Also, other scholars (e.g. Rønhovd (1993) and Kulbrandstad (2005)) define adjectives by prioritising their morphological property of being able to be inflected in comparisons. However, Kulbrandstad (2005: 116) maintains that also adverbs can be inflected in comparisons, whereas for Rønhovd (1993: 85) gradation is a feature characteristic only of adjectives, and which is also sufficient to define this part of speech. Among the morphological features of adjectives another type of inflection is also mentioned – agreement. Both the Faarlund et al. (1997), Rønhovd (1993) and Kulbrandstad (2005) consider the syntactic features of adjectives as subordinate to the morphological ones. At this point, Hagen (2008) gives a considerably different definition of adjectives, since he uses a different hierarchy of criteria, and prioritizes syntax. He agrees with Bhat (1994: 18) that morphological characteristics derive from functional factors. Because the adjective’s primary function is to modify the head noun by denoting a single, specific property (Bhat, 1994: 23), the attributive (and in the case of Norwegian also predicative) position can be regarded as fundamental, since in this position the adjective can fulfil its primary function. Because the topic of this study is adjective agreement, which is first of all a syntactic feature, some adjustments to the definition applied in the corpus had to be made. According to the NRG (whose definition of adjective is used in the ASK), words such as ofte (often), fort (fast, quickly) and lenge (long) are adjectives. However, these words are not compatible with the purposes of the present study, because they cannot be used in Norwegian with agreement markers that show adjective–noun agreement either in attributive or predicative position. For the present study, I operationalized the adjective class in the following way: a prototypical Norwegian adjective (1) denotes a single property, (2) occurs in both attributive and predicative position and (3) is able to be inflected for comparison and agreement (number, gender and definiteness). Adjectives that do not fulfil all these requirements (e.g. ordinal numbers) were not included in the analysis. Participles were also not included in the analysis, not even those that have become conventionally lexicalized as adjectives (e.g. komplisert = ‘complicated’). In addition, adjectives that are indeclinable were excluded from the study. This group includes adjectives in comparative form, which do not inflect for number, gender, or definiteness in Norwegian. Because adjectives were not tagged in the ASK in a way that was directly compatible with the way adjectives were operationalized for the

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

93

present study, all of the search results had to be manually checked and sorted. As far as the identification of errors was concerned, I was able to rely mostly on the existing ASK error tags, although some manual correction was needed. In the analysis, I ignored spelling errors as long as they did not involve inflectional suffixes. In cases where adjective inflectional suffixes were misspelled and I was not able to distinguish between spelling errors and inflectional errors, I categorized the incorrectly spelled/inflected adjectives as errors of inflection. The corpus data used in the analysis do not include any information about the informants’ thought processes during the writing of their essays, and there was no possibility to contact them afterwards. Adjectives used with the wrong meaning or in the wrong semantic context were not treated as errors. I analysed only whether the adjectives were inflected correctly or not. There were nevertheless some adjectives (only predicates) in the corpus that could not be categorized unambiguously as having correct or incorrect inflection because these adjectives have more than one possible form in spoken dialectal language, although the prescriptive grammar allows only one form. In these cases, the adjectives with ambiguous inflection were not excluded from the analysis. In order to verify which inflections are acceptable in Norwegian, I asked native speakers of different dialects of Norwegian to inflect the problematic adjectives in the same contexts as the learners had produced them. Where native speakers produced more than one inflection, I accepted multiple forms from the learners as well. The R statistical package was used to perform all statistical tests. This software works under General Public License (GPL) and is available through the Internet. In the R environment, it is possible to perform statistical analyses and prepare graphs and figures (Dalgaard, 2008: vii). The first test I used for my data was the ShapiroWilk test, which tests if the data are normally distributed. The test was run with data broken down by L1 group. Moreover, in every L1 group the data was further broken down by relative number of contexts for each of the inflection types used in the analysis (i.e. adjective agreement, adjective agreement in attributive and predicative position, and adverbial marking), and by relative number of errors in each of the contexts. Since the data were usually not normally distributed (there was only one case where it actually was normal – in the number of adjective agreement contexts in the L1 German group, but in the other cases it was non-normally distributed), I used non-parametric tests for three samples with no-repeated measurements, and these were: the Kruskal-Wallis test and the MannWhitney test as a post-hoc test. The Kruskal-Wallis analysis of ranks is a non-parametric equivalent of analysis of variance (ANOVA) used for multiple groups (Cantos Gómez, 2013: 45). The test was used in this study to determine with 95% probability (p = 0.05) whether the differences in the performance of the three L1 groups (English, German and Polish) were a constant feature of these groups, or whether they were random. Where a significant difference (p < 0.05) was noted in the Kruskal-Wallis test, a

94

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

post-hoc test needed to be performed to determine which pairs of groups the difference concerned, and which not, as the Kruskal-Wallis test does not give that information. The post-hoc test used in this study was the MannWhitney U-test, which is a non-parametric alternative to the t-test (Cantos Gómez, 2013). With this test, the L1 groups were compared pairwise (each pair separately). Because the test had to be repeated three times (three pairs were included in the comparison: ENG-GER, ENG-POL, POL-GER), a correction needed to be applied for multiple comparisons, in this case the Bonferroni correction. Another test was used in testing within-language differences – the Wilcoxon signed-ranks test. This test is a non-parametric repeated-samples test. It was applied here three times because two samples (relative number of errors in attributive and predicative position) from each of the L1 groups were compared. The Wilcoxon signed-ranks test is a nonparametric alternative to the t-test for paired samples (Cantos Gómez, 2013).

Results Data All the participants included in the investigation used 22,968 adjectives in total, meaning declinable adjectives in positive or superlative form which are marked for agreement in Norwegian, and adjectives marked as adverbials as well as indeclinable adjectives (e.g. bra = ‘fine, good’, annerledes = ‘different’), adjectives used in the comparative form (e.g. flere = ‘more’, bedre = ‘better’) and adjectives used as nouns (e.g. rødt = ‘red’). The last group amounts, however, to only 11 words. The present analysis included only the first two groups: adjectives used in positive and superlative form, and adjectives used adverbially. Declinable adjectives in positive and superlative form are marked for agreement in Norwegian, and occurrences of such adjectives in the present data will be referred to as adjective agreement contexts. Occurrences of adjectives used adverbially will correspondingly be referred to as adverbial contexts. The total number of adjective agreement contexts found in the data was 13,099, whereas the total number of adverbial contexts was 3509. The combined number of adjectives included in the present analysis was therefore 16,608. There were no significant differences (X2 = 1.6013, df = 2, p > 0.05) in the number of adjective agreement contexts produced by the different L1 groups, but there was a significant difference in the number of adverbial contexts produced by the L1 German and L1 English groups. The German-speaking group used significantly7 more adjectives than the Englishspeaking group, but there were no significant differences8 between the Polish L1 group and the other groups as far as the number of adverbial contexts is concerned. Each of the informants used at least one adjective in an adjective agreement context (in fact, at least three were used by native speakers of

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

95

English), but there were a few informants (15 in total) who did not use any adjectives with an adverbial function. Table 5.3 shows the number of occurrences and the relative number of errors in adjective agreement in relation to all adjective agreement contexts made by the different L1 groups. The first row of the table shows the size of the sample, which depends on the number of informants who used adjectives in agreement with nouns. Since all participants produced at least one instance of adjective agreement, the sample sizes are the same as the number of participants in each group. The table presents the mean and median for each of the three L1 groups. The penultimate row shows the number of texts (or number of participants) with no errors involving adjective agreement. The last row shows the total raw number of errors for each L1 group. Figure 5.1 shows the groups’ central tendencies and outliers in relation to their use of adjective agreement (compare with Table 5.3). These results Table 5.3 Errors in adjective agreement by L1 L1

Mean Median N texts with 0% TOTAL

ENGLISH (N = 200)

POLISH (N = 200)

GERMAN (N = 200)

no. of occurrences

% of all adj. agr. contexts

no. of occurrences

% of all adj. agr. contexts

no. of occurrences

% of all adj. agr. contexts

2.6 2 36

12.6 11.1

1.7 1 53

8.5 6.7

1.9 1 47

9.1 6.7

515

331

372

Figure 5.1 Adjective agreement errors as a percentage of all adjective agreement contexts produced by each learner group

96

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

reflect the fact that 164 (out of 200) English-speaking learners of Norwegian produced a minimum of one error in Norwegian adjective agreement. The same held true for 153 German and 147 Polish participants. In each of the groups, however, there were some outliers who produced many more errors than the average learner. Table 5.4 is very similar to Table 5.3, except that Table 5.4 deals with adverbial marking rather than adjective agreement. Thus, Table 5.4 shows the number of occurrences and the relative number of errors involving the adverbial marking of adjectives by each of the L1 groups. This time the samples are smaller because there were some informants who did not produce any adverbial contexts in their texts. As Table 5.4 shows, 116 English-, 115 German- and 95 Polish-speaking learners of Norwegian made no errors in adverbial marking. The same is seen in the boxplot below, which also illustrates the large number of outliers (which appear as circles in Figure 5.2). Table 5.4 Errors in adverbial marking by L1 L1

Mean Median N texts with 0% TOTAL

ENGLISH (N = 134)

POLISH (N = 131)

GERMAN (N = 158)

no. of occurrences

% of all advb. contexts

no. of occurrences

% of all advb. contexts

no. of occurrences

% of all advb. contexts

0.2 0 116

7.5 0

0.3 0 95

20.3 0

0.3 0 115

14.2 0

21

41

51

Figure 5.2 Adverbial marking errors as a percentage of all adverbial marking contexts produced by each learner group

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

97

Transfer analysis The results just presented were submitted to a transfer analysis based on Jarvis’s (2010) first two types of evidence (i.e. within-group similarities and between-group differences) together and the last two types of evidence (i.e. between-language differences and within-language similarities) separately. Inferential statistical tests that are designed to evaluate the significance of between-group differences generally do so by determining whether the observed between-group differences are reliably greater than the observed within-group differences. For the present analysis, I have chosen the Kruskal-Wallis test as a tool capable of providing evidence for within-group similarities and between-group differences simultaneously.

Within-group similarities and between-group differences Using the Kruskal-Wallis test, I compared the number of errors in adjective agreement produced by L1 speakers of English, German and Polish. The results of the test showed significant differences across the three groups (X2 = 18.5392, df = 2, p < 0.001***). To locate the differences more precisely, I performed pairwise Mann-Whitney tests using the Bonferroni correction. The results of these post-hoc tests indicate that the English L1 group produced significantly more adjective agreement errors than the two other L1 groups (ENG-GER: W = 23,863.5, p < 0.01**, ENG-POL: W = 24,616.5, p < 0.001***). No significant differences were found between the Polish and German groups (POL-GER: W = 20,682.5, p > 0.05). The English L1 speaking group also differs from the other two groups in relation to the adverbial marking of adjectives. A Kruskal-Wallis test combined with post-hoc Mann-Whitney tests indicated that the Englishspeaking learners of Norwegian produced significantly fewer errors in this area than the German and Polish participants (X2 = 11.8985, df = 2, p < 0.01**, ENG-POL: W = 21,739, p < 0.05*, ENG-GER: W = 17,526.5, p < 0.01** with Bonferroni correction). Once again, there was no significant difference between the Germans and Poles (POL-GER: W = 19193, p > 0.05 with Bonferroni correction).

Between-language similarities Between-language similarities should ideally be established by comparing the participants’ use of their native languages with their use of Norwegian, which is their L2. However, because I do not have samples of the participants’ use of their native languages, I have used translations of their Norwegian texts that are consistent with the grammar rules of their native languages. I translated the texts myself (being a native speaker of Polish), and discussed the translations with native speakers of English and German. I focused on how the texts would sound in English, German and Polish.

98

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Because there is no adjective agreement inflection in the English language, one can expect that English-speaking learners of Norwegian would avoid inflecting adjectives in the target language. Among the 533 adjective agreement errors the English test takers produced, 64.2% of them involve the lack of a suffix where one is obligatory in Norwegian. This means that nearly two-thirds of all the adjective agreement errors made by the L1 English group are due to the lack of any inflectional suffix where one is needed (which is also evidence of within-group similarities). The remaining errors involve the use of the wrong suffix. Below are some examples of errors in the L1 English learners’ performance in Norwegian adjective agreement. s01949: ‘Hofflig naboer… Vi har to nær naboer.’ (*hofflig->høflige, *nær->nære)10 (Kind neighbours… We have two close neighbours.11) s0014: ‘Alle disse tingene er viktig for oss som er innflytninger.’ (*viktig->viktige) (All the things are important for us who are immigrants.) s0018: ‘…er Norge det beste mulig land å bo på.‘ (*mulig->mulige) (Norway is the best possible country to live in.) s0309: ‘Jeg liker dette for vi har et aktiv sosial liv og…’ (*aktiv->aktivt, *sosial->sosialt) (I like it for we have an active social life and…) h0510: ‘Derfor er det absolute til vår langsiktig fordel å vise respekt og forsvarlighet…’ (*langsiktig->langsiktige) (Therefore it is absolutely for our long-term advantage to show respect and awareness…) h0271: ‘Over hele verden er forurensing et stor problem, men…’ (*stor->stort) (Throughout the world pollution is a big problem, but…)

The congruence between English’s lack of adjective inflections and the English-speaking participants’ omission of obligatory adjective inflections in Norwegian serves as a necessary piece of evidence for negative transfer from L1 English to Norwegian. By contrast, the English speakers’ incorrectly inflected adjectives, as in the following example, might be regarded simply as a failed attempt to inflect Norwegian adjectives, as the adjective ny has been correctly inflected in gender (neuter), but the definite form has been disregarded (nytt is a neuter indefinite form, whereas nye is definite form for all genders). s0309: ‘Selfolgelig må du prove å lære det nytt lands språk men…’ (*nytt->nye) (Of course you should try to learn the new country’s language but…)

As far as errors in the adverbial marking of adjectives are concerned, the L1 English learners of Norwegian make very few errors of this type. They

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

99

produced only 21 such errors in total, with an error-rate of 0.11%. The sentence below illustrates a correctly inflected adjective functioning as an adverb. h0585: ‘Det vil alltid være folk som kjører enten for fort eller for seint.’ (sein + -t) (There will always be people who drive either too fast or too slowly.)

Adjective agreement is characteristic of both Polish and Norwegian, and the adjective agreement system in Polish is in fact more complex than in Norwegian due to the diversity of suffixes that Polish adjectives can take, and also due to the multiple grammatical functions they represent. There are 12 different adjective suffixes in Polish (Wróbel, 2001: 127–128), and only three in Norwegian (Golden et al., 2014: 20). In both languages, adjective inflection is obligatory in attributive as well as in predicative position, and in both languages adjectives agree with their governing nouns in relation to three grammatical functions: number, gender (both languages), definiteness (Norwegian) and case (Polish). Considering the fact that Polish learners of Norwegian come from a more complex to a simpler system, it can be assumed that L1 Polish speakers will not find it difficult to inflect Norwegian adjectives, since they can rely on their knowledge of Polish adjective inflection while using L2 Norwegian. In reality, however, Polish-speaking learners also struggle with adjective agreement in Norwegian. Most of their errors (61.4% of all errors in adjective agreement) involve the lack of a suffix where one is obligatory in Norwegian for purposes of adjective-noun agreement. The degree of suffix omission is about the same as the English speakers show. That might raise questions about whether the English speakers’ tendency toward omission really is the result of L1 influence, and not e.g. a result of simplification. The essential difference, however, is that in the case of the English speakers this type of error is congruent with their L1, whereas it is not in the case of the Polish speakers. Suffix omission errors seem to be the most common type of error in all three L1 groups. In other cases, Polish-speaking learners use an incorrect suffix. The following examples illustrate some of the types of adjective agreement errors Polish-speaking learners of Norwegian make: s0772: ‘Det er veldig pen og moderne land.’ (*pen->pent) (To jest bardzo pi˛ekny i nowoczesny kraj. It is a very beautiful and modern country.) s0949: ‘…og det kan være en grunn for vår dårlig hølse.’ (*dårlig-> dårlige) (…i to moz˙e być przyczyna˛ naszego złego zdrowia. …and it can be a reason for our poor health.) s0940: ‘Vi må være veldig forsiktig.’ (*forsiktig-> forsiktige) (Musimy być bardzo ostroz˙ni. We must be very careful.) s0948: ‘Det er stort trafikk der.’ (*stort-> stor) (Tam jest duz˙y ruch. There is heavy traffic.)

100

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

h0442: ‘Jeg synes at foreldrene har et stor ansvar for barnas - særlig guttenes leseferdigheter.’ (*stor->stort) (Uwaz˙am, z˙e rodzice ponosza˛ du z˙a˛ odpowiedzialnos´ć za umiej˛etnos´ć czytania dzieci – zwłaszcza chłopców. I think that parents have a great responsibility for children’s – especially boys’ – reading skills.) h0081: ‘De har tid for å nytte den vakker naturen som vokser rundt de.’ (*vakker->vakre) (Maja˛ czas, by rozkoszować si˛e ta˛ pi˛ekna˛ przyroda˛, która ros´nie wokół nich. They have time to enjoy the beautiful nature that grows around them.) h0646: ‘Vi er glad når vi kan ringe kjærest eller…’ (*glad->glade) (Jestes´my zadowoleni, kiedy moz˙emy zadzwonić do ukochanego albo… We are glad when we can call a boyfriend or…)

In Polish, as well as in Norwegian, it is possible to use adjectives with an adverbial function after adding a suffix to the adjective’s stem. Polish-speaking learners of L2 Norwegian are therefore familiar with the process of adding suffixes to adjectives in order to render them as adverbs. However, the results show several cases (41 in total) where Polish participants produce errors related to the adverbial marking of adjectives. 18% of all Polish participants made one or more errors (mean = 0.21 errors) where they failed to inflect an adjective for adverbial function, as illustrated in the following example. h0646: ‘Ansatte i den bransjen bruker kroppen aktiv hele dagen.’ (*aktiv->aktivt) (Zatrudnieni w tej branz˙y uz˙ywaja˛ ciała aktywnie cały dzien´e. The employees in this sector use body actively the whole day.)

Adjective agreement is present in both German and Norwegian. However, there is a difference in the extension of adjective agreement in German and Norwegian: Norwegian adjectives are inflected in both attributive and predicative position, while German adjectives are inflected only in attributive position, whereas they remain uninflected in predicative position. The differences in the German L1 speakers’ performance in inflecting Norwegian adjectives in attributive and predicative position are described in the following section on within-language differences. In the meantime, the following examples illustrate the types of L1-congruent errors that the Germanspeaking participants produced in L2 Norwegian: s0460: ‘…alle mennesker skal ble syk av disse nye elektroniske tingene.’ (*syk->syke) (… alle Leute werden krank von den neuen elektronischen Dingen [werden]. … all the people will be sick of these new electronic things.) s0471: ‘…men vi kan være kritisk og ikke tror på alt…’ (kritisk->kritiske) (…aber wir können kritisch sein und nicht an alles glauben… …but we can be critical and not believe in everything) s0961: ‘Og hva med barna som er mellom 8 og 11 år gammel?’ (*gammel->gamle) (Und was mit den Kindern, die zwischen 8 und 11 Jahre alt sind? And what about children who are between 8 and 11 years old?)

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

101

h0396: ‘…og fordi livet blitt så ubetydningsfull for henne…’ (*ubetydningsfull-> ubetydningsfullt) (…und weil das Leben so bedeutungslos für sie ist… … and because life has become so meaningless for her…) h0557: ‘Jeg tror det er bare språket som er ulik…’ (*ulik->ulikt) (Ich glaube, dass es nur die Sprache ist, das unähnlich ist… I think it is only the language that is different…) h0421: ‘Vi spiser dårlig, beveger oss dårlig, blir ulykkelig…’ (*ulykkelig-> ulykkelige) (Wir essen schlecht, bewegen uns schlecht, sind unglücklich… We do not eat well, move too little, become unhappy…)

Because there is no adverbial inflection in German, and because German adjectives can be used as adverbs without the addition of any overt morpheme, it can be expected that German learners of L2 Norwegian will use the same pattern in their L2 as they use in their L1. That is, from the perspective of crosslinguistic influence, German-speaking learners of Norwegian can be predicted not to inflect Norwegian adjectives when they function as adverbs. The results show, however, that this is true only in some cases: 21.5% of the German participants did so, and they produced on average 0.26 errors of this type per individual. Following is an example of this type of error: h0402: ‘Norge er et land med en industri som er sterk eksportorientert og…’ (*sterk->sterkt) (Norwegen ist ein Land mit Industrie, die stark exportorientiert ist… Norway is a country with an industry, which is highly export oriented…)

Within-language differences To evaluate evidence for within-group differences, I have divided all the errors involving adjective agreement that the learners made into two categories: adjective agreement errors occurring in attributive versus predicative position. Tables 5.5 and 5.6 have the same structure as Tables 5.3 and 5.4. They show the number of occurrences and the relative number of errors in adjective agreement in attributive (Table 5.5) and in predicative (Table 5.6) position in relation to all adjective agreement contexts in respectively attributive and predicative position for each of the different L1s. Figures 5.3, 5.4 and 5.5 illustrate the differences in number of errors in both positions made by, respectively, English-, Polish- and German-speaking learners of L2 Norwegian. Using the Wilcoxon signed-ranks test, I tested whether any of the L1 groups showed significant differences in the relative number of errors in the two positions. In the case of both the English and the Polish L1 group, there is no significant difference in the number of errors between the attributive and predicative position (ENG: V = 5346.5, p > 0.05, POL: V = 5290.5, p > 0.05, Bonferroni correction). Native speakers of German who learn

102

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 5.5 Errors made in adjective agreement in attributive position by L1 L1

Mean Median N texts with 0% TOTAL

ENGLISH (N = 198)

POLISH (N = 200)

GERMAN (N = 199)

no. of occurrences

% of all adj. agr. contexts in attr. pos.

no. of occurrences

% of all adj. agr. contexts in attr. pos.

no. of occurrences

% of all adj. agr. contexts in attr. pos.

1.3 1 79

11.5 6.7

0.9 1 81

9.0 5.5

0.8 0 103

6.7 0

260

188

162

Table 5.6 Errors made in adjective agreement in predicative position by L1 L1

Mean Median N texts with 0% TOTAL

ENGLISH (N = 200)

POLISH (N = 199)

GERMAN (N = 200)

no. of occurrences

% of all adj. agr. contexts in pred. pos.

no. of occurrences

% of all adj. agr. contexts in pred. pos.

no. of occurrences

% of all adj. agr. contexts in pred. pos.

1.3 1 74

14.7 10

0.7 0 114

8.9 0

1.1 1 89

13.4 8.3

255

143

210

Figure 5.3 L1 English group’s adjective agreement errors as a percentage of all adjective agreement contexts they produced in attributive (left) and predicative (right) positions

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

103

Figure 5.4 L1 Polish group’s adjective agreement errors as a percentage of all adjective agreement contexts they produced in attributive (left) and predicative (right) positions

Figure 5.5 L1 German group’s adjective agreement errors as a percentage of all adjective agreement contexts they produced in attributive (left) and predicative (right) positions

Norwegian as a L2, however, make significantly more errors when they use Norwegian adjectives in predicative function than in attributive position (V = 3233, p < 0.001***).

Summary The between-group differences indicate that the English-speaking informants make more errors in adjective agreement than the German and Polish informants, and fewer errors in adverbial marking of adjectives than the two

104

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

other groups, who make more or less the same number of errors in both types of inflection. Furthermore, the results show that some of the errors made by the English and the German informants are congruent with their native languages (no-inflection in English and no-inflection in predicatives in German), whereas some are not. The errors made by the Polish learners of Norwegian are not congruent with the Polish language. Finally, native speakers of English and Polish make approximately as many errors in agreement in attributive position as in predicative. Native speakers of German make significantly more errors in predicative position than in attributive.

Discussion The summary above helps to clarify whether the analysed errors are caused by crosslinguistic influence. According to Jarvis and Pavlenko’s (2008) and Jarvis’s (2010) methodological framework, researchers should consider all four of the earlier mentioned types of evidence before making claims about the occurrence of transfer from one language to another. This strong methodological principle does not, however, imply that all four types of evidence must necessarily be found in order to verify transfer (Jarvis, 2010: 173). In accordance with the framework, one can argue that English learners of L2 Norwegian produce instances of negative transfer when they fail to inflect Norwegian adjectives for adjective-noun agreement. The English group is both internally homogeneous and different from the two other groups investigated (i.e. the German and Polish groups), as it produced significantly more errors in agreement. The agreement errors the English participants produced are congruent with the nature of the English adjective system, which does make use of agreement markers. Moreover, the English participants made as many errors in attributive position as in predicative position, and inflecting adjectives seems to be equally difficult for them. The hypothesis that English learners of L2 Norwegian will exhibit negative transfer in their (non)use of Norwegian adjective agreement is therefore confirmed, similar to the hypothesis that English speakers will exhibit positive transfer in their use of Norwegian adjectives as adverbials. The present study confirms that native speakers of English benefit from their mother tongue when they use Norwegian adjectives in their adverbial function. In this case, the L1 group is homogeneous as a whole and produced significantly fewer errors than the other two L1 groups. The performance of the English participants is, moreover, congruent with the standard use of English, where adjectives can be transformed into adverbs through the addition of a suffix (i.e. -ly). The hypotheses related to the German group are also confirmed. The German group is homogeneous in its inflection of adjectives in L2 Norwegian in relation to both agreement and adverbial marking, and this group also produced significantly fewer adjective agreement errors than the English group.

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

105

At the same time, the German group also produced significantly more adverbial marking errors than the English group. The adverbial marking errors the German participants produced are congruent with their native language, and this is also the case for some of the adjective agreement errors they made (i.e. those involving predicative adjectives). The German learners of Norwegian made significantly more adjective agreement errors in predicative position than in attributive, and this mirrors the fact that adjectives in German are not inflected when used as predicatives. This tendency, which appears to be a clear effect of crosslinguistic influence, might be strengthened by processability constraints. In accordance with Pienemann’s (1998) Processability Theory, the inflection of attributives may be easier to process (and learn) than the inflection of predicative adjectives, given that agreement in the former case occurs within a single phrase, whereas in the case of predicative adjectives, the scope of agreement marking encompasses a whole clause or sentence. The effects of transfer in the Polish group are less straightforward. Indeed, the hypotheses related to the Polish group were only partially confirmed. On the one hand, the Polish group was homogeneous in its adjective inflection in L2 Norwegian (both in agreement and in adverbial marking), and it made significantly fewer adjective agreement errors than the English group. The performance differences between these two L1 groups might plausibly have been caused by differences in their L1s, namely, whether their L1 has adjective inflection (Polish) or not (English). The adjective agreement errors they made in Norwegian are not congruent with Polish. The Polish speakers’ agreement errors also do not vary by position (i.e. attributive vs. predicative). On the other hand, I have not found any evidence for betweenlanguage similarities in the adjective inflections of the Polish participants. The errors the Polish-speaking learners produce in adjective inflection might have been a consequence of their not having acquired the category of definiteness, and the system of gender assignment in Norwegian, or of a negative influence from L2 English (as 58% of L1 Polish speakers have learned the language to at least the intermediate level12) or other L2s. The potential positive influence from L1 Polish on L2 Norwegian adjective agreement might therefore remain unrealized in this situation due to negative transfer related to definiteness and gender assignment, and / or due to negative transfer from L2 English (or other L2s). This explanation is, nevertheless, only speculative until it can be tested empirically. In the meantime, it is noteworthy that the Polish participants’ adverbial marking errors were also unexpectedly more numerous than those of the English group, even though the adverbial marking of adjectives is present in both Polish and English. But here, too, the significance level was low, and the high number of outliers in the Polish group is likely to have caused the difference between the two groups to become exaggerated. Even so, the real difference in the occurrence of errors in this area is not very large. In any case, the adverbial marking errors the Polish participants made are not congruent with the Polish language, where

106

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

adjectives, as in Norwegian, must also be morphologically marked. Despite the fact that there is no satisfactory evidence for claiming positive transfer from L1 Polish in the Polish participants’ adverbial marking of adjectives in L2 Norwegian, it might be a mistake to completely discount the possibility of transfer from Polish. Kellerman (1979) observed that learners’ judgements are affected by their proficiency in the L2 and by their perceptions and assumptions of what is transferable from the L1 to the L2. Importantly, in the present study, the English, German and Polish groups are evenly balanced in their L2 Norwegian proficiency.13 Thus, instead of differences in proficiency, it may be differences in their perceptions and assumptions of the distance between their L1 and the L2 that to some extent explain why the Polish learners alone do not rely on their L1 when dealing with adjective agreement in L2 Norwegian. They might perceive Polish and Norwegian to be such distant languages that they do not view the L1 patterns as relevant to the L2, and as a result they do not rely on their knowledge of inflecting Polish adjectives when they inflect adjectives in L2 Norwegian. On the other hand, the English and German speakers might perceive L2 Norwegian as so similar and close to their L1s that they more willingly rely on their knowledge of L1, even though it in many cases leads to errors. This possibility, as it relates to adjective inflection, should of course be tested in psycholinguistic research. In my analysis, I have not considered other forms of crosslinguistic influence than positive and negative transfer (i.e. avoidance, underuse and overuse), because such an analysis is difficult, if not impossible, to perform with the kind of data I have. On the one hand, all participants in the present study used adjectives, and they all exhibited an ability to use adjective-noun agreement, irrespective of their level of proficiency and the topic of their essays. On the other hand, the nature of the task – a composition – allowed them to choose ‘easier’ adjectives and use them in language structures that they already had mastered; they wrote the essays in a test situation, and it seems certain that most of them tried to score as highly as possible by using forms and structures that they were already quite proficient with. Perhaps a more qualitative analysis could find other forms of crosslinguistic influence than positive and negative transfer. The results of the present study provide support for Jarvis and Odlin’s (2000: 553) and Odlin’s (2003: 478) claim that crosslinguistic influence can occur in the domains of grammatical morphology and syntax. This might have already been clear as a general principle, but it certainly was not clear whether or how crosslinguistic influence would manifest itself in the area of adjective inflection. The results of the present study thus confirm the general principle as well as contribute detailed findings regarding the mechanisms and conditions through which crosslinguistic influence occurs in a specific and neglected area of grammatical morphology, which might be even regarded as redundant and devoid of meaning.

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

107

This study has not supported any evidence that might confirm Lichtman’s conclusions concerning the effect of distance between the adjective and its head noun on the ability of processing in the L2, as both the English and the Polish L1 group exhibited the same level of errors in the two adjective agreement positions: attributive and predicative. Only the German L1 group produced more errors in predicative position than in attributive, and this might be explained by crosslinguistic influence, perhaps in combination with processability constraints.

Conclusions The purpose of this study was to find evidence for positive and negative transfer in learners’ use of adjective inflection in Norwegian as a second language. Using data from the ASK corpus I have performed a transfer investigation using Jarvis and Pavlenko’s (2008) and Jarvis’s (2010) methodological framework. The results of this study confirm the occurrence of positive transfer in the adverbial marking of adjectives by English-speaking learners of Norwegian, and also in the use of adjective agreement in attributive position by German-speaking learners of Norwegian. The results also provide evidence for negative transfer in the adjective agreement errors of the L1 English group, and also in the adjective agreement errors in predicative position and in the adverbial marking errors of the L1 German group. The results also suggest traces of positive transfer in the L1 Polish group’s use of adjective inflection. I use the term traces because I cannot fully confirm crosslinguistic influence here. It is possible that potential positive transfer in Polish speakers’ use of adjective inflection is obscured by negative transfer related to definiteness and gender assignment, and / or negative influence from other L2s, e.g. English. This is an important question for future research. Another fruitful direction for future research would be a psycholinguistic investigation of Polish learners’ perceptions of the distance (or relationship) between Polish and Norwegian, as well as their perception of which types of forms, structures, and meanings are transferable between the languages. There is also a need to explore what other factors might affect the acquisition of adjective inflection in Norwegian.

Notes (1) The adjectival inflectional categories: SG – singular, N – neuter, INDEF – indefinite, POS – positive. (2) The other adjectival inflectional categories: PL – plural, F – feminine, DEF – definite. (3) Actually, Norwegian, Danish and Swedish were considered to be one language in this study, as adjective–noun agreement is generally the same in all of these Scandinavian languages. (4) Proficiency levels determined by the CEFR scale. (5) The apparent mismatch between the test levels and learners’ proficiency levels is caused by the fact that the learners’ texts have been reread and reassessed after adding

108

(6) (7) (8) (9) (10) (11) (12)

(13)

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

them to the ASK corpus. Each text has been separately linked to the CEFR scale by an experienced evaluator, so that it is possible to check the correlation between e.g. number of errors made by learners and their proficiency level measured in CEFR scale. My translation – MOJ. Kruskal-Wallis test: Kruskal-Wallis chi-squared = 17.8105, df = 2, p < 0.001*** & the post-hoc test – Mann-Whitney test: ENG-GER: W = 24,951, p < 0.001*** (Bonferroni correction). The post-hoc test – Mann-Whitney test: POL-GER: W = 22,237, p > 0.5 (Bonferroni correction), POL-ENG: W = 22,474.5, p > 0.05 (Bonferroni correction). The informant’s number. Should be read in this way: wrong/incorrect use of høflig/nær instead of correct form høflige/nære. All translations are mine –MOJ. The ASK corpus provides information about the learners’ proficiency in L2 English. These data are self-reported, and involve four levels: beginner, basic, intermediate and advanced. However, there is no information given about how these proficiency designations should be interpreted. Also, there is no information given about the learners’ knowledge of other L2s. Kruskal-Wallis test: Kruskal-Wallis chi-squared = 12.4584, df = 6, p > 0.05.

References Axelsson, M. (1994) Noun Phrase Development in Swedish as A Second Language. A Study of Adult Learners Acquiring Definiteness and the Semantics and Morphology of Adjectives. Stockholm University: Centre for Research on Bilingualism. Bhat, D.N.S. (1994) The Adjectival Category: Criteria for Differentiation and Identification. Amsterdam: John Benjamins Publishing Company. Cantos Gómez, P. (2013) Statistical Methods in Language and Linguistic Research. Sheffield: Equinox Publishing Ltd. Dalgaard, P. (2008) Introductory Statistics with R (2nd edn). New York: Springer. Ellingsen, E. and Mac Donald, K. (2004) På vei. Norsk og samfunnskunnskap for voksne innvandrere (2nd edn). [On the way. Norwegian and social studies for adult immigrants]. Oslo: Cappelen. Faarlund, J.T., Lie, S. and Vannebo, K.I. (1997) Norsk referansegrammatikk [Norwegian reference grammar]. Oslo: Universitetsforlaget. Glahn, E., Håkansson, G., Hammerberg, B., Holmen, A., Hvenekilde, A. and Lund, K. (2001) Processability in Scandinavian second language acquisition. Studies in Second Language Acquisition 23 (3), 389–416. Golden, A., Mac Donald, K. and Ryen, E. (2014) Norsk som fremmedspråk. Grammatikk. [Norwegian as a foreign language. Grammar]. Oslo: Universitetsforlaget. Hagen, J.E. (2008) Norsk grammatikk for andrespråkslærere [Norwegian grammar for second language teachers]. Oslo: Gyldendal Akademisk. Helbig, G. and Buscha, J. (1984) Deutsche Grammatik. Ein Handbuch für den Ausländerunterricht [German grammar. A handbook for teaching foreigners] (pp. 299–355). Leipzig: VEB Verlag Enzyklopädie. Janik, M.O. (2016) Adjektivbøyning i norsk – kongruenstilegnelse hos polskspråklige norskinnlærere [Inflections of adjectives in Norwegian – acquisition of congruence by Polish learners of Norwegian]. Poznan´: Wydział Neofilologii UAM. Jarvis, S. (2010) Comparison-based and detection-based approaches to transfer research. In L. Roberts, M. Howard, M. Ó Laoire and D. Singleton (eds) EUROSLA Yearbook 10 (pp. 169–192). Amsterdam: Benjamins.

Posit ive and Negat ive Transfer in the L2 Adjec t ive Inf lec t ion

109

Jarvis, S. and Odlin, T. (2000) Morphological type, spatial reference, and language transfer. In Studies in Second LANGUAGE Acquisition 22, 535–556. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York: Routledge. Kellerman, E. (1979) Transfer and non-transfer: where we are now. Studies in Second Language Acquisition 2, 37–57. Kulbrandstad, L.A. (2005) Språkets mønstre. Grammatiske begreper og metoder. [The language’s patterns. Grammatical terms and methods]. (3rd edn) Oslo: Universitetsforlaget. Leech, G. and Svartvik, J. (1991) A Communicative Grammar of English (pp. 172–176). Harlow: Longman. Lichtman, K. (2009) Acquisition of Attributive and Predicative Adjective Agreement in L2 Spanish. In M. Bowles et al. (eds) Proceedings of the 10th Generative Approaches to Second Language Acquisition Conference (GASLA 2009) (pp. 231–247). Somerville, MA: Cascadilla Proceedings Project. Odlin, T. (1989) Language Transfer: Cross-linguistic Influence in Language Learning. Cambridge: Cambridge University Press. Odlin, T. (2003) Cross-linguistic influence. In C.J. Doughty and M.H. Long (eds) The Handbook of Second Language Acquisition (pp. 436–486). Singapore: Blackwell Publishing. Papazian, E. and Helleland, B. (2012) Norsk talemål. Lokal og sosial variasjon. [Norwegian vernaculars. Local and social variation]. Kristiansand: Høyskoleforlaget. Pienemann, M. (1998) Language Processing and Second Language Development; Processability Theory. Amsterdam: John Benjamins. Pienemann, M. (2003) Language processing capacity. In C.J. Doughty and M.H. Long (eds) The Handbook of Second Language Acquisition (pp. 679–714). Singapore: Blackwell Publishing. Pienemann, M., Di Biase, B., Kawaguchi, S. and Håkansson, G. (2005a) Processability, typological distance and L1 transfer. In M. Pienemann (ed.) Cross-linguistic Aspects of Processability Theory (pp. 85–116). Amsterdam/New York: John Benjamins Publishing Company. Pienemann, M., Di Biase, B., Kawaguchi, S. and Håkansson, G. (2005b) Processing Constraints on L1 Transfer. In A.M.B. de Groot and J.F. Kroll (eds) Handbook of Bilingualism: Psycholinguistic Approaches (pp. 128–153). Oxford: Oxford University Press. Rønhovd, J. (1993) Norsk morfologi [Norwegian morphology]. Oslo: Ad Notam Gyldendal. Selinker, L. (1992) Rediscovering Interlanguage. London: Longman. Vigeland, B. (1995) Norske dialektar. Oversyn, heimfesting og normalisering. [Norwegian dialects. Overview, localization and standardization]. Oslo: Universitetsforlaget. Wróbel, H. (2001) Gramatyka j˛e zyka polskiego. Podr˛e cznik akademicki. [Polish Grammar. University handbook] Kraków: Spółka Wydawnicza, OD NOWA’ s.c.

Internet sources ASK’s homepage: clarino.uib.no/ask The Oslo-Bergen tagger – a grammatical tagger for Norwegian Bokmål and Nynorsk: http://tekstlab.uio.no/obt-ny/index.html

6

Gender Assignment and L1 Transfer in Norwegian Second Language Learners’ Written Performance1 Silje Ragnhildstveit Western Norway University of Applied Sciences

Introduction The primary focus of this study is how learners’ native language affects their acquisition and use of grammatical gender in Norwegian as a second language. It deals with two general questions: (a) whether having a native language (L1) with a grammatical gender system more similar to the target language (L2) facilitates the acquisition of grammatical gender, and (b) whether it is facilitative to have an L1 with grammatical gender when acquiring an L2 with grammatical gender. Similar questions have been investigated in other studies (e.g. Ellis et al., 2012; Sabourin, 2001; Sabourin et al., 2006; Spinner & Juffs, 2008). Results from these studies together do not exclusively indicate that it might be more facilitatiive to have an L1 with gender than to have an L1 without gender, but results from Sabourin (2001) and Sabourin et al. (2006) indicate that it might be more facilitative to have an L1 with a gender system more similar than less similar to the target language. A well-known method in the investigation of transfer is to compare the interlanguages of learners that have L1s that differ with respect to the linguistic phenomenon under investigation. The selection of L1s for the present study is based on whether the language is a gender language (German, Dutch and Spanish) or not (Vietnamese and English). The three gender languages were selected so as to represent different types of relationships with the Norwegian grammatical gender system: German has the same three genders as Norwegian (masculine, feminine and neuter), while Dutch and Spanish 110

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

111

Table 6.1 The gender system in the target language and the native languages Target language

Native language

Language

Norwegian

German (GER)

Dutch (DUT)

Spanish (SPA)

Vietnamese (VIE)

English (ENG)

Gender system

Masculine Feminine Neuter

Masculine Feminine Neuter

Common

Masculine Feminine

– Classifier language

–

Neuter

have only two genders (common/neuter and masculine/feminine, respectively). See Table 6.1. The German gender system is viewed as most similar to the target language system because the three-gender system is the most common gender system across the current varieties of Norwegian (see section ‘Gender and the target language Norwegian’). The Spanish gender system is viewed as most dissimilar to the target language in that it does not have a neuter. This study investigates gender assignment in adult second language learners’ written performance according to the use (correct and incorrect) and non-use (omission) of the indefinite articles en (masculine), ei (feminine) and et (neuter). The investigation is corpus-based, using ASK – Norsk andrespråkskorpus (The ASK corpus – a learner corpus of Norwegian as a second language). The texts were written by L2 learners of Norwegian as part of the intermediate-level test Språkprøven. All participants had an equal time limit and none of them used any examination support material. The written standard of Norwegian used by the informants is Bokmål (see section ‘Gender and the target language Norwegian’). Although the scope of transfer research includes questions about how any language a person knows influences his or her use of the target language (Odlin, 1989: 27), the present study focuses more specifically on how the participants’ L1 affects their acquisition and use of Norwegian as a second language. The ASK corpus does provide information about the participants’ knowledge of L2 English – though not of any other L2s – but since English does not have grammatical gender, the present study does not investigate how the participants’ knowledge of L2 English, in addition to for instance L1 German, might influence their acquisition and use of grammatical gender in L2 Norwegian.

Literature Review The literature review first gives a general introduction to grammatical gender and a brief contrastive analysis of the gender system in Norwegian

112

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

and the L1 gender languages. Subsequently, relevant theory and earlier studies are presented.

Grammatical gender One of the most frequent definitions of grammatical gender is Hockett’s: ‘Genders are classes of nouns reflected in the behavior of associated words’ (Hockett, 1958: 231). In this definition, two aspects of gender are included: nominal classification and agreement. According to Corbett (1991), agreement is a generally accepted defining criterion for gender (1991: 4), where the gender of a noun determines the form of ‘associated words’. Gender assignment, i.e. how gender is allotted to nouns, is a type of nominal classification. Gender assignment has been viewed as unpredictable in different gender languages, as claimed by Bloomfield (1933: 280) who said that ‘there seems to be no practical criterion by which the gender of a noun in German, French, or Latin could be determined’. However, this view is no longer prevalent. According to Corbett (1991: 7–8), ‘[gender] assignment may depend on two basic types of information about the noun: its meaning (semantics) and its form [morphological and phonological]’. The gender system in different languages consists of different combinations of this information and can have both semantic and formal rules for gender assignment. Furthermore, the number of nouns with rule-governed gender assignment differs between gender languages. Corbett also emphasizes that every gender system has a semantic core, meaning that a gender system is never purely formal (Corbett, 1991: 307).

Gender and the target language Norwegian Norwegian has three genders: masculine, feminine and neuter. In written Norwegian, there are two official written standards named Bokmål and Nynorsk. In Bokmål, there is freedom of choice when it comes to the use of feminine, i.e. it is possible to write Bokmål using masculine and neuter only, while Nynorsk uses the three-gender system. In spoken Norwegian, the three-gender system is the most common. According to Faarlund et al. (1997: 152), the three-gender system is used in the traditional dialects, except the dialect in the city of Bergen (in the south-western part of Norway), which only uses the two-gender system. Gender assignment in Norwegian may appear to be unpredictable, but there are some semantic and formal rules that govern it. There is a connection between biological sex and masculine and feminine gender. For instance, mann (‘man’), gutt (‘boy’), and bror (‘brother’) are masculine, whereas kvinne (‘woman’), jente (‘girl’), søster (‘sister’) are feminine. There are also formal rules for gender assignment that are connected to the derivational morphology of the noun. Examples of some of the most common relationships between grammatical gender and derivational morphology in Norwegian are given in Table 6.2 (adapted from Mac Donald (2009: 9)):

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

113

Table 6.2 Examples of relationships between derivational morphology and gender assignment in Norwegian Masculine

Feminine

Neuter

-sjon, e.g., stasjon (‘station’) -else, e.g., forståelse (‘understanding’) -ning, e.g., bygning (‘building’) -dom, e.g., barndom (‘childhood’) -nad, e.g., søknad (‘application’) -skap, abstract nouns e.g., kunnskap (‘knowledge’) -isme, e.g., kapitalisme (‘capitalism’)

-inne, e.g., venninne (‘female friend’) -erske, e.g., syerske (‘seamstress’) -ing, e.g., regjering (‘government’)

-ment, e.g., fundament (‘foundation’) -eri, e.g., bakeri (‘bakery’) -dømme, e.g., kongedømme (‘monarchy’) -skap, including persons e.g., fellesskap (‘fellowship’) -gram, e.g., program (‘program’) -um, e.g., publikum (‘audience’) -em, e.g., problem (‘problem’) -tek, e.g., bibliotek (‘library’) -mål, e.g., spørsmål (‘question’)

Regarding gender agreement, determiners and adjectives are the relevant categories. Table 6.3 shows determiners in Norwegian and their forms in the masculine, feminine and neuter gender. The indefinite articles are in bold. According to Norsk referansegrammatikk [Norwegian reference grammar], the indefinite articles can be viewed as the most important gender markers in Norwegian, in addition to the definite suffix on the noun itself (Faarlund et al., 1997: 150). The indefinite article in Norwegian is usually obligatory when a countable noun is in singular indefinite form and the noun phrase is referring to something that the speaker will say more about later in the discourse. The speaker can leave out the indefinite article when the countable noun in singular indefinite form is part of a holistic phrase with idiomatic content and the noun phrase is referring to something that the speaker will not talk more about (Hagen, 2000: 226–227). Table 6.4 displays examples of gender agreement in extended noun phrases. Tables 6.3 and 6.4 illustrate that there is a good deal of syncretism between masculine and feminine gender.

Gender and the L1 groups Of the L1s included, German, Dutch and Spanish have gender, while English and Vietnamese do not. A brief presentation of the gender system in German will be given, after which a brief presentation of English and Vietnamese is provided. Since there will be only limited attention paid to the L1 Dutch and Spanish groups in this chapter, brief descriptions of grammatical gender in these languages have been placed in Appendix 1.

114

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 6.3 Determiners in Norwegian and their gender forms* Demonstratives M

Possessives F

den (‘that one’) den denne (‘this one’) denne slik (‘such’) slik sånn (‘such’) annen (‘other’) hvilken (‘which’)

Quantifiers

N

M

F

N

M

F

N

det dette slikt

min (‘my’) din (‘your’) sin (2)

mi di si

mitt ditt sitt

noen (‘some’) ingen (‘no’)

noen inga

noe intet

ei éi all hver enhver

et ett alt hvert ethvert

en (‘a/an’) sånn sånt vår (‘our’) vår vårt én (‘one’) anna annet egen (‘own’) egen eget all (‘all’) hvilken hvilket hver (‘each’) enhver (‘every’)

*The English translation is not exhaustive

German has the same three gender categories as the Norwegian target language (masculine, feminine and neuter). However, the gender of equivalent nouns in German and Norwegian can be different. As in Norwegian, the semantics or the form of a noun can determine which gender a noun has in some cases. Askedal (1976: 30–31) and Reiten (2003: 47–51) point out some of these relationships. There is a certain connection between biological sex and masculine and feminine gender. For instance, Mann (‘man’), Sohn (‘son’), and Vater (‘father’) are assigned masculine gender, whereas Frau (‘woman’), Tochter (‘daughter’), Mutter (‘mother’) are assigned feminine gender. German also has formal rules associated with the derivational morphology of the noun. Nouns ending in -ig, -ich, -lich, -ing, and -ling are masculine. Nouns ending in -ei, -heit, -keit, -schaft, -ung are feminine, while nouns ending in Table 6.4 Examples of gender agreement in extended noun phrases in Norwegian Demonstrative

Quantifier

Adjective

HEAD

en a.MASC

fin nice.MASC fine nice.DEF fint nice.NEUT fine nice.DEF fin nice.FEM fine nice.DEF

bil car(MASC) bilen car.SING.DEF.MASC hus house(NEUT) huset house.SING.DEF.NEUT jente girl.(FEM) jenta girl.SING.DEF.FEM

den the.MASC et a.NEUT det the.NEUT ei a.FEM den the.FEM

Possessive

min my.MASC

mitt my.NEUT

mi my.FEM

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

115

-chen and -lein are neuter. As in Norwegian, German determiners and adjectives agree in gender. The scope of the present study is limited to gender assignment in the use and non-use of the indefinite article in Norwegian. Hence, this brief presentation of the gender system in German is limited to the indefinite article system. Examples of how the German gender system works with indefinite articles are shown in Table 6.5. (For demonstratives and possessives, see Appendix 2.) Table 6.5 shows that the indefinite articles in German do not differentiate clearly between masculine and neuter gender, except for the accusative form where the article has three different gender agreement forms. Feminine never overlaps with masculine and neuter. Vietnamese is a classifier language in which nouns are classified according to properties such as ‘animacy vs. non-animacy, properties of geometrical shape, and biological properties like “having a tail’’ or “belonging to a flora’’’ (Dyvik, 1983: 3). Examples based on Rosén (2001: 28–29) serve as an illustration: (1) ba three ‘three horses’ (2) hai two ‘two balls’ (3) sáu six ‘six flowers’

con CL-for-animate-beings

ngu.҆ a horse

trái CL-for-round-things

banh ball

đóa CL-for-flowers

hoa flower

As we can see from examples (1), (2) and (3), the classifier has its own lexical content and is chosen on the basis of the property of the object the classifier refers to. The classifiers ‘do not show variation of a formal property (as is the case when, say, an adjective marks agreement in gender)’ (Corbett, 1991: 136). Hence, the classifier system in Vietnamese represents another type of nominal classification found in analytical languages, while nominal classification in terms of a gender system is more typical of synthetic languages. Table 6.5 The indefinite article system in German Singular

Nominative Accusative Genitive Dative

Masculine

Feminine

Neuter

ein Mann (‘a man’) einen Mann eines Mannes einem Mann

eine Frau (‘a woman’) eine Frau einer Frau einer Frau

ein Kind (‘a child’) ein Kind eines Kindes einem Kind

116

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

The classifiers in Vietnamese do not only have a classifying function, but also an individualising function (Rosén, 2001: 28). The nouns in Vietnamese do not refer to individuals, but rather to concepts or qualities (Rosén, 2001: 29). The classifiers’ individualising function is to ‘pick out’ individuals from the qualities and make specific reference possible: (4) hai chó two dog ‘two types of dog’ (5) hai con chó two CL-for-animate-beings dog ‘two dogs’ In example (4), without the classifier con, the expression denotes ‘two types of dog’, not two individual dogs (Dyvik, 1987: 7). The classifier’s function is to refer to specific individuals, as in example (5), where the expression denotes two individual dogs (Dyvik, 1987: 6). Another important distinction between Norwegian and Vietnamese is that the former has an article system, while the latter does not. In turn, English has neither a grammatical gender system nor a classifier system. Like many languages, it does express natural gender in the sense that pronouns in English are gender-differentiated when the referent is human, but (most often) not when the referent is non-human. Crucially, English does not exhibit any of the lexical, morphological, or syntactic characteristics of a language having grammatical gender, and is consequently treated in this study as a genderless language.

Transfer and Second Language Acquisition From a theoretical point of view, there are two factors that make certain grammatical categories difficult to acquire, namely lack of transparency between form and meaning, and redundancy (DeKeyser, 2005: 3). Gender in Norwegian carries these characteristics to some degree. Hence, the acquisition of gender is notoriously difficult for second language learners (Hagen, 2000: 63). If gender in itself is so difficult for second language learners, it is interesting for instance to ask if gender assignment is easier for learners who have gender in their L1, compared to learners whose L1 does not have gender. The role of the native language is one among several expressions that refer to the phenomenon in focus, namely ‘the influence of a person’s knowledge of one language on that person’s knowledge or use of another language’ (Jarvis & Pavlenko, 2008: 1). The terms crosslinguistic influence, interference and transfer are terms that reflect different theoretical approaches to and understandings of the phenomenon (Ellis, 1994: 301; Jarvis & Pavlenko, 2008: 3). In the

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

117

literature, these terms have come to be used synonymously (Jarvis & Pavlenko, 2008: 3); in this chapter the term transfer is used. The current study acknowledges that transfer has come to be seen as an important factor in second language acquisition and that it is a complex phenomenon (Benson, 2002: 68). The theoretical assumption for learning in this study is that language is learned through general cognitive skills. It is assumed that learners can use all their knowledge, including their knowledge of the L1, and be active participants in their acquisition and production of a new language. Furthermore, this study adopts a usage-based perspective on language acquisition, i.e. that language is basically learned from usage (Ellis & Robinson, 2008: 5). This perspective acknowledges the impact of frequency in language acquisition, that frequent constructions have stronger mental representations (Ellis & Robinson, 2008: 13), or are more entrenched than less frequent constructions. According to Dahl (2004: 89), entrenchment is the idea that ‘the mental ‘trace’ of an experience becomes progressively strengthened each time an individual is exposed to it’. According to Arabski (2006: 12), learners’ knowledge of their L1 may ‘be an advantage if the two languages have features in correspondence, as there will be “positive transfer” (or “facilitation”)’. Whether correspondences lead to positive transfer may be governed by the learners’ subjective judgement of what they perceive as transferable. This has been pointed out especially by Kellerman (1977) who refers to the important notion of learners’ ‘psychotypology’. Their L1 knowledge may also be non-facilitative ‘where the patterns of the two languages do not coincide’ (Arabski, 2006: 12). This may lead to ‘negative transfer’, where L1 knowledge serves as an obstacle. Another situation involves cases where there is a zero relationship between the L1 and L2, for instance when the L2 has an article system (e.g. Norwegian) while the L1 does not (e.g. Vietnamese), or as in a study by Jarvis and Odlin (2000) where the L2 (English) has prepositions while the L1 (Finnish) generally relies on suffixes and postpositions to carry the same meanings and functions. Situations like this may lead to the omission of L2 function words that do not exist in the L1 because the learners disregard the relevance of the function words (Jarvis & Odlin, 2000: 550). This type of omission can be referred to as the transfer of an unoccupied position, where the absence of a function word in a particular position in the L1 is replicated in a corresponding position in the L2 (cf. Jarvis & Odlin, 2000). However, a zero relationship can also be beneficial because it does not typically lead to interference from the L1, as has been shown in Spinner and Juffs (2008) and Tokowicz and MacWhinney (2005) (Section ‘Earlier studies’). Nevertheless, the facilitative effects of crosslinguistic similarity tend to overshadow the beneficial effects of a zero relationship. For example, Palmberg (1977) and Ringbom (1987: 93–95) have pointed out that learners enjoy much greater advantages in acquiring English articles when their L1 has an article system similar to that of the L2, rather than when their L1 has no article system at all. Palmberg

118

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

(1977) found that L1 Finnish learners of English omit more articles than L1 Swedish learners of English, and this is explained by the fact that the article systems in Swedish and English are similar whereas Finnish does not have an article system per se. The present study deals with two general questions: (1) whether it is facilitative to have an L1 with gender when acquiring an L2 with gender, and (2) whether having an L1 with a gender system similar to the L2 facilitates the acquisition of gender. The German, Dutch and Spanish learners have gender systems in their L1s. So the task for them is to restructure an already existing set of categories. On the other hand, the English and Vietnamese learners are faced with the task of internalizing a new linguistic category, a whole new subsystem of the grammar, in their L2 that does not exist in their L1s. The question is whether the restructuring and internalization required of the learners with different L1s are of different degrees of difficulty. The German gender system is the most similar to the Norwegian system, while the Dutch and Spanish gender systems are more different. German learners of L2 Norwegian do not have to restructure their L1 gender category system; however, they do have to re-categorize some nouns that have different gender assignments between German and Norwegian. Dutch and Spanish learners have to restructure their gender system from a two-category system to a three-category system, and also need to re-categorize some individual nouns. Dutch learners re-categorize from common and neuter gender, and the Spanish learners from masculine and feminine, to the Norwegian system with masculine, feminine and neuter gender. That being said, all learners, regardless of their L1s, have the option of following the Bokmål system, which treats all feminine nouns as if they were masculine. Little use of the feminine can also be triggered by simplification, i.e. the elimination of redundant features, a process interlanguages share with pidgin and creole languages (see Trudgill, 1999). As mentioned, from a usage-based perspective, frequency of input is important. That is, frequent constructions become more entrenched than less frequent constructions. According to Trudgill (1999: 149), ‘Gender marking occurs with a very high degree of frequency indeed in those languages which have it, and is thus a feature with a very high degree of entrenchment’. The indefinite article in Norwegian is often obligatory, and according to Dirven and Verspoor (1998: 54), the indefinite article and the noun can be considered as a syntactic group, or according to Bybee (2008: 217), as a construction. Constructions like the indefinite article and a noun can be processed as a whole when this construction occurs with high frequency (Lieven & Tomasello, 2008: 174). This type of language processing is known as rote-learning or chunking (Bybee, 2008: 220; Mitchell & Martin, 1997: 6) and is recognized as an ability characteristic of both L1 and L2 learners (Bybee, 2008: 220). This type of language processing goes on without the person needing to analyse the input. It requires a large memory capacity, but

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

119

it contributes to fast and accurate acquisition (Mitchell & Martin, 1997: 7). The counterpart to rote-learning is language processing where the input is decomposed and analysed, which is more time-consuming and error-prone (Plunkett & Marchman, 1990: 1; Portin et al., 2008: 453). We can expect that the gender of nouns that are more frequent in the input will be acquired before the gender of less frequent nouns.

Other Factors The L1 is not the only factor that can influence L2 acquisition; many factors can affect it in a complex interaction. The ASK corpus provides personal data for each informant (text), such as L1, age, sex, home country, education, social contact with Norwegians, proficiency level (CEFR level), hours of lessons in Norwegian and years of residence in Norway. These data are factors that may influence language acquisition. A notable example can be seen in the German and Vietnamese groups having different profiles when it comes to CEFR level and years of residence in Norway (see section ‘Group profile for the L1 German group and the L1 Vietnamese group’). The current study is limited to controlling for these two factors. Both of them may have an explanatory value when it comes to the learners’ performance, and either or both could potentially confound transfer effects. For instance, if we observe different language performance by two learners with different L1s and different proficiency levels, it is not obvious whether it is transfer or proficiency level that can explain the differences in the language performance. The texts were given holistic ratings representing specific CEFR levels. The holistic rating for each text reflected the raters’ overall impression of its quality in relation to considerations included on multiple CEFR scales described in Carlsen (2012: 173): overall written production, reports and essays, creative writing, general linguistic range, vocabulary range, vocabulary control, grammatical accuracy, orthographic control, coherence and cohesion. Hence, the CEFR levels that the texts are placed at tell us something about the holistic proficiency based on the scales referred to above, at a certain point in time (when the informants took the test).

Earlier Studies The question whether it is facilitative or not to have a native language with gender when acquiring an L2 with gender has been addressed in several studies. Table 6.6 shows the selection of studies referred to in this chapter (including the current one). These studies all investigate L1 transfer by comparing learners who have an L1 with gender and learners who have an L1 without gender (except

Adults or children Number of participants

(N = 62)

Adults Dutch = 34 German = 9 Romance = 8 English = 11

Dutch German French Italian Portuguese Spanish Without English gender Type of study Comprehension: –Experiments –Grammaticality judgement tasks

Target language With Native language gender

Sabourin (2001)

Earlier studies

German Italian

Ellis et al. (2012)

(N = 104)

(N = 38)

Afrikaans English Comprehension: Comprehension: –Experiment(s) –Experiments –Picture naming task –Gender assignment task –Sentence completion task Adults Adults Dutch = 34 Italian = 6 German = 25 Afrikaans = 23 Romance = 21 English = 9 English = 24

English

Dutch German French Italian Spanish

Sabourin et al. (2006)

Table 6.6 Earlier studies and the current study

English

Spanish

(N = 2)

Adults Italian = 1 Turkish = 1

(N = 20)

Adults English = 20

Adults German = 100 Dutch = 100 Spanish = 100 Vietnamese = 100 English = 100 (N = 500)

Vietnamese English Production, written: –Test situation –Corpus study, cross section

Norwegian German Dutch Spanish

Tokowicz and Ragnhildstveit MacWhinney (2005)

Production, oral: Comprehension: –Interviews –Grammaticality –Longitud. study judgement –ERP

Turkish

German Italian

Spinner and Juffs (2008)

Current study

120 Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

121

Tokowicz & MacWhinney (2005)). However, they differ regarding target languages, native languages, number of participants and types of tasks. The types of tasks can be divided into comprehension and production, where the former elicit learners’ knowledge or understanding of gender, and the latter elicit learners’ ability to use gender in language production (oral or written). The comprehension tasks can further be divided according to whether they investigate explicit or implicit knowledge. For instance, explicit knowledge is investigated through a grammaticality judgement task where the informant is asked to judge the grammaticality of the stimulus. Such a task allows informants to rely on explicit knowledge while making judgements (Tokowicz & MacWhinney, 2005: 178). In turn, implicit knowledge is investigated through tools that measure unconscious or immediate reactions, including brain imaging methods such as Event Related Brain Potentials (ERP) (Tokowicz & MacWhinney, 2005: 178). Sabourin (2001) reports that adult L2 learners of Dutch who were native speakers of German or a Romance language (French, Italian, Portuguese or Spanish) (L1 with gender) performed better than L2 learners who were native speakers of English (L1 without gender) in two different grammaticality judgement tasks. This indicates that it can be facilitative to have an L1 with gender. The results of the study further showed that the German group performed better than the Romance group, which indicates that it is facilitative not only to have an L1 with a gender system, but to have one that is similar to the L2 gender system. Sabourin et al. (2006: 11–12) report that adult L2 learners of Dutch, irrespective of L1, perform well above chance in a gender assignment task. However, learners with German or a Romance language background performed better than learners with L1 English, and the German learners performed better than the Romance learners (Sabourin et al., 2006: 12). These results confirm Sabourin’s (2001) earlier findings that it can be facilitative to have an L1 with gender, and even more facilitative to have an L1 with a gender system that is similar to the L2 gender system. Ellis et al. (2012) found that adult Italian-speaking learners of German perform better than Afrikaans- and English-speaking learners in a picture naming task and a sentence completion task. Ellis et al. (2012: 17) say that ‘the results indicate that the acquisition of grammatical gender in an L2 is easier for learners whose L1 also expresses grammatical gender’. In a longitudinal study, Spinner and Juffs (2008) investigated how two adult L2 learners of German used gender in oral production, one learner with Italian and one with Turkish as L1. The results of this study showed that the presence of gender in Italian does not seem to facilitate the use of gender in L2 German. Overall, it seemed that the Turkish learner of German did better in the use of gender in oral production than the Italian learner of German, even though Turkish does not have gender.

122

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

In an investigation that made use of a grammaticality judgement task and ERP-measurements, Tokowicz and MacWhinney (2005) investigated both explicit and implicit knowledge and L1 transfer in relation to gender agreement. Their findings indicate that English learners of Spanish show high implicit sensitivity to grammatical gender (despite the absence of gender in their L1), but that their explicit judgements about which constructions have the correct gender assignment are barely better than chance. These results suggest that it is difficult for learners whose L1 does not have grammatical gender to establish explicit knowledge about which gender is associated with each noun in the L2.

Hypotheses Results from both Sabourin (2001) and Sabourin et al. (2006) suggest that it is facilitative to have a gender system in the L1 that is similar to the gender system in the target language. The first hypothesis is therefore formulated as a uni-directional hypothesis: Hypothesis 1: Learners whose L1 has a gender system that is the most similar to the gender system of Norwegian will perform more accurately than learners whose L1 has a gender system that is less similar to the gender system of Norwegian. In accordance with Hypothesis 1, the L1 German group is predicted to perform better than the L1 Dutch and Spanish groups, and the L1 Dutch group is predicted to perform better than the L1 Spanish group. Contradictory results are reported by the literature concerning the question of whether it is facilitative or not to have an L1 with gender when acquiring an L2 with gender. The second hypothesis is therefore formulated as a bi-directional hypothesis to avoid incorrect predictions (Field, 2009: 54–55). Hypothesis 2: Learners whose L1 has a gender system will perform differently from learners whose L1 does not have gender.

Method Group profile for the L1 German group and the L1 Vietnamese group As mentioned earlier, the present study focuses primarily on the L1 German group and L1 Vietnamese group. The informants in these two groups

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

123

are homogeneous with respect to the test situation, but they are heterogeneous with respect to other variables. Here we will look into essay assignments and topics, the number of words and median text length, as well as the proficiency level and the number of years in Norway. Information about the query strings used to find this information in ASK is provided in Appendix 3.

Essay assignments and topics The informants were given a range of essay assignments that were distributed across a range of different topics for the two groups (see Appendix 4, overlaps between the groups are marked in the tables).

Number of words and average text length The informants in the German group generally wrote longer texts than the informants in the Vietnamese group (see Table 6.7).

Proficiency level (CEFR) Table 6.8 shows that the informants in the German group were assessed to be at a higher proficiency level in general than the informants in the Vietnamese group.

Years of residence in Norway Table 6.9 shows that the informants in the Vietnamese group have generally lived for a longer period of time in Norway than the informants in the German group.

Search procedure In the following query strings, the German group is selected as the L1 (language = ‘tysk’), but all the L1 groups are alternately selected in the same query strings. Table 6.7 Number of words and median text length

Number of words Median number of words in each text

The German group

The Vietnamese group

29,128 270.5

25,480 240

Table 6.8 Proficiency level CEFR level

German

Vietnamese

A2 B1 B2 Total number of texts

11 69 20 100

55 45 0 100

124

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 6.9 Years of residence in Norway Years

German

Vietnamese

0–1 1–2 2–3 3–4 4–5 More than 5 Not reported

69 23 3 2 1 2 0

8 47 20 2 5 17 1

Indefinite articles were searched for individually using the Ord (‘word’) search field. First, all instances of the indefinite article en were searched for in the relevant subcorpora and were recorded not only with respect to correct or incorrect use, but also with respect to non-use (omitted article). Then, all instances of the indefinite article ei, and finally et, were searched for and quantified in the same manner as en. This is the query string (%c = case insensitive): ‘en’ %c :: language = ‘tysk’ & testlevel = ‘Språkprøven’. Since all feminine nouns in Norwegian can be masculine, feminine nouns used with the masculine article were recorded as correct. To account for omitted indefinite articles (in obligatory contexts in front of a noun) and misspelled indefinite articles, these were searched for using the Korreksjon (‘correction’) search field. This relied on the manually corrected version of the learner texts where omitted or misspelled articles were replaced with correct forms. This is the query string: [corr = ‘en’ %c] :: language = ‘tysk’ & testlevel = ‘Språkprøven’.

Recording of language data and personal data The raw data from the different queries contains the results from the query hits from Ragnhildstveit (2009), but the following analyses are new. The present analyses focus mainly on gender assignment using the gender agreement on indefinite articles as evidence for the assignment. Gender agreement in extended noun phrases will not be examined in the present study. In cases such as example (6), the use of the correct indefinite article will be counted as correct gender assignment even if the accompanying adjective is not in agreement with the controlling noun. (6) et åpen a.NEUT

øre open.MASC

ear(NEUT)

(s0065)

The query hits (the language data of use and non-use of the indefinite article) were recorded for each individual informant. ‘Native language’, ‘CEFR level’ and ‘Years of residence in Norway’ were also recorded for each individual.

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

125

Identifying L1 effects Despite substantial research on the nature of transfer, no generally accepted understanding of it exists. This might be due to the fact that transfer in itself is a very complex phenomenon (Odlin, 2003: 478) and that it is difficult to separate transfer from other factors that influence language acquisition (Ellis, 1994: 335). Additional reasons might be ‘inconsistencies or incompatibilities between the empirical methodologies of different transfer studies’ (Jarvis, 2000: 248–249). Jarvis (2000) put forward a unified framework for methodological rigor in transfer studies, which has three components (2000: 249). The first component is a theory-neutral working definition of transfer: L1 influence refers to any instance of learner data where a statistically significant correlation (or probability-based relation) is shown to exist between some feature of learners’ IL performance and their L1 background. (Jarvis, 2000: 252) The second component comprises three L1 effects in learners’ performance (Jarvis, 2000: 253): (1) intra-L1-group homogeneity in learners’ IL performance (2) inter-L1-group heterogeneity in learners’ IL performance (3) intra-L1-group congruity between learners’ L1 and IL performance. The third component is the requirement to control for variables that can obscure or be confounded with L1 effects. Such variables include age, educational background, language background, target language exposure, target language proficiency, and so forth (Jarvis, 2000: 260–261). The use of ASK makes it possible to carry out statistical analyses since ASK provides access to a relatively large amount of learners’ (written) performance data. It is to be emphasized that this study only looks at the first two effects of L1 transfer. Comparison between L1 performance and IL performance was not carried out in this study; hence, the third effect is not investigated. Language background is controlled for in the selection of L1 groups. In addition, two other variables will be investigated statistically by holding them constant: target language proficiency and years of residence in Norway.

Statistics The statistical tests used to identify differences between the L1 groups were (a) the non-parametric Kruskal-Wallis H test for more than two groups, and (b) the non-parametric Mann-Whitney U test for two groups. The general critical p-value was set to 0.05, a widely used p-value in second language research (Larson-Hall, 2010: 49). When the Kruskal-Wallis H test reports a significant difference, then the Mann-Whitney U test is used as a post hoc

126

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

test to determine where the differences lie. In these situations, a Bonferroni correction is used to avoid a Type I error, which involves falsely rejecting the null hypothesis (Field, 2009: 348). According to Gujord (2013: 285), the Kruskal-Wallis H test and the Mann-Whitney U test compare the group’s internal behaviour to the differences in behaviour between the groups. Consequently, a significant result indicates that the observations in one group are sufficiently similar, and sufficiently dissimilar from the observations in the other group, to claim the existence of L1 differences. Thus the Kruskal-Wallis and Mann-Whitney tests are appropriate for testing the data for the first two effects of L1 transfer (see above). For the Kruskal-Wallis H test, the H (X 2) and p-value will be reported, and for the Mann-Whitney U test the U, the Z and p-value will be reported. In addition to these test statistics, effect size (r) will be reported for each test. The effect size indicates how important a significant or a non-significant result is. According to Field (2009: 56), ‘An effect size is simply an objective and (usually) standardized measure of the magnitude of observed effect’. The effect size in the present study was always calculated after the MannWhitney U test by dividing the Z-value by the square root of the total number of informants (N) (Larson-Hall, 2010: 377–378). Effect size r < 0.1 = very small effect, r ≥ 0.1 = small effect, r ≥ 0.3 = medium effect, r ≥ 0.5 = large effect.

Analyses and Results This study carries out four different analyses. Analysis A: Article use and non-use by noun tokens, Analysis B: Article use and non-use by noun types, Analysis C: An overview of the number of noun types and Analysis D: Article use and non-use by shared noun tokens between the L1 groups.

Analysis A: Article use and non-use with respect to noun tokens Analysis A is a quantification of article use and non-use at the level of individual texts (Text-id) with respect to noun tokens. For example, in Text-id s1000, eight noun tokens were found (see examples 7–14), which represent seven noun types. The noun gård (‘farm’) occurred twice in this text, once with the correct article (example 11) and once with an incorrect article (example 12). (7) en plass (‘a place’) (8) en hybel (‘a bedsit’) (9) en jobb (‘a job’)

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

(10) (11) (12) (13) (14)

127

en bolig (‘a residential’) en gård (‘a farm’) et gård (‘a farm’) en hytte (‘a cabin’) et bofellesskap (‘a flatshare’)

To give another example, in Text-id s0131, five noun tokens were found (examples 15–19), which represent four noun tokens. In this case, the noun dag (‘day’) occurred twice, both times with the correct article. (15) en bok (‘a book’) (16) en dag (‘a day’) (17) en dag (‘a day’) (18) en dame (‘a lady’) (19) et brev (‘a letter’) Compound nouns were included in the analysis and were not merged with single nouns. For instance, vei (‘road’), blindvei (‘dead-end road’) and hovedvei (‘main road’) were treated as three different words because of their important semantic distinctions. Nevertheless, they all carry the same gender because, in Norwegian, the gender of compound nouns is determined by the gender of the word-final noun. The following three variables were tested statistically in an attempt to identify possible differences between the groups of learners: (1) Ratio of correct article use to total number of noun tokens (2) Ratio of total article use to total number of noun tokens (3) Ratio of correct article use to total number of articles used Analysis A includes all five L1 groups and tests Hypothesis 1: Learners whose L1 has a gender system that is the most similar to the gender system of Norwegian will perform more accurately than learners whose L1 has a gender system that is less similar to the gender system of Norwegian. Analysis A also tests Hypothesis 2: Learners whose L1 has a gender system will perform differently from learners whose L1 does not have gender.

Results Analysis A: Article use and non-use in relation to noun tokens See Table 6.10 for the raw data for Analysis A and Table 6.11 for statisitcs for Analysis A. While testing for differences in group means related to the ratio of correct article use to total number of noun tokens, the Kruskal-Wallis H test showed no significant difference between the learner groups (N = 488): H = 6.118, p = 0.191.

128

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 6.10 Raw data Analysis A L1 groups

Correct Incorrect Total article Article Errors total Tokens total article use article use use omissions (E = W + O) (T = A + O) (C) (W) (A = C + W) (O)

German Dutch Spanish Vietnamese English

648 607 493 445 624

66 90 72 21 74

714 697 565 466 698

14 17 28 50 19

80 107 100 71 93

728 714 593 516 717

Table 6.11 Group means and standard deviations Analysis A L1 groups

German Dutch Spanish Vietnamese English

Ratio of correct article use to total number of noun tokens

Ratio of total article use to total number of noun tokens

Ratio of correct article use to total number of articles used

Mean

Std. D.

Mean

Std. D.

Mean

Std. D.

88.5769 84.7676 82.3479 85.2949 87.4919

14.17295 18.16594 20.52187 22.01769 17.41790

97.9033 97.3420 95.0266 89.0452 97.0235

5.80390 6.76935 11.04616 20.43400 9.80660

90.5526 86.6663 86.7512 95.6750 90.0417

13.70405 16.55931 19.20155 10.73199 15.93561

While testing for group differences in the ratio of total article use to the total number of noun tokens, the Kruskal-Wallis H test did find a significant difference between the L1 groups (N = 488): H = 24.933, p < 0.001. The results of a series of Mann-Whitney U post hoc tests are presented in Table 6.12. Hypothesis 1 is uni-directional and predicts that the German learners will outperform the Dutch and Spanish learners, and also that the Dutch learners will outperform the Spanish learners. Because these predictions are uni-directional, the p-values for these comparisons are one-tailed and are marked with bold text in Tables 6.12 and 6.13. The results for Hypothesis 1 show that none of the bold p-values in Table 6.12 are below the critical 0.005 p-value (the criterion p-value required by the Bonferroni correction), indicating that there is no significant difference between the learner groups whose L1s have gender systems. On the other hand, Hypothesis 2 is bi-directional and predicts that the German, Dutch, and Spanish groups will perform differently from the English and Vietnamese groups. The p-values for these comparisons are two-tailed and they do indeed show a significant difference between the German and Vietnamese groups and also between the Dutch and Vietnamese groups. However, they also show a significant difference

Gender A ssignment and L1 Transfer in Nor wegian L2 Lear ners

129

Table 6.12 Mann-Whitney U post hoc tests for the ratio of total article use to total number of noun tokens L1 groups

U

Z

p

r

GER(N = 98) 4654.500 −0.601 0.074 DUT(N = 98) GER(N = 98) 4358.000 −1.695 0.045 SPA(N = 98) GER(N = 98) 3545.500 −3.943 måter] å 3SG.N be.PRS many.PL vogue.PL [C > way.PL] to være [C > vare] på sin egen helse be.INF [C > care] on one’s.M own.M health(M) ‘There are many ways to take care of one’s own health …’ (English, HN) Måte clause, complexity = 7

ta take.INF [C > på] … [C > on]

Results In this section, quantitative results are presented in the same sequence as the research questions above. They will be discussed in the next section. To what extent do learners and L1 Norwegians use PSCs at the two test levels, i.e. the lower-level language test (Språkprøven) and the higher-level language test (Høyere nivå)? As shown in Table 7.1a, out of 1400 learners, 496 use PSCs; 904 learners (64.6% of the learners) do not. More learners use PSCs in the higher-level test than in the lower level, respectively 307 (43.9% of the learners at the higher level) and 189 (27.0% of the learners at the lower level). Similar patterns are found in the L1 Norwegian part of the corpus although the differences between the test levels are smaller (Table 7.1b). Out Table 7.1a Present and missing prepositions in PSCs, number of learners. Total number of learners: 1400. Cor = learners with one or more only correct tokens. Inc = learners with one or more only incorrect tokens. Mix = learners with at least one correct and at least one incorrect token. None = learners with no tokens. Språkprøven

Høyere nivå

Both levels

Construction

Cor

Mix

Inc

None

Cor

Mix

Inc

None

Cor

Mix

Inc

None

Finite PSCs Måte PSC All PSCs

146 9 146

4 0 13

12 27 30

538 664 511

236 29 237

8 6 29

11 44 41

445 621 393

382 38 383

12 6 42

23 71 71

983 1285 904

Table 7.1b L1 Norwegian, number of L1 users. Total number of L1 users: 200. Cor = users with one or more only correct tokens. Inc = users with one or more only incorrect tokens. None = users with no tokens Språkprøven

Høyere nivå

Both levels

Construction

Cor

Inc

None

Cor

Inc

None

Cor

Inc

None

Finite PSCs Måte PSC All PSCs

32 4 35

0 0 0

68 96 65

48 4 50

0 2 2

52 93 48

80 8 85

0 2 2

120 189 113

Preposit ion Stranding in Nor wegian Lear ner L anguages

169

of 200 L1 Norwegians, 87 use PSCs, while 113 (56.5 %) do not. In the higherlevel test, 52 (of 100) individuals use the constructions, while the corresponding numbers at the lower level are 35 (of 100). We find more tokens of PSCs in the higher-level test than in the lowerlevel test for both learners and L1 Norwegians (Table 7.2a and 7.2b). In the learner texts, 64.5% of the PSCs are found in the higher level, while in the L1 Norwegian texts 59.3% are found in the higher level. The difference between test levels may partly be a task effect. However, most of the difference can be attributed to text length. The frequency of learner attempts per 1000 words is 1.337 at the lower level, and 1.397 at the

Table 7.2a Norwegian L2, present and missing prepositions in PSCs, number of tokens. Pres = cases of present prepositions. Miss = cases of missing prepositions. Att = attempts, i.e. sum of Pres and Miss Språkprøven Construction Relative clause Wh-clause Non-wh-fronting Wh-fronting Clefting Finite PSCs Måte PSC All PSCs

Pres 155 21 19 1 1 197 10 207

Miss 16 0 1 1 0 18 29 47

Høyere nivå Att 171 21 20 2 1 215 39 254

Pres 263 33 34 3 10 343 39 382

Miss 19 0 3 0 0 22 57 79

Both levels Att 282 33 37 3 10 365 96 461

Pres 418 54 53 4 11 540 49 589

Miss 35 0 4 1 0 40 86 126

Att 453 54 57 5 11 580 135 715

Table 7.2b L1 Norwegian, present and missing prepositions in PSCs, number of tokens. Pres = cases of present prepositions. Miss = cases of missing prepositions. Att = attempts, i.e. sum of Pres and Miss Språkprøven Construction Relative clause Wh-clause Non-wh-fronting Wh-fronting Clefting Finite PSCs Måte PSC All PSCs

Pres 19 5 12 5 3 44 4 48

Miss 0 0 0 0 0 0 0 0

Høyere nivå Att 19 5 12 5 3 44 4 48

Pres 44 7 12 0 1 64 4 68

Miss 0 0 0 0 0 0 2 2

Both levels Att 44 7 12 0 1 64 6 70

Pres 63 12 24 5 4 108 8 116

Miss 0 0 0 0 0 0 2 2

Att 63 12 24 5 4 108 10 118

170

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

higher level, whereas in the L1 Norwegian texts the frequency of use is 1.746 and 1.588, respectively.7 Unsurprisingly, finite PSCs were used by a higher number of learners, as well as L1 Norwegians, than måte PSCs at both test levels. But while the number of learners using finite PSCs is 57.4 % higher at the higher level than at the lower (255 compared to 162) the number of those using måte clauses is 119.4 % higher (79 compared to 36). However, the (relative) frequency of attempted finite clauses is actually higher at the lower level (1.126 per thousand words) than at the higher level (1.106 per thousand words), whereas the frequency of attempted måte clauses is 0.204 at the lower level and 0.291 at the higher level in the learner texts. For the L1 Norwegians, the same pattern shows up in the finite clauses; at the lower level it is 1.601, at the higher level it is 1.452. In the måte clauses it is different from the learners; at the lower level the relative frequency per 1000 words is 0.146 at the lower level and 0.136 at the higher level.

To what extent is the preposition present or missing in PSCs at the two test levels? As shown in Table 7.2a, out of a total of 715 attempted PSCs, there were 589 successful instances (82.4 %) and 126 examples of missing prepositions (17.6 %). The distributions are quite dissimilar for finite clauses and måte clauses: 93.1% of the attempts at PSCs in finite clauses are successful, while 6.9% are not; for måte clauses the numbers are 36.3% and 63.7%. The results in Tables 7.1a, 7.1b, 7.2a and 7.2b are only used descriptively. The following results are tested statistically. Because our sample is fairly small, it is acceptable to adopt an alpha level of 0.10. When we used α = 0.05, some of our results approached significance, leading to a conclusion that transfer effects are not present. In that case we are at risk of committing Type II errors (Jarvis, 2012: 7). We therefore choose α = 0.10 as our significance level. Larson-Hall (2010: 102) advocates such an increase in the alpha level.

Are attempted PSCs in the higher-level language test more complex than those in the lower-level language test? Table 7.3 shows that the median distance in terms of number of words between the locations of the nominal and the stranded preposition was 3 for both clause types and both levels. The difference between the levels is not significant (Mann–Whitney, p = 0.248). The mean word count distance is mostly around 3.80, with the exception of måte clauses at the lower level.

Preposit ion Stranding in Nor wegian Lear ner L anguages

171

Table 7.3 Complexity of attempted PSCs, number of words between complement and preposition (slot) (see details above) Språkprøven

Høyere nivå

Both levels

Construction

Mean

Median

Mean

Median

Mean

Median

Finite PSCs Måte PSC All PSCs

3.80 3.51 3.76

3 3 3

3.82 3.86 3.83

3 3 3

3.81 3.76 3.80

3 3 3

Is the preposition missing more often in long than in short attempted PSCs? As shown in Table 7.4, the word count distance is consistently higher in the unsuccessful attempts than in the successful attempts at using PSCs, in relation to both the mean and median values shown in the table. That is, the longer the stranding clause, the more likely it is that the preposition is missing. The difference in complexity between successful and failed use is clearly significant (Mann–Whitney, p < 0.001). Table 7.4 Complexity of successful and failed use of a preposition, number of words between complement and preposition (slot) (see details above). Av = average. Md = median Språkprøven

Construction Finite PSCs Måte PSC All PSCs

Høyere nivå

Present

Missing

Av 3.68 2.70 3.63

Av 5.17 3.79 4.32

Md 3 2.5 3

Md 4.5 3 4

Both levels

Present

Missing

Av 3.73 2.95 3.65

Av 5.27 4.49 4.71

Md 3 3 3

Md 5 4 4

Present

Missing

Av 3.71 2.90 3.64

Av 5.23 4.26 4.56

Md 3 3 3

Md 5 3.5 4

Is the complexity of attempted PSCs different between the L1 groups? As shown in Table 7.5, the Dutch L1 group has the highest median complexity with a word count distance of 4, whereas the L1 English group has the highest mean complexity of about 4.11. The median complexity is 3 for the lower-level test for the L1 English group while it is 3.5 for the higher-level test. Except for the values mentioned, all the other L1 groups have a median value of 3 at both test levels. The variation across L1 groups is not significant (Kruskal–Wallis, p = 0.645).

172

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 7.5 Complexity of attempted PSCs for the L1 groups, number of words between complement and preposition (slot) (see details above) and preposition (slot) Språkprøven

Høyere nivå

Both levels

L1 group

Mean

Median

Mean

Median

Mean

Median

English Dutch German Polish Russian BCS Spanish

4.11 3.66 3.89 3.63 3.87 3.77 3.37

3 4 3 3 3 3 3

4.12 4.06 3.66 3.59 3.64 3.90 3.68

3.5 4 3 3 3 3 3

4.11 3.93 3.75 3.60 3.72 3.86 3.55

3 4 3 3 3 3 3

Do learners with L1 English and learners from other L1 groups differ in the use of the finite PSCs with respect to correctness? In finite PSCs, the L1 English group differs from all the other L1 groups in the lower-level test in that it has the highest number of attempted PSCs, and in all of them the preposition is present (see Table 7.6). The difference is significant between the L1 English group and the other groups in the lower-level test (Mann–Whitney, p = 0.061). The difference is not significant with respect to the higher-level test, however (Mann–Whitney, p = 0.591), nor when the two tests are analysed together (Mann–Whitney, p = 0.392). In the higher-level test, it is the Bosnian-Croatian-Serbian group that has the highest number of Table 7.6 Present and missing prepositions in finite PSCs, number of tokens for the L1 groups Språkprøven L1 group

Dutch German Polish Russian BCS Spanish Non-Eng. English All L1s

Pres

24 32 28 16 29 29 158 39 197

Høyere nivå

Miss

Att

N

%

1 4 5 2 1 5 18 0 18

4.0 11.1 15.2 11.1 3.3 14.7 10.2 0.0 8.4

25 36 33 18 30 34 176 39 215

Pres

56 42 53 31 71 37 290 53 343

Both levels

Miss

Att

N

%

1 3 5 1 4 4 18 4 22

1.8 6.7 8.6 3.1 5.3 9.8 5.8 7.0 6.0

57 45 58 32 75 41 308 57 365

Pres

80 74 81 47 100 66 448 92 540

Miss

Att

N

%

2 7 10 3 5 9 36 4 40

2.4 8.6 11.0 6.0 4.8 12.0 7.4 4.2 6.9

82 81 91 50 105 75 484 96 580

Preposit ion Stranding in Nor wegian Lear ner L anguages

173

Table 7.7 Present and missing prepositions in måte PSCs, number of tokens for the L1 groups Språkprøven L1 group

Dutch German Polish Russian BCS Spanish Non-Eng. English All L1s

Pres

2 0 2 1 3 1 9 1 10

Høyere nivå

Miss

Att

N

%

5 2 3 4 2 6 22 7 29

71.4 100.0 60.0 80.0 40.0 85.7 71.0 87.5 74.4

7 2 5 5 5 7 31 8 39

Pres

4 4 8 5 7 7 35 4 39

Both levels

Miss

Att

N

%

3 10 4 5 8 12 42 15 57

42.9 71.4 33.3 50.0 53.3 63.2 54.5 78.9 59.4

7 14 12 10 15 19 77 19 96

Pres

6 4 10 6 10 8 44 5 49

Miss

Att

N

%

8 12 7 9 10 18 64 22 86

57.1 75.0 41.2 60.0 50.0 69.2 59.3 81.5 63.7

14 16 17 15 20 26 108 27 135

attempted PSCs as well as the highest number with the preposition present. The Polish group ranks second on attempts, while the Dutch group ranks second when it comes to PSCs with the preposition present. The English group has the same number of attempts as the Dutch group, but a lower number of successful attempts.

Do learners with L1 English and learners from other L1 groups differ in their use of måte clauses with respect to correctness? In måte clauses, the L1 English group is different from all the other groups in that it has the highest number of attempted måte PSCs as well as the highest number of failed attempts. The same pattern is found in the higher-level test, except that the Spanish group has the same number of attempts as the English (see Table 7.7). The difference between the L1 English group and the other groups approaches significance for the higher-level test (Mann– Whitney, p = 0.116) and the two tests analysed together (Mann–Whitney, p = 0.119), but this is not the case for the lower-level test (Mann–Whitney, exact significance = 0.725).

Do learners with L1 English differ in their use of the finite PSCs and måte PSCs with respect to correctness? The L1 English group has a much higher share of successful attempts in finite clauses than in måte clauses. Since the number of learners with a given L1 who have attempted both finite and måte PSCs is very low, we choose to analyse these data only for the group levels, i.e. based on the number of

174

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

correct and incorrect tokens of use. One should be aware of the limitations of this approach (not all observations being independent). For the English L1 group, the skewing in distribution is significant (two-tailed chi square) in the lower-level test (χ2 = 33.491), the higher-level test (χ2 =35.579) and the two levels analysed together (χ2 = 70.998). All L1 groups show a similar tendency, but none as clearly as the English L1 group. When summing up the results, we can note that PSCs are attempted by only 35.4 % of the learners, and that 27.4% of the learners who use PSCs do not have any instances of a missing preposition. Out of the 715 attempted PSCs in the learner texts, 82.4% are successful in that the preposition is used. These patterns show that PSCs are used by a minority of the learners only, and further that the constructions are fairly infrequent even in the texts that do contain PSCs. Pied piping constructions are virtually absent in the learner texts as well as in the L1 Norwegian texts (see Appendix). Low frequency applies not only to the learner groups, but to the L1 Norwegian group as well – both with respect to the number of texts containing PSCs and the frequency of the construction in individual texts. To a certain extent, the number of learners using PSCs mirrors the L1 Norwegian group. We also observe parallels between the learner group and the L1 Norwegian group when it comes to types of finite PSCs most frequently used, namely the relative clause construction, while the non-finite måte PSC represents a greater share of all the PSCs in the learner group than in the L1 Norwegian group. There is some difference between the test levels. We find more tokens in the higher than in the lower test level; furthermore, more learners use PSCs in the higher than in the lower test level. The same patterns apply to the L1 Norwegian group. As mentioned, the relative frequencies indicate that these differences may be attributed primarily to text length. Finally, the proportion of PSCs with the preposition present in learner texts is higher in the higherlevel than in the lower-level texts. This may indicate that overall proficiency affects the use of correctness in PSCs. This applies, not unexpectedly, to the learner groups only, with the exception of two tokens involving a missing preposition in the måte PSC in the L1 Norwegian texts. Although this is infrequent, it may indicate a process of change in the use of this construction in Norwegian. Although we do find differences between test levels, we do not find any clear correlation between the learners’ assessed CEFR proficiency levels and their (in)correct use of PSCs. When complexity is measured as median word count distance, the analysis shows no significant difference between test levels for either the finite clauses or the måte clauses. Importantly, though, complexity turns out to play a major role in the distribution of successful and failed use of the preposition. The longer the stranding clause is, the more likely it is that the preposition will be missing.

Preposit ion Stranding in Nor wegian Lear ner L anguages

175

Turning to differences between L1 groups, we find no significant difference with respect to complexity. With regard to correctness in finite PSCs, however, the L1 English group demonstrates more successful use than the other groups in the lower-level test, and this difference is significant. This pattern is, however, not found in the higher-level texts. As regards the nonfinite måte construction, the L1 English writers leave out the preposition more often than the other groups at the lower level; the difference is, however, not significant, and neither are differences at the higher level. The L1 English group stands out when we compare the use of finite and måte PSCs. Recall that except for English, none of these constructions are present in the L1s of the learners. English, on the other hand, has the finite PSCs, but not the non-finite måte PSC. The finite PSCs which are present in both English and Norwegian, respectively the source language and the target language, represent linguistic similarity in the two languages, whereas the Norwegian måte PSC has no stranding counterpart in English, hence it represents linguistic difference. In the lower-level test, we find that it is the L1 English group that uses both the finite and non-finite PSCs most frequently. But the most striking result is that in the constructions that represent similarity between the source language and the target language, all the tokens are correct – the stranded preposition is present – while in the construction that represents difference between the two languages, in all but one token, the preposition is missing. When we compare the use of these two different types of PSCs measuring correctness in the L1 English group, we find that the finite PSCs and the måte PSC differ significantly between the two test levels (cf. Table 7.8). In the picture drawn by this short summary, there is no doubt that the L1 English group stands out. Below, we will give a short presentation of the kinds of evidence that count as transfer effects. We will start our discussion with the findings that may serve as evidence for L1 transfer, and finally point to other factors that may influence the use of PSCs.

Discussion In the field of SLA today, there is a broad consensus that the question of transfer or crosslinguistic influence is central to the understanding of how an L2 is learned. But the studies conducted in this area show conflicting results and there is a high level of confusion concerning ‘when, where, in what form, and to what extent L1 influence will manifest itself in learners’ use or knowledge of a second language (L2)’ (Jarvis, 2000: 246). Jarvis confronts this situation by presenting his first version of a methodological framework for identifying L1 influence in L2 learning. This framework (2000) has had a great influence on more recent transfer studies. In 2010, Jarvis offered some important refinements to his framework, and we will

176

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 7.8 Present and missing preposition in finite clauses and måte clauses, number of tokens Finite clauses

Måte clauses

Språkprøven

Present

Missing

Present

Missing

χ2

English Dutch German Polish Russian BCS Spanish

39 24 32 28 16 29 29

0 1 4 5 2 1 5

1 2 0 2 1 3 1

7 5 2 3 4 2 6

33.5 12.2 5.6 2.9 6.4 3.4 11.5

Finite clauses

Måte clauses

Høyere nivå

Present

Missing

Present

Missing

χ2

English Dutch German Polish Russian BCS Spanish

53 56 42 53 31 71 37

4 1 3 5 1 4 4

4 4 4 8 5 7 7

15 3 10 4 5 8 12

35.6 11.6 22.4 3.4 10.1 20.9 16.3

discuss our findings in relation to the latest version. The framework includes a theory-neutral definition, an account of the types of comparison-based evidence that are required to argue for or against L1 transfer and a list of outside variables to be controlled for in any rigorous investigation of transfer (Jarvis, 2000: 249). The following working definition is proposed: L1 influence refers to any instance of learner data where a statistically significant correlation (or probability-based relation) is shown to exist between some feature of learners’ IL performance and their L1 background. (Jarvis, 2000: 252) Since the definition includes an indication of how instances of transfer are to be identified, it serves as a heuristic. The comparison-based evidence types are presented in a two-parameter classification scheme, in which the parameters are within versus between, and group versus language. The four evidence types are: (1) Intragroup homogeneity which entails comparison within a (language) group; the premise for identifying this evidence type is group-representative behaviour. (2) Intergroup heterogeneity entails comparison between (different language) groups; the premise for this evidence type is group-specific behaviour. (3) Cross-language congruity

Preposit ion Stranding in Nor wegian Lear ner L anguages

177

entails between-language comparison; the premise for this evidence type is source (language)-like behaviour. (4) Intralingual contrasts entails comparison within- (a recipient) language; the premise is source-stratified behaviour. The premises for the first two evidence types are group-based phenomena, while the premises for the latter two evidence types are source-based phenomena. The table below, reproduced from Jarvis (2010: 182) summarizes the four types of evidence, the comparisons they entail, and their individual and combined premises. Evidence

Comparison

Premise

Intragroup homogeneity

Within-group

Group-representative behaviour

Combined premise

Group-based phenomenon Intergroup heterogeneity Cross-language congruity Intralingual contrasts

Between-group

Group-specific behaviour

Between-language

Source-like behaviour Source-based phenomenon

Within-language

Source-stratified behaviour

Our present study is based on data that makes it possible to meet the requirements of three of the four evidence types in Jarvis’s framework. The third type of evidence, cross-language congruity, requires comparison between L1 performance and L2 performance by the same individuals. The ASK corpus does not contain L1 performance data except from L1 Norwegian, and as far as we know very few corpora have L1 performance data from learners. In addition to the lack of data for one evidence type, there is another characteristic in the learners’ performance data that could possibly have represented a problem in our study if we did not have L1 Norwegian performance data. This is connected to the fact that only a minority of the learners use PSCs in their L2 Norwegian performance, and although some use them, the constructions are generally infrequent in the individual texts. This applies to all L1 groups in our study. Importantly, though, it applies to the L1 Norwegian group as well. The limited use in the learner texts mirrors, to a great extent, the frequency patterns we find in the L1 Norwegian texts. Since we know of no former study of PSCs that is based on language use data, collected from a learner language corpus, we have no possibility to compare our frequency results with results from other studies in this respect. The frequency patterns are methodologically interesting for SLA studies more generally, not only for transfer studies, but in light of the notion of comparative fallacy: ‘the mistake of studying the systematic character of one language by comparing it to another’ (Bley-Vroman, 1983: 6). This fallacy has been discussed repeatedly in the field of SLA (Tenfjord et al., 2006). But comparing L1

178

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

target performance data with learner data may in fact be necessary in the analysis of learner language. In Nistov’s (2001) study of referential choice in narratives of L1 Turkish learners of Norwegian, she reviews Klein and Perdue’s (1997) discussion of a restricted referential system in the ‘basic variety’. Eager to avoid the comparative fallacy, they characterize the referential system in the learner languages as restricted. Nistov (2001) finds the same pattern as Klein and Perdue (1997) in learners’ production of L2 Norwegian, but it also shows up in the L1 Norwegians’ performance. Nistov (2010) concludes that it may be necessary to establish ‘what it is not’ to be able to establish ‘what it is’. Hence, if we did not compare the learner texts with those of the L1 Norwegians, we could have risked concluding that only a minority of the learners have acquired PSC. In our study, however, we have not identified lack of acquisition of PSCs; what we have identified is the learners’ use of these constructions which to a large extent conforms to the L1 Norwegians’ frequency pattern. The use of performance data from native speakers of the target language has been suggested by Jarvis (2000) as one of the possible types of between-group comparisons in the original framework. Consequently, the language use data that we base our L1 transfer research on are the data from the learners who use the constructions – with the preposition present or missing – as well as from the L1 Norwegians’ use of the constructions.

Identifying L1 Transfer In what follows we will compare the L1 English group’s performance with that of the other L1 groups, in relation to the evidence types in Jarvis’s 2010 framework. We start with the first two evidence types for finite PSCs, and conclude with a discussion of Jarvis’s fourth type of evidence and a concomitant comparison of the L1 English group’s use of both types of PSCs. The finite PSCs represent objective similarities between English and Norwegian, and the reader is reminded that this similarity between source and target language only exists for the English L1 group in the study. The måte PSC is present in Norwegian only and represents objective difference for all the L1 groups. For the lower-level test we found that in the texts of the L1 English learners, the preposition is present in all 39 tokens of finite PSCs, whereas in the higher-level test the preposition is missing in only four out of 57 tokens. This behaviour meets the requirement of intragroup homogeneity for the L1 English group. When we compare the behaviour of the L1 English group with the behaviour of the non-English L1s group, we find a significant difference with respect to correctness (preposition present) for the lower-level test. The difference is, however, not significant for the higher-level test. Accordingly, only the results of the lower-level test provide evidence of the second type – i.e. intergroup heterogeneity.

Preposit ion Stranding in Nor wegian Lear ner L anguages

179

Although the måte PSC is not present in any of the L1s represented in the study except for Norwegian, its inclusion in the present study nevertheless reveals important differences between language group behaviour in the use of the construction. It is the L1 English group that has the most tokens of attempted måte PSCs, as well as the most tokens where the preposition is missing. This pattern is found at both test levels, and as reported above, the difference between the L1 English group and the group of non-English L1s approaches significance in the higher-level test and in the two tests analysed together, but not in the lower-level test. The difference we find in the use of the måte PSC is interesting in the discussion of transfer effects and how transfer works, and we will come back to these results. Although we do not have the ideal L1 learner data, the hypothesis we test in our study is grounded in knowledge of the different L1s included in our study compared with Norwegian. By comparing the L1 English learners’ use of the finite PSCs and the måte PSC, we have the possibility to identify source-stratified behaviour. In English, the finite PSCs are present while the måte PSC is not. We find a significant difference between the use of the two PSC types. The contrast in the source language replicates itself in the L1 English learners’ use of the target language, a behaviour which meets the requirement of Jarvis’ fourth evidence type intralingual contrasts. Here it should be recalled, though, that the statistical method used (two-tailed chi square) may have limitations (see above). We analysed these data only for the group level and found that the unevenness in distribution is significant for both test levels as well as for the two levels analysed together. Jarvis (2010) emphasizes that the four types of evidence are interrelated and therefore they support one another, but also ‘they can sometimes be used to substitute for one another when it is impossible to collect data relevant to all four types of evidence’ (Jarvis, 2010: 182). The evidence of intralingual contrasts may be the strongest evidence of transfer effects in our study, because it demonstrates source-stratified behaviour at both test levels. Remember that in our between-group comparison we found evidence for intergroup heterogeneity, but only for the lower level, not for the higher-level test. Jarvis argues specifically for the evidence type of intralingual contrasts as an appropriate substitute for between-group comparisons (Jarvis, 2010: 182). In our study, we do not need it as a substitute, but as an additional evidence type that demonstrates transfer effects that remain hidden in the between-group comparison of the higher-level test. How may this be explained? When comparing the performance of the L1 English group with the other L1 groups, the picture is different for the two test levels. This may be attributed to the more advanced language abilities that the higher-level test measures. L1 transfer effects may be less beneficial for learners at a more advanced level of learning. Transfer effects can even function as a constraint on learning if there exists a subjective assumption of similarity that is different from the objective similarity. The L1 English

180

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

writers’ performance seems to support the assumption that similarities between the source and recipient languages are the driving force behind L1 transfer. Their L1 has very similar finite PSCs, and when this is noticed by the learners, it may lead them to conclude that PSCs in Norwegian are the same as in English, the result of what Jarvis and Pavlenko (2008: 176–182) and Ringbom and Jarvis (2009: 107) refer to as subjective similarity, perceived or assumed. Ringbom (1987: 60–61) additionally pointed out that established subjective similarities may obscure for learners the fact that there is something to learn (cf. Gass & Selinker, 2008: 127). Therefore, they may not pay attention to the stranded preposition in the Norwegian måte PSC. Another factor that may strengthen the assumed similarity between the source and recipient language is connected to variation in the use of this construction in Norwegian. As mentioned before, we observed a tendency for the learners to drop the preposition in the måte PSC. As reported in Table 7.1b, there are two tokens of the måte PSC with the preposition missing produced by L1 Norwegians. If the learner input contains similar variation, it may strengthen the learner’s assumption of similarity between the L1 and L2. Regarding the higher-level test, the gap between the English L1 group and the other groups has narrowed concerning their use of finite PSCs. This may be because the non-English L1 group at this more advanced stage has mastered the construction to a greater extent, which results in non-significant heterogeneity between the groups. This fact may obscure whatever transfer is still operating in the L1 English group as is evidenced in our study by Jarvis’ fourth evidence type, intralingual contrast. Regarding the måte PSC, the L1 English group has the highest number of tokens at both levels, but also the highest percentage of missing prepositions. This difference between the groups is significant in the higher-level test but not at the lower level. Thus, the non-English L1 groups appear to improve more quickly than the L1 English group concerning the måte SPC at the higher level. This suggests that the assumed subjective similarity obscures for the learner the fact that there are differences as well. The evidence of intralingual contrasts in our study strongly supports Jarvis’s recommendation that one use as many types of evidence as possible; the different types of evidence are interrelated and they may not only serve as substitutes for each other, but as evidence of a more comprehensive picture of the transfer phenomenon.

The Influence of Saliency, (In)Frequency and Complexity The main aim of this study has been to identify possible L1 influence. By using Jarvis’s methodological framework (2010) we have identified L1 influence in the recipient language of the L1 English group. Before concluding our

Preposit ion Stranding in Nor wegian Lear ner L anguages

181

study, we will give some attention to other factors that may explain some more general findings in the use of PSCs in our data. Although our study differs from earlier studies when it comes to L1 influence, our data correspond to more general empirical results in earlier studies in that missing prepositions (‘the null preposition phenomena’) decrease from the lower-level test to the higher level, the use of PSCs increases and PSCs are more frequent than pied piping constructions. One explanation of the fact that PSCs are acquired before pied piping is the one proposed by Bardovi-Harlig (1987). She suggested that salience may explain why the marked structure, PSCs, were acquired before the unmarked, pied piping. ‘The data for the two rules are unequally distributed. If salience is defined in terms of availability of data, we see that the two rules have unequal salience. […] the data available to the learner do not seem to be enough to trigger the acquisition of pied piping before stranding …’ (Bardovi-Harlig, 1987: 401). The central meaning of availability of data we read as frequency, i.e. stranding is a more frequent construction than pied piping in the input of the learners (cf. Gass & Selinker, 2008: 145), but Bardovi-Harlig does not claim that PSCs are frequent in English. In the present study, it is the stranding construction that is in focus, not the relation between the two constructions. We have argued, founded on the use of the construction in the L1 Norwegian data, that PSCs are fairly infrequent in Norwegian. Further, we have argued that the question of the frequency of a syntactic construction must be addressed by comparing it with other syntactic constructions in the actual language. Gass and Selinker (2008: 146) point out different ways in which increased salience can be brought about. ‘Among these is frequency of input (possibly at both ends – that is, highly frequent and highly infrequent items/structures)’. For Gass and Selinker (2008: 482) infrequency brings about saliency particularly at a relatively advanced stage of learning: Something which is very frequent in the input is likely to be noticed. On the other hand, particularly at more advanced stages of learning, stages at which expectations of language data are well established, something that is unusual because of its infrequency may stand out for a learner. Ellis (2002: 178) states that ‘frequency is not a sufficient explanation, otherwise we would never get beyond the definite article in our speech’, and he also pointed out that frequency effects cannot be considered in isolation (Gass & Mackey, 2002). Of course, it is not the infrequency itself that causes salience; it must in one way or another be connected to other characteristics of the input as well as general cognitive mechanisms. As already noted, when comparing the English L1 group with the nonEnglish L1 groups, the former benefits from similarities between finite PSCs in the source and target languages at the lower level but not in the higherlevel test. In the L1s of the non-English groups there is nothing that

182

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

facilitates the learning process of PSCs, neither for the finite nor the måte PSCs. Nevertheless, in the use of PSCs, the gap between the English L1 group and the non-English L1 groups has narrowed in the higher-level test. How can this fact be accounted for? We have observed that learners with no similar PSCs in their source language at a more advanced level are approaching the L1 English group in the use of the complex constructions of finite PSCs (significant intergroup heterogeneity is no longer in evidence in the higher-level test). The måte PSC, which none of the L1s has, is used with a much lower success rate by the L1 English group than by the others. This fact may also be interpreted as a result of how much attention is paid to the PSCs in the input. The L1 English learners at more advanced levels may pay less attention to PSCs because the assumed similarity between the source and target language obscures for them that there are differences as well (cf. Ringbom, 1987). The non-English L1 learners, on the other hand, may be more observant with respect to the input since there is nothing in their L1 that facilitates the learning process. The complexity measure we have adopted points to the operation of a general processing mechanism: the longer (more complex) the stranding clause is, the more likely it is that the preposition is missing. Although missing prepositions are found not only in the lower-level texts, but in the higherlevel texts as well, we find that the complexity of constructions where the preposition is missing is higher in the higher-level texts than in the lowerlevel texts. We also find that the share of missing prepositions is lower in the higher-level than in the lower-level texts. We use a simple measure of complexity of PSCs; nevertheless, the results are interesting. The difference in complexity between PSCs with and without prepositions is clearly significant. This is a result that shows that a heavy processing load influences the acquisition and use of PSCs. This factor may as well be one that contributes to the infrequency of the constructions, not only by the learners but the L1 Norwegians as well.

Conclusion In this study, we have focused on the typologically unusual construction of preposition stranding, present only in some Germanic languages, English and the Nordic languages in particular. The construction is fairly infrequent in Norwegian, and as shown in this study, it is not frequent in learner languages either. Although such constructions have been the subject of a number of SLA studies conducted in the context of universalist paradigms, results concerning transfer are inconclusive. As far as we have been able to ascertain, there has been no earlier study based on natural language use data – not unexpectedly, since the theoretical frameworks most commonly used are

Preposit ion Stranding in Nor wegian Lear ner L anguages

183

more interested in competence than performance. Another reason for the lack of studies based on natural language use data may of course be the need for large amounts of data to be able to extract the relevant constructions, which only learner corpora can offer. For many of the studies we have reported on, this was not an option since there was no corpus available. In this study, on the other hand, we have extracted data from the learner corpus ASK. We have been able to identify L1 transfer effects in PSCs, and we have been able to exemplify facilitating effects of L1 transfer as well as constraining ones. In our study, we have put into practice Jarvis’s (2010) refined methodological framework developed to address problems associated with conflicting results and confusion concerning ‘when, where, in what form, and to what extent L1 influence will manifest itself’ (Jarvis, 2000: 246), as well as his suggestion to use as many evidence types as possible.

Notes (1) In these examples, and many of those in our materials, there is a close semantic relation between the preposition and the verb (here tenke på ‘think of/about’). Some would therefore consider the verb and preposition to be a transitive lexical unit and regard the complement of the preposition as the object. However, there is firm syntactic evidence that the preposition and its complement form one constituent when the complement follows the preposition, i.e. a prepositional phrase. For instance, other constituents can intervene between the verb and the preposition (Nå tenker jeg aldri på henne [now think.PRS I never on her] ‘Now I never think of her’), and the whole prepositional phrase can be topicalized (På henne tenker jeg aldri [on her think. PRS I never] ‘I never think about her’). (2) In our corpus data, these PSCs tend to cluster around a few lexical collocations (e.g. sted å bo på ‘place to live in’, hyggelig å snakke med ‘nice to talk to’. These would have been interesting to study since lexical collocations may be assumed to facilitate acquisition of the constructions that they are used in. However, some of the corpus collocations appear to be reproduced from test assignments and are therefore not well suited for study. This is not the case for måte å … på. (3) In fact, the preposition på is sometimes omitted in måte PSCs in L1 Norwegian too, hence presumably in the learners’ input, and this may reflect ongoing change. As far as we know, there has been no research on this. (4) In addition, she also had child L2 learners. (5) R-pronouns are fundamentally locative adverbs (waar ‘where’, daar ‘there’), but are understood as pronouns in stranding constructions (waar ‘who, which, etc.’). These are quite different from what we find in Norwegian and English. (6) Some researchers prefer to restrict the notion of ‘construction’ to syntagmatically complex (i.e. polymorphemic) expressions, excluding monomorphemic expressions, but the issue is not directly relevant here. (7) The number of words in the learner texts is 190,960 for the lower level and 330,070 for the higher level. In the L1 Norwegian texts it is 27,490 for the lower level and 44,093 for the higher level. (8) The older relative pronouns hvem ‘who’ and hvilken ‘which’ may still be encountered, but such usage is stylistically strongly marked as formal and old-fashioned. It is probably extremely infrequent in learner input.

184

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

References Bardovi-Harlig, K. (1987) Markedness and salience in second language acquisition. Language Learning 37 (3), 385–407. Bley-Vroman, R. (1983) The comparative fallacy in interlanguage studies: The case of systematicity. Language Learning 33 (1), 1–17. Ellis, N.C. (2002) Frequency effects in language processing. Studies in Second Language Acquisition 24 (2), 143–188. Ellis, N.C. and Larsen-Freeman, D. (2006) Language emergence: Implications for applied linguistics – introduction to the special issue. Applied Linguistics 27 (4), 558–589. doi:10.1093/applin/aml028 Gass, S.M. and Mackey, A. (2002) Frequency effects and second language acquisition: a complex picture? Studies in Second Language Acquisition 24 (2), 249–260. Gass, S.M. and Selinker, L. (2008) Second Language Acquisition. An Introductory Course (3rd edn). New York and London: Routledge. Goldberg, A.E. (2003) Constructions: a new theoretical approach to language. TRENDS in Cognitive Science 7 (5), 219–224. Haspelmath, M. (2006) Against markedness (and what to replace it with). Journal of Linguistics 42 (1), 25–70. Hokari, T. and Wakabayashi, S. (2009) Null prepositions in Wh-questions and passives. Paper presented at the The 10th Generative Approaches to Second Language Acquisition Conference Sommerville. Jarvis, S. (1998) Conceptual Transfer in the Interlangual Lexicon. Bloomington, IN: Indiana University Linguistic Club Publications. Jarvis, S. (2000) Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning 50 (2), 245–309. Jarvis, S. (2010) Comparison-based and detection-based approaches to transfer research In L. Roberts, M. Howard, M. Ó Laoire and D. Singleton (eds) EUROSLA Yearbook 10: John Benjamins Publishing Company. Jarvis, S. (2012) The detection-based approach: an overview. In S. Jarvis and S.A. Crossley (eds) Approaching Language Transfer through Text Classification. Explorations in the Detection-based Approach (pp. 1–33). Bristol: Multilingual Matters. Jarvis, S. and Pavlenko, A. (2008) Crosslinguistic Influence in Language and Cognition. New York and London: Routledge. Kaplan, T. and Selinker, L. (1997) Book review article: Toward second language acquisition: A study of null-prep by E.C. Klein. Second Language Research 13, 170–186. Kellerman, E. (1977) Towards a characterization of the strategy of transfer in second language learning. Interlanguage Studies Bulletin 2, 58–145. Kellerman, E. (1983) Now you see it, now you don’t. In S.M. Gass and L. Selinker (eds) Language Transfer in Language Learning (pp. 112–134). Rowley, MA: Newbury House. Kellerman, E. (1985) Dative alternation and the analysis of data: A reply to Mazurekewich. Language Learning 35 (1), 91–101. Kinn, T. and Tenfjord, K. (2013) Missing prepositions: A report from an explorative corpus-based study. Paper presented at the Learner Corpus Research Conference, Solstrand, Norway. http://lcr2013.b.uib.no Klein, E.C. (1993) Towards Second Language Acquisition: A Study of Null-Prep. Dordrecht: Springer Science + Business Media B.V. Klein, E.C. (2001) (Mis)construing null prepositions in L2 intergrammars: A commentary and proposal. Second Language Research 17 (1), 37–70. Klein, E.C. and Casco, M. (1999) Optionality in English non-native grammars: Difference between L1 and L2 acquisition. Paper presented at the 23rd annual Boston University conference on language development, Sommerville, MA.

Preposit ion Stranding in Nor wegian Lear ner L anguages

185

Klein, W. and Perdue, C. (1997) The basic variety. Second Language Research 13 (4), 301–347. Larson-Hall, J. (2010) A Guide to Doing Statistics in Second Language Research Using SPSS. New York and London: Routledge. Law, P. (2006) Preposition stranding. In M. Everaert, H. v. Riemsdijk, R. Goedemans, and B. Hollebrandse (Series Ed.) The Blackwell Companion to Syntax: Vol. III (pp. 631–684). Oxford: Blackwell. Liceras, J. (1985) The role of intake in the determination of learners’ competence. In S.M. Gass and C. Madden (eds) Input in Second Language Acquisition (pp. 345–373). Rowley, MA: Newbury House. Mazurkewich, I. (1984a) Dative questions and markedness. In F.R. Eckman, H.B. Lawrence and N. Diane (eds) Universals of Second Language Acquisition (pp. 119–131). Rowley, Massachusetts: Newbury House Publishers, Inc. Mazurkewich, I. (1984b) The acquisition of the dative alternation by second language learners and linguistic theory. Language Learning 34 (1), 91–108. Mazurkewich, I. (1985) Syntactic markedness and language acquisition. Studies in Second Language Acquisition 7 (1), 15–35. Nistov, I. (2001) Referential choice in L2 narratives. A study of Turkish adolescent immigrants learning Norwegian. (Doctor Artium), University of Oslo, Oslo. Nistov, I. (2010) Med målspråket for auga – kontrastivt basert analyse på ny. In H. Johansen, A. Golden, J.E. Hagen and A.-K. Helland (eds) Systematisk, variert, men ikke tilfeldig. Antologi om norsk som andrespråk i anledning Kari Tenfjords 60-årsdag. Oslo: Novus Forlag. Ohba, H. (2003) Pied-piping and stranding in oblique relative clauses in Japanese EFL learners’ interlanguage grammars. In S. Wakabayashi (ed.) Generative Approaches to the Acquisition of English by Native Speakers of Japanese. Berlin: Mouton de Gruyter. Rezai, M.J. (2006) Preposition stranding and pied-piping in second language acquisition. Essex Graduate Student Papers in Language and Linguistics 8, 110–128. Ringbom, H. (1978) The influence of the mother tongue on the translation of lexical items. Interlanguage Studies Bulletin 3, 80–101. Ringbom, H. (1987) The Role of the First Language in Foreign Language Learning. Clevedon: Multilingual Matters. Ringbom, H. (2007) Cross-linguistic Similarity in Foreign Language Learning. Clevedon: Multilingual Matters. Sadighi, F., Parhizgar, M.R. and Saadat, M. (2004) Preposition pied-piping and preposition stranding constructions in the interlanguage grammar of Iranian EFL learners Asian EFL Journal. Retrieved from http://www.asian-efl-journal.com/Dec_04_FSandMP andMS.pdf Tenfjord, K., Hagen, J.E. and Johansen, H. (2006) The hows and whys of Coding categories in a learner corpus (or ‘how and why an error-tagged learner corpus is not ipso facto one big comparative fallacy’). Rivista di Psicolinguistica Applicata (RiPLA), VI(3), 198–208. van Buren, P. and Sharwood Smith, M. (1985) The acquisition of preposition stranding by second language learners and parameric variation. Second Language Research 1 (2), 18–46. van Riemsdijk, H. (1978) A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht: Foris. White, L. (1986) Markedness and parameter setting: Some implications for a theory of adult second language acquisition. In F. Eckman, E.J. Moravacik and J. Wirth (eds) Markedness (pp. 309–327). New York: Plenum Press. White, L. (1987) Markedness and second language acquisition. The question of transfer. Studies in Second Language Acquisition 9, 261–286. Wode, H. (1976) Developmental sequences in naturalistic second language acquisition. Working papers on Bilingualism.

186

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Appendix Pied piping in Norwegian Pied piping is not possible in Norwegian relativization since the language does not have any relative pronouns but uses a relative subordinator (som), which cannot be the complement of a preposition.8 For the other finite PSCs illustrated in the main text, Norwegian does have pied piping as a pragmatically marked and much less frequent alternative to stranding, as exemplified in (27) for wh-fronting. In the L1 Norwegian part of the ASK corpus, there are very few cases of pied piping (see below). (27) a matrix clause where the preposition is pied piped with its fronted wh-complement: Med hvem snakker du with who talk.PRS you(SG).NOM ‘With whom are you talking now?’

nå? now

The corpus was searched for examples of the six prepositions (see main text) followed by hvem ‘who’, hva ‘what’ or hvilken ‘which’ (including its inflectional forms hvilket and hvilke). The search string for Språkprøven was as shown in (28), those for Høyere nivå were identical except for stating that ‘testlevel = ‘Høyere nivå’’. (28) ‘i|på|til|for|med|av’ %c [lemma = ‘hvem|hva|hvilken’ %c] \ \\ :: language = ‘engelsk|nederlandsk|polsk|russisk|serbokroatisk|s pansk|tysk’ & testlevel = ‘Språkprøven’ This search and the corresponding search in Høyere nivå yielded 51 and 117 matches, respectively. In these sets of matches, those that were actually examples of pied piping were identified. Pied piping in the L1 Norwegian part of the corpus was searched for in the same way. Table 7.9 shows that 25 examples of pied piping were found in the L2 parts of the ASK corpus. Five instances of pied piping in Språkprøven and nine in Høyere nivå involved either på hvilken måte ‘in what way’ or i hvilken grad Table 7.9 Pied piping in the learner texts in ASK Pied piping construction

Språkprøven

Høyere nivå

Both levels

Wh-clause (marginal) Wh-fronting (marginal) Relative clause (incorrect) Sum

4 4 2 10

12 2 1 15

16 6 3 25

Preposit ion Stranding in Nor wegian Lear ner L anguages

187

‘to what degree’. In these cases, pied piping is a normal option in Norwegian – they are fairly lexicalized interrogative phrases. Apart from these expressions, then, there were only 11 instances of pied piping in the learner parts of the corpus. Of these, three were unsuccessful attempts at relative clauses involving the use of a preposition plus an interrogative clause-initially, where Norwegian requires stranding. In Norwegian, the simple hvordan ‘how’ is often preferred over på hvilken måte, and the L1 Norwegian texts in ASK had no examples of på hvilken måte. There were three instances of i hvilken grad in the lower-level test for L1 Norwegian, two in the higher-level test, and these five examples were all there was to find of pied piping in the L1 Norwegian texts.

8

Emotions Negotiated in L2 Texts: A Corpus Study of Written Production by Adult Learners on a Norwegian Test Anne Golden University of Oslo

Introduction1 Learning a new language has traditionally been seen as something mainly related to a learner’s linguistic enterprise. Hence, research has been centred on the acquisition of a new system. The new sound system, the new vocabulary, the new grammatical structures, and later on also the new use of the different language parts (new pragmatics), have been in focus. In recent years, with the growing influence of the sociocultural framework in language research in general, there has been an awareness of the learner in the language learning process. Second language learning is not only about the acquisition of a new set of grammatical, lexical and phonological forms, but as Pavlenko and Lantolf (2000: 155) claim, may also be seen as a struggle of concrete socially constituted and always situated beings to participate in the symbolically mediated lifeworld […] of another culture. These individuals have intentions, agency, affect, and above all histories and are frequently though not always known as people. Part of these beings’ (or people’s) struggle to participate in a new society consists of presenting themselves on official tests to document their language proficiency. In Norway, this is done through different versions of Norskprøven (‘the Norwegian test’), which is related to the Common European Framework of Reference for Languages (CEFR) scales. In these tests, one of the main 188

Emot ions Negot iated in L2 Te x t s

189

components is a written essay. The ASK Norwegian learner corpus consists of such essays – texts that were written for and received passing scores on the following two tests of Norwegian as a second language: Språkprøven and Test høyere nivå, as these test levels were called at the time of data collection. The question that will be explored in this chapter is how the test takers present themselves, or to be more specific, what kinds of emotions the adult learners consider appropriate to negotiate in essays written in a testing situation. Because native Norwegian raters would later evaluate these texts and decide if the learners’ use of the Norwegian language was good enough to pass the test at a certain level, and because the raters’ decision represented a high-stakes situation, the test takers probably wrote in a manner they had learned to consider appropriate for such situations. Learners’ linguistic skills are often difficult to distinguish from their writing ability and the content of their language use (Weigle, 2002, 2013), and it is thus possible that the ways they express emotion will affect how their writing is evaluated. Another question is how these expressions of emotion are related to the learners’ L1s. As the learners come from different cultural backgrounds and have different first languages, there is reason to believe that the kinds of emotions they choose to relate, as well as their way of referring to them, might be different. Since gender might also play a role (Newman et al., 2008), possible gender variation is also taken into consideration. Dewaele (2010) claims that emotions have mainly been studied in connection with language learning motivation in SLA, and he calls for more studies in this area because ‘emotions have received relatively little attention in the SLA literature’ (Dewaele, 2010: 20). He points to Scovel, who in 1978 had already stated that the role of affect in SLA is probably one of the least understood. Although Dewaele’s focus is related more to the affect that students experience while learning a new language, the negotiation of emotions related to different issues in the texts learners have to write will also add to this understanding. The topics given to test takers for the purpose of documenting their language proficiency often prompt learners to convey personal values or attitudes toward different aspects of life or to give accounts of activities and happenings that most migrants are involved in.2 Hence, expressions of emotion are very likely to occur. Kramsch (2009: 67) claims that ‘rational cognition, judgments, agency, and moral value, that are usually associated with the brain, could not exist without emotions, usually associated with the body, for it is emotions that guide us in our decisions on what to select from the onslaught of information we receive’. Furthermore, she underscores learners’ particular awareness of their emotions (2009: 66) when they are in the process of learning a new language. To find appropriate expressions of emotion in a test situation might therefore be demanding as this choice relates to hidden ideologies as well as genre expectations on two sides, both from the test takers and from the raters. Pavlenko and Driagina

190

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

(2007: 213) maintain that ‘the task of interpreting, communicating, and describing emotions in a second language (L2) is even harder [that in L1] because different languages have distinct emotion vocabularies and ways of expressing emotions.’ The emotion vocabulary used on a test while learners document their competence will therefore give an account not only of their linguistic competence, but also of the ways learners negotiate feelings in this special situation. This might be a valuable perspective for researchers investigating the production and rating of official language tests. In addition, the expressions used by different L1 groups might reveal potential influences of L1 emotion concepts on the test takers’ linguistic practices in Norwegian which will be of interest to language teachers.

Emotion Concepts In order to study expressions of emotion, it is necessary to define the object of study. Different studies seem to take for granted what should be considered expressions of emotion and thus delimitations are seldom discussed. In comparing the use of emotion terms in different languages, particularly terms used by learners of a new language, it is important to discuss the concept of emotion itself. Feelings and emotions are often used interchangeably, but Kramsch (2009: 68–69) follows Damasio (1999) in referring to EMOTIONS3 as unconscious patterns that come before (or independently) of any FEELING. Damasio argues that FEELINGS are mental experiences of body states. They arise as the brain interprets emotions, which are themselves purely physical signals of the body reacting to external stimuli. Damasio also distinguishes between primary and secondary emotions. HAPPINESS, SURPRISE, SADNESS, ANGER , FEAR and DISGUST are considered primary; they are unconscious representations of the body and are not related to any particular object in the external world. Over time, they become integrated into the body’s memory, e.g. patterns of neural activity that can be activated when the body encounters stimuli it has encountered before. Secondary emotions are changes in the body’s state (e.g. a racing heartbeat) and are caused by some external object or remembered or imagined objects. Feelings associated with secondary emotions are differentiations of the basic primary ones. For example, (a) IRRITATIONS and FRUSTRATION are differentiations of ANGER, (b) EXITEMENT and EUPHORIA are variations of HAPPINESS, (c) DEPRESSION and BOREDOM are forms of SADNESS, (d) PANIC and SHYNESS are aspects of FEAR and (e) lack of INTEREST or INDIFFERENCE are indicators of DISGUST. Social emotions, such as emotions that require the representation of the mental states of other people, are also included among secondary emotions (Damasio, 1999). These include emotions such as SYMPATHY, EMBARRASSMENT, SHAME, GUILT, PRIDE, EMPATHY, JEALOUSY, ENVY, GRATITUDE, ADMIRATION, INDIGNATION and CONTEMPT.

Emot ions Negot iated in L2 Te x t s

191

However, not all languages distinguish between feelings and emotions. In Norwegian, emosjon ‘emotion’ is rarely used in everyday language, and when it is, it is used as a synonym of følelse ‘feeling’ and typically in connection with the science of emotions (most often in psychology and related sciences). It might also be used as a characteristic of a person or situation, such as somebody or something being emosjonell ‘emotional’ or full of feelings. According to Wierzbicka (1999), the word emotion does not have a counterpart in every language, but all languages appear to have a word for the concept of FEEL. Therefore, this word is included as one of the semantic primitives in The Natural Semantic Metalanguage (NSM) developed by Wierzbicka and colleagues over several years (see Goddard, 2010; Wierzbicka, 2015). The activity of emotions therefore seems to be an integral part of the universal folk model of a person, which means that in all cultures, people attribute feelings to other people, as well as to themselves. This concept of FEEL probably includes what other languages define as the concept of EMOTION. Henceforth, the concepts of FEEL and EMOTION will not be distinguished as these concepts overlap in the language of the studied texts – i.e. Norwegian. Following the insights of the current literature on emotions, an attempt is made in this chapter to avoid confusing emotion concepts with emotion terms. Even though some emotion terms in two languages seem synonymous, the concepts behind those seemingly similar terms might not be identical. One challenge in comparing emotions is finding an appropriate metalanguage to describe them. Since the widespread use of English emotion labels often conceals the differences, these labels will easily be confused with the terms and even the concepts in English. Both Norwegian and English terms will be used in this chapter because emotion terms in Norwegian are the object of study and relate to the concepts in Norwegian, and English is the language used for communicating these concepts in this study. However, the reader needs to be aware that there are differences between the concepts in the two languages, even if the terms are similar, and the same goes for the concepts in the writers’ L1s. In line with Pavlenko (2008), emotion words are seen as ‘words that directly refer to particular affective states (“happy”, “angry”) or processes (“to worry”, “to rage”), and function to either describe (“she is sad”) or express them (“I feel sad”)’. According to the ‘embodied cultural prototype view’ presented by Kövecses (2000), which is a reconciliation of the ‘prototype view’ and ‘the social constructionist view’, emotion terms relate to emotion concepts, where the latter are prototypical scripts that are formed as a result of repeated experiences with emotions and talking about them (Kövecses, 1990, 2000; Lakoff, 1987; Pavlenko, 2005). This approach has been advanced as an alternative to a componential or feature-based view of emotion concepts (Pavlenko, 2008) and it is based on experimental studies. The repeated experiences with emotions lead to language, and ‘language guides the acquisition of

192

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

concepts, and concepts influence the interpretation of bodily states’ (Pavlenko, 2005: 80). Hence, there is reciprocity between experience, language and bodily state, and the formation of concepts occurs in a culture-specific way. [C]oncepts involve the ability to categorize events and phenomena in language- and culture-specific ways, […] and to respond to these phenomena in linguistically and culturally appropriate ways, comprehensible to other members of the speech community in question. (Pavlenko, 2005: 81) People with experience from one culture have formed emotion concepts according to their experiences and the language they have used to talk about them. Communicating emotions and talking about emotion-related phenomena in an L2 might therefore be influenced by their already established concepts as well as the vocabulary available to them in the new language. In this enterprise, the learners will rely on what they perceive or simply assume to be similar or different in the two languages (Jarvis, 1998; Ringbom, 2007), e.g. to what is referred to as subjective similarities and differences (Ringbom & Jarvis, 2009). The acquisition of emotion terms is critical to learning a new language. The study of the use of emotion terms might also provide us with information about conceptual transfer in the learning process. Pavlenko (2005: 85–86) underscores that the formation of L2-based concepts is a lengthy process, as learners have to learn to pay attention to distinctions not encoded in their L1 in order to be able ‘to determine the prototypicality of particular events or displays of emotions’.

Researching Emotions Emotion language may be categorized as expressive or descriptive. In English, exclamations such as wow, ouch, mmm are expressive and reveal emotions, but exclamations do not name them. Descriptive terms may name the emotion (as in the English nouns happiness, anger, guilt), but they may also belong to different parts of speech pointing to this feeling (ex. sad, disgusting, to mourn). Descriptive emotions are sometimes expressed more indirectly as there is ‘a well established feature of emotion language that it is highly figurative; that is, it is dominated by metaphorical and metonymic expressions’ (Kövecses, 2008: 380). Metaphorical expressions ‘denote various aspects of emotion concepts’ (Kövecses, 2000: 4) and, hence, indicate certain qualities of emotions but do not ‘name’ a particular emotion. In addition, there are somatic markers of emotions (also called ‘emotion-related words’; Pavlenko, 2008), which involve descriptions of bodily reactions that index or are associated with a particular emotion, like crying, singing and sweating. In this study, somatic markers will be referred to as representing the practice of emotions.

Emot ions Negot iated in L2 Te x t s

193

This study investigates the emotion vocabulary used by adult learners from different language backgrounds who have passed written tests of Norwegian. The focus is on the learners’ use of emotion terms – the vocabulary they use when they refer to something that has made them happy, sad, and so on. Thus, it is the statements and descriptions of emotions in the learners’ texts that are analysed. However, in a study dealing with the negotiation of emotions in texts, it is also necessary to delimit emotions from other cognitive activities such as evaluations, beliefs, and attitudes. Even if feelings are difficult or impossible to define, Wierzbicka (1999: 13) and Wierzbicka and Harkins (2001: 14) claim that ‘“ordinary people” generally assume that the way one feels can be described and that one can tell other people how one feels,’ and the way people do this can be reduced to four basic modes: (1) One can tell other people that one feels good or that one feels bad (e.g. I feel wonderful/awful). (2) One can tell people that one feels the way a person would feel in a certain situation and then identify, in one way or another, that prototypical situation (e.g. I feel like a motherless child/I feel lost/abandoned). (3) One can tell people what seems to be happening inside one’s body, e.g. bodily images (e.g. my heart is breaking/my heart is heavy). (4) One can link thought-related feeling and ‘felt’ bodily processes, e.g. link cry to something bad happening like being unhappy or afraid. Three categories of emotion expressions – namely ordinary descriptive, figurative and somatic – link well with the above modes proposed by Wierzbicka and her colleagues. The descriptive category encompasses mode 1, the figurative category modes 2 and 3, and the somatic category mode 4. However, in contrast to the modes proposed by Wierzbicka, these three categories also apply to emotions attributed to other people or to other situations.

Categorization of Emotions Within the ordinary descriptive emotion category, there are several emotion types in a language that point to different concepts. Emotion types are often regarded as more or less prototypical or basic (Kövecses, 2008). Basic emotions might refer to (a) emotions in a middle level of a vertical hierarchy of concepts (e.g. FEELING -> FEAR -> PANIC) and to (b) those emotions judged to be more prototypical (Kövecses, 2000: 3) in accordance with prototype theory (Lakoff, 1987; Rosch, 1975). Several attempts have been made to create catalogues of general and possible universal categories of emotions, and the criteria used for defining them vary according to the disciplines of the researchers and the purposes for assembling them. The list of emotions proposed by the linguist Kövecses includes the English basic emotions of ANGER, FEAR, SADNESS,

194

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

and JOY and less basic ones including ANNOYANCE, WRATH, RAGE, INDIGNA(for ANGER) and TERROR, FRIGHT and HORROR (for FEAR) (Kövecses, 2000: 3). The basic emotions in this list overlap with those proposed by the psychologists Frijda et al. (1995), where JOY is replaced by HAPPINESS. The researchers claim that these five emotions are present in the languages of 11 countries. The neuroscientist Damasio includes DISGUST as a primary (or universal) emotion in his list. Other researchers mention emotions somewhat overlapping with these, and the linguist Stefanowitsch (2006) claims that ANGER, FEAR, SADNESS, HAPPINESS and DISGUST are the basic emotions most often agreed upon. In psychology, the six basic emotions often mentioned are ANGER, FEAR, SADNESS, HAPPINESS, DISGUST and SURPRISE, which were proposed by Ekman and his colleagues in their cross-cultural study of 1972. Ekman (1999) explains that there are particular characteristics associated with each of these emotions, allowing them to be expressed in varying degrees. However, some researchers are sceptical of the notion of the basicness and universality of emotions, as the research does not discuss in depth what these concepts refer to. The existence of basic emotions is therefore questioned as is the universality of ‘emotion’ (Wierzbicka, 1999; Wierzbicka & Harkins, 2001). Kövecses (2000: 13) suggests that several of these emotions have various prototypical cognitive models associated with them – e.g. they have multiple prototypes, which makes the question of basicness even more complicated. Pavlenko (2005: 78) claims that, so far, there is no consensus as to the relations between bodily experience, words, and concepts related to emotions, nor of the appropriate methods for selecting and analysing emotion terms. Nonetheless, since there is a good deal of overlap between cultures in relation to the emotions that are considered to be common, basic or prototypical, these emotions are probably the most frequently referred to in daily life as well as in the language classroom; hence, the labels for these emotions are likely to be learned first even if the semantic ranges of such words might differ somewhat across languages. In some languages and cultures, they might refer to coarser categories of emotion, and in others they might be finer-grained (Pavlenko, 2005). In addition, some of the emotions referred to most frequently in a particular language also seem to be common in other languages and are often used as translation equivalents even if they do not really point to identical concepts. Where they differ, the categories may involve somewhat different sensations and perceptions, attributed to different causes, placed in different body parts and classified and evaluated in different ways (Pavlenko, 2005). Moreover, emotion terms are manifested through different parts of speech (e.g. through a verb such as to enjoy, a noun such as joy or an adjective such as joyful). Languages differ in terms of which parts of speech are available to cover a particular emotion concept and which parts of speech are used most frequently to refer to that emotion (e.g. through a noun such as Angst in German or the adjective afraid in English and redd in Norwegian). LOVE

TION

Emot ions Negot iated in L2 Te x t s

195

Accordingly, the particular conceptualization inherent in these parts of speech is evident. To bypass the problem of non-overlapping emotion concepts across languages, Mota and Santos (2015) used 19 emotion groups (each containing several different words and concepts) to classify emotions found in a wide range of Portuguese corpora. This study was inspired by previous work in Portuguese and English by Maia (1996). As an example of how the data were classified, the emotion group MEDO (FEAR) contained words corresponding to concepts such as COWARD, FEAR , DREAD, FRIGHT, FRIGHTENED, etc. An even broader categorization of emotion concepts is presented in Frijda et al. (1995), where emotions are assembled into groups or ranges such as ‘unspecified positive emotion’ (the JOY and HAPPINESS range), ‘unspecified negative emotion’ (the SADNESS range), ‘an emotion of strong affection’ (the LOVE range), ‘an emotion of threat’ (the FEAR range) and an ANGER range. This categorization system recognizes the differences in concepts across languages as well as the tendency to characterize emotions as positive or negative. However, a categorization into positive and negative terms does not constitute a dichotomy or even a gradation and there will be elements that are impossible to grade. Emotion terms are complex in meaning, and some emotions might be considered neutral – being positive and negative at the same time (e.g. å lengte [to long for] might be negative for the one that is longing, but positive for the one longed for). However, a general categorization into positive, negative and neutral emotions, in particular among the emotions considered prototypical, might be illustrative in the description of the use of emotions by second language learners and in particular in relation to gender, as research has indicated that there are differences: males tend to display more negative emotions and females more positive (Newman et al., 2008). From a learning point of view, the categorization of emotions into simple and complex terms is also interesting as there are differences in complexity in the formation of a word, even among common terms. This may have an impact on the recall as well as the use of certain word types, as challenges related to orthography might cause a learner not to use a complex word in a text that will be evaluated.

Research Questions The main questions explored in this study are the following: •

To what extent do adult learners reveal emotions in essays written in Norwegian in a test situation and how do they express them? Do their expressions of emotion reflect individual tendencies (resulting in a great deal of within-group variation)? ○

196

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Are the most frequent expressions of emotion in L1 Norwegian also favoured by the learners? What types of emotion vocabulary are present in the texts? Type 1: direct versus indirect (referring specifically to an emotion vs involving a metaphorical expression or somatic marker) Type 2: parts of speech (adjectives/adverbs, nouns, verbs) Type 3: positive or negative (or neuter). Type 4: complex or simple form

○

•

○

○ ○ ○

In addition, a more detailed analysis is performed for the texts written by three L1-groups: •

Are there any differences between the different L1 groups or similarities within one L1 group in their choice of emotions displayed in the texts and the emotion concepts used? If so, could this be related to the emotion concepts in their L1s and considered a form of transfer?

The hypothesis is that the learners’ display of emotions in texts written for a test is more an individual choice as there will be an autobiographic component in their affective socialization (Pavlenko, 2005: 82). This is certainly expected to be the case in the use of metaphoric expressions. However, it is anticipated that the learners will use some types of emotion expressions more often than others – especially those representing more basic types and simple forms. Gender differences are also anticipated, and I expect that learners with different L1s will exhibit a preference for emotion concepts that they perceive or assume to be similar to those in their L1s.

Data The data come from the ASK corpus, a Norwegian learner corpus consisting of texts written by learners from ten different language backgrounds as part of an official test of Norwegian (see Chapter 1 of this volume). As these texts are authentic texts, the topics are varied. The text prompt selected for this study is the following: Skriv en tekst om vennskap [Write a text about friendship), which henceforth will be referred to as Friendship, a prompt presented in the intermediate language test of Norwegian (or Språkprøven), which is designed for learners at approximately the B1 level on the CEFR writing scale. The length of the texts was supposed to be circa 200 words. The reason for selecting texts on Friendship is because this topic invites learners to use a variety of emotion expressions. The emotions expressed might be expected to be positive because friends are generally viewed in a positive light. However, like other human relations, friendships may end, which creates the context for expressing negative emotions or something in

Emot ions Negot iated in L2 Te x t s

197

between. The choice of this topic is also based on another study involving the ASK corpus (Golden, in progress), where metaphors encountered in the texts written on this topic were analysed and compared with texts written in response to a prompt about Norwegian nature. This study yielded a difference relating to the amount of both conceptual metaphors and metaphorical expressions used in the texts with these two topics. The reason for the difference is explained by the differing demands of these two text types and genres: the Friendship prompt calls for a more evaluative text, with the requirement that it also be reflective – hence, more metaphors were used – while the Nature prompt calls for a more narrative or descriptive text, which triggered less metaphorical language. As emotions are frequently expressed through metaphorical language, the current choice of topic seems promising. Texts written on this topic by 68 learners representing nine different L1s (Albanian, Bosnian-Croatian-Serbian, Dutch, English, German, Polish, Russian, Somali and Vietnamese) were extracted from the ASK. These texts have been analysed in order to explore the emotions negotiated in a test situation. The analysis also examines the number and types of emotions expressed. Table 8.1 shows the distribution of texts and words by L1 and gender. The texts from the three largest L1 groups (German, Russian and Bosnian-Croatian-Serbian) were selected for quantitative analysis in order to determine whether there are any differences between these L1 groups, and whether there are strong within-group tendencies in the emotions the learners refer to and the vocabulary they use to do so. The goal of this study is to investigate to what extent the writers’ choice of emotion terms reflects similarities and differences between emotion concepts in the target language (Norwegian) and their L1s. There were 45 texts written on Friendship by learners from these three L1 groups. Table 8.1 Number of texts, average number of words and gender of the participants in each language group L1

Number of texts

Total words

Average words PT4

Min/max words

Gender

Russian German BCS5 Albanian Polish Vietnamese Somali Dutch English Total

17 15 13 8 5 4 3 2 1 68

4307 4481 3846 2022 1471 861 724 553 193 18,458

253 299 296 253 294 215 241 277 193 271

162/383 182/455 138/475 183/303 173/333 177/279 223/259 162/391 193 138/475

16 F, 1 M 11 F, 4 M 9 F, 4 M 7 F, 1 M 5 F, 0 M 4 F, 0 M 2 F, 1 M 0 F, 2 M 1 F, 0 M 55 F, 13 M

198

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Method Since the total number of lemmas in the texts written on the topic of Friendship was 1352 lemmas, it was possible to manually choose potential emotion terms from a list of all the lemmas in the texts. The words belonging to these lemmas were displayed with their immediate contexts in a kwick-format available in ASK and the concordances were manually analysed in order to decide which items should be considered emotion terms. The expression of emotions falls into three types. (a) Words denoting the emotion itself (happy, anger, love). (b) Somatic markers of emotions, the doing of an emotion, e.g. a bodily reaction implicating an emotion (get red in the face, tremble, sing or cry). (c) Metaphorical expressions (be alone, have a good heart). In selecting words potentially belonging to category (a) a delimitation of emotion words had to be made. This involved deciding where to draw the line between words indexing emotions and words indexing other related mental activity such as evaluation or opinion formulation. Making such decisions was not without problems. Because a prototypical approach was taken to the task of categorizing these expressions, it was necessary to recognize fuzzy boundaries between categories (Taylor, 2003) and to include a large set of words during the initial selection. Subsequently, the emotion terms referred to most frequently were singled out. This included all terms that have been identified in past research as being basic or non-basic, on the one hand, and primary and secondary, on the other. This means that category (a) was divided into two subcategories: (a1) words typically denoting an emotion (i.e. prototypical terms) and (a2) words that are more peripheral, but still related to emotions in the given context. The latter subcategory also includes words describing situations or people on an emotional level (e.g. When a person or situation is described as behagelig [agreeable/pleasing/enjoyable], or when the writer reacts emotionally to a situation by saying å bry seg [to care/ to bother]), and words indicating a consequence of an emotion-filled experience (e.g. deilig [delicious/superb]). Subcategory (a2) therefore consists of what are called peripheral terms, and it overlaps somewhat with category (b) – somatic markers of emotions, which also involve consequences or reactions that implicate emotions. However, category (b) is more physically oriented as it describes bodily reactions, which is why it is described as involving the practicing of an emotion. Category (c) encompasses metaphorical expressions, as defined in conceptual metaphor theory in the work of Lakoff and Johnson (1980) and further developed by Lakoff (1993), Gibbs (1994), Lakoff and Johnson (1999) and Kövecses (2002). In this framework, a metaphor is seen as a cross-mapping of two domains: a source domain and a target domain. It is a way of conceiving one thing in terms of another. In order to identify metaphorical expressions

Emot ions Negot iated in L2 Te x t s

199

indexing emotions for the present study, I selected all lemmas from the previously described list of lemmas that were frequently used to refer to source domains in metaphorical expressions. This included lemmas referring to body parts and inner organs (hode [head], hjerte [heart]), family members (mamma [mum]), words related to weather (mørkt [dark], sol [sun]) and adverbs and prepositions often used in metaphorical expressions, (opp [up], ned [down]). These words were also displayed with their immediate context and manually analysed in order to determine whether the lemmas found through the search protocol were indeed part of an expression of emotion. When they were, they were categorized in accordance with the lemma(s) considered to be the core element(s) of the metaphorical expression.

Analysis The total numbers of emotion terms in these texts are displayed in Table 8.2. Most emotions are expressed through words denoting or related to emotions, both in numbers of lemmas (86 different) and in numbers of tokens (338 altogether). The prototypical emotion terms are more frequent (46 different lemmas appearing 198 times) than those more peripheral (40 different lemmas appearing 140 times). This gives an average of 5.0 emotion tokens per text (2.9 of them prototypical and 2.1 of them peripheral) and a frequency of 18.3 emotion tokens per 1000 words (10.7 prototypical and 7.6 peripheral) of this type. The more indirect ways of expressing an emotion are less used, particularly through somatic markers (an average of 0.7 per text and 2.5 per 1000 words). On average, metaphorical expressions are used more Table 8.2 Emotion terms used in the texts about Friendship and distributed into different categories

Different lemmas Tokens (total 18,458) Mean PT8 N = 68 Mean PTW9

Total emotion vocabulary (min–max)

Cat. A) Total terms denoting emotion (min–max)

Cat. A1) Prototypical terms6 (min–max)

Cat. A2) Peripheral terms (min–max)

Cat. B) Somatic markers (min–max)

Cat. C) Metaphorical expressions7 (min–max)

135 (0–17) 466 (0–23)

86 (0–13) 338 (0–16)

46 (0–10) 198 (0–14)

40 (0–7) 140 (0–7)

19 (0–4) 47 (0–4)

30 (0–4) 81 (0–5)

6.9

5.0

2.9

2.1

0.7

1.2

25.2 (0–55.6)

18.3 (0–55.6)

10.7 (0–37.0)

7.6 (0-20.8)

2.5 (0–17.2)

4.4 (0–16.4)

200

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

often than somatic markers (an average of 1.2 expressions per text and 4.4 expressions per 1000 words (PTW)). The most frequent lemmas denoting an emotion are the following, in decreasing order of frequency: glad [glad/happy] 39 adj, trist [sad] 25 adj, like [like/love] 25 verb, ensom [lonely] 17 adj, stole [trust] 14 verb, savne 10 [miss] verb, hyggelig [nice] 10 adj, kjærlighet [love] 9 noun, snill [kind] 9 adj, ærlig [honest] 8 adj. With the exception of ensom [lonely], these are all within the 2500 most frequent lemmas in the largest native Norwegian corpus NoWaC10 and should therefore be considered part of everyday vocabulary.

Parts-of-speech Emotions in Norwegian are mainly expressed through adjectives, but nouns and verbs are also used. A search in NoWaC indicates that between 2/3 and 3/4 of the emotion words retrieved are adjectives.11 A breakdown of the different parts of speech used in words denoting an emotion in the texts about Friendship is presented in Table 8.3.12 As shown in Table 8.3, emotion-related terms in the text about Friendship mainly involve adjectives. 2/3 of the terms denoting emotion are adjectives or participles functioning as adjectives, in terms of tokens (66.6%) and a bit less in terms of different lemmas (59.8%). The peripheral terms have an even higher tendency to be adjectives than the prototypical terms, but the reason for this might be the inclusion of more adjectives in the peripheral emotion terms. As adjectives are the preferred means for expressing emotions in Norwegian, several adjectives might have been included simply because the Table 8.3 Prototypical and peripheral lemmas and tokens denoting emotions and used by the writers of the 68 texts – categorized into parts of speech Category A) Total denoting emotion (min–max) Lemmas tot Adjectives Participles Adj + part Verbs Nouns Tokens tot Adjectives Participles Adj + part Verbs Nouns

86 40 (0–8) 11 (0–2) = 51 (0–9) 17 (0–4) 18 (0–3) 338 202 (0–9) 23 (0–2) = 225 78 (0–8) 35 (0–6)

Category A) Distribution among POS 46.5% 12.8% 59.3% 19.8% 20.9% 59.8% 6.8% 66.6% 23.1% 10.4%

Category A1) Prototypical terms (min–max)

Category A2) Peripheral terms (min–max)

46 12 (0–5) 10 (0–1) = 22 (0–6) 12 (0–3) 12 (0–4) 198 102 (0–7) 14 (0–1) = 116 53 (0–7) 29 (0–4)

40 28 (0–6) 1 (0–1) = 29 (0–6) 5 (0–3) 6 (0–2) 140 100 (0–6) 9 (0–1) = 109 25 (0–3) 6 (0–2)

Emot ions Negot iated in L2 Te x t s

201

researcher interpreted them as implying an emotion. Of the ten most often used lemmas, six are adjectives (glad, trist, ensom, hyggelig, snill, ærlig), three are verbs (like, stole, savne) and one is a noun kjærlighet. The favouring of adjectives when displaying an emotion is hence in accordance with the use of emotion lemmas in NoWaC.

Values of emotion terms (positive, neuter and negative) Emotion terms are very often viewed as denoting either positive or negative emotions. However, some emotions might also be considered neutral because they are difficult to grade or because they denote a feeling in a relationship between people, as mentioned earlier. The distribution between the terms denoting a positive, neutral and negative feeling is presented in Table 8.4. The terms are also categorized in accordance with the writers’ gender. A further analysis of the use of positive emotion lemmas indicates a higher use of positive ones (45 positive emotion lemmas used 216 times Table 8.4 Positive, neutral and negative lemmas and tokens denoting emotions and occurring in the 68 texts – categorized by gender Tot words women = 15,132 Tot words men = 3326

Cat. A) Lemmas/ tokens denoting emotions (min–max)

Cat. A) Lemmas/ tokens denoting emotions women N = 55 (min–max)

78 86 Different lemmas 229 338 Tot tokens 4.2 5.0 Tokens PT 15.1 18.3 Tokens PTW Positive terms 42 (0–9) Different lemmas 45 (0–9) 216 (0–12) 191 (0–12) Total tokens 3.5 3.2 Tokens PT 12.8 11.7 Tokens PTW Neutral terms 3 (0–1) 6 (0–1) Different lemmas 5 (0–2) 8 (0–2) Total tokens 0.1 0.1 Tokens PT 0.3 0.4 Tokens PTW Negative terms 33 (0–7) Different lemmas 35 (0–7) 114 (0–9) 101 (0–9) Total tokens 1.8 1.7 Tokens PT 6.7 6.2 Tokens PTW

Cat. A) Share of emotion terms women

Cat. A) Lemmas/tokens denoting emotions men N = 13 (min–max)

Cat. A) Share of emotion terms men

25 41 3.1 12.3 55.0% 64.7%

15 (0–3) 25 (0–8) 1.9 7.5

60.0% 61.0%

3.8% 1.7%

3 (0–1) 3 (0–1) 0.2 0.9

12.0% 7.3%

41.3% 33.7%

7 (0–3) 13 (0–3) 1.0 3.9

28.0% 31.7%

202

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

compared to 35 negative emotion lemmas used 114 times). This gives an average of 3.2 positive tokens per text and 11.7 positive tokens per 1000 words compared to an average of 1.7 negative tokens per text and 6.2 negative tokens per 1000 words. In the ASK corpus there is an abundance of texts written by women (Friberg & Golden, 2014), and the texts dealing with Friendship are no exception. 55 of the texts are written by women and 13 by men (see Table 8.1), and this makes a comparison between genders difficult. We can only say that the 13 men in this study have used fewer emotion terms than the women. On average, the frequency of tokens denoting emotion is 4.2 per text for the 55 women compared to 3.1 for the 13 men. Expressed as frequency per 1000 words, this gives an average of 15.1 tokens per 1000 words for the women compared to 12.3 for the men. The positive as well as the negative tokens have a higher frequency among the women, but the share of the positive and neutral tokens are higher for the men, and their share of negative tokens are smaller.

Simple and complex forms Emotions may be expressed through simple root words or through words that are derived from other parts of speech. In Norwegian, it is common to convert vocabulary into other parts of speech through derivation. This is typically done through the addition of an easily recognisable suffix. Table 8.5 presents the distribution of emotion terms expressed through root forms and derived forms. Overall, root forms, which are simpler and consist most often of only one syllable, are used more often than derived forms, which include at least two syllables. However, this is not the case for peripheral terms; here, derived forms are used just as often as root forms. Prototypical emotion terms, on the other hand – in particular, the ones often considered to be basic emotion terms – are mainly root words in the preferred part of speech (usually adjectives). As these words are among the most frequently used words in Norwegian,14 this is probably the reason why the root forms appear most often in these contexts. Table 8.5 Prototypical and peripheral lemmas and tokens denoting emotions, used by the writers of the 68 texts, and categorized into root forms and derived forms

Tot lemmas/tokens Root lemmas/tokens Derived13 lemmas/tokens

Category A) Total denoting emotion (min–max)

Category A1) Prototypical terms (min–max)

Category A2) Peripheral terms (min–max)

86/338 53/227 33/111

46/198 36/159 10/39

40/140 17/68 23/72

Emot ions Negot iated in L2 Te x t s

203

Concepts of emotion Most adjectives in Norwegian have noun counterparts that denote the same concept.15 However, there is little overlap in the use of both adjectives and nouns to denote the same concepts in the texts written on the topic of Friendship. Some verbs also overlap semantically with words representing other parts of speech, but these are few in number in Norwegian, and mainly involve verbs of Latin or Greek origin. In the texts on Friendship, only six of the 40 emotion concepts expressed belong to different parts of speech. The concepts that include ten or more tokens are displayed in Table 8.6 with an approximate English translation. The eight most frequent concepts (encompassing ten lemmas) cover almost half of all the emotion tokens used in the texts (49.1%). Five of the eight concepts are positive concepts (as are more than 2/3 of the tokens they encompass). The use of adjectives is the most favoured way of expressing these concepts. Two of the tokens are complex in their forms, these being the derived adjectives hyggelig [nice] and ensom [lonely]. However, both of these adjectives are much more frequent in Norwegian than the root forms hygge (noun or verb) and ene (adverb or verb). Table 8.6 The emotion concepts with the highest frequency of occurrence in the texts dealing with Friendship Concepts used in the 68 texts (approximate English conceptual counterpart)

Tokens

Amount by each POS

POS

Value

GLEDE17 (JOY)

49

Adjective Noun Verb Adjective

Positive

25

39 glad 7 glede 3 glede 25 trist

TRISTHET (SADNESS) Å LIKE (LIKE/LOVE) ENSOMHET (LONELINESS) GOD (GOOD) Å STOLE (TRUST) HYGGE (NICE) SAVN (MISS) Total

25

25 like

17

17 ensom

Root/ derived form16

% of the tokens in total (N = 338)

14.5%

Negative

R R R R

Verb

Positive

R

7.4%

Adjective

Negative

D

5.0%

7.4%

16

16 god

Adjective

Positive

R

4.7%

14

14 stole

Verb

Positive

R

4.1%

10 hyggelig 10 savne

Adjective Verb 107 Adjectives 52 Verbs 7 Nouns

Positive Neutral 114 Positive 42 Negative 10 Neutral

D R 8R 2D

3.0% 3.0% 49.1%

10 10 166

204

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Individual use of Emotion Vocabulary One of the research questions is to what extent the number and types of emotion terms seem to reflect individual differences rather than group tendencies. In Table 8.7, the number of emotion words used by individuals is displayed, with information given about the numbers of texts representing varying numbers of emotion-word in types and tokens. All but one of the participants used emotion terms in their texts, 39 participants used five or more lemmas (45 used five or more tokens). As the texts are of different lengths, a more accurate method of comparison is to use the number of tokens per 1000 words, which indicates that only nine (13.2%) participants used fewer than 10 tokens, and 40 participants (58.8%) used 20 tokens or more per 1000 words. The test takers, as a group, favoured the use of adjectives when they used emotion terms in their texts. The distribution of the different parts of speech among the individuals, broken into sequential increments representing the numbers of texts where they were encountered, is presented in Table 8.8. Most of the emotions expressed in the texts are referred to with adjectives, and there are rather few test takers who did not use emotion adjectives at all. If we include the use of participles, which function as adjectives in these cases, there are only eight texts without an emotion word with an adjective function, and there are three texts in which emotions are expressed through adjectives ten or more times. Shown in PTW, more than

Table 8.7 Emotion vocabulary used in the 68 target texts, grouped in sequential increments of number of texts and tokens per 1000 words (PTW) 18,458 Total texts 68

Total amount of emotion vocabulary

(A) Total terms denoting emotion

(B) Somatic markers of emotion

(C) Metaphoric expressions18 of emotion

Test takers with emotion terms Test takers without emotion terms Test takers with 1 lemma/token 2–4 lemmas/tokens 5–9 lemmas/tokens 10 or more lemmas/tokens Test takers with fewer than 10 tokens PTW 10–19.9 tokens PTW 20–29.9 tokens PTW More than 30 tokens PTW

67 1

66 2

25 43

42 46

5/5 23/17 28/30 11/15

12/10 29/24 19/21 6/15

10/9 15/16 0 0

21/23 20/20 1/1 0

9 18 20 20

14 28 16 8

21 4 0 0

35 7 0 0

Emot ions Negot iated in L2 Te x t s

205

Table 8.8 Terms denoting emotion in the target texts, categorized into parts of speech and grouped in different increments of text and tokens per 1000 words (PTW)

Test takers with emotion terms Test takers without emotion terms Test takers with 1 lemma/token 2–4 lemmas/tokens 5–9 lemmas/tokens 10 or more lemmas/tokens Test takers with Fewer than 10 tokens PTW 10–19.9 tokens PTW 20–29.9 tokens PTW More than 30 tokens PTW

Adj

Part

Adj + Part

Nouns

Verbs

58 10

22 46

60 8

18 50

40 28

16/15 30/24 12/29 0

21/21 1/1 0 0

12/12 30/25 18/20 0/3

10/8 7/9 1/1 0

25/21 14/17 1/2 0

24 25 7 2

22 0 0 0

24 25 8 3

16 1 0 1

32 7 1 0

half of the participants (36 out of 68) use emotion adjectives (and participles) with a frequency of at least 10 tokens per 1000 words. The part of speech least used is nouns, with only 18 texts containing emotion nouns, and in almost half of these only one noun is used. Emotion verbs are also not used very frequently; slightly fewer than 60% of the texts (40 out of 68) contain emotion verbs; only two texts include five or more occurrences of emotion verbs.

Summing-up One of the aims of this study was to investigate whether learners of Norwegian, when attempting to present themselves in the best light on a language test, choose to express emotions, and if they do, how and how frequently they express these emotions. The analyses above showed that altogether there were 135 different lemmas (or expressions) used in these 68 texts written by the learners who had chosen the prompt about Friendship. These emotion terms were mainly terms directly denoting an emotion. There was a particularly high use of prototypical emotion terms, which have been described in past research as being basic or primary emotion terms. In the present study, 40 different prototypical emotion concepts were identified in texts written on Friendship. Of the eight concepts characterized most frequently as basic or primary by Frijda et al. (1995), Damasio (1999), Kövecses (2000), Stefanowitsch (2006) and Kramsch (2009), seven were used in the 68 texts. These include GLEDE [JOY/HAPPINESS], TRISTHET [SADNESS], ENSOMHET [LONELINESS], LYKKE [HAPPINESS], REDSEL [FEAR /DREAD], FRYKT [FEAR], OVERRASKELSE [SURPRISE] AND SINNE [ANGER]. Of these concepts, all but GLEDE [JOY/HAPPINESS] and

206

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

LYKKE [HAPPINESS] were expressed exclusively through adjectives. The most frequent ones (here defined as >10) were GLEDE [JOY/HAPPINESS], TRISTHET [SAD NESS], and ENSOMHET [LONELINESS]. The concept of DISGUST, which is also mentioned as one of the basic emotions by some researchers, was not found in these texts, and OVERRASKELSE [SURPRISE], and SINNE [ANGER] were mentioned only once. Of these eight most frequent concepts (encompassing ten lemmas), six are among the most frequent ones in NoWaC.19 In other words, the frequency of a concept in the language to be learned seems to play a role in learners’ choice of emotion terms. However, there are examples of non-equivalent concepts between Norwegian and other languages, such as English, and there are probably additional non-equivalent concepts between Norwegian and the learners’ L1s. For the present study, it may have been preferable to begin with a categorization of emotions similar to the one developed by Mota and Santos (2015). This would have involved grouping together lexemes such as elske and like in Norwegian since both may be translated as love in English, as well as redsel and frykt, which can both be translated as fear. However, this would have required a comparison between Norwegian and each of the individual L1s. The eight most frequently expressed emotion concepts in the target texts account for almost half of all emotion tokens. The 135 different lemmas (or expressions) identified as expressing emotions included 466 tokens. Although this amounts to only 2.5% of the total tokens in the 68 texts, the representation of emotions in the text is much higher than this because one emotion word (or one emotion expression) colours the entire sentence it occurs in – and often even the immediately following sentence(s). On the other hand, it is sometimes the case that more than one emotion word appears in the same sentence, so it is difficult to measure the overall emotional weight of a text without analysing it in detail. To answer the question of whether the use of emotion terms in the present study should be considered to be frequent, a comparison needs to be made with other studies or corpora, but to my knowledge there are no studies of texts written as part of a test nor any studies of texts written by learners that have analysed the use of emotion vocabulary. Bednarek (2008) has, however, conducted a study on the occurrence of emotion terms in a corpus consisting of 19.5 million words. In this study, the researcher compared what she referred to as emotion talk across four subcorpora representing four genres: causal conversation, news reportage, fiction and academic discourse, where all genres except conversation consisted of written texts. Her definition of emotion talk was similar to the one used in this study (category A), which includes all sorts of human behaviour that signal emotion with recourse to linguistic expressions that explicitly denote emotion ‘whether they refer to self or the other’ (Bednarek, 2008: 11). Bednarek’s analysis focused on 1060 words20 identified as emotion terms, and on their frequency in the four genres. The frequencies of these words were calculated as

Emot ions Negot iated in L2 Te x t s

207

occurrences per 1000 words (PTW), and the analysis rendered the following results, in decreasing order of frequency: fiction 10.8, conversation 9.6, news 6.1, and academic discourse 2.2. In the present study, the frequency of emotion words PTW was found to be 18.3, which is rather high compared to all four genres in Bednarek’s study. However, Bednarek excluded the terms sorry, want and like as they also appear in contexts that do not denote an emotion per se, such as where they are used to maintain ‘conventionally polite’ conversation (Bednarek, 2008: 32–33).21 Accordingly, the frequencies of emotion terms would have been higher in Bednarek’s study if these terms had been included. All but two of the test takers in the present study used terms denoting an emotion, but the frequencies across texts varied considerably. On the other hand, almost 80% of the test takers included two or more different emotion lemmas in their texts, and more than a quarter included five or more different lemmas. Measured in occurrences PTW, this means that approximately 85% of the participants used more than 10 emotion tokens PTW. The most frequent way of denoting an emotion is through the use of an adjective, and 58 of the participants used this part of speech to do so. Fortytwo of the participants used two or more adjective lemmas in their texts, and 12 of these used five or more adjective lemmas.22 Of the 10 most frequently used emotion terms, six of them were adjectives, three were verbs23 and one a noun. The emotion style of these texts might therefore be characterized as an adjectival emotion style. This also seems to be in line with the search performed on the texts in the NoWaC, albeit the NoWaC search involved a mix of genres. Bednarek found in her study that the conversation genre has a clear verbal-adjectival emotion style (six of the 10 most frequent were verbs, four were adjectives), academic discourse has a distinct nominal style (nine nouns), whereas both news reports (with five nouns, two verbs, three adjectives) and fiction (four nouns, four verbs and two adjectives) had a more varied style. The 68 writers in the present study expressed positive memories or attitudes towards friendship in their exam essays more often than negative ones, with the women using more positive terms than the men. However, there were only 13 men represented in the sample. Root word forms also seemed to be preferred, but as root word forms are also among the most frequent emotion terms, frequency might be the main reason. As for the more indirect expression of emotions, both somatic markers and metaphorical expressions were less frequently used. Somatic markers consist of lexemes indicating an emotion activity, such as bodily experiences like laughing or crying. These practices of emotion are, however, present in only 25 texts, and only 15 texts (22%) include two or more different lemmas. The number of somatic markers used includes only 19 different lemmas and 47 tokens. The most frequent lemmas were the verbs le [laugh] and gråte [cry], and the adjective gøy [fun] (used in expressions like ‘to have fun’). Each of these lemmas was used only six to seven times.

208

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Metaphorical expressions pointing to emotions were used by more than 60% of the participants (61.8%), but fewer than 1/3 (30.1%) used more than one expression in their texts. The core metaphorical elements in the expressions most often used were the adverbs alene [alone], hjerte [heart] and drømme [dream], each of which were used between eight and 16 times. The conclusion at this point is that emotion terms are frequently used by test takers when they write essays on topics such as Friendship. Terms directly denoting emotions are favoured, and adjectives are especially preferred. Positive terms are used more than negative terms. Indirect ways of expressing emotion – especially somatic markers – are used much less often. Metaphorical expressions were used by most of the test takers, but usually only once per essay. The results also seem to suggest that learners’ specific ways of practicing emotion and their figurative ways of expressing it are a matter of individual preference, at least in the context of learner language in a test situation.

Emotion Vocabulary in the Texts Written by Three L1 Groups So far, the analysis has included all 68 texts written by test takers from nine different L1 backgrounds (see Table 8.1). In order to explore whether there are any traces of crosslinguistic influence from the learners’ L1s in their expression of emotions and in the emotion vocabulary they use, the three L1 groups with the most participants have been selected, e.g. 15 participants with German as their L1, 13 participants with Bosnian-Croatian-Serbian (BCS) as their L1, and 17 participants with Russian as their L1. These texts have been compared in order to examine if it is possible to detect any differences between or similarities within these L1 groups (cf. Jarvis, 2000) regarding their use of emotion vocabulary. Table 8.9 presents the total numbers of emotion terms in these texts. Comparing the use of emotion terms between the three language groups reveals a difference: The Russian learners use more emotion vocabulary – just barely more per text (PT) than the other two groups – and their texts are the shortest among the three groups in terms of mean text length (see Table 8.1), but the difference is bigger per 1000 words (PTW). This difference is, however, not significant between the three L1 groups.24 The difference between the Russian and the German groups nevertheless seems considerable.25 The Russian learners’ use of emotion vocabulary entails 29.0 tokens and 14.9 different lemmas PTW compared to 25.5 tokens and 12.5 different lemmas for the BCS-group, and 20.5 and 12.3 for the German group. It is the use of tokens denoting emotions where most of the differences are found; the Russian learners have an average of 22.8 tokens consisting of 10.9 different lemmas denoting emotions, and both prototypical and peripheral terms are

GER 4481 words tot Different lemmas/ total tokens BCS 3846 words tot Different lemmas/ total tokens RUS 4307 words tot Different lemmas/ total tokens GER (N = 15) Mean different lemmas/total tokens PT BCS (N = 13) Mean different lemmas/ total tokens PT RUS (N = 17) Mean different lemmas/ total tokens PT GER (4481 words) Mean different lemmas/ total tokens PTW BCS (3846 words) Mean different lemmas/ total tokens PTW RUS (4307 words) Mean different lemmas/ total tokens PTW

Total tokens in texts

38 (0–14) 68 (0–19) 27 (0–8) 63 (1–11) 47 (0–13) 98 (1–16) 2.5 4.5 2.1 4.8 2.8 5.8 8.5 (0–30.8) 15.2 (0–41.9) 7.0 (3.0–22.8) 16.4 (3.0–25.6) 10.9 (0–55.6) 22.8 (5.4–55.6)

48 (2–12) 98 (2–15)

64 (1–14) 125 (1–17)

3.7 6.1

3.7 7.5

3.8 7.3

12.3 (0–37.4) 20.5 (0–50.7)

12.5 (9.1–38.3) 25.5 (10.7–43.1)

14.9 (5.4–40.0) 29.0 (5.4–55.6)

Cat. A) Terms denoting emotion total (min–max)

55 (0–19) 92 (0–23)

Total emotion vocabulary (min–max)

4.9 (0–37.0) 13.7 (0–37.0)

9.6 (3.0–14.4) 11.7 (0–19.1)

4.7 (0–19.8) 9.8 (0–30.8)

1.6 3.5

1.2 3.5

1.4 2.9

27 (0–10) 59 (0–10)

15 (0–6) 45 (1–8)

21 (0–9) 44 (0–14)

Cat. A1) Prototypical terms (min–max)

3.9 (0–14.4) 9.1 (0–20.8)

4.4 (0–11.4) 4.7 (0–14.5)

3.8 (0–11.7) 5.4 (0–11.7)

1.2 2.4

0.9 1.4

1.1 1.6

20 (0–3) 39 (0–7)

12 (0–2) 18 (0–3)

17 (0–5) 24 (0–5)

Cat. A2) Peripheral terms (min–max)

1.4 (0–10.9) 2.3 (0–10.9)

1.6 (0–12.9) 3.1 (0–17.2)

1.3 (0–6.4) 2.0 (0–11.3)

0.5 0.5

0.5 0.9

0.4 0.6

8 (0–4) 10 (0–3)

7 (0–3) 12 (0–4)

6 (0–3) 9 (0–3)

Cat. B) Somatic markers (min–max)

Table 8.9 Emotion terms used in the texts written about Friendship by three L1 groups, quantified into different categories

2.6 (0–10.9) 3.9 (0–13.1)

2.9 (0–9.6) 6.0 (0–9.6)

2.5 (0–6.6) 3.3 (0–6.6)

0.5 1.0

1.1 1.8

9 (0–4) 17 (0–7) 0.7 1.0

14 (0–3) 23 (0–4)

11 (0–3) 15 (0–3)

Cat. C) Metaphorical expressions (min–max)

Emot ions Negot iated in L2 Te x t s 209

210

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

used more often than in the other two L1 groups. However, the Russian group’s use of both somatic markers and metaphorical expressions to express emotions is less frequent than in the BCS-group. The BCS group exhibits the highest frequency of metaphorical expressions used to express emotions, and this is true in relation to both PT and PTW. Still, there were only 14 expressions used a total of 23 times, and nobody used more than three different expressions. The most frequently used expressions contain the metaphorical core hjerte [heart], which was used once by four participants, and the adjectives alene [alone] and rik [rich], each of which was used once by three learners. The eleven remaining expressions were used by only one learner apiece. The BCSs also exhibited the highest frequency of somatic markers (i.e. seven), but gråte [cry] and le [laugh] were the only emotion terms that were used by more than two learners. Only one or two learners used the five remaining somatic markers. The German group displayed the highest average text length, which was 299 words PT (see Table 8.1), and the lowest number of emotion terms in relation to both PT and PTW. The German group also exhibited the lowest use of words denoting an emotion, as well as the lowest use of somatic markers and metaphorical expressions. A comparison of different lemmas reveals that there are only two lemmas that were used by a minimum of 20% of the participants within each of the three groups.26 Only the adjective glad [happy] and the verb like [like] crossed this threshold (see Table 8.11 (later) for a display of the shared and the unique concepts used in the three groups and the discussion that follows). There was, however, a good deal of individual variation within each L1 group, but less so in the BCS-group (even though they showed the greatest variation in text length, see Table 8.1). When the cut-off is set at 25 emotion tokens PTW, which is roughly the mean for these three L1 groups, only two of the 15 German test takers (13.3%) exceeded this cut-off, whereas seven of the 13 BCS (53.8%) and eight of the 17 Russian text takers (47%) did. A breakdown of the parts of speech used to express emotions is shown in Table 8.10.

Adjectives For all three groups, adjectives were the preferred part of speech used to denote emotions. Adjectives constituted 44.7% of the different emotion lemmas for the Russians (the lowest) and 55.6% for the BCS group (the highest). Measured in tokens, the proportion of emotion terms that were adjectives was 47.1% for the Germans (the lowest) and 71.4% for the BCS group (the highest). This difference is not significant.27 Despite the lack of significance, which is due to sample size, we see a clear pattern here: The BCS group exhibits the highest proportion of adjectives compared to the other two groups regardless of how it is measured, and the difference is considerable between the Germans and the BCS group.28 When participles, which

Tot diff lemmas Mean PT Adjectives (A) Participles (P) A+P Verbs (V) Nouns (N) Tokens tot Mean PT Adjectives (A) Participles (P) A+P Verbs (V) Nouns (N)

38 2.5 18 (0–8) 5 (0–1) = 23 (0–9) 10 (0–8) 5 (0–2) 68 4.5 32 (0–8) 6 (0–1) = 38 (0–9) 23 (0–8) 7 (0–2)

GER Cat. A) Total denoting emotion (min–max)

7.1/47.1% 1.3/8.8% = 8.4/55.9% 5.1/33.8% 1.6/10.2%

4.0/47.4% 1.1/13.2% = 5.1/60.5% 2.2/26.3% 1.1/13.1% 15.2/100%

8.5/100%

GER per 1000 words/ share of emotion terms 27 2.1 15 (0–4) 3 (0–1) = 18 (0–5) 7 (0–6) 2 (0–1) 63 4.8 45 (1–7) 5 (0–1) = 50 (1–10) 11 (0–3) 2 (0–1)

BCS Cat. A) Total denoting emotion (min–max)

11.7/71.4% 1.3/7.9% = 13.0/79.3% 2.9/17.4% 0.5/3.2%

3.9/55.6% 0.8/11.1% = 4.7/66.7% 1.8/25.9% 0.5/7.4% 16.4/100%

7.0/100%

BCS per 1000 words/ share of emotion terms 47 2.8 21 (0–6) 3 (0–2) = 24 (0–6) 11 (0–4) 12 (0–6) 98 5.8 53 (0–8) 5 (0–2) = 58 (0–10) 26 (1–8) 14 (0–6)

RUS Cat. A) Total denoting emotion (min–max)

12.3/54.1% 1.2/5.1% = 13.4/59.2% 6.0/26.5% 3.3/14.3%

4.9/44.7% 0.7/6.4% = 5.6/51.1% 2.6/23.4% 2.8/25.5% 22.8/100%

10.9/100%

RUS per 1000 words/ share of emotion terms

Table 8.10 Prototypical and peripheral lemmas and tokens denoting emotions used by the writers from the three language groups, categorized into parts of speech

Emot ions Negot iated in L2 Te x t s 211

212

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

have an adjectival function, are included in the analysis, the proportion is even higher. Nevertheless, it is the Russian group that shows the highest use of adjectives PTW for referring to emotions, even though their proportion of adjectives is not as high as it is for the other two groups. Interestingly, the Russian test takers used more adjective emotion terms than each of the other groups, whether this is measured as different lemmas (4.9) or tokens (12.3). When it comes to the number of learners in each group who used adjectives more than once to denote an emotion, we find that this applies to eight of the 13 learners (53.3%) in the German group, nine out of 17 learners (52.9%) in the Russian group, and ten out of 13 learners (76.9 %) in the BCS group. So, here, too, the BCS group shows a greater use of emotion adjectives in comparison to the other two groups. In other words, the learners with BCS as their first language exhibit both the highest proportion of adjectives and the highest proportion of learners (around three-quarters of them) who used adjectives more than once, whereas the learners with Russian as their first language show the highest mean frequency of emotion adjectives altogether, even though only slightly more than half of the Russian learners used adjectives more than once.

Verbs The German group used a slightly higher proportion of verbs in total (33.8%) and in relation to different lemmas (26.3%). The BCS group produced the smallest proportion of verbs in relation to tokens, whereas the Russian group displayed the smallest proportion of verbs in relation to different lemmas. Nevertheless, the Russian group produced the highest number of emotion verbs PTW when quantified either as different lemmas (2.6) or tokens (6.0). Six out of the 15 learners (40%) in the German group used verbs more than once, and four of them used at least two different verbs. The verb like [like] was used the most (nine times by five learners), followed by the verb stole [trust], which was used four times by four learners. Five of the 17 learners (29.4 %) in the Russian group used verbs more than once and all of them used different verbs. Only three of the 13 learners (23.1%) in the BCS group used verbs more than once, and these were different verbs. Similar to the German group, the verb like [like] was also the preferred verb for the Russian and BCS groups. In summary, the learners with German as their first language showed the highest proportion of verbs and the highest proportion of learners using verbs more than once (though less than half of them did so). The learners with Russian as their first language, in turn, showed the highest mean frequency of emotion verbs altogether, but only slightly over a quarter of the learners in this group used verbs more than once. The verb like [like] was the most preferred emotion verb by all three groups.

Emot ions Negot iated in L2 Te x t s

213

Nouns The Russian group produced the highest proportion of nouns from the perspective of both different lemmas (25.5%) and tokens (14.3%), but they produced only six emotion nouns in total, which were used by only five learners in this group. Also, only three of the nouns were used by more than one learner. The Russians also produced more occurrences of nouns as emotion terms PTW than the other two groups, and this is true for both different lemmas (2.8) and tokens (3.3). However, only three learners in this group used nouns more than once, and all three used different nouns. The preferred noun was glede [happiness], used three times, once apiece by three learners. In the German group, only two learners used more than one noun, both of whom used different nouns. The preferred noun was glede [happiness], used three times, once apiece by three learners. In the BCS group, only two nouns (kjærlighet [love] and sorg [sorrow]) were used, once by each of two different learners. In sum, the frequency of nouns was low in all three groups, and there were few learners who used a noun more than once.

Concepts As some of the lemmas point to the same concepts, a comparison between the concepts used by at least 20% of all the participants in each group29 reveals that only two concepts, GLEDE [HAPPINESS/JOY] and LIKE [LIKE/LOVE], were expressed by all three groups (see Table 8.11). The concept of TRISTHET [SADNESS] was used by all three groups, but it crosses the 20% threshold for only the German and BCS groups. The concept of GODHET [GOODNESS] was used by only the BCS and Russian groups. HYGGE [NICE] was used by all three groups, but crosses the 20% threshold for only the Russian group. ENSOMHET [LONELINESS], STOLE [TRUST] and ÆRLIGHET [HONESTY] were used by all three groups, but by a minimum of 20% of the learners in only the German group. FORNØYDHET [CONTENT|HAPPINESS] and REDSEL [FEAR] were used by two groups, but by 20% of only the BCS group. LYKKE [HAPPINESS] and MORO [FUN] were unique to the Russian group – i.e. not used by the other two groups. Thus, altogether, 12 Norwegian emotion concepts were expressed by at least 20% of the learners in at least one L1 group. These will be referred to as the concepts in focus. To answer the question of whether the emotion terms used in Norwegian seem to be influenced by concepts in the learners’ L1s, a comparison of the 12 concepts in focus is performed. As Pavlenko (2008, 2009), claims there are three possible relationships between concepts in two languages: (a) two concepts (one in each language) may be similar or identical; (b) one language may have a concept that has no counterpart in the other language; and (c) two or more concepts across the two languages may be in

214

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Table 8.11 Concepts shared and unique for the three L1 group used by at least 20% of one of the groups

Shared by all groups

Shared by 2 groups Shared but not by 20% in more than one group Shared by 2 groups, but only 20% in 1 group Unique

CONCEPTS30

GER Tot 34 lemmas

BCS Tot 26 lemmas

RUS Tot 43 lemmas

GLEDE [HAPPINESS/JOY]

4 glad (A) 3 glede (N) 1 glede (V) 9 like (V) 5 trist (A)

14 glad (A) 1 glede (V) 4 like (V) 9 trist (A)

12 glad (A) 3 glede (N) 1 glede (V) 6 like (V) (3 trist (A))31

(1 hyggelig (A)) 4 ensom (A) 4 stole (V) 3 ærlig (A)

5 god (A) (1 hyggelig (A)) (2 ensom (A)) (1 stole (V)) (1 ærlig (A))

4 god (A) 5 hyggelig (A) (3 ensom (A)) (2 stole (V)) (1 ærlig (A))

3 fornøyd (P) 3 redd (A)

(1 fornøyd (P))

(1 redd (A))

LIKE [LIKE/LOVE] TRISTHET [SADNESS] GODHET [GOOD] HYGGE [NICE] ENSOMHET [LONELINESS] STOLE [TRUST] ÆRLIGHET [HONESTY] FORNØYDHET [CONTENT/HAPPINESS] REDSEL [FEAR] LYKKE [HAPPINESS] MORO [FUN]

5 lykkelig (A) 1 lykke (N) 4 morsom (A)

partial overlap. The assumption is that if a concept in Norwegian is perceived or assumed to be similar or identical to a concept in the learners’ L1 (type a), it has a greater chance of being learned early and used more often. If a concept is perceived or assumed to have no counterpart (type b), or if two concepts are perceived or assumed to be in partial overlap across languages (type c), the assumption is that such concepts will take longer to learn and will be expressed less frequently at early and intermediate stages of learning. In this study, there is no possibility to gauge the learners’ own perceptions or assumptions about the concepts in focus in a way that would allow for the measurement of subjective similarity and difference (Ringbom & Jarvis, 2009). Therefore, a method of translation-retranslation has been adopted (Golden, 2005, 2010). This method involves the translation of words from one language to another by experts (translators, linguists, etc.), and then these words are translated back into the original language by another set of experts. The Norwegian terms used by the learners for expressing the concepts in focus were, thus, translated into each of the three languages by linguists whose native languages were German, BCS, and Russian32 and then retranslated into Norwegian by another set of linguists who were also native speakers of these L1s. The first set of translators was instructed to provide the best translation (or several if necessary) of the Norwegian emotion terms presented to them, to add other parts of

Emot ions Negot iated in L2 Te x t s

215

speech to the same word family if relevant, and to indicate whether they considered each of the translated L1 lexemes to be frequently used; if they did not consider these lexemes to be frequent in the L1, they were asked to supply alternate terms that referred to the same concept. The list of emotion terms in each of the three languages was then presented to the second set of linguists, who were instructed to translate these terms into Norwegian and indicate if the presented term was frequently used or not. The translations and retranslations were then compared in order to determine whether the concepts in focus correspond with multiple terms – and possibly multiple concepts – in the learners’ L1s. The results of this comparison are shown in Table 8.12. The method proposed by Jarvis (2000, 2010) for confirming cases of crosslinguistic influence from learners’ L1s requires evidence involving within-L1-group similarities in relation to the phenomenon being investigated, differences between L1 groups, and similarities between learners’ L1 and their use of the L2.

Similarities within groups The similarities within each group are somewhat difficult to examine because the frequencies of the words are rather low, but the threshold for similarities in learners’ expression of emotion concepts has been set at 20% of the participants in each group. Accordingly, all groups meet this requirement for six of the 12 concepts in focus. For the German group, these concepts are GLEDE [HAPPINESS/JOY], LIKE [LIKE/LOVE], TRISTHET [SADNESS], ENSOMHET [LONELINESS], STOLE [TRUST] and ÆRLIGHET [HONESTY], for the BCS group, these are GLEDE [HAPPINESS/JOY], LIKE [LIKE/LOVE], TRISTHET [SADNESS], GODHET [GOOD NESS], FORNØYDHET [CONTENT|HAPPINESS] and REDSEL [FEAR], and for the Russian group, they include GLEDE [HAPPINESS/JOY], LIKE [LIKE/LOVE], GODHET [GOODNESS], HYGGE [NICE], LYKKE [HAPPINESS] and MORO [FUN].

Differences between the three L1 groups Norwegian and German belong to the Germanic family of languages while BCS and Russian belong to the Slavic language family. One would naturally assume that crosslinguistic similarities – particularly those relating to the lexicon – are greater within language families than between them, but false friends are also common within language families. From our translation data, we see that there are certain areas of overlap between the BCS and Russian terms, which makes it difficult to isolate possible cases of L1 influence unique to one or the other of these groups. We also do not know whether the participants know any additional languages, although there is a strong likelihood that the BCS learners know Russian. Nonetheless, German and the two Slavic languages are clearly different, which means that differences in the use of Norwegian by the German learners, on the one hand, and

Used by GER test takers: 9 like (V)

Used in Friendship: like (V)

LIKE [LIKE/LOVE]

Used by GER test takers: 4 glad (A) 3 glede (N) 1 glede (V) Translations mögen (V) gefallen (V)

GLEDE [HAPPINESS]

Used in Friendship: glad (A) glede (N) glede (V)

Translations radost (N) radostan (A)

Translations froh (A) Freude (N) (er)freuen (V)

любить (lyubit’) (V)

Used by RUS test takers: 6 like (V) Used by BCS test takers: 4 like (V)

Used by RUS test takers: 12 glad 3 glede (N) 1 glede (V) Translations нравиться (nravit’sja) (V)

счастливый (scˇastlivyj) (A) счастье (scˇast‘e) (N)

Translations pадостный (radostnyj) (A) pадоваться (radovat’cja) (V) pадость (radost’) (N)

RUS

svid¯ati se (V) vol(j)eti (V)

Translations dopadati se (V)

Used by BCS test takers: 14 glad (A) 1 glede (V)

radovati (se) (V) srec´an/sretan (A) srec´a (N)

BCS

GER

NORWEGIAN

Table 8.12 Concepts of emotions used in the texts about Friendship written by learners with German (GER), Bosnian-Croatian-Serbian (BCS) and Russian (RUS) as L1 and compared to concepts in these languages

216 Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Used in Friendship: hyggelig (A)

HYGGE [NICE]

Used by 1000).33 The results are presented in Table 8.13 along with information about the status of these concepts in the three L1s as well as the expressions of these concepts used by learners from these L1 groups in their texts on Friendship. Concepts that according to the translation represent a perceived or assumed similarity between Norwegian and the respective L1 are marked as ‘1 concept,’ whereas those representing perceived or assumed differences are marked ‘>1 concept.’

Interpretation of why a concept is used or not used The results show that the concepts

GLEDE [HAPPINESS/JOY]

and

LIKE [LIKE/

LOVE] were used frequently by all three L1-groups. Even though the Norwegian

concept LIKE [LIKE/LOVE] does not seem to correspond to only one German concept, nor to only one in the two Slavic languages, these two concepts (and their associated lexemes) occur with high frequency in Norwegian. This might explain why these concepts were used frequently by the learners in all three groups. The concept GOD [GOOD] also has a very high frequency in Norwegian and has one translation equivalent in all three languages, but it is only used by the Slavic learners. This is, however, not a typical emotion term in Norwegian. Whether it is used less frequently as an emotion term in German that in the Slavic languages is impossible to determine with the available data, but this is a plausible possibility given the greater similarities that exist between German and Norwegian, in comparison to those that exist between the Slavic languages and Norwegian. Of the concepts with less marked frequencies in Norwegian (less than 100,000 in NoWaC), the concepts of TRISTHET [SADNESS], ENSOMHET [LONELINESS], STOLE [TRUST] and ÆRLIGHET [HONESTY] were all used by Germans, and the translation results suggest that they are each assumed to be equivalent to a corresponding concept in German. The concepts of TRISTHET [SADNESS] and FORNØYDHET [CONTENT/HAPPINESS] were used by the BCS learners, and they are each assumed to be equivalent to a single concept in BCS. LYKKE [HAPPINESS] was used by the Russians, and it is assumed to correspond to a single concept in Russian. The concepts of HYGGE [NICE] and MORO [FUN] did not meet the requirement for within-group similarities among the Germans, and both of these concepts are assumed to be different from or not overlapping with concepts in German. Additionally, REDSEL [FEAR] is not expressed through an adjective in German as it is in Norwegian (close to 87% are adjectives in NoWaC), and this may account for part of the reason why this concept was not expressed by at least

222

Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

20% of the German learners. The concepts of HYGGE [NICE], STOLE [TRUST], LYKKE [HAPPINESS] and MORO [FUN] were not used by the required number of BCS learners, and all of these concepts are assumed to be different from or not overlapping with concepts in BCS. Finally, the concepts of TRISTHET [SADNESS], STOLE [TRUST], ÆRLIGHET [HONESTY] and REDSEL [FEAR] were not used by the required number of Russian learners, and, once again, all of these concepts are assumed to be different from or not overlapping with concepts in Russian.

Conclusion The use of 12 emotion concepts by three L1-groups (with a total of 45 participants) has been analysed. This has resulted in a discussion of emotion terms occurring in texts written on the topic of Friendship by three groups of learners. I have also discussed and presented evidence of the effects of perceived and assumed similarities and differences between the target language and learners’ L1s. I have interpreted the results as showing that perceived and assumed similarities and differences in emotion concepts between the target language and the three L1s account for most of the patterns of use and non-use of emotion concepts in the data. The present study has also found that concepts that are expressed at very high levels frequency in Norwegian (i.e. GLEDE [HAPPINESS/JOY] and LIKE [LIKE/LOVE]) also tend to be expressed with high levels of frequency by language learners even when these concepts do not correspond closely with concepts in the learners’ L1s. This means that out of the 36 cells showing the results in Table 8.13, only 10 do not conform to predictions based on crosslinguistic similarities/differences or input frequency. Of these, seven are counterexamples or cases that may involve a combination of multiple factors. The use and non-use of GOD [GOOD] in the data is, however, difficult to explain. On the one hand, this is a concept that is very frequently expressed in Norwegian, but on the other hand, it is not typically used as an emotion concept per se, which means that the frequency with which it is used to express emotion is probably rather low. There appears to be a perceived or assumed similarity between this concept in Norwegian and in all three L1s, but it is used as an emotion concept only by the BCS and Russian groups. It might be that this concept is not perceived or assumed to function as an emotion term by the German learners, whose L1 is similar to Norwegian, but is used by the BCS and Russian learners to express a positive emotion in cases where they have not learned or cannot remember a more precise emotion term. An important question is whether the present study provides evidence of conceptual transfer. Conceptual meaning has to do with the ways humans represent experience in their minds (Bylund & Athanasopoulos, 2014), and the conceptual meanings of emotions reflect how emotions are represented in the mind. When this varies from one language group to

Emot ions Negot iated in L2 Te x t s

223

another – as it does due to people’s repeated experiences with emotions and talking about them (Kövecses, 1990, 2000; Lakoff, 1987; Pavlenko, 2005) in different ways – L1-related representations of emotion might influence individuals’ expression of emotion in the L2. In addition, learners have experience with the language to be learned, including emotion terms representing emotion concepts. Some of these terms, concepts, and the mappings between them are perceived as similar to their pre-existing L1 mappings, while others are not. Perceived or assumed similarities might lead to the early use of such terms, whereas perceived or assumed differences might lead to a reluctance to use them. If learners of a new language (such as Norwegian) hesitate to use certain terms when they write about emotions in Norwegian, this might be because their starting experiences with the new ways of expressing emotions do not match the old ones. The mental images, scripts, scenarios, and so forth associated with these terms and expressions may be different (Jarvis, 2009; Lakoff, 1987) between the two languages, and learners might not have learned which distinctions they need to attend to when expressing concepts in the L2. In other words, they might not have acquired L2-based concepts yet (Pavlenko, 2005). L1 conceptual transfer is described as involving a ‘reliance on concepts encoded in L1 when using a later learned language, [and this] takes place in the lexicons of beginning and intermediate L2 learners and in particular of classroom learners who have not had an opportunity to be socialized into the target language community’ (Pavlenko, 2008). Insofar as this applies to the data of the current study, conceptual transfer seems to be a plausible explanation of some of the patterns observed. However, there might also be other explanations, and the methods used to eliminate alternative explanations need to be rigorous. Jarvis (2016) calls for three types of empirical evidence for confirming conceptual transfer: verbal evidence of language-specific conceptual planning processes, verbal evidence of language-specific conceptualization and non-verbal evidence that corroborates the verbal evidence. A corpus of open-ended written texts does not contain these types of evidence as they relate to the use and non-use of emotion concepts, and the translation-retranslation method also does not offer such evidence. However, the present data do show differences in language behaviour between three L1 groups that can largely be explained in terms of source-language-specific patterns at a level well above chance. Several alternative explanations can also be eliminated due to the ways that certain variables were held constant across participants: they produced their texts under similar testing conditions, they wrote their texts within the same limited timeframe, they have all received formal Norwegian instruction, and so forth. The interpretation that the learners’ L1 has had an effect on their use (or non-use) of emotion concepts is therefore likely and should be further investigated in future controlled studies where the same participants are asked to write texts (or tell stories) and to provide evidence of the

GER

Used Similarity in translation (1 concept)

Used Difference in translation (>1 concept)

Used Similarity in translation (1 concept)

Not used Similarity in translation (1 concept) Used 1 concepts)

Freq in NOR

VHF

VHF

MF

VHF

MHF

CONCEPTS

GLEDE [HAPPINESS/JOY]

LIKE [LIKE/LOVE]

TRISTHET [SADNESS]

GODHET [GOOD]

HYGGE [NICE]

Diff to GER

Diff to RUS

Used 1 concept)

Diff to GER and BCS

Used 1 concept) Used Similarity in translation (1 concept) Diff to RUS

No

Used Difference in translation (>1 concept)

No

Diff to GER and BCS

Diff to GER

No

Used Difference in translation (>1 concept)

No

Different use in RUS group vs other L1 groups?

RUS

Different use in BCS group vs other L1 groups?

Used Similarity in translation (1 concept)

Used Difference in translation (>1 concept) Used Difference in translation (>1 concept) Used Similarity in translation (1 concept)

BCS

Diff to BCS and RUS

Diff to RUS

No

No

Different use in GER group vs other L1 groups?

Table 8.13 Comparison of emotion concepts between Norwegian and German (GER), Bosnian-Croatian-Serbian (BCS) and Russian (RUS) in relation to the expression of these concepts in the texts on Friendship

224 Crosslinguist ic Inf luence and Dist inc t ive Pat ter ns of L anguage Lear ning

Used Similarity in translation (1 concept) Used Similarity in translation (1 concept)

Not used Similarity in translation (1 concept) Used 1 concept)

Used Similarity in translation (1 concept)

MLF

ENSOMHET [LONELINESS]

Diff to RUS

Diff to RUS

Diff to BCS

Diff to BCS

Diff to BCS and RUS

Diff to BCS and RUS

Diff to BCS and RUS

Not used No direct translation

Used Difference in translation (>1 concept) Not used No direct translation

Used